Pro Audio Design Forum

I had an odd idea that might have merit, but I don't have a digital platform worthy of testing.

The conundrum with trying to null test digital audio paths is the conversion time delay, even a slight delay will degrade null depth at high frequency.

My proposal is to add some delay to the digital signal between A/D and D/A such that the analog output is delayed exactly 1/2 or one whole input sine wave period (plus the conversion delays) so the analog sinewave input can be precisely nulled against the analog sinewave output.

Delaying the signal 1/2 wavelength seems like it would also provide a free signal inversion, but upon reflection this would also reveal asymmetrical distortions in the original sine wave source, so one full period with the inversion provided in the analog side null circuitry seems prudent.

It is probably easier to tweak the frequency of the sine wave for best null instead of the digital path delay which may not have fine enough time resolution.

So for example an approximately 1kHz sinewave delayed just enough (roughly 1mSec but slightly less) so that when added to the conversion delay the signal comes out exactly in phase with the input (actually 360' out) for nulling.

It seems any DAW with rudimentary delay capability could do this, with a sine wave source that is fine adjustable frequency. With a full wavelength delay the purity of the sinewave should not matter as long as it is stable.

Hopefully this would confirm that digital distortion or artifacts are vanishingly small in a decent codec path.

JR

I think what you're thinking of may be similar to Audio Diffmaker: http://www.libinst.com/Audio%20DiffMaker.htm

I have a copy and have been entertaining comparing an analog RIAA EQ'd playback to a biquad-based DSP RIAA vst plugin.

Audio DiffMaker is a freeware tool set intended to help determine the absolute difference between two audio recordings, while neglecting differences due to level difference, time synchronization, or simple linear frequency responses.

mediatechnology wrote:I think what you're thinking of may be similar to Audio Diffmaker: http://www.libinst.com/Audio%20DiffMaker.htm

I have a copy and have been entertaining comparing an analog RIAA EQ'd playback to a biquad-based DSP RIAA vst plugin.

Audio DiffMaker is a freeware tool set intended to help determine the absolute difference between two audio recordings, while neglecting differences due to level difference, time synchronization, or simple linear frequency responses.

No... That compares two streams after conversions to digital, so it ASSumes the digital conversion is perfect.

My proposed null is performed in the analog domain and directly looks for differences between the input and output after the A/D and D/A conversions. The secret sauce is delaying the digital path just short of one wavelength to time align them. This only works for sine waves or repetitive signals.

JR

The conundrum with trying to null test digital audio paths is the conversion time delay, even a slight delay will degrade null depth at high frequency.

Which is why people who do "in the box" MS but want to process Mid, Side or both in analog by looping out to a converter and EFX always complain about having a phase hole.
For the same reasons you cite very small latency, phase and level errors prevent hybrid ITB/Analog OTB MS from being practical.
At 44.1, a one sample offset is a huge phase error when reconstructing MS.

I'm still not sure why Audio Diffmaker couldn't be used.

The stimulus would feed the D/A looped back to the A/D and the converted file compared to the original stimulus.
The delta would reflect the conversion errors.

People take imperfect CD rips of files and compare the original wav to the flac versions.
flac is lossless and people use Diffmaker to prove it to themselves.

What you propose would be a useful utility.
What do you use to analyze the null? Another PC or sound card? Or just listen to it?

mediatechnology wrote:

The conundrum with trying to null test digital audio paths is the conversion time delay, even a slight delay will degrade null depth at high frequency.

Which is why people who do "in the box" MS but want to process Mid, Side or both in analog by looping out to a converter and EFX always complain about having a phase hole.
For the same reasons you cite very small latency, phase and level errors prevent hybrid ITB/Analog OTB MS from being practical.
At 44.1, a one sample offset is a huge phase error when reconstructing MS.

I'm still not sure why Audio Diffmaker couldn't be used.

The stimulus would feed the D/A looped back to the A/D and the converted file compared to the original stimulus.
The delta would reflect the conversion errors.

Yes, apparently if you start in the digital domain you can compare the trip out and back in again.

People take imperfect CD rips of files and compare the original wav to the flac versions.
flac is lossless and people use Diffmaker to prove it to themselves.

What you propose would be a useful utility.
What do you use to analyze the null? Another PC or sound card? Or just listen to it?

Null could be done in any console with polarity flip... result is simple measurement of product level.

Listening to the null product could reveal the nature of the error.

JR

PS: I was not familiar with diffmaker but it seems it could be used to characterize D/A - A/D loops.. once quantified, A/D or D/A could be swapped out to ID which is the larger source of error.

I've found listening to and measuring nulls can be very instructive.
Sometimes you can see needles in haystacks.

Do you have a way to run Diffmaker?
I recall that you use a Mac.

mediatechnology wrote:I've found listening to and measuring nulls can be very instructive.
Sometimes you can see needles in haystacks.

Do you have a way to run Diffmaker?
I recall that you use a Mac.

I don't have any A/D D/A loops to test... '

just sharing my mental masturbation.

JR

I think I might try Diffmaker on the Dual Class-A II.

I could use the D/A to A/D loopthrough as the reference file.
Inserting the Dual Class-A II between D/A and A/D provides the comparison file.

That nulls out the conversion path. Music can be used as the test source.

I did some null testing here using both Audio DIffmaker and also using simple invert/summation to compare DSP-based RIAA EQ to Analog RIAA EQ: viewtopic.php?f=15&t=885

I shared this idea with a friend of mine who is pretty into this kind of stuff, and he said an A/D/A path has enough intrinsic delay to work without having to add any extra delay, just pick the right frequency sine wave and null away...

I don't know if he tried it yet.

JR