I follow it close enough, John.
I actually own a BenchMark ADC1/DAC3 and dont use a computer and soundcard at all for recording and playback. Other than to download HD files. But my questioning is for the wider audience. Even the average consumer. And, sometimes, i want to be the average consumer and not have to worry about all this to get the highest music accuracy I can.
That does not seem possible today at prices most would be willing to pay. So, I dont see a path for average next gen peoples to get to a higher audio sound quality.
All the people trying to make the digital part of the system work smoothly and best quality is exactly what we did when the analog was all. I look forward to their discoveries and improvements. Digital POOGE?
-Richard
Last edited:
It works right here, since I can see the format being sent out from Foobar ASIO in the Lynx mixer. Foobar ASIO sets the sample rate and bit depth according to what is required for each file as the file is being queued up to play. That way a mix of file types can be put into one playlist and each file will play correctly.
Last edited:
I'd love to enjoy Eric Dolphy with people If you consider this is what "average joe" will do are you really saying that Foobar ABX is a 'test'? I don't usually see null results queried as to the setup used for listening, but do see this done when positive results are reported. Is this not proving the point about the issues in using Foobar ABX as a 'test' - it didn't even occur to me - I just naturally assumed people doing a 'test' would use the best quality playback (an external USB DAC) not the computer's in-built soundcard
Good god, it's incredible that this level of variability is considered worthy of being called a 'test'
Last edited:
Just about two decades ago I was working with a guy developing a feedforward error correction system to linearize analog fiber optic transmission of the CATV spectrum. This fellow was really interesting - his undergrad degree was in EE, but his PhD was in physics. (He was mostly interested in optical systems.)
One day, he said to me, "Hey! Look at this." It was a MATLAB simulation of an audio amplifier. He was using it as a simple analog of the more complicated optical and RF system we were trying to build in order to test out a part of his system model.
His simulation showed that amplifiers with overall loop feedback worked almost perfectly with sine waves, but could have a problem with signals that didn't have long phase coherence times. This was a big deal with the optical system, since a lot of practical (cheap) lasers don't have really good phase coherence. But, the model showed that the audio noise floor in the audio amplifier would modulate when the audio signal phase coherence didn't hold.
He went on to explain that, in his mind anyway, this would be like what might happen when an actual human played a trumpet. Over the course of a couple seconds, the natural phase modulation of the tones produced by the trumpet would vary. The trumpet frequencies might stay perfectly constant, or might vary, but neither really mattered. It was the phase coherence that mattered.
He went on to say further that the more non-linear the input stage was, the worst the effect. Overall feedback would correct the harmonic and IMD generation to the extent that was expected pretty much, but the phase modulation variation would be translated into a noise floor variation.
When I thought back on that, I realized that the overall block diagram of an audio amplifier using loop feedback is similar to that of an FM detection system or a phase noise detection system. There's a non-linear mixer at the input, and a time delayed version of the input signal is fed to another input port of this mixer. The output of the mixer is then a demodulated spectrum of the phase modulation of the input signal. The math is pretty straightforward.
That make any sense the way I explained it?
Of course, the characteristics of the non-linear input stage, aka mixer (mathematical multiplier), vary from circuit to circuit and device type to device type, as does the characteristics of the "delay line" (the voltage amplifier stage and the output buffer.) This all affects the "demodulator" characteristics.
*If* you subscribe to this at all, that *might* be an explanation for noise floor modulation and why multi-tone tests don't show any results.
I've been using Reaper for several years now and it does all resolutions/clock rates the hardware supports in the record and playback modes. Internally, all operations are carried out with 64 bits. Files can be saved in any of the standard formats and/or resolutions.
I paid once for a personal license and have been updating it continuously ever since.
The hardware is MOTU Audio Express, running under the MOTU ASIO driver. I also use the same setup to acquire data for further analysis (THD, etc.).
Regards,
Braca
Very interesting, CG
Has anybody analysed real music signals for phase coherence?
In other words, is it a property of the signal itself or a property of electronic stage non-linearities or an interaction of both?
Noise floor modulation would seem to be a tricky distortion to measure both in terms of test signal & actual measurement
It certainly isn't among the standard measurements but may well be audible?
Has anybody analysed real music signals for phase coherence?
In other words, is it a property of the signal itself or a property of electronic stage non-linearities or an interaction of both?
Noise floor modulation would seem to be a tricky distortion to measure both in terms of test signal & actual measurement
It certainly isn't among the standard measurements but may well be audible?
Last edited:
I've not found an appliance that does everything I want. At some point I have to give in and either try and write some code myself (unlikely) or put a PC in the way which for me is a fail on about 3 counts. But my requirements are unusual, but not unique.
Not sure how many people actually need ADC though for domestic replay. People either are all digital or all analog. It's a limited number who will consider a digital stage in the way who play LPs.
As for SRC. Surely the answer to that is to process all your files off line to get them into the same format so SRC is minimal? ( I should note I appear to be insensitive to the gremlins of SRC so its low on my personal worry list).
That's correct, unfortunately.
I made a mistake in my previous post. Instead of "Audacity and Cool Edit Pro 2.1" I wanted to write "Adobe Audition and Cool Edit Pro 2.1". Audacity is free, which is good, but it is not that reliable tool.
I am testing Foobar + card by playing 24bit test files (like -120dBFS as well), recording it and analyzing either by AA or CE.
And an unvalidated simulation that we don't have our hands on to actually understand. It may be totally right for whatever topology was modeled. But it raises eyebrows in that it'd wreak havok on hundreds of thousands of applications where this purported noise modulation would materially affect the measurements being made, especially cases where the signal sits barely out of the noise.
So, I'm not necessarily buying it (especially as a major effect). I've perhaps made too many well-intended-but-garbage simulations in my life that did not survive contact with the validation experiment.
So, I'm not necessarily buying it (especially as a major effect). I've perhaps made too many well-intended-but-garbage simulations in my life that did not survive contact with the validation experiment.
And how do you know if it was done when there are no internal controls in the 'test' to verify the 'test' is sensitive enough?
I've never seen this issued as a warning along with the challenge to someone to 'prove' what they hear using Foobar ABX. It would seem that a null result is unquestioned as to it's validity- only positive results get scrutinized, yet accumulated null results are regularly used as 'evidence' of inaudibility
No, but in general an FFT of music would have energy in all bins so if you pass the music through a non-linear process you will get IM products in all bins in other words "noise". A different section of music and a different noise. With one or two sine waves there are only limited number of harmonics and all other bins are 0. A 32 tone multi-tone hits a lot of bins and if you shuffle the amplitudes of the multi-tone you will see a change of the "grass". The FFT process itself is noiseless.
This is all easy to do with simulation.
Well... I have no idea if anybody has analyzed music for this. I'd suspect that zillions of people have looked at the spectra of music to see if the notes coming out of instruments (including voice) are phase modulated or not. I just don't know. The point I'm suggesting is that there is a possible mechanism for why an amplifier noise floor could be modulated based on amplifier non-linearity of a sort not usually looked at, based on some naturally occurring characteristics of actual sound that isn't tested for in traditional distortion analysis. And, I'll be the first to suggest that it's quite likely that someone has tread this ground before - somebody certainly smarter than I am.
Measuring this kind of thing is probably challenging, but I'd think that perhaps using a "peak hold" measurement to capture the noise floor over a number of seconds might be interesting. Most of the time, audio measurements average for several seconds in order to reduce the effect of the noise floor. Here, we might want just the opposite. (In the telecom world, where complex modulation like QAM and OFDM are used, it's not uncommon to do this. A single bad constellation sample caused by who-knows-what gets averaged out, barely showing a change in MER. But, a bunch of bits are lost...)
A simple test to try might be to just apply a phase modulated tone and capture the noise floor over a few seconds. See how much of the modulation gets "demodulated" by the amplifier.
Maybe this is nothing. Maybe it's something but so low in level that it may as well be nothing.
It's tempting to hope that it may be something, since a lot of what people have been suggesting based on anecdotal observation - for better or worse - could be explained by this. But, if it's nothing, at least I've given everybody something else to heap ridicule upon.
- Status
- Not open for further replies.