How do you perfom intermod tests to check high sample rate performance of converters?

[simplersounds]

https://web.archive.org/web/20200202124 ... young.html


Does anyone know how to perform these tests?

Do I monitor at the same sample rate as the files to check? example: file is 96kHz, monitor at 96kHz to check for intermod OR file is 96kHz, monitor at 48 to check intermod ?

[BertKoor]

Not having read the archived page yet, I trust RMAA (RightMark Audio Analyser) for measuring all sorts of possible abberations in digital audio. If RMAA doesn't find anything wrong, then nothing is wrong.
https://audio.rightmark.org/products/rmaa.shtml

[simplersounds]

bump!

[BertKoor]

Does anyone know how to perform these tests?

Now having read some parts of the article: download the wav samples from the above archive and play them back. You should hear absolutely nothing.

If the files are not available, their description is enough to reconstruct them. Take any capable audio editor that has a sine tone generator. New empty file of 96kHz sampling rate, add sine tones of 30kHz and 33kHz. Play it back, and you should hear silence.

And just as I thought, RMAA tests for IMD as well and not just in the ultrasonic region of the spectrum:

A new test of intermodulation distortion has been added. From RMAA 5.1 the common IMD test uses a standard SMPTE test signal. In a new Swept sine IMD test, we use a set of 2 harmonics with a frequency difference of 1 kHz (CCIF standard), sweeping the whole audible range of frequencies. Thus, the dependence of intermodulation from frequency is tested. This test allows detecting high-frequency intermodulations that are specific to internal sample rate convertion (SRC) algorithms of a sound card.

Do I monitor at the same sample rate as the files to check? example: file is 96kHz, monitor at 96kHz to check for intermod OR file is 96kHz, monitor at 48 to check intermod ?

If you let your audio interface first resample it down to 48kHz, then you are testing the brick-wall filter in the resampling of the interface (all content above 24kHz must be filtered out) , not any possible intermodulation that occurs later in the chain.

[simplersounds]

Appreciate the reply! Just to clarify for myself, I set my DAW and interface to 96k while monitoring the 96k file?

[BertKoor]

Yep.

[simplersounds]

My converters produce a click at all sample rates, although the tone/noise is only audible at regular listening levels at 96k, improves at 48k. What is the take away here you think? I guess I'm wondering if these tests are actually revealing a problem or not? Does anybody have a converter/interface that is able to pass the signal without audible noise, clicks, tones etc. ?

My supposedly high end interface failed, but I won't name. If anyone 'passes' would you mind sharing what interface and/or converters you're using?

Thanks!

[BertKoor]

Which of the test files did you use?

It might not be due to the interface, it might come from your speakers.

You can do a test with just a 30kHz sine. That tests not intermodulation, but a burst of ultrasonic sound.

How realistic do you think this test is? Have you ever looked at the spectrum of real music? It usually has Iess content in each higher octave. If you buy commercial 96kHz tracks, it might even have nothing above 21kHz. Because people are suckers for the bigger numbers.

[simplersounds]

Which of the test files did you use?

Intermod Tests:
30kHz tone + 33kHz tone (24 bit / 96kHz) [5 second WAV] [30 second FLAC]

Song clip shifted up by 24kHz (24 bit / 96kHz WAV) [10 second WAV]

^ I used these two ^

[BertKoor]

I might have a try at it, but that requires first dusting off my desktop PC that has a capable 192kHz audio interface but also has been dormant for 5 years. I think it still runs Windows XP and it is 3 floors away from the nearest ethernet cable.

Instead I'll do a mind experiment. If you are serious about testing something, you follow a structured approach. What component are you testing? Is there anything else in the chain that could influence it? Can you exclude that? What exactly are the observations? Are there any plausible explanations for the observations? Can we do further experiments that proof whether my hypothesis is right or wrong?

So you hear a click. Is that only at the start, or also at the end of the file? Do you also hear a click when playing an audio file containing absolute silence? (yes I have seen/heard that happen although it was decades ago, but you must rule it out)

Now my mind experiment. Suppose I have an interface I want to test, hooked up to an "ideal" amp and a speaker that can faithfully reproduce frequencies up to 100kHz.

Side line: you will have trouble finding proper specs of your amp & speakers beyond 20kHz. So how your amp & speakers really behave is unknown. Around 1989 I bought a simple Sony hifi amp, and in the technical specifications they bragged it could reproduce frequencies up to 100 kHz. Manufacturers don't do that anymore. Also for an amp high frequencies are no problem (as long as you stay in kiloHertz range, not gigaHertz, and it is designed for it) but for speakers it's a whole different story. There are specialised ultrasound transducers for a reason. My point is that common hifi equipment might pose some limitations in these ultrasonic experiments. And that's also what RightMark Audio Analyser does: take out all the rest of your equipment from the chain, just hook up the outputs of your audio interface into its inputs and go measure the interface in isolation.

Back to the experiment: ofcourse I cannot hear ultrasonic. But if I put my finger on the speaker membrane I should be able to feel whether it is moving or stationary. If you switch the ultrasound on & off periodically, would you then hear that? This is in principle modulating the amplitude of an ultrasound carrier wave. Me thinks given the right circumstances and non-ideal equipment you can hear that. Lower the frequency, go lower and lower, you are left with just the switching on of the ultrasound wave.

If you move a membrane forward instantly and do nothing more than that, it will sound like a click. I think that's what you are hearing, not any intermodulation effects. You need to do more experiments to prove or disprove that. You might for instance make a fade-in & fade-out on the audio files. That should get rid of switch-on & switch-off effects and you'd be left with the pure intermodulation things.