Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Scott Holtzman]

I wish I had held on to my Technics 10 1/2" reel deck for nostalgia.  Even at 15 IPS there was audible wow/flutter/hiss.

Sarcasm aside, A Studer A820 Mastering deck is about the best you can get for analog.  1/2" tape, 2-track at 30ips.

-and it is probably the most expensive as $100 barely gets you 20min worth of stock!

Most consumer decks are 1/4" 4-track (2 "sides") gave you you double the track width and speed of a cassette -but with usually better wow&flutter.  And much less compromise in head design and tape handling due to no size constraints.

I have an older Sony deck that's all I have left from that time in my listening life.  All the pro machines were sold off in the ADAT era.
Was just listening to some old tapes over the holidays and discovered that it's developed a brake-pad issue so its easy to dump tape everywhere if stopping after fast winding!

[Stephen Swaffer]

One thing often forgotten in this conversation is processing time ( Latency ). Never mind hearing 96khz. But it processes twice as fast. In todays times when everything goes digital. Desk, Processor, Wireless, etc the accumulation of several units in the signal pass can create Problems.
If you for example take a digital wireless mike, run it into a digital console with a plugin and then into a digital In-Ear to a performers ear, you could possibly create a latency in excess of 5ms.This could pose a problem to performers. Run that in 96khz  and its half that. Which is a massive improvement for the Singer or Drummer. That in return will yield a better performance and therefore a better End-result. There is a good place for higher sample rates in live sound. But that obviously has nothing to do with the test mentioned here.

I'm probably missing something obvious, but it seems to me that 96 kHz audio would have twice as much data to process, so all else being equal should take twice as long to process?  Why would doubling the sampling rate speed up the processing?

[Scott Holtzman]

I'm probably missing something obvious, but it seems to me that 96 kHz audio would have twice as much data to process, so all else being equal should take twice as long to process?  Why would doubling the sampling rate speed up the processing?

No the clock runs twice as fast so more samples acquired  in less time.

A to D takes three steps.  Acquisition, Quantization and Storage.  For our purposes storage is writing to the data bus.

[Tim McCulloch]

I'm probably missing something obvious, but it seems to me that 96 kHz audio would have twice as much data to process, so all else being equal should take twice as long to process?  Why would doubling the sampling rate speed up the processing?

Along comes faster *processing*.

Scott points out that it's the faster acquisition of data that is relevant to word clock speed.  The CPU is another matter, but it's safe to assume that we can get data through the CPU much faster than we can acquire audio samples (at rates that make sense for audio).

[Scott Holtzman]

Along comes faster *processing*.

Scott points out that it's the faster acquisition of data that is relevant to word clock speed.  The CPU is another matter, but it's safe to assume that we can get data through the CPU much faster than we can acquire audio samples (at rates that make sense for audio).

CPU speed seems irrelevant in today's world where the wordclock runs at 1/10 of  a MHZ and CPU/FPGA speeds are measured in GHZ.

There was an interesting discussion of hardware floating point on CISC uP's vs. using fixed point registers and shifting the data left and right through some type of oversample buffer.

I looked for that discussion to link to it, I thought it was in the last quarter but could not find it.

[Peter Morris]

Stirring the pot is most likely the case- which is totally fine however, Ethans research and tests tend to try and prove something that one cannot possibly prove given the type of test he is offering. In other words and with all do respect- don't waste your time. Without going on a huge rant here about the benefits of hi res audio, in my opinion one must understand what they are listening to. It's not much different than a person listening to a cheap violin and a Strad. There's a lot of people that prefer one over the other and a lot of people that don't have a preference etc. but if you play the Strad and explain what makes a great violin a great violin, most likely they will be able to pick out the great violin in any kind of test or at the very least pick out the violin that sounds like the Strad. when the Strad is played.
 We did many tests in the studio when I was at Sony when DSD and hi res audio was being developed. At the end of the day, was the DSD and high res audio (96K 24 and 192) same or different compared to the live mixer output with the band playing? Hearing it for myself, first hand, the higher res audio and DSD was way closer to same as the live feed. Obviously there is a lot in play including filters etc (which might make the biggest difference) but the same or different really helps me sort this stuff out. For me, the DSD and 192 were almost identical to the live output. Doing tests for these kind of perceptual things is really really really difficult and Ethans test is just a farce to make that kind of conclusion.

That being said, there are many studies on this  and I welcome you to read up on it if you are interested. The last real paper is fairly current and can be found here on the AES website  http://www.aes.org/e-lib/browse.cfm?elib=18296
The paper is authored by Josh Reiss, a professor at Queen Mary Univ. of London. Check it out!

Remember- in our "live" world, hi res audio has other benefits that may not have anything to do with sounding better.

Regards to the list...

Roger
 

Exactly, and thanks for posting the link to the AES paper.  I tried to find it but had no luck.

There are a couple of things that people miss in this argument - 16 bits will give you a lot of detail, but in practice the gain structure of the recording or live signal is often far from perfect. There can be lot less than 16 bits defining the waveform.
   
The quality of 44.1 khz Vs 96khz has a lot to do with the impact of the anti-aliasing filters, not our ability to hear 20khz +; 96 khz makes it a lot easier to get good results. 
 
To my ear the difference is very very small - as an analogy it’s like looking through a clean glass window compared to having no window, you can tell.  24/96 is very natural and less fatiguing especially the HF, which I suspect may have something to do with the phase shift of the anti-aliasing filters... but as I said the difference is small and often not perceivable, especially with Ethan's  test methods I suspect.

[David Morison]

There was an interesting discussion of hardware floating point on CISC uP's vs. using fixed point registers and shifting the data left and right through some type of oversample buffer.

I looked for that discussion to link to it, I thought it was in the last quarter but could not find it.

This one maybe?

[TJ (Tom) Cornish]

I'm probably missing something obvious, but it seems to me that 96 kHz audio would have twice as much data to process, so all else being equal should take twice as long to process?  Why would doubling the sampling rate speed up the processing?

The mathematical operations in doing the A -> D conversion -> whatever you do to the signal in the desk - EQ, dynamics, etc. -> D -> A conversion are all dominated by the wait time between samples; as Scott says, a 0.1Mhz clock signal isn't particularly fast relative to CPU speed.  Some mixing consoles' latency is actually specified in samples. 

The Yamaha 01v96 is a desk that I'm familiar with, and the spec sheet lists latency as being less than 1.6ms at 48KHz and less than 0.8ms at 96KHz.  Doing some quick math, this puts the desk's processing latency at about 70 samples, which is pretty amazing.

Sometimes latency increases with desk complexity.  Schemes like latency management systems that time align every path through the mixer to be the same length - i.e. channel -> output is the same latency as channel -> group -> output - do so by adding time to the shorter paths, which slows the desk down to the longest possible audio path.

I don't work in environments where perceived sound quality differences from moving to 96KHz will be apparent (live sound is dominated by many other factors - stage talent, room acoustics, speaker placement, speaker quality/tuning, mic quality, etc. - these factors being several orders of magnitude more important than desk sound), but as more and more things become digital, managing the total latency budget does start to matter when using in-ear monitors.  This is the main benefit of 96KHz to me.

[ProSoundWeb Community]