A Little Correction to "THE LOWDOWN ON HI-FI"

[Bigt]

In this part of the commentary, reader Aaron Drabbitt says:

This brings me to bi-wiring and bi-amping. Don't bother. It really doesn't make a difference that is in proportion to the cost of doing it.


Mr. Drabbitt has stated this wrong. Bi-wiring, the practice of splitting a single input terminal into two and using two wires for connection, has absolutely no advantage over a single wire.

Bi-amping, on the other hand, has some real advantages. Bi-amping is widely used in professional applications and has come into common use in home theaters, mostly whenever a subwoofer is included. The normal consumer speaker uses a passive crossover to split the audio band into frequency ranges that are appropriate for each driver. In other words, treble to the tweeter, bass to the woofer and so forth. This passive crossover is located after the amplifier and contains no active components. Designing and implementing a passive crossover is no easy task. Drivers are not perfect and combinations of drivers are not always ideal. Compensation for these imperfections are designed into the crossover. Describing what must be done would take a book (and there are books that do just that). Changes are difficult. Overall efficiency suffers with a passive crossover. And there is the possibility of crossover failure.

A bi-amped speaker uses an electronic crossover ahead of the amplifier, or sometimes within the amplifier itself (but before amplification). An electronic crossover does not have to contend with the speaker directly - the amplifier does that. This really simplify things. Some setups allow the use of computers for crossover design. With an electronic crossover, slopes and shapes can be easily changed. Things like level compensation, time delays and equalization are a button, dial or keystroke away. With an electronic crossover, frequency response can be made almost dead flat.

The downside to bi-amping is the cost of the electronic crossover, which can be quite expensive, and the extra amplifiers needed, along with more cabling. It's not for anyone on a budget. Is it worth it? There certainly can be a performance advantage over passive crossovers that is of value to the enthusiast but I don't think the average Joe will care.

My dream speaker, if ever I win the lottery, will be tri-amped, quad-amped with the subwoofer. It will be big and super-efficient. It will sound great. It will make no sense at all since I will never be able to play it full-out, but that don't matter at all to me.

[Hellbound]

In this part of the commentary, reader Aaron Drabbitt says:


Mr. Drabbitt has stated this wrong.

The downside to bi-amping is the cost of the electronic crossover, which can be quite expensive, and the extra amplifiers needed, along with more cabling. It's not for anyone on a budget. Is it worth it? There certainly can be a performance advantage over passive crossovers that is of value to the enthusiast but I don't think the average Joe will care.

This brings me to bi-wiring and bi-amping. Don't bother. It really doesn't make a difference that is in proportion to the cost of doing it.

So you agree with him, is that it?

[pipelineaudio]

If volume is a big concern, active crossovers can make a HUGE difference. Not forcing a speaker to speak of and an amplifier to amplify frequencies that are irrelevant to it can reach major gains, and in that case the average joe would hear it

Bi wiring though? I put that in the realm of shakti stones and magic IEC cables

[jj]

So you agree with him, is that it?

Ok, there is one thing that pros call bi-amping. That's using an active crossover and a different amplifier for different drivers.

This is a GOOD IDEA for the applications it's used for.

The other thing called biamping is using two amplifiers, without an active crossover, and simply fastening one driver, with passive crossover, to one amp, and the other crossover input (and driver) to another. This is usually silly.

[Bigt]

So you agree with him, is that it?

Yes, other than this small quibble, I agree with the author.

[pipelineaudio]

His emphasis on accoustics is spot on! But really getting good at reccomending or predicting results in that realm can be more than should be asked of the average joe, and unfortunately there are all manner of accoustic woo-ers out there

I disagree with him about cat-5 being the best for speaker wire. I like the regular old "lamp cord" or "zip-line" you can get from the hardware store, or grandma's old lamp. Its easier to solder and c-c-c-c-cheap

[jj]

His emphasis on accoustics is spot on! But really getting good at reccomending or predicting results in that realm can be more than should be asked of the average joe, and unfortunately there are all manner of accoustic woo-ers out there

I disagree with him about cat-5 being the best for speaker wire. I like the regular old "lamp cord" or "zip-line" you can get from the hardware store, or grandma's old lamp. Its easier to solder and c-c-c-c-cheap


Well the first rule of how to fix bad acoustics is: FIX THE ACOUSTICS

And I use Home Depot #14 (#12 doesn't fit in many speaker and amp connectors) zip cord.

[Jeff Wagg]

I have a home theater, and I was living out of state when it was built. When I got there to check out the construction (after dry wall, unfortunately), I found it wired with braided bell wire. The guy said, "Yeah, it costs three times as much, but it gives you 5% more sound than regular cable." And by regular cable he meant monster cable of some sort.

I made him eat that extra cost. And now that I've lived with it, it's a PITA to work with. I hate it. It's a stupid, brittle, hard to run, hard to hook up piece of woo crap. And I doubt that 5% part. In fact, I'll bet it's gives less "sound," whatever that means, because it's so damn hard to hook it up to the amp.

Zip wire. Accept no substitute.

[Hellbound]

Ok, there is one thing that pros call bi-amping. That's using an active crossover and a different amplifier for different drivers.

This is a GOOD IDEA for the applications it's used for.

The other thing called biamping is using two amplifiers, without an active crossover, and simply fastening one driver, with passive crossover, to one amp, and the other crossover input (and driver) to another. This is usually silly.

Well, my point was that neither said the techniques "Do nothing". The commentary stated it was "too little boost for the cost", while our OP stated that it was "expensive, but can make a difference". In other words, I objected to the OP calling the commentary "wrong", as this did not seem a mistake, but a simple difference of opinion on cost vs. benefits.

[jj]

Well, my point was that neither said the techniques "Do nothing". The commentary stated it was "too little boost for the cost", while our OP stated that it was "expensive, but can make a difference". In other words, I objected to the OP calling the commentary "wrong", as this did not seem a mistake, but a simple difference of opinion on cost vs. benefits.


The problem is with the fact that high-enders usually mean the "use two amps and no low-level crossover" version, and that pros mean "two amps with low-level crossovers".

The pro version is entirely sensible. The audiophile version is usually silly.

[jj]

I have a home theater, and I was living out of state when it was built. When I got there to check out the construction (after dry wall, unfortunately), I found it wired with braided bell wire. The guy said, "Yeah, it costs three times as much, but it gives you 5% more sound than regular cable." And by regular cable he meant monster cable of some sort.

I made him eat that extra cost. And now that I've lived with it, it's a PITA to work with. I hate it. It's a stupid, brittle, hard to run, hard to hook up piece of woo crap. And I doubt that 5% part. In fact, I'll bet it's gives less "sound," whatever that means, because it's so damn hard to hook it up to the amp.

Zip wire. Accept no substitute.


You should have made him replace it with #14 zip from Home Despot.

[jimlintott]

My car stereo is bi-amped. The deck provides the crossover and is fully adjustable for frequency and Q. It can do three way if I added another mono block amp for the sub. Right now it uses a four channel amp in a three channel mode. Right, left and subwoofer.

I wouldn't run zip cord in wall.

[Jeff Wagg]

I wouldn't run zip cord in wall.

And why not?

[scotth]

You guys are overlooking a real issue with large full-range speakers. Their moment-to-moment (instaneous) impedence can vary tremendously from their stated 'nominal' impedence. How does this change the calculations? During the attack (first 1/4 wave) of a bass drum hit, the impedence seen by the amp can be easily dip below .25 ohm for a few milliseconds (5-20ms is reasonable). That is because, for speakers on the very high end, nearly all of the impedance for the bass driver is reactance rather than resistance. If the driver isn't moving yet (prior to the bass pulse), its reactance is (near) zero.

If the amplifier has a beefy enough 'final' to output the full voltage dictated by the input signal, even into a fraction of an ohm, it is obvious the current will briefly go sky high (40V into .25 ohm gives 160A). You now have enough current to drop significant amount of voltage across the wire used to connect the speakers.

By doing nothing more than bi-wiring you do a few things...
1) If it is two runs of the same exact wire, at least you will cut the effective resistance of the wire in half, cutting the voltage lost across the wire in half.
2) If the bass driver is the specific cause of periodic drops to extremely low impedance, you'll only drop the voltage across the wire driving the bass driver. The wire feeding the midrange drivers and above will see the 'undistorted' signal as it appears at the output of the amp.

If the amp doesn't have finals that can support short bursts in excess of 100A, doing the passive bi-amping give similar benefits. I'll certainly agree though, active bi-amping (using a x-over before the amps) is more effective.

To be sure, the improvement in total performance is slight from bi-wiring and passive bi-amping, but it is certainly there. It can be measured with test equipment, and it can be heard. So, when and where does it make this (appearently mysterious) improvement happen?

Drums, drums, drums. It puts the 'snap' back in the drums. When one listens to live unamplified drums there is a distinct character to them. You feel the crack of a rim shot in the breastbone and teeth, for example. You feel a real 'kick' from the kick drum. This is something that is lost during playback in the vast vast majority of sound systems. It is also something that even casual listeners immediately notice and comment when they hear a system that can actually reproduce it. They can't tell ya what the difference was, only that the drums "sounded so real". The difference is that the attack was preserved and reproduced.

Notes and definitions.....

When I say 'attack', I mean the first instant of any sound. That is followed by the 'sustain' and then finally the 'decay'. We are looking at the amplitute vs time plot of any sound.

For human voice or string instruments played with a bow, the attack is pretty much non-existant. These sounds are much easier to reproduce, especially for the amp and wires. Most problems that are audible reproducing these sounds are directly cause by the speakers, unless the equipment earlier in the chain is really poor.

For percussive instruments (including piano, bells, etc.. not just drums) the amplitude of the sound goes from zero to its loudest point virtually instantly and then quickly drops to some fraction of the attack level and then more slowly decays back to silence.

Edited to add: There is plenty of woo in the audiophile ranks. There is plenty of wire woo, specifically. But putting out the blanket declaration that bi-wiring and passive bi-amping is useless in all circumstances is simply not true.

Re-Edited to add: If any of you are in the North Texas area, I can demonstrate all this quite readily. I can switch through sets of cable and show that difference can clearly be heard. If you have a sampling/recording o-scope (mine burnt up in our house fire last year), we can do the measurements that will show the difference numerically.

[jj]

And why not?

Safety. You should use something intended to put inside a wall.

But #12 or #14 stock wire is good. If it's going a long distance, use a low-inductance wire. Belden and others make such stuff for not that much.

[jj]

You guys are overlooking a real issue with large full-range speakers. Their moment-to-moment (instaneous) impedence can vary tremendously from their stated 'nominal' impedence. How does this change the calculations? During the attack (first 1/4 wave) of a bass drum hit, the impedence seen by the amp can be easily dip below .25 ohm for a few milliseconds (5-20ms is reasonable). That is because, for speakers on the very high end, nearly all of the impedance for the bass driver is reactance rather than resistance. If the driver isn't moving yet (prior to the bass pulse), its reactance is (near) zero.


Sorry, no, you clearly don't understand what a reactive load means.

If a load is quite capacitive, you would see exactly that kind of behavior, BUT THEIR IMPEDENCE IS NOT VARYING AT ALL.

If the impedence varies, then the device is nonlinear or time-varying. Period.

Impedence can be a function of frequency indeed, and in the case you discuss you're talking about something for which the magnitude of the impedence drops at high frequencies.

But that doesn't mean its nonlinear or time-varying.

"short-term impedence" is a meaningless term unless you do have a substantially nonlinear system, in which case it's a lot more complex than you make it. All you describe is something that looks like a reactive load with a capacitive behavior at high frequencies.


If the amplifier has a beefy enough 'final' to output the full voltage dictated by the input signal, even into a fraction of an ohm, it is obvious the current will briefly go sky high (40V into .25 ohm gives 160A). You now have enough current to drop significant amount of voltage across the wire used to connect the speakers.


Sorry, that depends on the dv/dt of the signal. THAT depends on the bandwidth of the signal. For a given amplitude and maximum frequency, you set a limit on the maximum dv/dt. Period.

Yes, if the cable is bad, that can matter, but in fact, very few speakers have anything like that kind of impedence.


2) If the bass driver is the specific cause of periodic drops to extremely low impedance, you'll only drop the voltage across the wire driving the bass driver. The wire feeding the midrange drivers and above will see the 'undistorted' signal as it appears at the output of the amp.


Bass drivers behave inductively, not capacitively. I have measured a few hundred, at least, and that's how they are to the last. Sorry. That's how it is.

Now, even if the bass driver DOES this for some reason, raising the current level at 100Hz will not affect the current at 10,000Hz, unless you get the wire warm enough to cause nonlinear conduction.

And you don't. The insulation would melt.

You're forgetting about linear superposition, and you're assuming that a "distortion" arises from a linear effect, when in fact it does not. What's more, your reasoning about woofer "instantaneous current" is wrong.



If the amp doesn't have finals that can support short bursts in excess of 100A, doing the passive bi-amping give similar benefits. I'll certainly agree though, active bi-amping (using a x-over before the amps) is more effective.


If you're overloading the amp, get a bigger amp. Not two amps. Sheesh. Lots cheaper.


To be sure, the improvement in total performance is slight from bi-wiring and passive bi-amping, but it is certainly there. It can be measured with test equipment, and it can be heard.


It can be measured with test equipment? Biwiring? Would you like to point me to a published example, please?


When I say 'attack', I mean the first instant of any sound. That is followed by the 'sustain' and then finally the 'decay'. We are looking at the amplitute vs time plot of any sound.


It appears that you are, actually, looking at the AMPLITUDE ENVELOPE of the sound, not the amplitude. Could you confirm?


For percussive instruments (including piano, bells, etc.. not just drums) the amplitude of the sound goes from zero to its loudest point virtually instantly and then quickly drops to some fraction of the attack level and then more slowly decays back to silence.


You do realize that this says that such sounds are mostly high-frequency, yes? That completely removes the woofer from the issue.


Re-Edited to add: If any of you are in the North Texas area, I can demonstrate all this quite readily. I can switch through sets of cable and show that difference can clearly be heard. If you have a sampling/recording o-scope (mine burnt up in our house fire last year), we can do the measurements that will show the difference numerically.

Can you show this in a blind test? Do the wires have equivelent gauges and interconductor capacitance?

[jimlintott]

And why not?

Like JJ said, safety it's against most building codes.

The nicest in wall speaker wire I ever worked with had Monster labelled all over it. It was so soft and supple. It was also expensive and I doubt it sounded better but sure was a pleasure to pull.

[Jeff Wagg]

Like JJ said, safety it's against most building codes.

The nicest in wall speaker wire I ever worked with had Monster labelled all over it. It was so soft and supple. It was also expensive and I doubt it sounded better but sure was a pleasure to pull.

But...wait. My electrician pulled zip cord in the rest of the house, and it passed inspection.

[scotth]

jj, think about this... a bass driver behaves much like a dc motor.

Also note, impedance is the sum of resistance and reactance. Reactance can be capacitive, inductive, and (often overlooked) mechanical (by physically moving a wire or coil of wire through a magnetic field as in the case of motors and conventional speaker drivers).

When a DC motor is not spinning, you see only its dc resistance. When power is applied to a stationary motor it will draw a huge currect to get started. It is called the surge current, generally. Once the motor is up to speed, the back EMF (voltage generated by motor spinning) increase the effective impedance tremendously.

A bass driver is pretty much a linear motor placed in a spring pendulum arrangement.

The parallel is just about perfect. During a percussive attack (yes, I was referring to amplitude envelope) you get a giant voltage spike applied to a driver that isn't moving (generating no back EMF), and until that speaker gets a movin' you will see enormous currents, exactly analogous to the surge current when starting a dc motor. It will just be for a much shorter duration.

Does that illuminate it for you, or do I need to go back and work all the figures, in detail, from top to bottom? I don't mind, but don't feel like typing that much unless I really have to.

And skipping to your very last inquiry: Yes, I can show this in a blind test even using casual listeners as the judge. I have never seen a wire that has enough capacitance that it can't be ignored for short runs (50 feet or less) at audio freqencies. Same for stray wire inductance. Wire resistance on the other hand can be a big player provided the rest of the playback chain isn't current limited in some other fashion.

[jj]

jj, think about this... a bass driver behaves much like a dc motor.


Yes and no. A DC motor is not very linear in a lot of ways (iron losses, etc, other factors that have no bearing here) and has no resonance.

A speaker is more linear, and has a specific resonant frequency. They are different in very basic ways.

There is no restoring force for a DC motor.

There is a restoring force for a speaker.


Also note, impedance is the sum of resistance and reactance. Reactance can be capacitive, inductive, and (often overlooked) mechanical (by physically moving a wire or coil of wire through a magnetic field as in the case of motors and conventional speaker drivers).


Wrong, really. Mechanical things like mass are represented in the electrical impedence. Impedence is impedence. Saying it's the sum of resistance and reactance implies that it's the sum of a real plus an imaginary component, which is of course true, but tautological.

The way in which the electrical reactance comes about is irrelevant if the system is quasi-linear, which it is for a loudspeaker.


When a DC motor is not spinning, you see only its dc resistance. When power is applied to a stationary motor it will draw a huge currect to get started. It is called the surge current, generally. Once the motor is up to speed, the back EMF (voltage generated by motor spinning) increase the effective impedance tremendously.


Irrelevant in its entirety. The motor has no restoring force.

The back EMF in the motor is controlled by the speed of the motor, which has no restoring force.

The back EMF of the speaker is ALSO controlled by the speed of the cone, yes, but there is restoring force, and no "rest state" velocity (ignoring friction in both cases).

Since there is no storage of energy in (rotational for the motor) velocity, you have a different situation. Turn off the current in the DC motor, you get a voltage, and one that sticks around for a long time. The "rest position" involves continuous rotation.

Turn of the current in a voice coil, it returns, resonantly, to center, and resonates. Energy is radiated to the atmosphere, indeed, but the rest position of the speaker is zero.

If you add damping to both, you get what really happens. The motor slows slowly, very slowly, and the speaker moves back and forth rapidly decaying to zero. One has a defined rest position and restoring force, the other doesn't.

The motor also has a much larger ratio of DC resistance to turning impedence. Which also is something you're not considering.


A bass driver is pretty much a linear motor placed in a spring pendulum arrangement.


No, it's a current driver placed in a mechanical resonant circuit. The inclusion of a pendulum is not really correct, as a pendulum is a nonlinear gravititational resonator, and a spring/mass is the mechanical resonant circuit in question.

This mechanical resonant circuit reflects EXACTLY in the impedence of the driver.

In the case of the motor, you have energy storage in rotational velocity. You have an integrator that you do not have in the speaker.


The parallel is just about perfect. During a percussive attack (yes, I was referring to amplitude envelope) you get a giant voltage spike applied to a driver that isn't moving (generating no back EMF), and until that speaker gets a movin' you will see enormous currents, exactly analogous to the surge current when starting a dc motor. It will just be for a much shorter duration.


Stuff and nonsense. First, the crossover prevents any such "giant voltage spike" from happening at any time in the woofer. There is a FREQUENCY DEPENDENT FILTER in front of the woofer. Ergo, there is no spike.

Second, woofers have a DC resistance of about half of their nominal impedence. You NEVER GO BELOW THAT RESISTANCE in any normal woofer designed according to any sort of modern Theile-Small (or later)-derived alignment. Period.

Third, the lack of back EMF can not generate any current that is larger than that set by the DC resistance of the coil.

Starting from rest, you can not draw more current instantaneously than that of the coil resistance. The only way to do that is to have energy stored in opposing phase in the speaker movement, and even then it's simply a consequence of the FIXED (in linear range of operation) speaker movement.


Does that illuminate it for you, or do I need to go back and work all the figures, in detail, from top to bottom? I don't mind, but don't feel like typing that much unless I really have to.


Don't bother, unless it will help you to realize that your physics and your claims about impedence are completely mistaken.

You're wrong. Flatly, outright wrong. I really don't know what more to say to you.

You can't draw more current when the speaker is in rest state, period, end-of-discussion, than would be expected from the VIN/DCR where VIN is the instanteous voltage and DCR is the DC resistance of the speaker. Period. End of discussion.

It's not anything like a motor, because the motor stores energy in its moment of inertia WITHOUT RESTORING FORCE.

A speaker has restoring force. The equivelent of a motor is a speaker with infinite displacement and no surround or spider.

There are subwoofers that use rotary actuators. They have immense rotational momentum in terms of loudspeakers, are violently inductive (their ratio of DC resistance to impedence is much higher, too), but they also require special amps that can drive them, too.


And skipping to your very last inquiry: Yes, I can show this in a blind test even using casual listeners as the judge. I have never seen a wire that has enough capacitance that it can't be ignored for short runs (50 feet or less) at audio freqencies. Same for stray wire inductance. Wire resistance on the other hand can be a big player provided the rest of the playback chain isn't current limited in some other fashion.

I have easily seen wires whose impedence can be ignored for sensible speaker cable run lengths.

I've also see ones that can't.

The ones whose impedence can be ignored can be got at Home Depot. The ones that can't don't generally come from there, unless one gets #22 bell cord or something.


But, just remember, the DC resistance is key here. The DC resistance of a motor is teeny-tiny compared to its impedence (at operating speed).

For a speaker, first, there is no "operating speed". Second, the DC resistance is typically half of the nominal impedence. Third no "big spikes" get through the crossover into the woofer anyhow, and the crossover inductor DCR in any crossover I've ever seen is bigger than the wire DCR anyhow.

[gruk]

And why not?

I know we only run some types of Cat5 in under-floor spaces. Essentially only cable sheated in low-smoke, fire-retardant plastic. I can well imagine that the cheapest zip-cord you can buy don't have those properties and I'd use the same rough guidelines for "in walls" as I'd use for "under data centre flooring".

[Harlequin]

I think the issue with "under data centre flooring" is that the amount of cable is so high that a fire could result in significant toxic fumes if you don't use low-smoke insulation. The walls in your house are unlikely to have this problem.

[Jeff Wagg]

When I investigated cabling for a new house, I came across plenum rated cable. This is the stuff you're talking about..doesn't give off fumes. It's required in commerical building codes.

When I inquired about putting this in my house, the cabling company said I was crazy, because the romex used in homes isn't plenum rated, and will produce 100x the fumes that CAT 5 would.

Based on that, I don't see how zip wire would be unsafe in the walls of a home, unless they could some how CAUSE a fire.

[jimlintott]

When I investigated cabling for a new house, I came across plenum rated cable. This is the stuff you're talking about..doesn't give off fumes. It's required in commerical building codes.

When I inquired about putting this in my house, the cabling company said I was crazy, because the romex used in homes isn't plenum rated, and will produce 100x the fumes that CAT 5 would.

Based on that, I don't see how zip wire would be unsafe in the walls of a home, unless they could some how CAUSE a fire.
I think there might be a confusion in terms here. When I say zip cord I am referring to a plain two conducter stranded wire encased in a plastic or rubber insulation. Lamp cord is zip cord in 14 guage. While it wouldn't necessarily cause a fire it is not rated for in wall use because of how it might behave in a fire. Also for electrical wiring there is no ground wire. In wall electrical wire is usually solid strand multi conductor, at least three, white. black, green and then sometimes a bare ground wire. It is insulated with a fire proof insulation and often reinforced so it won't be damaged while pulling.

In wall speaker wire is usually stranded, available in a variety of guages. It is reinforced for pulling and the pair is twisted and often shielded to help eliminate picking up any hum from electrical wires. Of course it has acceptable in wall insulation.

I would be really surprised if your house is wired with zip cord but I wouldn't be surprised if there is some confusion over a slang term.

[Jeff Wagg]

I pulled out a the wall plate, and sure enough it's zip cord. But I still fail to see how this is any less safe than the non-plenum rated CAT 5 and RG-6 I have everywhere.



I think there might be a confusion in terms here. When I say zip cord I am referring to a plain two conducter stranded wire encased in a plastic or rubber insulation. Lamp cord is zip cord in 14 guage. While it wouldn't necessarily cause a fire it is not rated for in wall use because of how it might behave in a fire. Also for electrical wiring there is no ground wire. In wall electrical wire is usually solid strand multi conductor, at least three, white. black, green and then sometimes a bare ground wire. It is insulated with a fire proof insulation and often reinforced so it won't be damaged while pulling.

In wall speaker wire is usually stranded, available in a variety of guages. It is reinforced for pulling and the pair is twisted and often shielded to help eliminate picking up any hum from electrical wires. Of course it has acceptable in wall insulation.

I would be really surprised if your house is wired with zip cord but I wouldn't be surprised if there is some confusion over a slang term.

[jimlintott]

I pulled out a the wall plate, and sure enough it's zip cord. But I still fail to see how this is any less safe than the non-plenum rated CAT 5 and RG-6 I have everywhere.
What's it for? Plugging in your toaster? Or speakers.

Sorry I don't know enough to answer your question. I just know that here zip cord is non-code.

Electricians are good with their high voltage stuff but have been known to do terrible jobs of running low voltage A/V cabling. (Not all, I'm sure.)

ETA: Plenum rated is for use where it may be laid in ducting of HVAC systems. Something that is likely never done in residential wiring. Cat5 and RG6 are both rated for in wall use.

[macgyver]

Hi all,

I'm the original author of the "lowdown on Hi-Fi" post. I've had some difficulty getting subscribed to the forum, but all is well now.

First of all, consider the context of my message - this was a letter written specifically to somebody that I had already discussed audio with at some great length and was basically designed to 're-direct' his money/time efforts towards a more satisfying result.

I can see that this thread has done what most audio threads do, which is to become a bit of a battleground :-)

On the issue of bi-amping and bi-wiring:

By bi-wiring, I specifically mean to use multiple runs of wire from a single power amp output to a standard passive network loudspeaker. Sometimes the X-over allows separation of the lo/band/hi pass stages with the removal of jumpers or a buss bar. My point is that IF you are using sufficient quality and quantity of conductor to begin with, the idea that adding even MORE wire in parallel is not worth the money. The woo-woo aspect of this application, IMHO, is that somehow the signals will "route" themselves down the wire most suited to their specific frequency. While this is to some extent true due to resistance being a function of cross sectional area and things like "skin effect", this is generally not significant at audio frequencies. The cost factor is dependant too on the type of speaker wire we're talking about. I'm used to seeing handmade silver foil speaker and interconnect wire marked with "directional arrows" at several hundred feet per meter...my advice stands; don't bother. (I'm still waiting for somebody to explain how "signal direction" is relevant in this application?)

By Bi-amping I'm specifically referring to using multiple amplifiers and speaker wire to feed the same speakers listed above. The additional amplifier is superfluous. In the case of the speaker allowing separation of x-over sections, the idea is that somehow the reduced bandwidth and altered load that the amplifier is now responsible for will result in an "improvement" of sound quality. However, it's more likely, IMHO, that mismatches in gain between active components will result in the original loudspeaker engineer's hard work being thrown out the window. Worst case scenario is a system "tuned by golden ear" that places a solid state amplifier for the low pass section, and a transformer coupled tube amplifier for the highs. The owner of such a system will definitely hear a "difference" - my point is that it's not necessarily an "improvement".

Finally, my statement of 'best' speaker wire being Cat5 needs some explanation as well. Google "Aaron Drabitt" for some measurements that I made a few years ago. I'm not allowed to post the URL (yet).

My definition of "best" is basically that it does the job extremely well for the least cost. Other benefits of Cat5, other than electrical, are that it is easily found in various "FT" ratings, so that it can be used in new construction and easily pass inspections. Also, the biggest problem that I find with speaker wire is that over time the copper will corrode and, unlike silver, cu oxide doesn't conduct very well. This corrosion can migrate quite some distance up the copper strands and into the jacket of the cable as well. Corrosion WILL have significant effects on sound quality. Cat5, however, has each conductor lovingly wrapped in it's own individual weatherproof (some) jacket, so corrosion is limited to the stripped portion. You could even use "flooded" Cat5 if you really wanted to put the kibosh on corrosion (and rodents too).

One caveat to Cat5 is with the use of unusually wideband amplifiers. The additional parallel capacitance that helps to flatten the inherent inductive rise of a conductor can become dominant at higher supersonic frequencies and essentially become a short circuit. This would cause a high frequency feedback loop and oscillation that could blow the output stage of the amp. As one P.Eng once told me "if you want to build an amplifier try to design and oscillator, if you want to build an oscillator try to design an amplifier" Murphy's law is always in effect.

Hope that clears things up a bit?

[atari24]

Well, I was curious as to why you said cat5 was the best, and no one wanted to answer in the other thread, so thanks!

[macgyver]

I just thought I'd wade in here to speak to the issue of impedance, and loudspeakers.

Impedance is very much misunderstood, so for the layman I'll give a few non mathematical ways to differentiate impedance and resistance:

pure resistance is constant regardless of signal frequency and measures the same in both DC and AC systems.

resistance does "work" by converting electrical energy to heat, pure reactance does not consume any energy. Reactance changes with respect to signal frequency. I use the term "reactance" here, because real world impedance is most often a blend of resistance, capacitive reactance, and inductive reactance.

Without alternating current/voltage impedance is meaningless.

With respect to loudspeakers:

We have to define what type of loudspeaker driver we're talking about before too much discussion can be meaningful. In this case I'll stick to the typical "dynamic" voice-coil based cone driver.

This type of driver is an electromechanical system, and not easily mathematically modeled. Some brilliant research by the founder of LinearX resulted in the system I use called LEAP. His model was able to very accurately describe the dynamic driver in terms of impedance when he discovered that many of the mechanical properties such as the compliance of the basket/spider/surround appeared very much like a capacitor or other easily modeled "electrical entity". I'm summarizing here, so accuracy isn't my goal.

What's interesting to a loudspeaker designer, like myself, however is how well his software does at modeling the changes that driver loading in a cabinet will have on impedance. A typical bass driver, for instance, will show an impedance spike at it's self resonant frequency, sometimes reaching 20 - 40 ohms. It will then show a "nominal" impedance somewhere close to it's rating for an octave or two, and then you begin to see a significant impedance rise as the reactance of the voice coil becomes dominant. Put the speaker into a sealed box, and that impedance spike will shift. Shift it to the "right" place, and you can extend the bass response. Put that same speaker in a box with a hole in it and suddenly you have two impedance spikes, one for the speaker and one for the "virtual speaker" you just created with that hole. Again, shift them to the "right" places, and you can mix the out of phase output of the port with the output of the speaker and extend frequency response smoothly to a higher order roll off than the sealed box (due to the band pass nature of the port). You've just made a completely mechanical X-over by coupling a single speaker to a box of a specific volume to a resonant port of specific dimensions (Helmholtz resonator, in fact).

In design it's all about the impedance, and how you can manipulate it. X-overs extends this far beyond the scope of this post (but LEAP does an admiral job of that too!) I don't get any kickbacks from LEAP I assure you. I paid full pop for that!

One last note, however, about active "line level" X-overs and multiple amp systems. They're fantastic if done "right". The problem is that the same issues with x-over design apply, and it's not something that can be done effectively by ear. In the case of this type of design, impedance will matter with initial driver placement in the correct enclosure type. Once that's done, you're working with capacitive feedback networks and op-amps to generate the x-over slopes you desire. This can't be done without knowing also the inherent "frequency signature" and "phase relationship" between the individual drivers. Some FFT analysis to determine acoustic "center" of the driver, and frequency response measurement is essential to match the active x-over phase and slope and integrate all the drivers, amps and filters into a seamless system. Again LEAP will do this too, but I couldn't afford the active filter component....nor could I afford to multi-amp a system.

cheers!

[macgyver]

Originally posted by jj:

"You do realize that this says that such sounds are mostly high-frequency, yes? That completely removes the woofer from the issue."

This brings up another pet peeve of mine regarding the often touted concept of "attack" when referring to bass drums, or other percussive instruments in sound reproduction.

I find that poor old Jean Baptiste Joseph Fourier is often forgotten but, as much as I hated my Fourier Transform mathematics course, he has to be given credit.

On the surface it seems that a bass drum would be a low frequency sound that would be entirely the responsibility of the subwoofer to reproduce. For those of you with subwoofers in your system, try disconnecting everything BUT the subwoofer, then tell me that all you need is a "faster" subwoofer for better sound. A subwoofer rarely reproduces more than a single octave so even the concept of "harmonic distortion" becomes somewhat irrelevant in this application. What determines the sharp rise of the so called "fast attack" of the bass drum is phase coherent high frequency energy, most likely reproduced by all the other speakers in your system OTHER than the subwoofer. The subwoofer provides the fundamental, the rest of the speakers provide the "attack".


Take this further into the realm of digital audio reproduction and you'll find something interesting: standard PCM audio CDs can't reproduce ANY harmonics for a fundamental above 11.025Khz. Since a square wave represents a fundamental with an infinite series of odd harmonics, this means that if you tried to record a 12kHz square wave (or any repetitive waveform other than a pure sinusoid) the best you could hope to play back is a 12kHz sine wave.

What does this mean for audio reproduction? Not much if you agree with the fact that the human ear couldn't hear the 22.05kHz first (even) harmonic anyway....

It's often high frequency hearing loss that causes older folks to complain of "not understanding speech clearly" for the same reason. The aged ear can hear the fundamental frequency of the voice quite easily with no impairment but, without the high frequency information, it becomes a muddied mess of mumbling noise. It's this reduced high frequency bandwidth that can make it difficult to recognize a familiar voice on the telephone as well.

In closing, it's the high frequency information that adds the character or "timbre" to sound reproduction, while the low frequency will provide warmth and even visceral impact through "shaking the walls".

[jj]

I find that poor old Jean Baptiste Joseph Fourier is often forgotten but,


And abused as often as he's forgotten. too.


as much as I hated my Fourier Transform mathematics course, he has to be given credit.


Along with Laplace, Helmholtz, etc, indeed. I'm not sure who initially invented 'z' transforms, really, and but them, too. :)


On the surface it seems that a bass drum would be a low frequency sound that would be entirely the responsibility of the subwoofer to reproduce. For those of you with subwoofers in your system, try disconnecting everything BUT the subwoofer, then tell me that all you need is a "faster" subwoofer for better sound. A subwoofer rarely reproduces more than a single octave so even the concept of "harmonic distortion" becomes somewhat irrelevant in this application. What determines the sharp rise of the so called "fast attack" of the bass drum is phase coherent high frequency energy, most likely reproduced by all the other speakers in your system OTHER than the subwoofer. The subwoofer provides the fundamental, the rest of the speakers provide the "attack".


Exactly, which shows that in addition to all the other mistakes in the idea that one needs a huge current to start a subwoofer moving (like forgetting its DC resistance), the idea that the voltage driving the woofer changes rapidly is just plain bogus. Those frequencies never get to the woofer, well, unless the crossover is shorted. :D


What does this mean for audio reproduction? Not much if you agree with the fact that the human ear couldn't hear the 22.05kHz first (even) harmonic anyway....


Well, it's well-established that the ear is a frequency analyzer, that everything above 16kHz or so is detected on the first part of the cochlea (right next to the stapes/windows), and that that gets blasted real early in life nowadays.

So except for nonlinear effects, which only occur at pretty high levels, higher than you should listen, yes, that's quite right, you can't hear the second harmonic (2*f) (have to watch out for that 'first harmonic' thing, btw) unless you're really young, in the first place.

Now, you MIGHT hear a difference if you use a nonlinear speaker (which most are) and something above 20kHz or so intermods down into the audible range, which does happen, and which has seriously confounded more than one experiment that claimed to show supra-20kHz hearing in older folks...

But that's hearing a bug, not hearing something above 20kHz.


It's often high frequency hearing loss that causes older folks to complain of "not understanding speech clearly" for the same reason. The aged ear can hear the fundamental frequency of the voice quite easily with no impairment but, without the high frequency information, it becomes a muddied mess of mumbling noise. It's this reduced high frequency bandwidth that can make it difficult to recognize a familiar voice on the telephone as well.


Well, most speech information is in the range of 300Hz to 7kHz. At least for most western languages. The most common problem with speech understanding is that of midrange noise loss around the ear canal resonance, which removes mid-frequencies in which a great deal of articulation is present.

But you're not altogether wrong, either. Hearing differences between fricatives can be an HF loss problem for older folks, too.


In closing, it's the high frequency information that adds the character or "timbre" to sound reproduction, while the low frequency will provide warmth and even visceral impact through "shaking the walls".

Well, consider, information is bandwidth * bit depth. The receptors on the basilar membrane have about 30dB (5 bit) accuracy at best. The more bandwidth, the more information. But of course, you have to weight this on a critical band scale, since each critical band contains roughly an equal number of inner hair cells.

[macgyver]

(have to watch out for that 'first harmonic' thing, btw)

I knew somebody would call me on that :o

Thanks for the feedback, BTW

[jj]

I knew somebody would call me on that :o

Thanks for the feedback, BTW

Suffice it to say I've done it too.

[Bodhi Dharma Zen]

Interesting thread. Its always amazing to see how some audiophiles are prone to believe almost anything. Even the understanding of some physical laws are "twisted" to fit audiophile myths.

But the best are the ones who believe all the cable paraphernalia and are willing to do blind tests!!! just to fail miserabily, of course :)

[Paulhoff]

Geee, did I miss the fun.

JJ, it is late for me now, and I will have to reread your post tomorrow.

Paul

:) :) :)

[pipelineaudio]

Countdown till this thread self destructs in 96khz vs 44.1khz flame fest .....

[Paulhoff]

OK, I'll bite, you will not hear the difference.

Paul

:) :) :)

[pipelineaudio]

OK, I'll bite, you will not hear the difference.

Paul

:) :) :)

this post has been reported






























































to both Nika Alrditch and Slipperman at the same time

[Paulhoff]

Golly gee. :D

Paul

:) :) :)

[jj]

Ok, I'll join the party...

Somebody explain to me when any signal over 30kHz matters to a human being, if presented at any level below 110dB SPL.

Proceed, please.

[Bodhi Dharma Zen]

OH, but there are studieees!! besides its more acurateee, it represents better what the director/enginer/artist wanteeed

[jj]

OH, but there are studieees!! besides its more acurateee, it represents better what the director/enginer/artist wanteeed

ARE YOU SURE! :p

[Paulhoff]

I still remember years ago, my audiophile friend saying after hearing a violin on his speakers, "That is how it is suppose to sound". With me saying, "Yea, they all sound alike, if you haven’t heard it live, how do you know how it sounded to begin with, in that room with that player". :rolleyes:

Paul

:) :) :)

[jimlintott]

I don't pretend to know enough to argue about the higher sampling rate but some of my DVD Audio discs sure sound nice. Now I'm sure it could be a number of things: I'm fooling myself, they may have less compression being targetted to a different audience and to be fair I don't have the same regular CDs to compare to.

I know for sure that CD sampling rate is way more than double my own frequency response. :D

Is it possible that it isn't the higher sampling rate but the higher bit rate (24 vs 16) that is bringing the most improvement? Didn't I read on a similar thread that somene said that the higher sampling rates are better for mixing?

I'm just curious.

[Bodhi Dharma Zen]

There is more information, that much is for sure. Still, the question in hand is if the difference is audible. The same goes for the other side, this is, MP3s. Most people cant tell if they are listening to the MP3 or the CD if the MP3 its encoded at 192kbps.

Now, sure, I have listened pretty impressive things in DVD-A and SACD, but the whole equipment was top notch. I would like to listen to MP3 against a high resolition format, in a double blind condition.

BTW, I happen to listen almost all my music in MP3, and I was an audiophile. I learned that its not the equipment, nor the source, but the equipment location in the room, what matters the most.

[jj]

I don't pretend to know enough to argue about the higher sampling rate but some of my DVD Audio discs sure sound nice. Now I'm sure it could be a number of things: I'm fooling myself, they may have less compression being targetted to a different audience and to be fair I don't have the same regular CDs to compare to.

I know for sure that CD sampling rate is way more than double my own frequency response. :D

Is it possible that it isn't the higher sampling rate but the higher bit rate (24 vs 16) that is bringing the most improvement? Didn't I read on a similar thread that somene said that the higher sampling rates are better for mixing?

I'm just curious.

Well, mostly they use up space and processing power, if you want my opinion. I can justify going to something like 64kHz (with 20kHz bandwidth) if I want to reach way, way far into possible problem mechanisms.

But the real problem is that we need exactly the same material, EXCEPT for the storage mechanism, in order to compare things. That's hard.

[Bodhi Dharma Zen]

well, any SACD have two layers...

[jj]

well, any SACD have two layers...

Did they use exactly the same processing? At least in some cases it seems evident not.

[Paulhoff]

There is a point that it becomes the "Princess and the Pea" syndrome.

Paul

:) :) :)

[Bodhi Dharma Zen]

Did they use exactly the same processing? At least in some cases it seems evident not.

mm dunno, Im not an audiophile anymore! ;)

[Paulhoff]

Audiophile around here does have a bad name.

Paul

:) :) :)

[pipelineaudio]

there are plenty of real reasons for using 96khz

hearing past 20khz is not usually one of them

[jj]

there are plenty of real reasons for using 96khz

hearing past 20khz is not usually one of them


Ok, let's hear them, then.

[wybili]

Ok, let's hear them, then.

Basically, higher sample rates make it easier to design digital signal processing algorithms.

A couple examples:

1. If you are designing an anti-aliasing filter, you have to make a tradeoff between stop band rejection, transition bandwidth, and phase shift. At 96k, you can let the transition bandwidth get pretty large (as a lot of the transition band will be inaudible) and worry about the other constraints.

2. Suppose you wanted to model an analog audio circuit that had feedback in it. To get feedback in a digital application, you have to use a one frame delay, which induces a phase shift. For the sake of argument let's say 15k is the highest frequency a person can hear. At 44k, a one frame delay of a 15k signal is 34% of a cycle; at 96k, the one frame delay is 15% of a cycle. The phase shift will still be awful near Nyquist, but at 96k we only care about the behaviour well below Nyquist.

[pipelineaudio]

holy crap the dudes first post here, took the words out of my mouth AND opened the potentially suicidal can of worms of death sparing me!

AWESOME!!!

[jj]

Basically, higher sample rates make it easier to design digital signal processing algorithms.


Ok, I want to make an equalizer. I want to make a 6dB peak in the bass range that is centered on 40 Hz and that has a 3dB bandwidth of 20Hz and 60Hz. Please explain how it is easier to do this at 96kHz than it is at 44100. Please address the number of bits required in the filter state variables and filter coefficients.


A couple examples:

1. If you are designing an anti-aliasing filter, you have to make a tradeoff between stop band rejection, transition bandwidth, and phase shift. At 96k, you can let the transition bandwidth get pretty large (as a lot of the transition band will be inaudible) and worry about the other constraints.


True. Hm, where's that PDF? Is this for linear systems reasons or for other reasons related to human perception of below 20kHz artifacts?


2. Suppose you wanted to model an analog audio circuit that had feedback in it. To get feedback in a digital application, you have to use a one frame delay, which induces a phase shift. For the sake of argument let's say 15k is the highest frequency a person can hear. At 44k, a one frame delay of a 15k signal is 34% of a cycle; at 96k, the one frame delay is 15% of a cycle. The phase shift will still be awful near Nyquist, but at 96k we only care about the behaviour well below Nyquist.

Of course, you can't model an analog circuit in 'z' space. 'z' space and 's' space have different properties. You can, however, make a filter that peaks at some high frequency near the nyquist rate. It will have a frequency and phase response you can control just like the one you can get from 's' space, although you have to control it differently.

Are you planning on using Match-z, bilinear-z, or what kind of mapping of the s domain to the z domain? Your point would seem to be a (rare) case of using the bilinear mapping when you should have used either IIZ or match-z. Could you be more specific here? Bear in mind that you only have to match the phase response of the previous circuit BELOW fs/2.

I will presume that you are aware that it's possible to have a sub-sample delay, right?

[pipelineaudio]

Conventional wisdom, though by someone who sills 192krule8 says that theoretically 44.1 wouldbe perfect *filterwise* if done right for all 20khz and below, but who has made this perfect one?

[jj]

Conventional wisdom, though by someone who sills 192krule8 says that theoretically 44.1 wouldbe perfect *filterwise* if done right for all 20khz and below, but who has made this perfect one?

Err, no it doesn't say that, actually.

There is no question of the linear-systems issues. The question is how the linear-systems issues interact with the ear, which is anything but linear.

http://mue.music.miami.edu/AES/adc.ppt

See slides 68 to 73 and thereabouts. You may have to step back a bit in the presentation to get a good feel for the plots.

[pipelineaudio]

Im not considering how these things interact with the ear, Im talking about a linear low pass filter that is able to pass all the way up to 20k without screwing up

[wybili]

jj, I can see that you have quite a bit more experience in DSP than I do (just started learning about a year gao), so I may be just plain wrong here. :D

Ok, I want to make an equalizer.

Sorry, I guess I should have said that a *few* DSP algorithms are easier at higher sample rates. I agree, I don't see anything to gain by oversampling in a case like that.

True. Hm, where's that PDF? Is this for linear systems reasons or for other reasons related to human perception of below 20kHz artifacts?

I'm not quite sure what "linear systems reasons" are. But, to put it in a sort of wishy-washy way, if you are designing for example a linear phase FIR filter, you have a certain number of "constraints" you can impose-- the number of taps in the filter. With more taps you can impose more constraints, but the group delay also increases, since that is proportional to the number of filter taps. At least that is how I think of it.

Are you planning on using Match-z, bilinear-z, or what kind of mapping of the s domain to the z domain? Your point would seem to be a (rare) case of using the bilinear mapping when you should have used either IIZ or match-z. Could you be more specific here? Bear in mind that you only have to match the phase response of the previous circuit BELOW fs/2.

Hmm, I don't really see how this applies to the exact problem I was working on. I'm not actually familiar with IIZ and match-z mapping; can you point me to a reference? (Like I said, I am relatively new to this stuff.)

Basically, I was trying to make a digital filter with a similar sound to the Moog ladder filter, found in the Minimoog and other classic synthesizers. For the actual filter section, I used a chain of four one-pole lowpass filters. The phase shift at the cutoff frequency happens to be pi.

Then the resonance knob controls the amount of negative feedback which is mixed back into the input of the filter. Since the feedback is negative, turning up the resonance boosts frequencies where the phase shift is about pi and cuts frequencies where the phase shift is about 0. If you turn the resonance all the way up, it will oscillate at the cutoff frequency (which is considered a very desirable property among synthesizer buffs).

So the extra one sample delay will make it resonate at the wrong frequency, and you have to compensate for that in one way or another. Oversampling is an easy way to improve this (at the cost of computation time).

Now if I understand you, are you saying that there's another way to model this circuit so that you don't have to use digital feedback?

I will presume that you are aware that it's possible to have a sub-sample delay, right?

I was not aware. Well, I know you can get non-integer delays, but I don't see how you can get a sub-sample delay in this case. You get x as input; then you compute f(x). For the next frame you get x' as input, and you compute f(x' - r*f(x)). How can you feed f(x) back into f with less than a one frame delay?

[jj]

Sorry, I guess I should have said that a *few* DSP algorithms are easier at higher sample rates. I agree, I don't see anything to gain by oversampling in a case like that.


Grin, now consider doing that at SACD rates. But only briefly, please.



I'm not quite sure what "linear systems reasons" are. But, to put it in a sort of wishy-washy way, if you are designing for example a linear phase FIR filter, you have a certain number of "constraints" you can impose-- the number of taps in the filter. With more taps you can impose more constraints, but the group delay also increases, since that is proportional to the number of filter taps. At least that is how I think of it.


Better to say that the delay increases, since most FIR designs are constant-delay, but yes.

It is easier to make a filter, because the filter is shorter. See the file I pointed pipeline audio at for why you MIGHT (notice, MIGHT) want a shorter filter.



Hmm, I don't really see how this applies to the exact problem I was working on. I'm not actually familiar with IIZ and match-z mapping; can you point me to a reference? (Like I said, I am relatively new to this stuff.)


Ugh. I'm old. If you can find a copy of Rabiner and Gold, check out the discussion on different ways to convert S to z domain.


Then the resonance knob controls the amount of negative feedback which is mixed back into the input of the filter. Since the feedback is negative, turning up the resonance boosts frequencies where the phase shift is about pi and cuts frequencies where the phase shift is about 0. If you turn the resonance all the way up, it will oscillate at the cutoff frequency (which is considered a very desirable property among synthesizer buffs).


Try just using a delay line for your feedback, and perhaps an allpass in the feedback path.


Now if I understand you, are you saying that there's another way to model this circuit so that you don't have to use digital feedback?


You need a different model. I'm not sure offhand exactly what, but you need a model that does not just translate analog into digital.


I was not aware. Well, I know you can get non-integer delays, but I don't see how you can get a sub-sample delay in this case. You get x as input; then you compute f(x). For the next frame you get x' as input, and you compute f(x' - r*f(x)). How can you feed f(x) back into f with less than a one frame delay?

First, one would usually say "one sample" not "one frame". The question is not how you get less than one sample delay, but how you generate the feedback taps to get the delay you expect. The question is actually quite complex. Try calculating the way the poles and zeros move, and map them to the z domain, maybe. (match-z mapping)

There are a variety of options, but most often it's better to redesign the whole thing from scratch in the digital domain.

[jj]

Im not considering how these things interact with the ear, Im talking about a linear low pass filter that is able to pass all the way up to 20k without screwing up

Problem is, a long enough filter WILL interact, and one that cuts off between 20kHz and 22kHz MIGHT, it's not frankly clear.

So, if you consider only the filtering, you're leaving out the part that matters, which is the actual final reciever.

[wybili]

Just for the record, I fixed the problems with that particular filter, using oversampling and another trick. It's not technically "correct," but I like the sound, the CPU hit is reasonable, and no one using the software has complained.

Ugh. I'm old. If you can find a copy of Rabiner and Gold, check out the discussion on different ways to convert S to z domain.

Thanks, I will look for that book (I do need a good reference).

First, one would usually say "one sample" not "one frame".

Sorry, I got in the habit of saying "frame" because the software I was writing also uses "sample" in the sense of a short audio clips.

[jj]

Just for the record, I fixed the problems with that particular filter, using oversampling and another trick. It's not technically "correct," but I like the sound, the CPU hit is reasonable, and no one using the software has complained.


Probably the best way.


Thanks, I will look for that book (I do need a good reference).


"Digital Signal Processing" by Larry Rabiner and Ben Gold.
It is, annoyingly, out of print. My copy, even more annoyingly, will probably have to take a trip to the bookbinders in order for it to stick together. It has been used.


Sorry, I got in the habit of saying "frame" because the software I was writing also uses "sample" in the sense of a short audio clips.

I've met that as well. Unfortunately, when talking about sampling, etc, 1 sample is one time instant, not a bit of material to be repeated, etc...

Yes, it would surely help if the terminology was consistant. It's like
"compression" (is that level compression or bitrate compression?) or "coding" (is that bit-rate reduction, as in signal coding, or is that writing computer code? or is that hiding information?).

[Ducky]

jj, I can see that you have quite a bit more experience in DSP than I do (just started learning about a year gao), so I may be just plain wrong here. :D


Ask him what experience he has.

Seriously.



As usual I find this thread late and all the good points were made.

Carry on. I have popcorn.

[Bodhi Dharma Zen]

JJ,

Now that Fowlsound mentions it, and that I did my research ;) I wonder, have you ever tried (or the people arround you) those famous doble blind tests between the CD and MP3?

I have made my own tests, of course, but just with some rock and electronica. I believe that around 192Khz the ear have a pretty impossible task in trying to differenciate between them.

Now, for classical music, I have not done the tests yet, but I have the feeling that the MP3 sounds "noisy" and that resolution lacks. Since I hate the usual audiophile terminology I wont try to describe the differences anymore, but suffice to say that I believe that you can hear the difference between both formats with this kind of music.

(I know! typical audiophile claim!) Im ashamed :)

What do you think?

Of course, the best answer would be to do the tests myself, but I find time lacking lately. Thats why I ask the expert :)

[Paulhoff]

Well I know how I feel about MP3's. When I remember recording on 1,800 ft. tape and getting 45 minutes in one direction using 7 1/2 ips for a total of 1 1/2 hours for both directions. Or using 3 3/4 for a total of 3 hours and remember the problems of finding the song that I want etc. Now on the CD I can get 5 hours using 320 kbps, with no problem with finding the song I want, and if I want I can put information about the song on the file with a picture too. I have record radio shows that are 5 hours long using MP3’s that would be a pain using Tape and you couldn’t use a wave file that you could put on a CD for that long. Also using tape I would know it wasn’t the original at 3 ¾ ips because the highs dropped off, but you dealt with it to get the time. Now I am hard press to hear the different the even 160 kbps and I am only using the music for mainly background so it doesn’t matter that much anyway. Now I have a Sony 5 disk CD/DVD player that will play DVD’s with MP3’s so now you can play music for over a day at 320 kbps. So having used tape years back, I can appreciate MP3’s in a big way.

Paul

:) :) :)

[Ducky]

Well I know how I feel about MP3's. When I remember recording on 1,800 ft. tape and getting 45 minutes in one direction using 7 1/2 ips for a total of 1 1/2 hours for both directions. Or using 3 3/4 for a total of 3 hours and remember the problems of finding the song that I want etc. Now on the CD I can get 5 hours using 320 kbps, with no problem with finding the song I want, and if I want I can put information about the song on the file with a picture too. I have record radio shows that are 5 hours long using MP3’s that would be a pain using Tape and you couldn’t use a wave file that you could put on a CD for that long. Also using tape I would know it wasn’t the original at 3 ¾ ips because the highs dropped off, but you dealt with it to get the time. Now I am hard press to hear the different the even 160 kbps and I am only using the music for mainly background so it doesn’t matter that much anyway. Now I have a Sony 5 disk CD/DVD player that will play DVD’s with MP3’s so now you can play music for over a day at 320 kbps. So having used tape years back, I can appreciate MP3’s in a big way.

Paul

:) :) :)


Even at 44.1 in uncompressed .wav it sounds great and uses relatively low real estate on teh hard drive. Hell one service I offer is 16 tracks at once for live shows, which usually means at least one set ot 45 minutes. This takes about 3.5gig of space on the drive. Considering I have a 120g drive I record to, and can scale if needed, it very much beats out the cost of 2" tape. Often when a new client hires me they are a bit surprised to see me show up to a gig with a laptop, a 4 space rack, 16ch XLR isolated splitter and two firepods. I split before the FOH mixer and get a great recording to work with.

Once again, Paul, we agree on something :)

[Paulhoff]

More on MP3's

http://www.iis.fraunhofer.de/amm/projects/mp3/index.html

Also look up J. D. Johnston - JJ - on this site. He may know something, Ha Ha, like a lot.

Paul

:) :) :)

[jj]

JJ,

Now that Fowlsound mentions it, and that I did my research ;) I wonder, have you ever tried (or the people arround you) those famous doble blind tests between the CD and MP3?

Well, I was in a number of the tests that MPEG-audio did, designed some of them, selected the material for some of them, and so on, so, yes, I tried them.

If you read the MPEG results you will find out that there was always a distinction between original and coded, even though it was often evaluated as "small" it was still quite statistically significant at all of the bit rates for MP3 that are below 192kb/s at least.

Even MPEG-2 AAC, which is substantially better, was not transparent at either 128kb/s or (barely) at 160kb/s.

In blind tests, ABC/hr, yes, yes.


jj - live from Toulouse, http://www.icassp2006.com/

[Bodhi Dharma Zen]

In blind tests, ABC/hr, yes, yes.

So, above 192khz is basically impossible to discern among them. Where can I find such studies? in the fraunhofer site? I will look for them

And I didnt understood your last sentence. Are the "yes, yes" answers for my last questions? This is, yes, above 192 they are practically indiscernible and yes with classical music there is some background noise and lack of resolution??

Guess I will have to run my own tests some day, maybe with a group of audiophiles I know ;)

[jj]

So, above 192khz is basically impossible to discern among them. Where can I find such studies? in the fraunhofer site? I will look for them


that's 192 kb/s first, not kHz.

And the results were not definitive at 192kb/s. There were some MPEG tests, I don't think that FHG has the info up, but I'll ask one of their guys tomorrow where I might re-find the document. They should be here for the plenary tomorrow.


And I didnt understood your last sentence. Are the "yes, yes" answers for my last questions? This is, yes, above 192 they are practically indiscernible and yes with classical music there is some background noise and lack of resolution??


The "yes, yes" was to the use of ABC/hr. It's going to be hard to discern differences. I can't say "impossible".


Guess I will have to run my own tests some day, maybe with a group of audiophiles I know ;)

It's hard, but you're welcome try. Check out BS1116 test protocols for starters.

[Bodhi Dharma Zen]

Thank you. My guess is that the differences Im hearing right now are because of the expectations. Even when I OBJECTIVELY know that Im unable to discern between CDs and 192kb/s MP3 (khz was a typo) using DBTs, I still have this horrible "feeling" that with classical music I can discern the difference. Its horrible not to know when our senses are playing tricks on us.

I will chech the BS1116 that you mention, and I will check this ABC/hr thing.

[Paulhoff]

Its horrible not to know when our senses are playing tricks on us.

Not the senses, but ourselves. Expectation is the key word here. I have seen it with my audiophile friend time and time again. I remember when we were listening to a CD that was hooked up to the receiver by coax for the digital input, the next thing I know he is changing the 3 foot coax to another 3 foot coax of a different type. Next thing he is saying is that the bass is stronger and more pronounced. I can not get him to understand how digital is not the same as analog. He is still thinking in the old phone cartridge ways of thinking, if that is a good term for it. I have seen him get up and move a speaker an inch or two, and sit down and say, "Yea, that better now". I will not get into the speaker wire and other cable wire things that he buys into (and pays a pretty penny for). Our Expectations and ourselves are the biggest problem when using only ourselves for the BDT.

Paul

:) :) :)

[Bodhi Dharma Zen]

Ok, read the protocol, or the bits I found, and I already have an ABX software. I will do some tests this week.

[Bodhi Dharma Zen]

I just did a small test. No need for statistics at all. I used Cool Edit Pro to rip a Genesis song, and then listened to the original and the 192 "high quality" version. I can easily pick the CD everytime, just by hearing about three seconds of the song. I need something more challenging, maybe a different coder?

[jimlintott]

In my car I have a CD wallet full of discs that are all MP3. Most are encoded at 320kbps and I would be hard pressed to tell them from the original CD, especially in the car.

The other day I grabbed a disc and was listening and thought that it didn't sound good and checked the encoding rate on the deck and sure enough it was 128kbps. Some of the content was the same as other high bit rate discs I have. If I had to describe some of the differences I would say that the bass sounded boomy and boosted but without definition. Overall everything sounded thin and lacking. It sounds better than FM but I noticed that it didn't sound all that great. I find that on FM radio the sound is acceptable but that symbols often just sound swishy. Similar to the sound that you get from crappy tweeters.

My car stereo is what is known as an SQ system rather than DB. Sound quality vs. LOUD. Still the higher bit rate MP3s are perfectly fine even for long periods at max volume.

[macgyver]

Just my two bits on this whole MP3 thing:

I'm not entirely sure why we're bothering with lossy compression algorithms at all, when digital storage density is getting greater, and the prices are getting lower?

In addtion there is LPAC and such that offer lossless compression of PCM sources?

http://www.true-audio.com/

IMHO, I just don't see the point for high-end audio reproduction purposes. You certainly wouldn't want lossy compression for archival storage either.

I guess I'm just of the opinion that if you don't HAVE to do it, then don't!

[Paulhoff]

I just did a small test. No need for statistics at all. I used Cool Edit Pro to rip a Genesis song, and then listened to the original and the 192 "high quality" version. I can easily pick the CD everytime, just by hearing about three seconds of the song. I need something more challenging, maybe a different coder?

Did you do a DBT, on this.

Paul

:) :) :)

[Paulhoff]

Just my two bits on this whole MP3 thing:

I'm not entirely sure why we're bothering with lossy compression algorithms at all, when digital storage density is getting greater, and the prices are getting lower?

As I have said before, it is good for recording very long radio programs. Also I am a ham operator, and it is very good for recording contacts that I have made on the short-wave bands. And the smaller size still can come in handy.

Paul

:) :) :)

Why used jpg when you can use BMP

[macgyver]

As I have said before, it is good for recording very long radio programs. Also I am a ham operator, and it is very good for recording contacts that I have made on the short-wave bands. And the smaller size still can come in handy.



Fair enough, but that's why I specified for high-end audio reproduction purposes. It can also be argued that it is best for copying the files over networks, but even to that end - the bandwidth gets cheaper by the day...

Small size = convenience - but I'm addressing more the issue of file size vs. quality and whether there's really a need for lossy compression in this case?

[Paulhoff]

Well, I would not use MP3 for high end, like my camera has Tiff if I want a truer picture over jpg. I am amazed of how good MP3 sound for the compression size that they are.

Paul

:) :) :)

[Bodhi Dharma Zen]

Did you do a DBT, on this.

Paul

:) :) :)

No need. And Im a fan of DBTs. When you get several in a row, with just listening some seconds, its because the difference is large. Well, large is not an appropriate word, but its clearly noticeable.

I can do it 100 out of 100 with 100% certainity. DBTs are handy when in the preliminar you miss one out of three or more.

Damn... Im sounding like an audiophile. :(

;)

[Bodhi Dharma Zen]

Well, I would not use MP3 for high end, like my camera has Tiff if I want a truer picture over jpg. I am amazed of how good MP3 sound for the compression size that they are.

Paul

:) :) :)

In fact I listen almost just MP3s. With my MP3 player (Zen Touch, 20Gbs) I have all the music I could want to hear right in my hand. And not only in my computer (equiped with fairly good MM speakers, the Gigaworks S750), but even in my large "High End" equipment.

[macgyver]

Well, I would not use MP3 for high end, like my camera has Tiff if I want a truer picture over jpg. I am amazed of how good MP3 sound for the compression size that they are.



The MP3/JPG analogy is a good one, actually. I'm a semi-pro photographer (schooled to be a pro, 'semi' because I don't make any money at it), and for my 'family snapshot' pics I use a Panasonic Lumix (amazing Leica lenses on the Panasonic) which offers TIFF.

I've stopped using TIFF and opted for lowest compression JPG for a couple reasons (as I jump completely off-topic):

TIF files are not RAW, and actually have all the same post processing (except compression) applied to them as JPGs do. So I find the resulting images virtually identical.

I would expect that if I were to post process the JPG, then I would see a difference after a few extended generation copies, but this is of little concern to me, since I "edit in camera" and generally print exactly what I shoot.

Lastly (but a minor point on my 1Gig SD card) the TIFs are very large for no apparent advantage. They would quite likely be even larger than a RAW file of the same image. A RAW file would offer much more advantage to the advanced photographer, and likely in a smaller package.

I don't understand why some manufacturers offer TIF, but not RAW...

[Paulhoff]

I took a few pictures of a big blow up pumpkin on someone's lawn with jpg and tiff, when I enlarged the pictures, the jpg have a little off colored green where it was orange, where the tiff did not and there are other features too, but again for the compression it is fine. If one knows the compromises all is fine with me. The know the trade offs with MP3 and jpg and I am OK with it.

Paul

:) :) :)

[Paulhoff]

No need. And Im a fan of DBTs. When you get several in a row, with just listening some seconds, its because the difference is large. Well, large is not an appropriate word, but its clearly noticeable.

I can do it 100 out of 100 with 100% certainity. DBTs are handy when in the preliminar you miss one out of three or more.

Damn... Im sounding like an audiophile. :(

;)

Sorry, here I don't agree, you must do a DBT; we can fool ourselves much easier then you think.

Also I use MusicMatch 7.1, {I didn't upgrade, even if it is free for me, I don't need all the new S**T). My son used the window’s version, and I can tell it is a bad version of MP3 converter, and have done the DBT with it.

Paul

:) :) :)

[macgyver]

If one knows the compromises all is fine with me. The know the trade offs with MP3 and jpg and I am OK with it.


True enough. I still use low ISO transparency film for my "good" work, partly for the "distortions" that film inherently has, and partly for the resolution.

[Bodhi Dharma Zen]

Sorry, here I don't agree, you must do a DBT; we can fool ourselves much easier then you think.

Also I use MusicMatch 7.1, {I didn't upgrade, even if it is free for me, I don't need all the new ****). My son used the window’s version, and I can tell it is a bad version of MP3 converter, and have done the DBT with it.

Paul

:) :) :)

And you are right. I didnt explain well myself. I didnt do a DBT, but I did a blind test. 10 out of 10 I can easily pick up the differences with just a few seconds of playing.

My wife selected either the MP3 or the CD track and turned of the monitor, then I click to select either the first or second sample. True, at first I was not sure of which was which, but after several seconds I noted a difference, and now I can pick it all the time, its that clear.

Without turning on the monitor I know which is the original and which the 192kb/s MP3 (obviously I confirmed it turning on the monitor after the trials).

Maybe the encoder in Cool Edit Pro is not that good, because I did some tests in the past using LAME and I remember it was a lot more difficult to find a difference.

[Paulhoff]

MusicMatch is a very good encoder, it is licensed by the Fraunhofer Institute.

Paul

:) :) :)