32 bit stereo ADC, wow

[living sounds]

If you can reliably hear it your should be able to measure and quantify it. If you can measure it you can often improve it. While that is less of an option with large scale ICs.

Over the decades I have found that I can measure things I can't hear, but never heard something I couldn't measure.

JR

PS: One technique to enhance generational degradation from conversions between a/d and back d/a again, is to loop the same signal through a conversion path multiple times. After enough passes, path weaknesses can reveal themselves (or not).

Sure, but the problem seems to be how to measure it. We are talking about differences easily audible within a 44,1 khz / 16 bit format.

 

[Newmarket]

Well that's all very well. But what sort of Floobydust ? Analogue ? Digital ? Discrete ? Jazzophonic ? Valve ?

 

[JohnRoberts]

Sure, but the problem seems to be how to measure it. We are talking about differences easily audible within a 44,1 khz / 16 bit format.

I've told this story so many times I'm bored by it... Back in the 70s/80s I was not satisfied with my bench equipment to parse out circuit deviation from ideal. Long story short I modified an old school (too easy) SMPTE IMD analyzer (Heathkit of course) to use 19kHz and 20kHz tones instead of 7kHz and 60 Hz. This was very effective at showing flaws in my phono preamp designs that looked great using conventional THD+N.

Of course those tests are now available in competent bench test equipment. For testing digital conversion may some kind of null testing against a reference path. Digital conversion complicates things with tiny delays that can degrade nulls.

Measure it and then you can manage it.

JR

 

[cyrano]

Benefit ? How ?

When I recorded to tape, I needed to fit nature in less than 60 dB of dynamics. Hiss was often a problem with quieter sources.

The 96 dB range my first digital handheld provided was a godsend. No more hiss. Still, unexpected loud sources produced awful distortion. It took a while before I realised I could record at let's say -44 dB average and avoid that problem. That alleviated the problem but didn't solve it completely.

With 148? dB of dynamic range, it will reduce the problem even further.

Have I been had by Sound Devices' marketing?

 

[Newmarket]

When I recorded to tape, I needed to fit nature in less than 60 dB of dynamics. Hiss was often a problem with quieter sources.

The 96 dB range my first digital handheld provided was a godsend. No more hiss. Still, unexpected loud sources produced awful distortion. It took a while before I realised I could record at let's say -44 dB average and avoid that problem. That alleviated the problem but didn't solve it completely.

With 148? dB of dynamic range, it will reduce the problem even further.

Have I been had by Sound Devices' marketing?

Okay. So with enough bits/dynamic range you can set a level and get both sources recorded at useable levels without saturation. But the noise of the jet will still reduce intelligibility of speech.
It seems that for that particular scenario some pre-conversion analogue multiband dynamics processing, or a noise cancelling approach, would be a better path to a solution.

 

[cyrano]

It would be, but nobody makes it...

 

[davezawadi]

I agree with John Roberts above. I remember the great noise about certain types of cables at one time when I was at SSL. Certain people were certain that one type or another had a signature that was audible. Measurements showed nothing at all as a difference even at very high signal levels, so we went for proper double-blind testing. This means that the judges know nothing about the test setup, or what they were hearing. There was no statistical difference between the results, and quite a few people realised that preconception was the problem. I also did the same testing with various microphones and a musician. Could anyone reliably identify various types of microphones although all were very familiar with the whole range, from U87, KM184, and lots of others? Various characteristics could be identified, but nothing we could not also measure with an Audio Precision System 1D. The ear-brain system is very curious indeed!

 

[davezawadi]

That Newmarket is not the audio system, it is your ears-brain system! There is nothing you can do about it, and if that worries you a perceptual codec must be instant death but seems to be acceptable to almost everyone today. Using 32 bits for audio is going too far, nothing else in the chain can have this dynamic range, it is miles below the noise level of your speaker power amplifier or even any recording studio. Certain microphones have noise levels around that from air molecules hitting the diaphragm, this limits the real dynamic range that can be used to around 100dB, which is approximately 16 bit audio, so 24 bits is more than enough. You also need to realise that an ADC or DAC is not accurate to its resolution, those are two completely different numbers.

David CEng etc.

 

[Newmarket]

That Newmarket is not the audio system, it is your ears-brain system! There is nothing you can do about it, and if that worries you a perceptual codec must be instant death but seems to be acceptable to almost everyone today. Using 32 bits for audio is going too far, nothing else in the chain can have this dynamic range, it is miles below the noise level of your speaker power amplifier or even any recording studio. Certain microphones have noise levels around that from air molecules hitting the diaphragm, this limits the real dynamic range that can be used to around 100dB, which is approximately 16 bit audio, so 24 bits is more than enough. You also need to realise that an ADC or DAC is not accurate to its resolution, those are two completely different numbers.

David CEng etc.

Not understanding why I am referenced in that comment. Or what I am (not) "worried" about ?

 

[abbey road d enfer]

Not understanding why I am referenced in that comment. Or what I am (not) "worried" about ?

That's a common problem, when people post an answer to an unquoted question.
God will recognize his own...

 

[davezawadi]

Because one sound hidden from your hearing by another is not the "recording", nothing can be done about it. You will not succeed with "noise canceling" perhaps with another microphone to pick up just the "noise", it cannot work well although you might get some improvement of pure tones, perhaps the whine. You may be able to EQ some of the noise away, assuming the whole record chain is linear, and perhaps cancel a bit more, and this is where a very wide recording dynamic range can be helpful. To get full cancellation you need several things, each of which is largely unachievable. These are exact level matching to a tiny fraction of the level difference (say to -50 dB), exact delay matching over the frequency range, and exactly the same frequency response, again to huge accuracy. Try some experiments with white noise fed to two channels, one inverted, on a mixer. Tiny changes in gain, eq etc lead to little cancellation, which should be perfect. Add 2 microphones in different positions to the picture and you are lost!

 

[Newmarket]

Because one sound hidden from your hearing by another is not the "recording", nothing can be done about it. You will not succeed with "noise canceling" perhaps with another microphone to pick up just the "noise", it cannot work well although you might get some improvement of pure tones, perhaps the whine. You may be able to EQ some of the noise away, assuming the whole record chain is linear, and perhaps cancel a bit more, and this is where a very wide recording dynamic range can be helpful. To get full cancellation you need several things, each of which is largely unachievable. These are exact level matching to a tiny fraction of the level difference (say to -50 dB), exact delay matching over the frequency range, and exactly the same frequency response, again to huge accuracy. Try some experiments with white noise fed to two channels, one inverted, on a mixer. Tiny changes in gain, eq etc lead to little cancellation, which should be perfect. Add 2 microphones in different positions to the picture and you are lost!

But "full cancellation" wasn't the matter being discussed. No one was positing that perfect cancellation could be realistically achieved.

 

[abbey road d enfer]

But "full cancellation" wasn't the matter being discussed. No one was positing that perfect cancellation could be realistically achieved.

I believe cyrano in post #22 hinted at that. Of course he didn't suggest "perfect" cancellation. I believe the subject is what practical improvement results from 32-bit conversion. The jury is still out...
Most DAW's take the 24-bit flow from the converters and re-format it to 32-bit float, which allows much more accurate processing inside. It does not improve conversion in the least.
My understanding is that Audio Devices do it in their mixer-recorders, which makes sense, taking advantage of the already existing 32-bit AES/EBU format.
Where their mktg dept is laying it on thick is when they suggest that the improvements in post-processing result in improvements in conversion.

On the subject of cancellation, it is clear that no basic processing that can be done in a DAW can achieve it, but there are adaptive algorithms that work wonders, but require enormous DSP resources.
A hint can be seen with Melodyne, where a single note in a full track can be bent/pitched/autotuned...

 

[Newmarket]

I believe cyrano in post #22 hinted at that. Of course he didn't suggest "perfect" cancellation. I believe the subject is what practical improvement results from 32-bit conversion. The jury is still out...
Most DAW's take the 24-bit flow from the converters and re-format it to 32-bit float, which allows much more accurate processing inside. It does not improve conversion in the least.
My understanding is that Audio Devices do it in their mixer-recorders, which makes sense, taking advantage of the already existing 32-bit AES/EBU format.
Where their mktg dept is laying it on thick is when they suggest that the improvements in post-processing result in improvements in conversion.

On the subject of cancellation, it is clear that no basic processing that can be done in a DAW can achieve it, but there are adaptive algorithms that work wonders, but require enormous DSP resources.
A hint can be seen with Melodyne, where a single note in a full track can be bent/pitched/autotuned...

Yes - although I didn't think they meant some perfect solution rather simply some benefit in the example used - and see his reply to my question in post 46.
I mentioned cancellation in passing as we know that such techniques are used to benefit in various cases - headphones, mobile phones, 'Zoom' type comms etc.
Actually in the example used - speech with a loud background noise - then a hypercardiod mic deployed effectively and bandwidth restricted to a "speech" bandwidth will go a long way

 

[cyrano]

Just anecdotal musings from an old geezer:

The BBC spent a lot of time and effort performing listening tests in the fifties, mainly with classical music. They tried to find the best reproduction system, as close as possible to live performance. It was the beginning of stereo, the era the BBC monitors were started

The tests yielded very little results besides what was already known. I just remember three conclusions:

A - The best test results came from ordinary people who where regular audience in live performances.
B - The worst results came from musicians.
C - Engineers seemed to sometimes perform better, sometimes worse. What was amazing, was that they seemed to concentrate and agree on a different problem every time they were asked to perform a test.

I already knew about A. It's not a surprise, as these are the people who know how instruments sound.

B wasn't a surprise either, as most professional (classical) musicians are pretty much deaf anyway.

C was very surprising to me. Not the fact that the engineers agreed, but that they seemed to concentrate on a different aspect every time.

 

[abbey road d enfer]

B wasn't a surprise either, as most professional (classical) musicians are pretty much deaf anyway.

Not only. The sound they hear is not the sound that is intended to project to the audience.

C was very surprising to me. Not the fact that the engineers agreed, but that they seemed to concentrate on a different aspect every time.

I've often found that one particular SE claims hearing things nobody else hears.
Is it a more or less deliberate attempt at creating a discriminating perception that allows them to market a unique talent?
Quite often, when I was participating in vinyl mastering sessions, I found that some EQ or dynamics tweaks were not entirely justified, meaning that without the tweaks, the result was equally good as with tweaks, of course a tad different, but not clear cut. ME's have to justify their salary and create an aura of mysticism around their job in a competitive business.
Of course, listening is a talent based on hours and hours of practising.

 

[cyrano]

That mysticism certainly isn't exclusive to audio engineers, I think. I used to be in IT. Lots of the same stuff, including the worshipping of certain brands of gear.

 

[abbey road d enfer]

That mysticism certainly isn't exclusive to audio engineers,

Probably. However, there are many domains where I'm far from being an expert, so I refrain commenting on them, but when it comes to audio I have a good BS detector.

 

[JohnRoberts]

I joke that life is an IQ test and critical thinking is a useful skill.

It seems like some disciplines attract more magical thinking than others. Audio equipment certainly has more than average. Some consumers support this by rewarding those hyperbolic marketing claims, with purchases.

JR

 

[living sounds]

I was never happy with converters vs the analog source but couldn't put my finger on it. Ironically, it was the hifi people (not synonymous with audiophools) who got me to check out R2R DACs.

I also think there frequently is a kind of blindness with experts of the "if you cannot measure the problem with the tools you've got, it doesn't exist" variety.

As for magical / wishfull thinking - yes, it's everywhere, in hifi audio as well as in pro audio. Lot's of highest profile grammy winning recording engineer geniuses (I mean that) who believe some absurd stuff. It's human nature to fill the gaps somehow, and in audio, due to its intangible nature, there are a lot more gaps than in many other fields, I think.