KVR Audio

Cordelia · Post by **Cordelia** » Tue Jun 05, 2007 9:32 am

Where is thorkz?
He can hear it.

tetsuneko · Post by **tetsuneko** » Tue Jun 05, 2007 9:54 am

can I get a unibin version for testing?

Kingston · Post by **Kingston** » Tue Jun 05, 2007 10:25 am

JonHodgson wrote:As for this test, 64 bit ints versus 64 bit floats? Come on, considering that no D/A converter manages better than 20/21 bits worth of signal to noise at full scale if a programmer can't make any differences between the two inaudible, then fire the programmer.

indeed. it should be remembered that the most expensive high end nasa grade DA converters can handle roughly 120dB s/n ratio, which is close to the theoretical limit of AD/DA electronics.

It baffles my mind how differences happening at -180dB levels could somehow propagate into the audio stream. not even magic could make it audible, unless there's a fundamental flaw with the signal processing theory as we know it today.

aciddose · Post by **aciddose** » Tue Jun 05, 2007 10:42 am

"not even magic could make it audible"

errors are cumulative in a majority of dsp processing.

camsr · Post by **camsr** » Tue Jun 05, 2007 10:45 am

wtf are "nasa grade" convertors? dB is a relative term and at higher voltage its possible to get a snr better than 120dB. Why not?

JonHodgson · Post by **JonHodgson** » Tue Jun 05, 2007 10:46 am

aciddose wrote:"not even magic could make it audible"

errors are cumulative in a majority of dsp processing.

That's a HELL of a lot of cummulation.

Kingston · Post by **Kingston** » Tue Jun 05, 2007 10:46 am

aciddose wrote:"not even magic could make it audible"

errors are cumulative in a majority of dsp processing.

yes I agree if these filters were used as part of neural network with literally thousands of connections the s/n ratio limit would probably become an issue.

Kingston · Post by **Kingston** » Tue Jun 05, 2007 10:48 am

camsr wrote:wtf are "nasa grade" convertors? dB is a relative term and at higher voltage its possible to get a snr better than 120dB. Why not?

do a bit of googling of thermal and other (usually) secondary electronic effects that start to physically limit AD/DA conversion and the s/n ratio. oh dearie, we're actually limited by the physical reality.

Christian Budde · Post by **Christian Budde** » Tue Jun 05, 2007 10:49 am

camsr wrote:wtf are "nasa grade" convertors? dB is a relative term and at higher voltage its possible to get a snr better than 120dB. Why not?

Yes 'dB' is realtive, so why should it be any better at higher voltage? Keep in mind, that it's also logarithmic. AFAIK it's independent of the input voltage. If you gain your signal you'll get additional noise and the SNR stays the same :-/

Give me some minutes and I'll try to post a plot with the residual distortion. Then I also calculate the SNR...

aciddose · Post by **aciddose** » Tue Jun 05, 2007 10:49 am

it's easily an issue with 32 bits - with 64 bits it becomes a non-issue.

that said, Christian; you must provide the source code in order for this "challenge" to be valid. i suspect that i'll be able to point out major flaws in your implementations - for example, that both pieces of code implement exactly the same thing only with data type changed.

Kingston · Post by **Kingston** » Tue Jun 05, 2007 10:51 am

aciddose wrote:it's easily an issue with 32 bits - with 64 bits it becomes a non-issue.

yeah. in case you didn't notice we're fully 64bit here. since the first post.

aciddose · Post by **aciddose** » Tue Jun 05, 2007 10:52 am

"If you gain your signal you'll get additional noise and the SNR stays the same :-/ "

this is completely incorrect - you must understand how noise works in electronic circuits to understand why. in most amplifying circuits, a circuit with a gain of 1 will have noise at a voltage level equal to that of a circuit with a gain of 100. the difference will be that in the circuit with a gain of 100, the relative noise level will then be 1/100th.

a correctly designed amplifier is one where the noise level generated inside the amplifier itself is much lower than the noise level that will ever be encountered at the input. your statement assumes that the input signal contains noise which is then amplified - however the truth is that in most circuits the output noise is mostly composed of noise which has been generated inside the circuit itself.

camsr · Post by **camsr** » Tue Jun 05, 2007 10:56 am

Christian Budde wrote:
camsr wrote:wtf are "nasa grade" convertors? dB is a relative term and at higher voltage its possible to get a snr better than 120dB. Why not?
Yes 'dB' is realtive, so why should it be any better at higher voltage?

Heat won't have the same effect on a higher voltage because it just dominates with energy. And were not talking about gain, this is about DA conversion or have you already forgot? Why do you think studio equipment runs on a higher voltage than consumer grade equipment? The increased energy factors out other sources of noise.

whyterabbyt · Post by **whyterabbyt** » Tue Jun 05, 2007 10:57 am

camsr wrote: Why do you think studio equipment runs on a higher voltage than consumer grade equipment?

whit?

JonHodgson · Post by **JonHodgson** » Tue Jun 05, 2007 10:58 am

camsr wrote:wtf are "nasa grade" convertors? dB is a relative term and at higher voltage its possible to get a snr better than 120dB. Why not?

It seems it's not this simple, ask the DAC designers for an explanation, but the simple fact of the matter is, NOBODY does any better than this. It's pretty irrelevant though, because the point at which it top grade converter loses its linearity is way below the noise floor of the rest of your system.

Integer is King? - the challenge