NOS DAC with oversampling - anybody done it?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I know this is an oxymoron statement. But the basic arguments against oversampling itself is the introduction of the digital filter. Does anyone out there know of any DAC which oversamples but does not introduce digital filtering?

I think this would probably be the best implenmentation for this case.

Oon
 
Dear Philpoole,

I agree that oversampling is a form of digital filtering, but its not the one that supposedly cause the stuff that is really bad which is the brickwall 20KHz filter which I believe uses some form of DSP to recreate the original signal.

Dear Bernhard,

Is there any reason, why hardware/linear interpolation doesn't work well? Could you point me to the specific article or thread or review.

Oon
 
Hi Oon,

The brickwall filter is an analogue filter, not a digital filter which technically should be used with a NOS DAC, but it was discovered that this could be omitted and it wouldn't sound too bad. It was omitted because it could often be a 5th, 6th or 7th order filter, with lost of phase shifts going on which can alter the sound. Hence the introduction of the use oversampling to simplify the requirements of the analogue filter. My 8xOS can get away with a 3rd order filter, maybe even 2nd order- although I haven't tried.

Phil
 
But these filters (ed. oversamplers) add pre- and post ringing to signal. The sound is not so precise and clear. It is slushier. In the
HB-DAC1704 no traditional oversampling filters are used.The upsampling method is done by a very precise interpolation in the SRC and no downsampling filtering is needed when using the direct downsampling feature of the new high end chip (SRC4392)

They're using a sample rate converter, which some argue is worse than an oversampling digital filter. According to some, being more versatile (can convert to a range of sample frequencies and bit depths as opposed to just one) can compromise on performance.
Either way, its still a digital filter.

I don't think there's such a thing as a free lunch.

Have you actually tried oversampling?

I started with a NOS DAC, and progressed to oversampling and haven't looked back.
I think the benefits outweigh the concerns - if implemented well.
 
Dear Philpoole,

I believe the brickwall filter is a digital filter. That is why it couldn't be fixed in the DAC part per se, it's in the digital receiver section. But quite often you couldn't disable the digital filter without disabling the OS feature. I have some experience in this matter looking at how the different DAC respond o this. On an oscilloscope, if you were to feet in a sinewave at very high frequency, say 15KHz, you can actually see and hear the beating of the frequency with the sampling frequency. But with a normal DAC you would actually get a full sine wave. In other words, the receiver part actually reconstruct the missing signal. But when you try fitting in a normal music signal, it could confuse the reconstucting digital circuitry.

I think if look up TNT, you can find the interview with the original NOS DAC from 47 labs I think, he mentions it.

I'll try and look it up for you.

BTW Philpoole, I am a bit confused on your last statement, are you using a you using NOS DAC or an oversampling DAC now, when you say you haven't looked back.

Oon
 
Hi,

I vaguely recall some comment about a brickwall filter and a digital filter. But I still feel the most obvious brickwall filter is in the analogue stage.
Oversampling effectively keeps the transfer function the same (audio up to 22kHz, then nothing), perhaps that is interpreted as a brickwall? The point of oversampling is that you raise the sampling frequency, so it is much easier to filter in the analogue domain.

I'll investigate what you suggest.

With DSP, at the other side of the sampling frequency, you get an aliased signal. In NOS, this starts at 66kHz, so you ideally need to pass 22kHz, but have cutoff completely at 60kHz - that's tricky, and is often called a brick wall.

I have 8xOS now, and I much prefer it to NOS. That's what I meant. I've never looked back, I wouldn't return to NOS.

Cheers,
Phil
 
I have some experience in this matter looking at how the different DAC respond o this. On an oscilloscope, if you were to feet in a sinewave at very high frequency, say 15KHz, you can actually see and hear the beating of the frequency with the sampling frequency. But with a normal DAC you would actually get a full sine wave. In other words, the receiver part actually reconstruct the missing signal. But when you try fitting in a normal music signal, it could confuse the reconstucting digital circuitry.

What DACs were you comparing? Were you seeing the output of the DAC unfiltered? Maybe you were seeing IMI caused by aliasing that would be filtered out by the analogue filter?

I guess it was filtered. With an unfiltered NOS DAC, 15kHz would be attenuated.
 
I can't quite recall excatly where I was probing but would have been the analogue output. But when you put it so close together, frequency so close to the nyquist frequency, your sampling point in a sine wave would progress a little bit down the phase with every new cycle and hence it looks like a beating.

Oon
 
Analogue output of the DAC? If so, that's an unfair test. Testing after the filter would be better.
But when you put it so close together, frequency so close to the nyquist frequency, your sampling point in a sine wave would progress a little bit down the phase with every new cycle and hence it looks like a beating.
- Are you referring to NOS then? The idea of oversampling is to effectively increase the nyquist frequency, so wont have frequencies so close to the nyquist frequency.
I thought you were reporting this odd beating with oversampling.
 
Analogue output of the after the low pass filter of a NOS DAC. I downloaded some .wav file at 15kHz or so then send it through the DAC. Measured at the output of DAC. the point is the sampling point taken on each of the waveform is a bit behind the phase. So without the digital filter it won't get corrected.

For example if the first sample is taken at positive peak at 15kHZ. The next positive sample will be taken at positive peak minus 21us (66us -44us). And hence it gives the illusion the waveform is decreasing in amplitude and creates a beat.

The OS sampling DAC recreates the sine wave quite beautifully.

Oon
 
I think I'm confused by this whole discussion :xeye:
I now realise you saw issues with the NOS DAC (e.g. beating) rather than the OS DAC. Initially, I thought you were looking out for an improvement to NOS, because of problems like this beating (which, confusingly for me, I assumed you saw with OS).
I now understand that's all false.
You've seen the fundamental flaw of NOS. So, what's the matter with OS?
 
Hi Philpoole,

Yup, I was looking on a compromise between NOS and OS, maybe interpolation as was mentioned by thor66. The standard OS actually recreates the wave again from the signal input. That is why you get a nice sinewave instead of a beating signal. However, when you push in a non sinewave, the behaviour becomes unpredictable. For example as shown by the interview at sakura systems, if you put in a square wave, you get a pulsed sine wave.

So since music waveform is rather random, there could be issue in the accuracy of the reconstruction.

For one, i notices, that in a NOS, the ssss sound in a singer's voice tent to be more muted, and to me that sounds more natural, since I don't hear sss all the time during a conversation with some one.

Oon
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.