Does digital audio work like digital images...ie more bits for highend?

Things change a lot when the Nyquist Theorem is implemented in the physical world with actual electronic circuitry.

Without getting into the hairy mechanics of their operation, a DAC includes circuitry to form true sine waves i.e. smoothly curving waveforms based on the sample points it's given - otherwise the sound that would result would be completely unlistenable (a 20kbps MP3 file would sound like heaven in comparison).

BTW: analog gear also introduces it's own set of problems when it handles a music signal and like digital, what exits the other end is NOT exactly the same as what entered.

 

LeeS, that graph has been known to be misleading. A dirac spike is infinite (not a representation of a musical signal) thus this graph only shows the filters at play in each case.

 

I'm not so sure. I think it gets to the differences in approximating a wave.

 

Here's another question that may be asked:

 

That plot also presupposes that the analog chain has a bandwidth exceeding 300khz and a flat phase response throughout the passband. Which is trivially false for virtually every analog medium in existence. IIRC, given that DSD rolls off after 50khz, the DSD plot isn't terribly trustworthy either.

I agree with MeToo (heh), but even if the impulse response plots were accurately plotted, I don't think it would be all that important. The ear doesn't hear waveforms.

 

I'm not sure I follow. The ear hears an analog waveform that is being played back after recreation. It seems to me that better recreation of the waveform such that it is closer to the originally recorded event would be valuable.

 

The auditory system decomposes pressure waves into a filterbank with reasonably well defined frequency limits. There are complications involved with phase error sensitivity (or lack thereof) at high frequencies, but once the sound hits the cilia, it's more like a FFT than a waveform.

After all, that's why lossy codecs use filterbanks. It's how the ear hears, and manipulating the sound in that domain is much more efficient (and results in much better sound quality) than manipulating it in the time domain.

That waveforms are far easier to manipulate visually/computationally, and record onto analog/digital media, doesn't matter in the context of what actually goes on in the ear.

 

Now resample the sinewave sampled at 44.1kHz up to 192kHz and post a screenshot of the result...

 

LeeS, no need to post things defending HiRez. You'd be preaching to the quire as far as I'm concerned.

I am simply pointing out several things I have noticed and learned regarding solely the SACD format, not HiRez in general.

 

Quite a few players have a filter at around 50kHz for SACD playback. On some you can activate/deactivate it at will. Mine does not have any. I was told so by Spain's product manager and later have been able to confirm it when I have recorded the analog outs of my player when reproducing SACD at 192/24. In these cases, the high frequency noise can be seen up to near 100kHz.

To cut/filter the signal at around 50kHz would make it approximately the equivalent of a 96/24 LPCM's frequency output. This is surely the reason why many players opt to output DSD as 88.2/24 LPCM.

 

LeeS, again, the Dirac pulse is not a waveform: http://en.wikipedia.org/wiki/Dirac_delta_function

 

It seems to me that there is still value from recreating the waveform as accurately as possible.

 

"Now resample the sinewave sampled at 44.1kHz up to 192kHz and post a screenshot of the result..."

hehe yeah I guess resampleing does work there. no need it's exactly like the first gen 192.

 

But if it were a redbook format recording upsampled to 192kHz it would not sound like the original 192kHz recording. This one would always sound better because it contains more musical information.

 

Hi Metoo,

I agree. Seeing a coarse display of a sine wave and assuming this tells you how something will sound, i.e. in this case, that there would be no audible consequence is just plain silly, especially when the audible evidence says something completely different. And in audio, what counts other than the audible evidence?

This without even mentioning the artifacts engendered by most sample rate conversion algorithms.

Put another way, if you resample a 44.1k file to 192k, with the best SRC algorithm on the planet, the end result will still have no information beyond what is in the original. It might sound better on playback, with the right D-A than the original (because a different filter will be used) but it won't sound anything like a 192k original.

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

Right, instead of connecting the dots with straight lines like the display in your audio program, both the resampling algorithm and a DAC will do the right thing and filter the output so you get back the same 10kHz sine wave you put in. (Obviously this only works because 10kHz is well within the bandwidth of a signal sampled at 44.1kHz. No one's saying that upsampling will magically re-create stuff that's no longer there...)

 

I wasn't either. I didn't mean that they were actually identical. Just that the pictures or snapshots were.

Of course it wouldn't have any info above the nyquist frequency of 44.1 if it was upsampled. Now if you were to add a distortion or a reverb in a multitrack and then upsample that is a whole different animal.

 

Hi Barry,

I was a violinist for thirty years and quite honestly I always found the CD standard to be woefully inadequate when trying to reproduce (in particular) violin sound. I would have to say though, that my experience with even 24-192 audio tells me that not even it can reach the level of transparency to my ears compared to, say, a top flight analogue source such as high end open reel or 45 RPM "audiophile" vinyl played back on a top tier system.

I had never quite understood, for example, why I was never a Jascha Heifetz fan (everyone else seemed to be). To me his actual sound was very strident, dry, coarse and un-interesting. But then again all I had ever heard was an assortment from the - it must be thousands - of CDs. Once I heard him playing on an audiophile quality LP (and close miked as well!), I completely changed my opinion. It was just that CD was letting him down - and badly! And whilst there are certainly those who will argue that an audiophile LP is better mastered, I would have to say that I must be incredibly lucky: 25 years of listening to Heifetz CDs and all of them sound abysmal and then by an incredible stroke of luck of galactic proportions the very first LP of his I ever hear suddenly sounds "right". So I would simply say that given the number of "tries" there have been to get him to sound right on CD, surely some remastering engineer should have scored a bullseye by now after 25 years.

But I guess not, since you can't win Wimbledon with a ping-pong bat either. I really could not give a twopenny stuff what the digital textbooks say if it never actually happens in practice. And after 25 years of never hearing the theory work in practice, I went high res and have never looked back. It's not equal to the very best analogue however I feel the advantages outweigh the slight sonic compromises.

I certainly find listening to anything at CD standard to be both extremely boring and extremely fatiguing. I find with 24-192 I can listen for extended periods without any boredom and fatigue. I find that once I use 24-192, there is a holographic opening up of the entire 3D soundstage compared to lower sampling rates. Additionally, sections of an orchestra now sound like the group of individual members that they are, rather than sounding like an extended blanket so to speak.

To my ears 24-192 is still not up to best open reel or LP standards, but it has a lot of significant advantages too of course. And really, I only miss the extreme quality of those original analogue sources if I am ABing them simultaneously with a 24-192 copy.

I woud certainly be interested if higher sampling rate PCM ever became available in it's own right, especially as storage costs come down and processing power goes up. And also given that DSD seems to polarise the audiophile / engineering communities and at best is very proprietary.

 

Jon, on Sunday we did a violin recording in 24/176 and 16/44 of Olga Shpitko, a violinist with the Atlanta Symphony. I just listened to the 24/176 after driving in the car all day with the CD version. The instrument tone could not be more different. The CD is sharper, the DVDA much sweeter like what we heard in the church we recorded in. The "hall" was evident on the DVDA, mostly missing on the CD. The decay of the violin plucks were clean on the DVDA, almost missing.

Other factors are essential like mic placement and good acoustics...but in my opinion one really needs a hirez format to capture the performance.

 

I tend to find violins are more favourable to high res digital if the acoustic is a little dry with a bit less reflection. Whether this is an inherent characteristic of the venue itself or just the fact that it is filled with an audience does not seem to matter a whole lot in my experience.

I also think it is the type of mic which is critical - not just the placement which I agree is extremely critical. Vintage mics such as the M50 seem to go down great with close miked chamber stuff. And even something like a U49 can be great for a full string section even if the mic does have that slightly nasal and extremely warm and cuddly character to it.

There is another thing with violins these days too. Most modern players are using synthetic strings and I can tell you that these strings can't hold a candle tonally to the older wound gut strings. Sure, the modern strings are more reliable and can potentially push out more volume, but it's not so common knowledge that they don't necessarily last longer. The older wound gut strings also require more finesse to get them to respond, but to use an audio analogy, the wound gut strings are like an Ampex 350 and the modern synthetics are like an early 80s Sony CD walkman. There are far more subtle and extended overtones with the gut core strings too - something which you really need very high res digital (both in terms of sampling and bit depth) to have any hope at all of capturing.

Luckily there are still many players around these days who I would call the "audiophiles" of the violin world. Leila Josefowicz and the brilliant up and coming Anne Fontanella are the greatest examples of what I call "tonalists" - players who deserve to be put down on high res digital with their use of gut strings - the perfect antidote for the hundreds of ubiquitous CD standard recordings made with violinists equipped with those horrible perlon core strings.

So we have the double whammy these days of low res digital consumer recordings and synthetic strings, versus the good old days of wound gut strings and analogue. Given that you are already of sufficent hearing dexterity to appreciate high res digital, you would surely appreciate the sound of the older strings versus the modern ones all other things being equal.

 

Aren't the woods used totally different now as well? I remember hearing some story on NPR about a string player needing to carry a permit for his bow because it was made from a banned rainforrest wood.

I know there is a local Luthier around here who is sitting on some wood that I want to offer my first born child for.

 

Although you will find some rare deviations from the norm, nearly all modern makers will still use most of the same materials that were originally used when violin building reached it's peak in Europe in the early 18th century. So that remains pine for the belly and maple for the back. Sometimes there are some deviation in the interior linings and blocks (the highly respected Australian maker Cedric Clarke actually sometimes used interior linings from the willow in old cricket bats procured from a school!). I think there are a couple of makers in Australia today who use different woods for the back as well.

If you are talking "restricted" materials it tends to be more the bows. I'm not sure of the legal status of these materials but I thought the permanbuco used for bow sticks was in shorter supply these days and was being protected in some way. Whale bone and tortoise shell in bows are defintely on the outer these days for the frog and windings. Some bows are being made from carbon fibre these days both for some perceived technical advantages and possibly in respect of the permanbuco situation as well.

You'll always find advocates that the modern violins are just as good as the old ones but my personal experience is quite different. Back in the 70s a BBC radio program had a blind comparison of a Strad (Italy early 1700s), Guarneri (Italy early 1700s), Vuillaume (France, late 19th century) and a modern violin. They had Pinchus Zuckerman and Isaac Stern trying to pick out each instrument "blind" (ie they listened to it being playing it behind a curtain). I heard the show on a portable Sony radio (this was the late 70s) and even with the bedside radio I found it dead easy to pick which was which. Not only did the modern violin sound harsh and uninviting, but you even hear the player struggling with the "response" compared to the old Italian ones. I was surprised to find that neither Stern nor Zuckerman guessed correctly, though they probably spent far more time with these violins under their chins than as an audience. From 6 inches away the differences are far more obvious.

The best violin I ever played was a Nicolo Gagliano - the same make of violin that the famous Anne Sophie Mutter started her career with. Beautiful instrument and an absolute steal at a mere 40 grand (back then of course - add a zero to today's value). My own instrument was a Stefano Scarampella from 1889. The differences between it and a moden violin were extremely obvious. Again it was like comparing great analogue to cheap digital.

"Sensible" players today like Hilary Hahn chose what many would call second tier instruments. She plays a Vuillaume but then again many don't know that Kreisler also played a Vuillaume and lent that instrument to Josef Hassid who himself was one of the most briiliant and promising violinists in history but whose life was tragically cut short after mental problems and subsequent brain surgery. Anyway, Hilary Hahn is proof that you don't need a Strad or Guarnerni to be one of the top handful of violinists in your generation.

Back on the topic of violins being a great test of digital recording technology, I would still be able to tell apart violins, but when you hear them in the flesh, there is so much that is completely missing from the best recordings. There is a whole extra layer of complexity that just seems to elude the modern recording process. I wish everyone had a top end turntable and could hear this Living Stereo Heifetz Beethoven Sonata recording I have (the Cisco reissue). I have never heard a recording accurately capture a violin before hearing that. I've made copies of it at everything from 16-44 right up to 24-192. Only 24-192 even begins to cut the mustard, but even then when you compare it to the LP there is still something missing.

I would love (as intimated earlier) to capture violin sound at ever increasing resolution until such time as it sounded just like the real thing. But I think the reality is that microphones would be struggling to keep up with what super high resolution digital (say 768 Khz or similar) could deliver.

 

Hi Jon P,

I understand and your feelings and might have agreed completely up until not that long ago. I believe much of what 192 can deliver is very dependent upon the hardware. What converters were used in the 192 you auditioned?

Having done most of my recent work at 24/96, I've spent quite a long time experimenting with 24/192. As the sample rate goes up, more demand is put on the conversion and clocking circuitry. In my experience, a lot of converters with a 192 spec just don't sound their best when clocked that high. So 192 from one device does not equal 192 from another device. (We can say the same thing for 44.1 and more recently, 96 but those rates are relatively easy nowadays.)

It could be that you might still feel the same way, even if you auditioned the hardware I've been testing -which I am not at liberty to name at this time, as it has not yet been released. All I can say is not that long ago, I'd have agreed with you completely. In fact, this is so recent for me, that this week will be the very first time I use 192 "officially" in a recording session for my own label. (I tend to test things for a very long time before adding them to the tools I use in my work.)

Best regards,
Barry
www.soundkeeperrecordings.com
www.barrydiamentaudio.com

 

The same goes for 44.1.

 

Hi Robert,

Agreed. But things get even more difficult as the rate goes up.

Note, in the same post I also said:
"(We can say the same thing for 44.1 and more recently, 96 but those rates are relatively easy nowadays.)"

Best regards,
Barry
www.soundkeeperrecorrdings.com
www.barrydiamentaudio.com