XXHighEnd

Well, not really of course.

BUT... the other day, I thought I'd try something that I've wanted to try for a while. I played a 16/44.1 file through XX and captured the output of the NOS1 on the PM2 at 24/192. Comparing this recorded 24/192 file (non-upscaled) with the original 16/44.1 file (upscaled) shows that the recording has captured the output of the NOS1 really well. The only cables I had to hand to connect the NOS1 to the PM2 were pure silver, and these have added a bit of brightness to the original sound. But otherwise, it's pretty accurate.

Now, the file happens to be 'So What' from 'Kind of Blue'. Being a 1959 recording, this is now out of copyright - I can distribute it freely. If anyone is interested in hearing the 24/192 recording, I'm happy to place it on a free file-sharing site.

Of course, the 24/192 should NOT sound better than the original 16/44.1 (which I'm sure is in your collection) played back on whatever DAC you have. But I think you'd find that it sounds at least as good, if a little brighter though. And this is remarkable...

... considering that the 16/44.1 has been upscaled in XX, passed down a USB cable, through the NOS1's D/As and its output stage, into the PM2's input buffer and through its A/Ds and finally through the Weiss AFI1 back to a second PC which captured it.

Anyway, let me know if you'd like to hear how an async-USB NOS1 sounds... but not really of course!

Mani.

Still laughing. But also : usually you are more serious than you want to believe yourself. So somewhere somehow you must be talking about something inportant.

So, what if you post that again without the suggestions it won't be worth a thing ?

?

I'll give the virtual experience a try, send it on over......and thanks for taking the time to be so crazy but fun!

Well, I did set out to do something 'important', but soon realised that it was a stupid idea and a total waste of time.

What I wanted to do was to determine how 'perfect' the NOS1 really was. The method would be simple: record the NOS1's analogue output and compare the recorded file (played back on the NOS1) to the original file (played back on the NOS1). The idea was that any 'distortions' in the NOS1's output would be compounded on playing back the recorded file (because it would have passed through the NOS1 not once, but twice)... Therefore, the closer the original file and recorded file sound, the more 'perfect' the NOS1 must be. If they sound identical then the NOS1 must be perfect.

That was the idea. BUT... of course it assumes that the ADC is perfect! Well, I have a good ADC, but I'm not sure it's 'perfect'.

FWIW, the recorded file does sound remarkably similiar to the original file, but not quite. Interestingly, the hiss at the beginning sounds totally different - more 'grainy' and quite a bit brighter. I'm pretty certain that this has more to do with: 1) the analogue cables; 2) the ADC used; and 3) (and most importantly) the firewire interface used with the ADC... than the NOS1 itself.

So, really a total waste of a couple of hours. But to make 'something' of this time, I thought people might like to hear the recorded file anyway, just for the hell of it. This in itself is totally stupid though, because they'd just be hearing their DAC and not the quality of the NOS1!

Peter, you're usually correct on most points. But on this one, you're completely wrong - I was trying to talk about something serious, but failed miserably.

Mani.

Thanks for humouring me! I'll get the file to you once I'm back home (this weekend).

Would be interested in hearing your thoughts anyhow...

Mani.

Mani, I am sorry, but I just didn't understand it. As said, you are always serious, but in the mean time seemed to debunk your own "idea".

I guess (now) you were just not confident yourself about it ...
In the end though, you may have proven something. But, not sure indeed.

In any event, thanks. And must think about it a little ...
Peter

Like I said, it occurred to me that what I was trying to do was just full of flaws and a bit silly.

But as we say here in the UK, "In for a penny, in for a pound":

Find below three spectrum plots. These were derived from a 10 second 16/44.1 file of a 5KHz square wave.

1. This is the spectrum of the file 'upsampled' to 24/192 using the iZotope program without any anti alias filter applied. So this upsampling was done purely in software.

2. The second file is a 24/192 recording of the 16/44.1 file being played back on XX. The details here are:
- XX set to Double Octo (16x) Arc Prediction -> 705.6KHz
- PC connected to NOS1 via USB
- NOS1 set to 'DAC is 24/786'
- NOS1 analogue output connected directly to PM2 analogue input
- PM2 analogue-to-digital conversion set to 24/192
- PM2 digital output connected to Weiss AFI1 via dual-AES (PM2 master clock)
- AFI1 connected to second PC via firewire
- digital output captured on second PC with Sony 'Audio Studio'

3. This files shows the noise floor of the complete chain above.

So what's there to learn from all this? I'm not sure...

But firstly, all the peaks above 22.05KHz are really 'false' and unwanted - only peaks which are an integer multiple of 5KHz would be 'real', and there are essentially none above 20Khz because the original file was only 16/44.1.

However, I'm amazed at the similarity in the waveforms between 1. and 2. Something is telling me that what I'm showing is a great demonstration of the accuracy of the NOS1 - the fact that it can match, almost identically, the upsampling done in software.

What none of this shows, however, is how the NOS1 sounds with Double Octo Arc Prediction. I've compared a number of tracks upsampled with iZotope to 24/192 (with and without an AA filter) vs. DOAP. Upsampling using iZotope just 'kills' the music (and iZotope is supposed to be the best SW upsampler!).

What I set out to do was to 'prove' how good DOAP was. Again, I think I failed. But did I manage to show something of any value meantime? Maybe...

Mani.

Not all that easy to understand !

What I many can add to it, is that there's filtering active beyond, say, 20KHz. Because ... wat you see beyond that is not the "per 5 KHz" ranging of a next harmonic. No, what you see *there* the the "per 5KHz ranging "back" of aliasing with 192Khz as the mirror. Start looking back from 48KHz to see it more clear.


Meaning (I think !) ... there "are" essentially none in the reconstructed 44.1 file, because no more than 22.05KHz kan be in there; however, they show because no Anti Aliasing was applied.
(?)

I tried to make more conclusions, but it is a bit hard to compare the two first pictures, even if you have two open in two browser instances (because it really needs the whole spectrum to be compared. Also, it may be too confusing what is and what is not influenced by the "recording".

Lastly, maybe this is not all that much accurate for a test at all. It may proove that AP is not messing up badly, but the accuracy is at the sample level, and not in a wild FFT, again, if e.g. AP would show it completely different it would only tell that it is wrong all over. Now it's equalish, it doesn't tell all that much.

Peter

No, actually no 5KHz multiples show at all. Many peaks above 22.05KHz look very, very close to being 5KHz multiples, but they're really not.

There's probably nothing to be learned here. But I had fun playing with this all, and still remain impressed by the fact 2. is so 'clean', even having passed through a DAC, an analogue output stage, an analogue input stage and an ADC.

Mani.

Of course you are right. I said that myself ...

Still, think from 48Khz down. There they are again, but now as aliases.

:good: