How can a non-oversampling DAC sound good?

[manisandher]

In the Altman DAC is there a powerful analog filter in the I/V section?

 

But wouldn't some of the posted graphs seem to indicate little if any filtering?

Yeah, it's clear from it's output spectrum that the Altmann has no digital or analogue filter. What's not obvious to me is whether this is a good or bad thing. I mean, it sounds pretty good to my ears... better than I ever would have expected a NOS DAC could sound.

 

Mani.

[VenturaRV]

Was thinking about that. Audio Note DACs used a transformer as the filter. There's a pic of the Altmann board on their site if you want to look. But wouldn't some of the posted graphs seem to indicate little if any filtering?

 

 

Sent from my iPhone using Computer Audiophile

As far as I know, Audio Note top of the line Analog Devices chipset as DAC and double triode in the output stage uses transformers at the very end.

Inferior models uses Philips 1535 DAC chip, double triode and no transformers at all.

My Luxman DA-07 is a non-Shannon, non oversampling and filterless DAC. It was developed from pure math works made by professors at. the Tsukuba University, is really costly and its sound worths the price. No complains about high ultrasonics to my ears or my equipment.

 

 

VenturaRV

[Sunya]

Watts Up...? - Page 12

[Jud]

As far as I know, Audio Note top of the line Analog Devices chipset as DAC and double triode in the output stage uses transformers at the very end.

Inferior models uses Philips 1535 DAC chip, double triode and no transformers at all.

My Luxman DA-07 is a non-Shannon, non oversampling and filterless DAC. It was developed from pure math works made by professors at. the Tsukuba University, is really costly and its sound worths the price. No complains about high ultrasonics to my ears or my equipment.

 

 

VenturaRV

 

Except Luxman replaced it with an oversampling DAC. I am skeptical the Luxman engineers knew less in 2013 when they designed the DA-06 than they did in 1987 when they designed the DA-07.

[Jud]

Watts Up...? - Page 12

 

From that page (Rob Watts from Chord):

 

Basically, if you use an interpolation filter that had a sinc impulse response, then you would perfectly reconstruct the bandwidth limited waveform before it was sampled - so the sampled data would have all the missing bits in between perfectly represented - there would be absolutely no difference whatsoever from the original. So it would not matter whether you were sampling at 22 uS, or 22 fS, the bandwidth limited signal would be absolutely identical in both cases.

 

But to do this we of course need the analogue signal to be perfectly bandwidth limited (that's the job of the ADC - and this actually is not technically difficult to do, Davina for example will have > 200 dB of bandwidth limiting), but from the DAC's point of view we need an interpolation filter that infinitely oversamples and has infinite ringing. To do this we would need a sinc FIR filter that had an infinite number of taps; something clearly impossible to do. So to cope with a limited number of taps, I found that changing the algorithm (this determines the values of the coefficients, and that sets the shape of the impulse response, or how the ringing is defined). So hence how and why the WTA filter was created.

 

So I can see that maybe you are asking the question - the ideal filter rings and rings and rings for an infinite period of time, and it looks nothing like the original impulse, which is zero, a pulse, then zero. So how can one possibly say that this filter returns the original signal completely unchanged? After all everybody says ringing is bad and unnatural.

 

So how do we answer this paradox? Its easy - the original impulse is NOT a legal signal as it is not bandwidth limited. An impulse for CD has exactly the same level at 22.05 kHz as at DC - but sampling theory absolutely demands that the signal be bandwidth limited, and this means that at exactly 22.05 kHz and above the signal has exactly zero output - not the full output that a impulse supplies. So all this talk about ringing is fundamentally mistaken as you are basing a prejudice on a signal which a competent ADC would never be capable of supplying.

 

Here is the key idea - take an ideal impulse, then bandwidth limit it, so that for CD the level at 22.05 kHz and above is exactly zero, then use this as your test signal. What would happen here is that ideal filter, which rings and rings and rings with the illegal impulse, would return absolutely no difference from the un-sampled bandwidth limited impulse. And cruder filters that have short ringing will actually produce more ringing, and more changes to the original signal! So the idea that ringing of filters is bad is based on a false premise, and people simply not understanding the theory properly.

 

 

For me, there's some slipperiness to this description. He's essentially saying that

 

- "perfect" bandlimiting is easy to do (perfect is impossible, darned good is not all that difficult)

 

- The impulse signals everyone uses (Rob speaking) to test ringing aren't thus bandlimited

 

- Therefore everyone's just not understanding properly that their test signals are misleading them and filters that ring in response to illegal signals aren't bad.

 

 

I tend to think "everyone" isn't misunderstanding.

 

- First, though darned good bandlimiting is not difficult, in reality the creation of a lot of the music we all like to listen to did not have such good bandlimiting. For reproducing this music, filters that ring like bells in response to illegal signals may not be the best choice.

 

- Second, there are folks who think accurately reproducing the sounds of instruments requires reproducing a couple of things that *may* necessitate bandlimiting at higher than 22.05KHz. Not everyone thinks these things are necessary, but some people do: (1) reproducing ultrasonic harmonics so they can intermodulate with each other and with audible frequencies the same way they do live; (2) reproducing attack transients with rise times shorter than those of 22.05KHz sine waves.

[Sunya]

Recording in DXD allows 0 output at FS/2; then you can use iZotope to downsample which can offer over 250 dB of attenuation at FS/2.

[Jud]

Recording in DXD allows 0 output at FS/2; then you can use iZotope to downsample which can offer over 250 dB of attenuation at FS/2.

 

I'm assuming you have to downsample for your DAC. I would upsample to 768KHz and possibly sigma-delta modulate for mine.

 

 

Sent from my iPhone using Computer Audiophile

[Sunya]

Yes, downsample for a DAC that doesn't support the DXD rate. Using the DXD rate makes it easy to get a bandwidth limited signal out of the ADC.

 

The Chord's upcoming Davina ADC will support output rates up to 768kHz.

[VenturaRV]

Except Luxman replaced it with an oversampling DAC. I am skeptical the Luxman engineers knew less in 2013 when they designed the DA-06 than they did in 1987 when they designed the DA-07.

Luxman engineers designed, after the pioneering DA-07, other "Fluency" DACS, including one that offered the option to switch on the fly between Shannon and Fluency modes (its sampling rates and bitdepth went up to 196k/24, and had a DSD converter, too). It wasn't cheap, for sure.

Luxman stopped manufacturing such type of converters mostly because market reasons, not because improvement in technical knowledge (following your argumen, their engineers had more skills in 1987 than in years before, when they designed normal converters and CD players, so they never had change to a NOS, filterless model if that had not find it superior).

By the way, both Audio Note and Wadia continue making excellent NOS, filterless machines (at high prices, of course).

BTW.: Luxman models, filterless or not, including DSD decoding makes this last as any other DSD convertor.

 

 

VenturaRV

[Jud]

I had a "Fluency" DAC, partly self built. I liked it.

 

 

Sent from my iPhone using Computer Audiophile

[VenturaRV]

I had a "Fluency" DAC, partly self built. I liked it.

 

 

Sent from my iPhone using Computer Audiophile

Congratulations.

Wish I were able to do such a difficult task! [emoji28]

 

VenturaRV

[lmitche]

I had a "Fluency" DAC, partly self built. I liked it.

 

 

Sent from my iPhone using Computer Audiophile

Here is a nice write-up on a fluency DAC.

 

http://www.audiostream.com/content/sonore-simple-design-exd-dac#fha8IoK6sDyk50YK.97

[ted_b]

Yeah I had the Sonore/exD inhouse back in June of 2012 and asked Michael to do a write-up too. My stint was very brief but loved it.

http://www.computeraudiophile.com/f6-dac-digital-analog-conversion/new-sonore-digital-analogue-converter-direct-stream-digital-playback-12021/#post164509

[YashN]

Yash, I was just being an idiot! Here's the Altmann output again before the ADC:

And here are the expected and actual outputs after the ADC:

Don't say that, I think you're one of the most articulate and clever folks around here, and this is also having read your posts on other sites

 

I'm not sure if you're saying you now did these new measurements after calibrating the PRISM with REW though.

 

The idea is to perform this process first with the 'soundcard calibration' feature, and then save this file. Then, all subsequent measurements would use this file (you have to load it) to compensate some of the capture hardware's characteristics automatically by REW while capturing new gear. This will then provide a more accurate picture of the new gear captured, with less of the 'signature' of the capture gear.

 

Definitely check those Log-scale graphs too and let us know what you find in the 0-30Hz region.

 

Regards.

[YashN]

All we're seeing here is the ADC's 16 bit noise floor! I haven't done it, and have no inclination to do so, but if I were to switch the ADC to its 24 bit setting, I'm certain the expected and actual spectra would look identical.

 

Measurements can definitely be a lot of work. For instance, if you want to analyse this as precisely as possible, you would want to compare all the capture input interface possibilities (let's say Balanced vs Single-Ended), with various input attenuation positions, and soudcard calibration with REW before capturing.

 

And of course, you would want to check if any ground loop issues or Leakage Currents, PSUs can affect the measurements greatly and mitigate or reduce those.

 

Here's a recent capture with an M-Audio Fast Track Pro, itself connected to the iMac server with a normal USB cable, capturing (Audacity as well here) an Audacity-generated 1kHz Sine through the iDSD Nano, with the custom USB Connector and injection of cleaner power with my DIY Linear Regulated Power Supply. The LRP is rather clean, very little spurious signals on zoom if at all, but check the lower regions, it's quite astounding.

 

 

Of course, ideally here, I would want the M-Audio itself to be isolated with my custom USB Connection as well as injected with clean power (would require cloning or improving my Linear Reg PSU). I'd use it connected to a laptop if it actually had drivers working with the latest macOS (it doesn't, so that doesn't work), so I will just power it from a laptop battery but that still requires me building a second custom USB connector.

[audiventory]

Measurements can definitely be a lot of work. For instance, if you want to analyse this as precisely as possible, you would want to compare all the capture input interface possibilities (let's say Balanced vs Single-Ended), with various input attenuation positions, and soudcard calibration with REW before capturing.

 

For full picture need use sweep sine and learn time-spectrum diagramm in expanding different level ranges.

 

Otherwise we may miss distortions.

 

For interesting areas of the time-spectrum diagramm we can use 2-D spectrum.

[Miska]

By the way, both Audio Note and Wadia continue making excellent NOS, filterless machines (at high prices, of course).

 

Where does Wadia have NOS DAC's? All Wadia's I've seen have been OS models with their own digital filters...

[Miska]

Yes, downsample for a DAC that doesn't support the DXD rate. Using the DXD rate makes it easy to get a bandwidth limited signal out of the ADC.

 

The Chord's upcoming Davina ADC will support output rates up to 768kHz.

 

I'm using the RME ADI-2 Pro which has ADC and DAC capable of 768/32 and DSD256... Very nice device overall.

[Miska]

This, I don't understand at all. Surely the filtering needs to be before the digital to analogue conversion takes place to be effective. As audiventory points out, a non-oversampling DAC will be sending a whole bunch of ultrasonic grunge to the electronics downstream (its own output stage included).

 

Anyone want to chime in on this?

 

Yes, here's output of Metrum Musette with 0 - 22.05 kHz sweep. And it is not filter-less, it would be even worse if it wouldn't have any output filtering, although the filtering it has is quite minimal passive filter.

 

It cleans up quite a lot with upsampling to 384k. Although it doesn't have enough analog filter to remove remaining images at multiples of the 384k sampling rate.

 

 

For comparison, here's output of the same sweep from my DSC1 DAC running at DSD512:

[manisandher]

Miska, what effect would these ultrasonics have in practice? I mean, clearly, no one would be able to hear any of these ultrasonics directly. So, would it be how they might be interfering with the other electronics?

 

Mani.

[Miska]

Miska, what effect would these ultrasonics have in practice? I mean, clearly, no one would be able to hear any of these ultrasonics directly. So, would it be how they might be interfering with the other electronics?

 

You may get intermodulation back to audio range. Since those images are directly correlated with the source signal the intermodulation products are also directly correlated. This is emphasized for gear that has THD strongly increasing as function of frequency. For example with DSD/SDM and good modulator the high frequency noise, even if leaked, is decorrelated and intermodulation products sound like white noise.

 

(by the way, vinyl playback also tends to have strong amount of IMD)

 

Without any extra hardware, the Musette output IMD spectrum is cleaned up to some extent with upsampling to 8x rate.

 

Here's without upsampling at 44.1k:

 

And here's with upsampling to 384k:

 

 

As a side note, there's also notable difference in amount of jitter.

 

Without upsampling at 44.1k (Jtest24):

 

With upsampling to 352.8k (Jtest24):

[manisandher]

Without any extra hardware, the Musette output IMD spectrum is cleaned up to some extent with upsampling to 8x rate.

 

As a side note, there's also notable difference in amount of jitter.

 

Thanks! Really interesting...

 

Mani.

[Jud]

Yes, it's intermodulation that the "we can't hear ultrasonics anyway so our ears are filters" thinking doesn't account for.

 

Very interesting about vinyl. Wonder if ears trained to vinyl might hear some IMD as "right."

 

Of course not all IM with ultrasonics is distortion. Real instruments have ultrasonic harmonics and we hear the audible-frequency results of intermodulation those harmonics cause.

 

 

Sent from my iPhone using Computer Audiophile

[VenturaRV]

Where does Wadia have NOS DAC's? All Wadia's I've seen have been OS models with their own digital filters...

I wanted to say OLD Wadia "Audio Computers", as they called them.

If I am wrong about this particular brand, my apologies.

 

However, that doesn't change my statements about, at least, old (not too much) NOS Luxman models and Audio Note ones (these not old), and the fact that I have still to heard ANY machine (no matter its resolution and/or price) that can beat in SQ my all-redbook combo DP-07/DA-07. And I've heard a lot already.

 

I do really think there are a good amount of prejudices about the 16/44,1k format, one that suffered from both extended bad playback hardware and poor recorded material, massively selled (just go to Bob Katz considerations in YouTube about, among other issues, the famous "Loudness War").

Even better: if you can, just compare any standard edition of a redbook album with (if it's available) the Mo - Fi edition (redbook too). Night and day difference.

 

The full capabilities of CD standard rarely has been acomplished.

 

What leads me to point out a paradox: the same "audiophile" people that has been condemning 16/44,1k for decades as being the opossite to real good sound, now salutes it enthusiastically in the form of Tidal "Hi-Fi" streaming. Things to see (and hear)! [emoji41]

 

VenturaRV

[Miska]

I wanted to say OLD Wadia "Audio Computers", as they called them.

 

I'm not sure what is old, I remember only back to late 90's.

However, that doesn't change my statements about, at least, old (not too much) NOS Luxman models and Audio Note ones (these not old), and the fact that I have still to heard ANY machine (no matter its resolution and/or price) that can beat in SQ my all-redbook combo DP-07/DA-07. And I've heard a lot already.

 

I know about Audio Note. For Luxman I didn't quickly find info on DA-07. But DA-06 and the other current models are all normal oversampling DACs based on TI/BB DAC chips (mostly PCM1792).