Audio DiffMaker

[Ducatista47]

I also believe that listening to things with a microphone is instructive. Most measurement systems used at a hobby level employ a $50. omnidirectional mic. They nearly don't even care which way they are pointed.

Try listening to your system through a quality cardioid measurement mic and headphones in real time while moving and pointing the mic around the room. This will give you some idea of just how much filtering the gray matter computer between your ears is doing, and help you focus on what and where improvements can be made. This is much like soloing a channel on a mixing desk.

Barry.

Or even look at the plots of two measurements taken a few cm apart. Run them through the Diffmaker and you'll see/hear a huge difference and it is the same system in the same room.???

Widget

Total agreement. Small differences in listening location, like getting up to change something and sitting down again, have long been known to alter what is heard. A late friend who ran an audio dealership for years would say things like, "I have never heard an interconnect or speaker wire that didn't change the sound. It is not always a good change, but there always has been one."

I knew better than to propose that a movement of a centimeter would change a lot. He didn't want to hear that reasoning. He made a lot of money selling wire.

That's another reason DiffMaker seems like a useful tool to me. The microphone setup, unlike our ears, doesn't move between recordings.

[Ian Mackenzie]

I have no doubt differential can determine a change.

However as Floyd Toole has proven we largely listen with our eyes when it comes to sound reproduction!

This is why blind testing is important.

In Tooles paper on you tube he explains in laymans terms whi this is the case.

Going further, the research outlines what in fact the ear/brain is sensitive to in terms if changes in sound reproduction.

Some changes we are quite tolerant of while others are not .

This is particularly relevant to evaluation loudspeakers frequency response and impact of the listening room.

Correlate this study with enough data and you have grounds for determining what sounds good and what doe not!

Unfortunately the audio consumer is always on the look out for what could improve his listening experience and this is the motivation of the audio mafia to make money.

Twisting the circumstances with massive cables that are visually appealing to the eyes screws with the brains logic and you easily get a "yes" it sounded better.

However, what the listener (the customer) does not know the listener does not know in terms of the risks of good and audio reproduction beyond what is printed in the advertising spec sheet like 0.002 % distortion.

On the other hand there is no argument a JBL 1500 AL driver is superior to a JBL 2235 driver.

But in a double blind test what aspects of the audio reproduction from a range of completed loudspeakers using these drivers will the listen hear as audible likes and dislikes?

A system with good drivers but in a poorly executed design may end up not being preferred to lesser quality drivers in a well executed system and so it goes on.

This is why loudspeakers advertised with expensive tweeter may not be rated as appealing to a lower cost tweeter that is mounted with a correctly designed wave guide and crossover network.

But if the listener first visually sees and identifies the expensive tweeter he may indicate he preferred the system using that tweeter while a under a blind test he may prefer the other tweeeter with the wave guide.

With amplifiers for a long time the spec sheet ruled. These days research indicates that some 3rd harmonic distortion with increased amplitude(volume) is preferred but not higher odd order harmonics.

My own interpretation is that the listening experience is like a photographer looking through a lense of an SLR camera and taking a picture. The printed image is what you see in the end.

How do you judge what is better image between two or more photos taken by different photographers with a different lenses of the same subject?

Throw marketing claims in the mix about a particular lense and the visual appearance of a big chunk of glass and you might automatically prefer the image taken with that lense over another.

[Mitchco]

One can find several DiffMaker tests on Archimago's blog. Here is one: http://archimago.blogspot.ca/2013/05...site-dmac.html

I have used DiffMaker as well:
http://www.computeraudiophile.com/co...ty-comparison/
http://www.computeraudiophile.com/bl...t-results-156/
http://www.computeraudiophile.com/bl...l-results-155/

DiffMaker has a few bugs, but the workarounds are in the above posts. DiffMaker is fun and educational - especially listening to the difference file.

I should have mentioned the above links also include ABX listening tests, plus the difference files one can download and listen to.

Which is the crux of the matter, how does one correlate the visuals to the audibles?

This article http://www.computeraudiophile.com/co...ility-testing/ contains amplitude audibility testing and bit-depth audibility testing. From the article, "Note, if you are following along with the downloaded files, adjust the volume for your typical listening level and remember the setting. The idea here is to not only compare the files, but through the course of the listening tests, the difference levels will change, reducing to inaudible. It is educational to listen to the difference files over the course of the test using the same monitor level throughout.

This approach helps put into perspective how we perceive relative loudness. The only way to achieve this is to set the volume control once and leave it there. However, occasionally we will need to increase the volume to maximum to find the noise floor. Be sure to mark the original level so that it can be returned to."

From the article, in conclusion: "I set out to determine how far away from bit-perfect I can hear an audible difference in sound quality. In the case of adjusting amplitude with a digital eq, my audibility threshold is about 1 dB of change. 4dB of eq was easy to hear, but I could not detect a .1 dB eq change. In the case of adjusting bit-depth, my audibility threshold is around 10 to 12 bits, using rock music as the sample. At 8 bits I could easily detect a change, but not at 12 bits. In both the amplitude and bit-depth test cases, it appears that my audibility threshold level is around -70 dBFS when comparing an altered file to the original file."

I wrote those articles to satisfy my own curiosity as to what was or wasn't audible to my ears. If one looks for scientific research on psychoacoustics, my results seem to be consistent with other's, when comparing the original file with an altered file, listening at a reference level of 83 dB SPL C weighting, slow response, on a SLM. It would be interesting to see other forum members results if giving the files a listen.

A good online site for audibility testing: http://www.audiocheck.net/ in case one has not seen it.

With respect to DiffMaker, from Bill's AES paper: http://www.libinst.com/AES%20Audio%2...ng%20Paper.pdf "The difference test doesn’t detect just aurally relevant changes, it detects audio band changes of any kind."
I found the paper a good read to understand the algorithms used in the software.

Based on my own audibility testing, I concluded if I measure a difference file to be below -70 dbFS in DiffMaker or Audacity or whatever, then I am going with that it is not audible to my ears and carry on.

Enjoy the music!

[DS-21]

I understand that, and most quality amps or other line stage devices have little if any spectral differences or colorations. The differences I am listening for are not deviations in frequency or tonal colorations, I am listening for the much more elusive spatial cues in stereophonic recordings.

I really hope you don't actually expect "spatial cues" from a mere gain block. Unless you're talking about one with some sort of signal processing, such as a crossfeed circuit on a headphone amp. Or one with really poor channel separation.

To continue the speaker analog...

That is a mistake, because there is no reasonable analog between a mere gain block and a transducer, but going on...

compare a beryllium diaphragmed driver with a more conventional one. The frequency response could be identical, but the acceleration factor and distortion levels are not and the subjective difference is very real.

I don't subscribe to "magic materials" thoughts either. The bottom line is, drivers sound different because of either motor issues or diaphragm resonances. Nothing magical, and nothing not thoroughly explained by our current models. Even if one is made out of a crystallized fairy dust matrix and the other out of dog poop.

Be can push resonances higher than other current materials, if used well. A difference that's easily measurable. (That's why the TAD Reference line has bigger midranges than the TAD Evolution line; the Mg cone midrange in the Evos needs to be smaller to play as high cleanly, because of diaphragm resonances.) But otherwise there's nothing special about it, except that audio marketers have glommed onto the word as some sort of fetish talisman. Note that JBL themselves have moved away from it and towards cheap plastic rings in a novel configuration for their new flagship CD line. I've not read many complaints about the M2's top end, though unfortunately I still haven't had the opportunity to hear them for myself.

I have no doubt differential can determine a change.


However as Floyd Toole has proven we largely listen with our eyes when it comes to sound reproduction!

Yup. The difference is that it's much much easier to take two recordings with DiffMaker than to run a serious controlled listening test. Ultimately less resolving (just because DiffMaker gives you something other than silence, doesn't mean the difference is audible or significant) but a whole lot easier.

[Ian Mackenzie]

I understand that, and most quality amps or other line stage devices have little if any spectral differences or colorations. The differences I am listening for are not deviations in frequency or tonal colorations, I am listening for the much more elusive spatial cues in stereophonic recordings.

There are also qualities that are much harder to describe, but often lower quality (not lower cost) electronics sound congested and tend to sound more compressed with less dynamic detail... or detail in general. (Not to be confused with a slightly elevated midrange which can subjectively sound like an increase in detail... we see this often in speakers.)

Widget

I think Mr Widget is referring to transparency and resolution.

Just like listening to a nice Chesky recording and then a non Chesky producition. Assuming your playback system has the transparency and resplution those spatial quality will be audible they are otherwise lost or masked in a lesse system.

For example comparing the PSS 600 power amp l once raved about that had amazing specs on paper to the Pass Labs 250.5 the amount of low level spatial information is astonishing.

But l use a high resolution source and loudspeakers when making that statement.

This then becomes an audio education when one goes back to lower resolution recordings and equipment.

If l used ITunes as a source l would not be making test statement.

[Mr. Widget]

I really hope you don't actually expect "spatial cues" from a mere gain block...

I'm not expecting it to create them, but if you've actually listened to a variety of circuits you must have noticed that the perceived soundstage will vary... perceived musical instrument location within the sound field etc. Of course this presupposes that the listener is sitting in the center of a stereophonic playback system and are actually listening as opposed to casual listening off to the side of the system etc.

I don't subscribe to "magic materials" thoughts either.myself....

I was thinking of the basic physics of the materials... which as you pointed out affects relatively easily measured differences in resonance and distortion. The "analogy" I was trying to make was that by substituting materials it would be possible to have nearly identical frequency responses and yet different sound qualities.


Widget

[gasfan]

I'm not expecting it to create them, but if you've actually listened to a variety of circuits you must have noticed that the perceived soundstage will vary... perceived musical instrument location within the sound field etc. Of course this presupposes that the listener is sitting in the center of a stereophonic playback system and are actually listening as opposed to casual listening off to the side of the system etc.

I was thinking of the basic physics of the materials... which as you pointed out affects relatively easily measured differences in resonance and distortion. The "analogy" I was trying to make was that by substituting materials it would be possible to have nearly identical frequency responses and yet different sound qualities.


Widget

I can see this software being especially useful in identifying malfunctioning components. Perceived sound stage only varies among those circuits whose R and L signal levels are not identical. Differences are infinite. Accurate is always the same.

[Ian Mackenzie]

Channel separation is another factor as in signal to noise ratio.

There is a diminishing return on low distortion if the system noise floor mask distortion products.

Typically it's the volume control and the inherent noise floor of the circuit that effect inter channels balance and perceived sound stage.

Muse of Japan now have a good digitally control analogue volume control chip.

The chip has an internal resister ladder that is switched seamlessly.

This chip is finding its was into hi end preamps because it offers lower noise = spatual cues from low level detiail in recordings while holding exact channel balance.

[Ian Mackenzie]

I don't accept the comment of 1 db of change when evaluating the L Pad levels of 4 way loudspeaker.

Overall system level 1 db maybe an indication of the threshold of a change in amptitude.

However that is what the ear brain is acutely sensitive to.

If when adjusting L pad levels 0.5 db is discernible between bass/mid/high frequency bands.

This supports evidence that -+ 0.25 db for riaa phone equalisation is regarded as accurate riaa equalisation.

I think this would be a far more useful test for the software.

[gasfan]

Channel separation is another factor as in signal to noise ratio.

There is a diminishing return on low distortion if the system noise floor mask distortion products.

Typically it's the volume control and the inherent noise floor of the circuit that effect inter channels balance and perceived sound stage.

Muse of Japan now have a good digitally control analogue volume control chip.

The chip has an internal resister ladder that is switched seamlessly.

This chip is finding its was into hi end preamps because it offers lower noise = spatual cues from low level detiail in recordings while holding exact channel balance.

That's all well and good but the issue I was alluding to in Widget's post refers to channel balance at the output stage of a ss amp. If quiescent current is not held to within 1/10th of a mv between channels, nothing you can do afterward will restore what's been lost. Maintaining the phase relationship between channels is the most critical factor in reproducing the stereo image. Sonically it's similar to speakers out of phase to an extent. Especially bottom end which is often recorded in mono. You may not notice it because you are listening to two speakers with information coming from two sources. But as the mismatch increases, you hear less and less stereophonic spatial cues until all you have is
basically two sources. It's kind of like stereo. But you're being fooled. If 95% of us would simply go out of our way to truly match quiescent current in our amps, we'd stop

trying to tweak the life back into the music because that's where it gets lost.

[Ian Mackenzie]

well that's all well and good too but l doubt the crossover networks (after the amp) are any where near that tolerance and given this impacts on the phase response of the left and right channel the mismatch in production of the amp left and right channel is in significant, like 100/1.

Pass labs amps have super symmetry but that is to cancel distortions by a factor of 20 db within the balanced output stage of arch channel.

Some people say let an amp warm up for two hours to ensure everything is stable and work in the right temperature.

Perhaps that is why leaving a preamp on indefinitely is common practise.

But l don't think this relates to bias imbalance.

Take this that most crossover caps are 5%, L inductors 5% in production l find it unlikely the networks would be hand matched with a precision LCR meter.

Then look at the gross room in balance due to interaction above the Schroder frequency due to lack of room symmetry in s typical domestic room (critical according to Floyd). Compared to any phase mismatch up stream else where this is like 1000/1.

The work by Floyd and Geddes pin points the critical aspects of phase and amplitude in audio reproduction.

It's quite intuitive to read their papers.

What l am saying is in electronics it's all relative.

My own experience is that increased bias current in preamps and power amps increases apparent image depth.a typical audio chip is biased to 0.1 Ma. Compared to a discrete opamp l have built biased to 10 Ma the image depth and micro dynamics are superior.

Most commercial power amp run tiny amounts of output stage bias current and typically rely on feedback in various topologies to arrive a marketable specifications.

Most of them sound terrible compared to an ML, Passlabs or similar that make few concessions in heat sink real estate due to bias current.

The better desks run discrete opamps for this kind of audio performance where the signal would otherwise be scrubbed by numerous chip opamps stages in the recording process.

The Audio Diff maker might be useful in recording QA when Compared to the acoustics domain specific aspects of The human ear/brain make some issues in significant and other critical. That stuff is where the IP is useful .

Put it another way in a optimum acoustic environment the electronics are not on the spectrum, it's the loudspeaker.

[gasfan]

Theory is one thing and as important as it is, practical observation is what counts. I would love to have a shoot out over this exact issue. I can guarantee heads will spin. In any case I challenge anyone to prove me wrong. Just be sure the settings haven't drifted by the time you put the cover back on. This is exactly why a wide margin of tolerance is considered acceptable. Everybody knows it's going to drift. But has anyone ever made sure it didn't just to see? Hence two identical amps will almost always sound different. No one questions it because on paper it shouldn't matter. On a graph, approaching equilibrium, sonic improvement would look like an exponentially ascending line that quickly straightens out on either side within 1/10th of a mv of difference. Test it for yourself.

Btw, I sold a Meridian 559 a while back because it's 'auto-bias' function was no where near precise enough.

[hsosdrum]

I get your overarching point and in general tend to agree, however are you certain about the best mic/best ear comparison? The unit of one decibel is by definition the smallest sound increment that a human with normal hearing will perceive. A quality microphone will register small fractions of a decibel.


Widget

True, but measurement mics don't pick-up sound the same as human ears—only a binaural head does that. Once you're talking about a mic sitting on a stand it's an apples and oranges comparison with human hearing, which taints any result. (This is one of many reasons why we can hear differences we can't yet measure.)

And this doesn't even get into the brain's ability to process the nerve impulses it receives, which is far greater than any audio measurement system extant.

[hsosdrum]

How about amplifiers? There are many that seem to measure essentially identically but in my experience do not sound the same. Now I know people who I hold in high esteem who claim there is no sonic difference in quality amplifiers. I have to respectfully disagree with that statement.

I have experienced numerous cases where people both interested and disinterested in high fidelity have recognized differences in amps when no one present knew which test subject was being auditioned.


Widget

Amps that measure identically can (and do) sound different because once you connect them to a speaker they become part of a larger circuit, and the speaker part of that circuit absolutely influences how the amplifier part performs (and visa-versa). It's impossible to perform measurements on an amp when it's connected to a speaker because you can't be sure if you're measuring the amp or the speaker's influence on it. That's why they measure amps with them connected to test resistors: the resistors eliminate variables. But when you measure an amp connected to resistors you're measuring an incomplete circuit that doesn't behave like the one that's formed when you connect the amp to a speaker. So the measurements, like the circuit, are also incomplete: they cannot tell us enough about how the amplifier performs in the real world.

[Ian Mackenzie]

Originally Posted by Mr. Widget

I'm not expecting it to create them, but if you've actually listened to a variety of circuits you must have noticed that the perceived soundstage will vary... perceived musical instrument location within the sound field etc. Of course this presupposes that the listener is sitting in the center of a stereophonic playback system and are actually listening as opposed to casual listening off to the side of the system etc.

I was thinking of the basic physics of the materials... which as you pointed out affects relatively easily measured differences in resonance and distortion. The "analogy" I was trying to make was that by substituting materials it would be possible to have nearly identical frequency responses and yet different sound qualities.


Widget

I can see this software being especially useful in identifying malfunctioning components. Perceived sound stage only varies among those circuits whose R and L signal levels are not identical. Differences are infinite. Accurate is always the same.

I am not sure I agree that your hypothesis is correct and would related the the pshyco acoustic perceptions as you indicate.

I agree that if there is a channel imbalance in amplitude this will certainly effect sound stage shifts and the volume control is documented conclusively to mess up the sound stage over its range of adjustment. Scientists like Doug Self have spent a lifetime looking a solutions.

My take on this is The Widget was referring to different circuits in different amplifiers, ie the gain stage and associates stage like the volume control and or balance control.(not differences in amplifers over time or the same model and one location not performing as well as another due to mal adjustment.)

What I think The Widget was saying is that one preamp or power amp can by design have an impact on the sound stage.

If the volume is static and the levels are perfectly matched the sound stage can perceptually seem different.

Other factors suporting this including independant power supplies and earthing and PCB paylout fundermentally impact on the apparent percieved sound stage. This is acknowledge by professionals like John Curl.

The selection and type of semiconductors can also have a significant impact on sound quality as does how the devices and the surrounding circuits are set up.

Lets talk practically for a moment:

A while ago I re build a faulty SAE 50 watt / channel amp with modern semi conductors and new PS supply caps.

Some of the parts had previously been replaced with the incorrect substitute transisters!

According to the owner, the subjective improvement was night and day!

Does this correlate to perceived changes in sound stage?

It was while ago and all I can say was it performed far better than it had any right to.

The improvement in clarity and definition was remarkable but I let the owner draw his own conclusions.

In a rare show I faith I support The Widgets comments.

In theory the CMRR (the ability of a stage to reject noise or chatter on the supply rails) in a circuit designs can be good or bad depending on the design desicisons and budget of the production. In theory this might mess up pereived soundstage.

In practise bais current can drift with temperature but it normally difts in a linear manner with both channels as the VAS is effected by temperature.

The hotter the amp gets the higher the bias.

In a class A/B amps they use elaborate bias contols to ensure quiestent current is stable, sensors on heat sinks and some transisters now have inbuilt thermeral sensors to ensure rapid bias control. All this does is turn on the power transisters by a certain amount to control crossover distortion at low power levels 1-3 watts and to stop the amplifer destoying itself as the quiestent current would otherwise continue to rise and till the power transisters blew.

In your comment on the bias per mv, this will depend on the value of the emmiter resister and their and the matching of the output devices.

Most class A/B amps run around 50 ma of queisent current in the output stage.

Assuming 0.2 ohm emmiter resisters and say 2 parrellel devices, then 0.005 (5 millivolts) across each of the resisters that would be 50 ma bias on the output outstage.

What you are saying is 1/10th accurancy in bias current is required for stable stereo image.

That would be like 0.0045 mv or 0.0055mv across the emmiter resisters. This would decrease the output stage bias down to 45 ma or increase to 55 ma.

Thats 10% change in bias current but its more about the effect on the amps crossover distortion at low power levels, like 1-3 watts.

At higher power levels the amp goes into class B.

While I get it that matching quiesent current between L + R is desirable I would suggest this is not likely to impact on sound stage.

The other factors I mentioned above are going to swamp this.

That said the matching or non match of devices is going to be a far the most significant issue effecting quiestent current in a production amp than incorrect adjustment. Throw in 40-60 db of feedback, that is there to flatten out any wrinkles.

However, I will fire the question to Doug Self and see what he comes back with?

In Class A amps this is important and they are usually matched within 5%.

I know Pass Labs adjust for the on-lid bias condition after a long burn in.