Multiple A/D D/A Conversions - AVS Forum
Forum Jump: 
 
Thread Tools
post #1 of 16 Old 07-16-2013, 11:14 AM - Thread Starter
Member
 
alexwgoody's Avatar
 
Join Date: Dec 2011
Location: San Jose
Posts: 54
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 13
I am wondering if anyone can speak to or link me to any information regarding the harm caused by multiple A/D to D/A conversions (assuming relatively high-grade components). I am imagining using up to 3 conversions (does more matter?). I ask because I run multiple types of listening formats- both digital and analog, 2 channel & multi-channel currently all in multiple rigs- and have found success in my current DSP use in my two-channel all digital computer audio system (primarily in DRC from REW & digital crossovers on my speakers/subs). I've seen very little useful information with basic Google searches and would love to read some information/graphs/papers. I am no engineer, but have done my best to become knowledgeable and am able to understand some of the math, interpret graphs and understand the research.

I have multiple reasons for wanting to know the possibly harmful effects to see if it is worthwhile to use all of my sources in my digital rig and am interested in running a surround processor (and turntable) into my miniDSP 10 x 10 to be able to run multi-channel surrounds with my current digitally crossed-over room corrected speakers and subs (2).

Thank you.

(I would just try it on my own and listen- but I do not want to spend any more money/time than is necessary. Living on a relatively tight budget, buying new gear - especially the surround processor just to be disappointed with the sound quality would devastate my bank account)
alexwgoody is offline  
Sponsored Links
Advertisement
 
post #2 of 16 Old 07-16-2013, 12:13 PM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 134
This is exactly what you are looking for:

Converter Loop-Back Tests

--Ethan
alexwgoody likes this.

RealTraps - The acoustic treatment experts
Ethan's Audio Expert book

Ethan Winer is offline  
post #3 of 16 Old 07-16-2013, 12:29 PM - Thread Starter
Member
 
alexwgoody's Avatar
 
Join Date: Dec 2011
Location: San Jose
Posts: 54
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 1 Post(s)
Liked: 13
Thank you so much, that was exactly what I was looking for and expected. Just to add and confirm; some of the gear samples at different rates for A/D and DSP (16/44.1 audio from computer, lossless blu-ray audio + SACDs, analog signals, etc.- the miniDSP runs at 24/48 for all DSP effects, ). Do you have any advice for the rate at which I should sample the signals (which some of my gear lets me control? Does it make a difference?
alexwgoody is offline  
post #4 of 16 Old 07-17-2013, 02:18 PM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 134
Quote:
Originally Posted by alexwgoody View Post

Do you have any advice for the rate at which I should sample the signals (which some of my gear lets me control? Does it make a difference?

I don't know what you have in mind, so it's tough to be specific. I can tell you that for music, 44.1 KHz at 16 bits (standard CD quality) is sufficient for audible transparency. Anything greater is not needed for a playback medium, no matter what you might read in magazines and audio forums. However, music for video production is typically recorded at 48 KHz because that's the standard for DVDs. So if you are recording something that will end up on a DVD, it's better to record at 48 KHz initially rather than convert it later.

--Ethan

RealTraps - The acoustic treatment experts
Ethan's Audio Expert book

Ethan Winer is offline  
post #5 of 16 Old 07-17-2013, 05:42 PM
AVS Special Member
 
MarkHotchkiss's Avatar
 
Join Date: Mar 2007
Location: Long Beach, California
Posts: 1,476
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 82
^^^ This.

Hi Alex,

The only thing that I would like to add is that recording at a 96kHz (and maybe 192kHz) sample-rate might be called for if you intend to do some DSP audio post-processing. Some DSP effects algorithms work better or introduce less degradation when they have additional data to work with. (Studios often master their CDs at 352.8kHz).

But your final mix need be no more than 48kHz.
MarkHotchkiss is offline  
post #6 of 16 Old 07-18-2013, 01:27 AM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,387
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 763 Post(s)
Liked: 1178
Quote:
Originally Posted by MarkHotchkiss View Post

^^^ This.

Hi Alex,

The only thing that I would like to add is that recording at a 96kHz (and maybe 192kHz) sample-rate might be called for if you intend to do some DSP audio post-processing. Some DSP effects algorithms work better or introduce less degradation when they have additional data to work with. (Studios often master their CDs at 352.8kHz).

But your final mix need be no more than 48kHz.

To clarify, if you add strongly nonlinear effects in the digital domain there can be aliasing which you can circumvent with high sample rates. This would not be equalization or delay. I don't know of any of the normal kinds of digital processing that is done in home audio that would be nonlinear in this way. It could happen in a synthesizer used to produce electronic music or some kind of a guitar fuzz box.
arnyk is offline  
post #7 of 16 Old 07-18-2013, 01:32 AM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,387
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 763 Post(s)
Liked: 1178
Quote:
Originally Posted by alexwgoody View Post

I am wondering if anyone can speak to or link me to any information regarding the harm caused by multiple A/D to D/A conversions (assuming relatively high-grade components).

Actually, the opposite. I did some demonstrations of back-to-back digital conversions of very clean musical recordings back around 2002. Using a high quality professional audio interface (LynxTWO) I was able to do 20 back-to-back conversions that could not be detected by ear in a blind, level matched, time synched listening test. Level matching in a test like this has to involve level matching and gain adjustments because 20 back-to-back conversions is 40 total conversions. If there is a 0.1 dB error in every conversion, it adds up to a 4 dB error after 40 conversions, and a 4 dB error is clearly audible in an ABX test.
arnyk is offline  
post #8 of 16 Old 07-18-2013, 11:22 AM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 134
Quote:
Originally Posted by arnyk View Post

if you add strongly nonlinear effects in the digital domain there can be aliasing which you can circumvent with high sample rates.

When I asked about this at Hydrogen Audio a while ago, I recall that the consensus was plug-ins that benefit from higher sample rates simply up-sample internally. So there's no need to actually record at a sample rate higher than the final distribution medium. Are there exceptions to that?

--Ethan

RealTraps - The acoustic treatment experts
Ethan's Audio Expert book

Ethan Winer is offline  
post #9 of 16 Old 07-18-2013, 01:25 PM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,387
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 763 Post(s)
Liked: 1178
Quote:
Originally Posted by Ethan Winer View Post

Quote:
Originally Posted by arnyk View Post

if you add strongly nonlinear effects in the digital domain there can be aliasing which you can circumvent with high sample rates.

When I asked about this at Hydrogen Audio a while ago, I recall that the consensus was plug-ins that benefit from higher sample rates simply up-sample internally. So there's no need to actually record at a sample rate higher than the final distribution medium. Are there exceptions to that?

--Ethan

I really don't know anything about the internal code in plug-ins. It makes sense that the authors of plug-ins that applied nonlinear effects would upsample internally in the interests of avoiding the nasty audible effects of aliasing.
arnyk is offline  
post #10 of 16 Old 07-18-2013, 03:24 PM
AVS Special Member
 
MarkHotchkiss's Avatar
 
Join Date: Mar 2007
Location: Long Beach, California
Posts: 1,476
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 82
Hi Arny,
Quote:
Originally Posted by arnyk View Post

To clarify, if you add strongly nonlinear effects in the digital domain there can be aliasing which you can circumvent with high sample rates. This would not be equalization or delay. I don't know of any of the normal kinds of digital processing that is done in home audio that would be nonlinear in this way. It could happen in a synthesizer used to produce electronic music or some kind of a guitar fuzz box.
Agreed, to a point. I had made the wrong assumption that Alex was recording and mixing. Looking back at his original post, I don't see where I got that idea, but that is why I mentioned "recording".

Higher sampling rates can decrease the group-delay through FIR filters (and also IIR filters, to a lesser extent). But as you say, that doesn't really matter when just listening. It effects the relative phase of the speakers, but the higher sampling rate offers no advantage.
MarkHotchkiss is offline  
post #11 of 16 Old 07-18-2013, 03:49 PM
AVS Special Member
 
MarkHotchkiss's Avatar
 
Join Date: Mar 2007
Location: Long Beach, California
Posts: 1,476
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 82
Hi Ethan,
Quote:
Originally Posted by Ethan Winer View Post

When I asked about this at Hydrogen Audio a while ago, I recall that the consensus was plug-ins that benefit from higher sample rates simply up-sample internally. So there's no need to actually record at a sample rate higher than the final distribution medium.
I find it curious that they would say that - It must have something to do with their architecture (I don't know much about Hydrogen Audio).

To up-sample the stream would be perfectly valid. But if we talk about a standard VST plug-in, the plug-in doesn't necessarily know at what sample-rate that it is being fed samples, so it can't easily determine whether or not to up-sample. It could probably make a determination of the sample-rate, but that sounds needlessly complex to me, and goes against the modular, minimalist philosophy of VST plug-ins. It seems more practical for the user to make the determination on whether or not she wants to up-sample prior to feeding the plug-in, and place an up-sampling plug-in in the chain in front of the "effects" plug-in involved.

Quote:
Are there exceptions to that?
I can think of one. It has to do with digitizing vinyl, and utilizes content up to 48kHz (but that's a topic for another thread). I suspect that the only exceptions would be applications that have content above 22kHz. With content that is truly limited to 20kHz, up-sampling should give you identical results as recording at the higher sample-rate.
MarkHotchkiss is offline  
post #12 of 16 Old 07-18-2013, 05:40 PM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,387
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 763 Post(s)
Liked: 1178
Quote:
Originally Posted by MarkHotchkiss View Post

Hi Ethan,
Quote:
Originally Posted by Ethan Winer View Post

When I asked about this at Hydrogen Audio a while ago, I recall that the consensus was plug-ins that benefit from higher sample rates simply up-sample internally. So there's no need to actually record at a sample rate higher than the final distribution medium.
I find it curious that they would say that - It must have something to do with their architecture (I don't know much about Hydrogen Audio).

To up-sample the stream would be perfectly valid. But if we talk about a standard VST plug-in, the plug-in doesn't necessarily know at what sample-rate that it is being fed samples, so it can't easily determine whether or not to up-sample

If I'm reading this reference correctly, the host sample rate can be obtained by the plug-in

http://jvstwrapper.sourceforge.net/vst20spec.pdf

page 52

"
virtual double updateSampleRate ();
// gets and returns sample rate from host (may issue setSampleRate() )
"
arnyk is offline  
post #13 of 16 Old 07-18-2013, 06:15 PM
AVS Special Member
 
MarkHotchkiss's Avatar
 
Join Date: Mar 2007
Location: Long Beach, California
Posts: 1,476
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 82
Hi Arny,
Quote:
Originally Posted by arnyk View Post

If I'm reading this reference correctly, the host sample rate can be obtained by the plug-in

http://jvstwrapper.sourceforge.net/vst20spec.pdf

page 52
Quote:
virtual double updateSampleRate ();
// gets and returns sample rate from host (may issue setSampleRate() )
Thanks for that - I have never enjoyed finding anything in that manual.

My understanding is that the plug-in can indeed query the sample rate, but can't do much about it, since the sample-rate is "global". I'm not saying that what Oxygen Audio describes cannot be done, just that it is awfully awkward, and goes against the plug-in philosophy.

As I see it, the main issue is that a plug-in is normally called once for each sample. The plug-in would therefore need to buffer enough samples in order to perform the up-sample, then do any "magic" it was meant to perform, and then decimate back down to the native sample-rate. Add to that the desires of the user, who might have specific sample-rates in mind for her processing.

The more intuitive approach would be to up-sample first, if you deem that it is necessary. Then you can run your entire plug-in chain at that sample-rate, considering that other plug-ins within the chain might also benefit from the higher sample-rate. The final decimation down to 48kHz (or 44.1kHz) doesn't need to occur until your processing is finished.

This is how I understand the VST philosophy: Plug-ins are minimal building-blocks that allow you to build complex chains of effects without creating unnecessary overhead. Having a single plug-in take it upon itself to re-sample seems to go against that philosophy. That is why I found Oxygen's response to be odd.
MarkHotchkiss is offline  
post #14 of 16 Old 07-19-2013, 12:49 PM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 134
Quote:
Originally Posted by MarkHotchkiss View Post

It must have something to do with their architecture (I don't know much about Hydrogen Audio).

Hydrogen Audio is a science-minded audio forum:

http://www.hydrogenaudio.org/forums/index.php?act=idx

Here's that specific thread from about a year ago:

http://www.hydrogenaudio.org/forums/index.php?showtopic=95472

--Ethan

RealTraps - The acoustic treatment experts
Ethan's Audio Expert book

Ethan Winer is offline  
post #15 of 16 Old 07-22-2013, 08:09 PM
AVS Special Member
 
MarkHotchkiss's Avatar
 
Join Date: Mar 2007
Location: Long Beach, California
Posts: 1,476
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 82
Hi Ethan,
Quote:
Originally Posted by Ethan Winer View Post

Hydrogen Audio is a science-minded audio forum:

http://www.hydrogenaudio.org/forums/index.php?act=idx

Here's that specific thread from about a year ago:

http://www.hydrogenaudio.org/forums/index.php?showtopic=95472

--Ethan

Thanks for that. I had no idea what Hydrogen Audio was, and assumed it was a software or manufacturing company. I'm impressed by the caliper of participants, and now have them bookmarked.

I just now finished reading your thread, and have a few comments:

First, although your question was quite general, the responses were more narrowly focused on IIR filter-based equalizers. IIR filters make for a good example, based on their non-linear properties, but it does tend to slant the responses. The equalizer software mentioned, particularly the Sonoris Equalizer software, are more "applications in a plug-in" rather than a single "effect" in a plug-in. In that type of situation, it is indeed desirable to have the sample-rates controlled by the plug-in, as the final result is very dependent on the sample-rate. So, in that context, I would agree that up-sampling within the application would be preferred.

There was only one response that referred to non-linear effects unrelated to EQ (post 4: compressor/expander and wah-pedal), and that was more the application I was considering when I mentioned that sample-rate conversion should be done outside the plug-in: a series of effects in a chain, typically used as inserts.

I also noticed that no responses addressed FIR filters, where benefits can be derived from down-sampling as well as up-sampling. Bilinear transforms are unnecessary with FIR filters, because they have no frequency-warping, so that benefit of up-sampling doesn't exist for FIR filters.

One thing that the thread pointed out, but could have stressed more, was that resampling would not be inferior to having the original recorded at the higher sample-rate. Although resampling can add a insignificant amount of noise, so can recording at a higher sample-rate. This is a fact that I often forget.

All around, it was a very informative thread.
MarkHotchkiss is offline  
post #16 of 16 Old 07-23-2013, 11:26 AM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 134
^^^ Okay then!

RealTraps - The acoustic treatment experts
Ethan's Audio Expert book

Ethan Winer is offline  
Reply Audio Theory, Setup, and Chat

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off