96Khz vs 192Khz

[stevendive]

Because they can... The Chicago and Doobies Quadio release were 192kHz I believe. The 4.0 files of the Chicago double albums were outrageously large files.

But isn't there some advantage to using the high sample rates when doing mixing, editing, etc?

From discussions on a specialist mailing group hosted at Virginia Tech (mainly ambisonics focus), high res files are valuable for multi channel mixing as the noise floor can be raised too far if ‘normal’ res were to be used.

 

[quadaholic]

Funny that you'd mention this...I just had to prepare Chicago's "25 or 6 to 4" for radio use and started with the 192k version. When I opened it in Audition there was a really annoying whistle/whine/squeal that made me think I had a hardware issue. But that noise was completely gone after SoX downconversion.

Atrocity, I've had that same problem with 96k files. In my case, there was tape/electronics related noise up around 40k that was being downsampled badly by my computers electronics and as a result weird noise/artifacts was resulting. As soon as you downsample files to 48k noise is gone because sample rate doesn't go up that high to pick it up. I bought some king biscuit reels on ebay a while back, and when I transferred them, we found a problem with the recording bias tone on one of the reels, and as a result was generating a huge noise at around 38k. It's been on the tape all along, but none of us could hear it until the downsampling cheez turned it into static/hiss. As far as old tapes and tape machines go, there is often a lot of weird stuff that turns up in the inaudible range range if you look at things on a computer. Back then, if you couldn't hear it , it didn't exist, if you get what I'm saying.

 

[MidiMagic]

You have to divide by the number of audio channels.

 

[jimfisheye]

You have to roll off high frequencies above the Nyquist limit before downsampling or you get aliasing.
Some sample rate conversion apps might do that as a feature. Best to check though.

 

[ted_b]

There is a direct relationship between the sample rate (and bit depth) you send to your DAC (whether it be a dedicated outboard DAC, a part of your cd player or AVR, etc) and the quality of its playback. We upsample to give our DACs a break, to send them music that is most compatible with their clocks, filters, etc.

Most DACs upsample internally. And the often use $.50 parts to do it. When we upsample prior to the DAC we can, theoretically, lessen the load on the DAC and have it produce more accurate and musical playback. In my case, I have two main DACS, one for 2 channel (Holo May) and one for multichannel (exaSound s88), along with the dac(s) that are in the pre/pro, etc. When going direct to my Holo or exaSound I use a 3rd party upsampling engine called HQPlayer that allows me to choose sample rate, bit depth, filtering and noise shaping. For example, with my Holo May, it is measured as most linear at 20 bits, so I send it 20 bits (and the Holo has an NOS mode where it doesn't upsample internally). And through trial and error I have found the perfect mix of filter/noise shaper and sample rate (32fs or approx 1.4 Mhz for 44k-based material and 1.5Mhz for 48k based material). It's where the sweet spot is. And it has nothing to do with wasted ultrasonic noise, it has to do with using the gentlest and most linear filtering, dithering/noise shaping and bit depth that I can use. The results are breathtaking. Same logic is true for upmodulating DSD (to say, DSD256).

So, net/net, don't assume upsampling is a waste. Find your dac's sweetspot and send it the best combo you can. It might even be at a different clock frequency than your 96k/192k comparisons (48k clock might be worse than the 44k clock).

 

[jimfisheye]

Hmmm...

I don't like adding noise to my music. eg. dither and noise shaping

That's vestigial from the days of 12 bit converters. (When they combined 12 bit + 4 bit chips in early days circuits. And the bottom 4 were pretty useless. And they got desperate trying to resolve low level signals in essentially a 12 bit system.)

24 bits gives you a ground floor of 8 bits you never need to touch and then a whole 16 bits on top of that for the program. That's 96db of dynamic range on top of a noise/resolution floor with the lowest signal at -96db having 8 bits of resolution. If you need to add noise signals to boost up the resolution of something lower down than that... What in the Sam hell are you doing?! Trying to reproduce gun blasts from point blank range and releasing a "Go Deaf!" series of recordings? (More like a "Blow up your speakers" series probably.)

I typically leave the system set to 96k and let the OS audio system upsample anything lower on the fly.

Yeah, some of the mystery sample rate shuffles in AV receiver type products. Probably why I don't like those kind of "combo" products. Give me a straight audio interface I can matter of fact set. The modern generation of sample rate conversion algorithms are pretty robust. To the point where you could do multiple lossy conversions between SD and HD and not hear any damage at a glance even on a reference system. But there are still a few old apps floating around and a few hardware devices that do on the fly conversions really poorly. Causing aliasing is operator error.

 

[atrocity]

Atrocity, I've had that same problem with 96k files. In my case, there was tape/electronics related noise up around 40k that was being downsampled badly by my computers electronics and as a result weird noise/artifacts was resulting.

I hadn't even considered the possibility that it was an artifact of something in the recording! I just assumed that my cheap hardware (or Audition) was doing a poor job of downconverting on the fly.

Might be interesting to play the same track via different software to see if the noise is consistent.

 

[quadaholic]

You have to roll off high frequencies above the Nyquist limit before downsampling or you get aliasing.
Some sample rate conversion apps might do that as a feature. Best to check though.

Thanks for this insight. It will come in very handy for me in the future!

 

[quadaholic]

I hadn't even considered the possibility that it was an artifact of something in the recording! I just assumed that my cheap hardware (or Audition) was doing a poor job of downconverting on the fly.

Might be interesting to play the same track via different software to see if the noise is consistent.

I expect it will come and go, depending on what gear you are using. When I was doing transfers, it sounded fine coming through the protools system and then started up when I played it through my laptop speakers.

 

[quadaholic]

Hmmm...

I don't like adding noise to my music. eg. dither and noise shaping

That's vestigial from the days of 12 bit converters. (When they combined 12 bit + 4 bit chips in early days circuits. And the bottom 4 were pretty useless. And they got desperate trying to resolve low level signals in essentially a 12 bit system.)

24 bits gives you a ground floor of 8 bits you never need to touch and then a whole 16 bits on top of that for the program. That's 96db of dynamic range on top of a noise/resolution floor with the lowest signal at -96db having 8 bits of resolution. If you need to add noise signals to boost up the resolution of something lower down than that... What in the Sam hell are you doing?! Trying to reproduce gun blasts from point blank range and releasing a "Go Deaf!" series of recordings? (More like a "Blow up your speakers" series probably.)

I typically leave the system set to 96k and let the OS audio system upsample anything lower on the fly.

Yeah, some of the mystery sample rate shuffles in AV receiver type products. Probably why I don't like those kind of "combo" products. Give me a straight audio interface I can matter of fact set. The modern generation of sample rate conversion algorithms are pretty robust. To the point where you could do multiple lossy conversions between SD and HD and not hear any damage at a glance even on a reference system. But there are still a few old apps floating around and a few hardware devices that do on the fly conversions really poorly. Causing aliasing is operator error.

I would add that I recently heard some archivists talking about how the standard sample rate for digital transfers is all ready expected to go up to 384k (double 192) ! The smaller the "steps" the better I guess.

 

[ted_b]

Where did you hear this? DXD (aka 24 or 32 bit at 352.8k) is often an editing sample rate, especially for DSD in Pyramix, but to say 384k is going to be a standard is not something I've ever heard.

 

[quadaholic]

Where did you hear this? DXD (aka 24 or 32 bit at 352.8k) is often an editing sample rate, especially for DSD in Pyramix, but to say 384k is going to be a standard is not something I've ever heard.

I should check to just make sure I'm not having a false memory (which happens with me from time to time), but I think it was a recent webcast with Kelly Pribble. He's one of the best transfer people in the world, and works at Iron Mountain.

 

[Owen Smith]

I would add that I recently heard some archivists talking about how the standard sample rate for digital transfers is all ready expected to go up to 384k (double 192) ! The smaller the "steps" the better I guess.

Anyone talking about the smaller steps being better does not understand sampling theorem and needs to read Shannon and Nyquist. You can draw pretty pictures that make it look like smaller steps should be better, but that's rubbish it isn't how it works.

 

[dadregga]

When we upsample prior to the DAC we can, theoretically, lessen the load on the DAC and have it produce more accurate and musical playback.

For me to take that assertion seriously I'm gonna need you to explain that again in a way that makes sense in terms of basic electronics.

I'm not sure what you mean by "load" (or "accuracy", or "musical", really).

 

[LuvMyQuad]

Yeah, "musical" is a snake oil term. Its meaningless as a descriptor.

 

[4-earredwonder]

192/24 has been around since the advent of DVD~Audio when a lot of the included Stereo layers were indeed at that sampling rate. I don't download or load my discs onto a hard drive so I, personally, welcome 192/24. Most of the audiophile companies, domestic and Japanese, have also included 192 on their DVD~A stereo discs as well as BD~A discs and to my ears they are some of the best sounding discs in my collection. And who can quibble with the recent spate of Rhino QUADIO BD~A releases [The Best of the Doors, Chicago and the Doobies] at that higher sampling rate?

Whether it's overkill as compared to 96/24, I have NO qualms with 192 ..... IMO, anything beats that atrocious 44.1 sampling rate which continues to plague the RBCD.

And while I'm at it, I wish ALL RBCDs would be MQA encoded .... as unfolded, they include 24 bit resolution with high sampling rates and really sound magnificent. DON'T KNOCK IT UNLESS YOU'VE ACTUALLY TRIED IT! Reading about it and experiencing it FIRSTHAND are two different animals altogether!

 

[MidiMagic]

There are too many cooks in this overkill bush. They earn a worm.

 

[4-earredwonder]

There are too many cooks in this overkill bush. They earn a worm.

LEAVE US OUT OF THIS?

 

[jimfisheye]

For me to take that assertion seriously I'm gonna need you to explain that again in a way that makes sense in terms of basic electronics.

I'm not sure what you mean by "load" (or "accuracy", or "musical", really).

I'll take a shot.

The "load" would be on DACs sample rate filter eq at SD sample rates. ie. The fact that said eq is in use at all.

In SD sample rates (especially 44.1k) the sampling frequency is RIGHT next to the audio band. So close that the sampling frequency needs to be eq filtered out with a steep lo pass or it will roll into the audio band. This is in the analog domain after the AD chip. This kind of steep eq is a hard circuit to build! It turns out that the factor you are comparing between different DACs at SD sampling rates is that analog eq circuit.

Meanwhile HD sample rates put the sampling frequency miles away from the audio band. The margin is so wide that eq filtering is not needed. Like tape bias whistle.

If your DAC sounds better at 96k because of this, upsampling SD program is an excellent workaround. The music signal is in there fully. The analog filter eq is your problem. We remove it from the equation. And yes, it's gross compared to any generational loss from upsampling. So much so that it's not fair to mention.

Right, so 96k puts the sampling frequency miles above the audio band. Not sure what anyone thinks they need 192k or above to do. I mentioned my tests. There may be some processing that benefits from extreme HD sample rates (I doubt it). When you're done with that, the full program can be put into 96k with no loss.

 

[GaryA]

Anyone talking about the smaller steps being better does not understand sampling theorem and needs to read Shannon and Nyquist. You can draw pretty pictures that make it look like smaller steps should be better, but that's rubbish it isn't how it works.

I am sure you are correct, but that’s not a useful comment. Rather than refer people to somebody else, why don’t you simply explain in a few words how it really does work, so that everybody can easily understand why you are saying that smaller steps are not necessarily better?