ORIGINAL: bitflipper

Since I always record at 44.1, I'd like to pose a (maybe dumb) question to those of you who record at higher sample rates: if you bring up a 96KHz wave on a spectrum analyzer, do you actually have anything in there above 20KHz? Or does the converter band-limit everything at 20KHz regardless of the sample rate? In other words, are you actually recording ultrasonic material? (OK, so that's 3 questions, sorry)

Here maybe is an example (analog master -> DVD-A (vs CD vs LP)) but, are those analog masters processed when digitalized ... ?

Junski

Thanks, tarsier. That's exactly what I was looking for. You even used the same interface that I have (828 MkII), so your observations are personally relevant. It doesn't answer the bigger question, which is do you really need to record those frequencies in the first place, but it does satisfy my curiosity.

Regarding the Earthworks microphones: Sweetwater sent me the Earthworks demo CD, which sat on my desk for a long time because I couldn't imagine that you could demonstrate the benefits of extended-frequency mikes on a CD. But the demo offers compelling evidence that recording over 20KHz has some benefit. The differences are not so subtle that you need "golden ears" to hear them. I assume you use these for drum overheads, that's what most people buy them for -- do you use them for other stuff as well, e.g. acoustic guitar?

When I mentioned the Earthworks mikes in an earlier post, somebody pointed out that they are very power-hungry, a fact that I verified via the manufacturer's published specs. So demanding, in fact, that using several of them at once could easily cause a drop in your phantom voltage. You'd have a loss in quality and sensitivity and not know why. I guess there's no free lunch.

Interesting pictures, thanks for the link.

There really isn't very much up there (above 20K), is there?

There aren't any numbers, as it's a spectral display, but it looks like an Audition screenshot and my experience using Audition's spectral display is that when the picture's that faint and purple that it's barely audible even when it's well under 20KHz.

bitflipper: It makes sense that there is stuff above 20Khz when recording at higher sampling rates because the filters only start at arround 20Khz and go up (with a gentle slope) to half nyquist.

UnderTow

Yes, that does make sense.

The big question, I guess, is does it matter? We've established that ultrasonic frequencies do combine to form valid audible components, and therefore contribute to sonic fidelity. But that combining action happens in acoustical space while recording. Once it's been converted to digital, the ultrasonic components have already done their work and we should only care about audible frequencies from that point on.

Preserving them through the processing phase might give them further opportunities to subtly contribute to the sound, but ultimately they're definitely going to be truncated prior to distribution at 16/44.1. In the meantime, they are as likely to degrade sound as enhance it. Seems to me that preservation of ultrasonics is a concern in the initial analog realm only, through the use of high-quality microphones and preamps.

And here's another consideration that hasn't been touched: jitter becomes a much bigger problem at higher sample rates. At 44.1, jitter isn't really a concern with the fairly high-quality prosumer converters most of us use - the built-in clocks are more than adequate. The people I talk to who swear by high-end wordclocks and claim noticeable improvement are all recording at high sample rates - maybe not a coincidence.

I could be wrong but I don't think that jitter is any different at different sample rates for the same converters. For ADCs the converter just taps off the data stream from the actual sampling/decimation chip at different speeds but as you yourself pointed out, the actual sampling happens at X times that rate. As far as I know, the actual sampling happens at the same rate reguardless of the hand off rate. The same goes for the DACs. The oversampling part that generates the analogue signal will be running at the same rate reguardless of the incoming data rate.

As for external clocks, from my understanding, they don't make any sense unless you are clocking different converrters to each other. In the end, the converters are actually running off their local crystals wether the clock is local or the PLL's are syncing to external clocks. As far as I know and from what I have read from people much much more knowledgable than me, the local clock will always be the tightest. Or rather, the money spent on the external clock would be best spent on better converters. Another marketing gimmick it seems...

The people buying into higher sample rates might be the same people that will buy into external clocks and expensive aftermarket power cables.

Edit: There are legitimate reason to have external clocks but I doubt that less jitter is one of them.

UnderTow

ORIGINAL: bitflipper
And here's another consideration that hasn't been touched: jitter becomes a much bigger problem at higher sample rates. At 44.1, jitter isn't really a concern with the fairly high-quality prosumer converters most of us use - the built-in clocks are more than adequate. The people I talk to who swear by high-end wordclocks and claim noticeable improvement are all recording at high sample rates - maybe not a coincidence.

Dan Lavry (of Lavry Engineering who makes top-notch converters) has said that if your sound improves when you run a converter with an external clock, then your converter is defective. Or to put it another way, there's no excuse for the circuit that accepts the incoming clock to do a better job than the circuit that generates the clock internally in the first place--best case scenario is that they should sound equivalent.

I once asked Greg Hansen (who developed the Zefiro ZA2 digital I/O card, and the Zefiro Inbox) why the Inbox didn't have a wordclock input. He said that it originally was going to have one, but they took it out since it increased the noise of the box unacceptably.

As far as high sample rates, remember that virtually all converters nowdays (since about year 2000 or so) are low-bit depth, oversampling, delta-sigma converters. Meaning that they sample at a very high rate using only a few bits (like 2 or 4 bits and 7.68 MHz sample rate) and then they sample rate convert to whatever bit depth and rate are requested. Whatever jitter there is in the clock, the distortion should be about equivalent regardless of the ultimate sample rate.

Awesome guys, keep the info coming . As far as I can tell, it is convenient to have the ultrasonic material recorded from the beginning as to avoid upsampling and downsampling during mastering. Of course, the drawback would be the resources spent. But, to me, it just makes sense to record at higher sampling rates on material that would benefit from it (acoustical instruments) and avoid SRC as much as possible.

As for jitter, I have a question. If higher sampling rates don't affect jitter then why is it possible to hear a better quality recording from prosumer cards at higher sampling rates?

I have checked several technical references, and they are consistent in this: if you can hear a difference in sound quality at higher sample rates, it is due to inadequacies of your hardware, specifically of the anti-aliasing filter.

They are equally consistent in saying that most of the commonly-held misconceptions about the subject can be traced back to older converter designs, and that any "pro" converter manufactured over the last 10-15 years will not suffer the problems that led people to resort to higher sample rates in the first place.

As far as I can tell, it is convenient to have the ultrasonic material recorded from the beginning as to avoid upsampling and downsampling during mastering. Of course, the drawback would be the resources spent. But, to me, it just makes sense to record at higher sampling rates on material that would benefit from it (acoustical instruments) and avoid SRC as much as possible.

True, if you know you'll be required to deliver the final product at 48KHz, say for DVD, then it makes sense to record it that way from the getgo if you have that option, simply to avoid a conversion. But if your ultimate target is CD or MP3, then it makes no sense to burn up the extra disk space.

If higher sampling rates don't affect jitter then why is it possible to hear a better quality recording from prosumer cards at higher sampling rates?

This is what I was responding to in my previous post. It's because of the analog filters, not the sample rate.