Review and Measurements of Totaldac d1-six DAC

[simbloke]

This you cannot get with the OS delta-sigma...

Because those steep rise times mean components throughout the frequency spectrum. You can't listen to square waves. An audio system is not supposed to be able to reproduce square waves!

 

[zalive]

This neither.

It's actually almost irrelevant. Staircase part of the waveform is above Nyquist, so inaudible. Ears filter this out.
Higher IM and that's pretty much it.
What I don't understand is why designer implemented FIR instead of making anti-sinc (treble roll off compensation) in the analog domain.
He got this by using FIR:

He can tell as much as he wants it's not much ringing, this is still a lot of pre and post ringing. I suspect he'd get less with the analog anti-sinc.
NOS design principle is to avoid processing, yet he chose to process the treble...unlogical choice which doesn't go along with what should be the core principles when designing a filterless NOS R2R.

 

[Pluto]

ABX is hard and stressful

I have participated in several well-conducted ABX-style tests. They are neither hard nor stressful. I suspect they only become stressful when you are so desirous of a particular outcome that you perceive yourself to be under pressure to “achieve” the result that, for one reason or another, matters so much to you.

In professional circles, ABX is widely regarded as being the gold standard of test methodologies when you need to compare a new process or method with an established approach. I really do not see why audio amateurs bear such a grudge to this reliable and proven method.

Actually I do see why, and it's rather sad.

 

[Deleted member 2348]

This […]

… is an excellent example for an "illegal" signal that makes NOS without reconstruction filter shine. Unfortunately this kind of signal cannot be created by properly (Shannon, Nyquist, bandwidth limited to less than fs/2, …) sampling an analog square wave signal.

 

[daftcombo]

It's actually almost irrelevant. Staircase part of the waveform is above Nyquist, so inaudible. Ears filter this out.
Higher IM and that's pretty much it.
What I don't understand is why designer implemented FIR instead of making anti-sinc (treble roll off compensation) in the analog domain.
He got this by using FIR:
View attachment 30457
He can tell as much as he wants it's not much ringing, this is still a lot of pre and post ringing. I suspect he'd get less with the analog anti-sinc.
NOS design principle is to avoid processing, yet he chose to process the treble...unlogical choice which doesn't go along with what should be the core principles when designing a filterless NOS R2R.

I agree with you. Putting a FIR in a NOS DAC smacks of amateurism.

 

[TomatoTBone]

Last posts on HCFR by Vincent from Totaldac, about Amir's measurements :
"il faut juste ne pas laisser de boucle de masse ni poser l'alim trop près du DAC"
> you have to not let ground loop and to not put the power supply too close to the DAC.
"Non il n'a pas rectifié son erreur, il laisse une bosse sur la réponse en fréquence à 20kHz. [...]"
> no he [Amir] did not correct his mistake, he leaves a bump on the frequency response at 20kHz.

I think he'll never admit his DAC measures poorly.

 

[VintageFlanker]

Last posts on HCFR by Vincent from Totaldac, about Amir's measurements :
"il faut juste ne pas laisser de boucle de masse ni poser l'alim trop près du DAC"
> you have to not let ground loop and to not put the power supply too close to the DAC.
"Non il n'a pas rectifié son erreur, il laisse une bosse sur la réponse en fréquence à 20kHz. [...]"
> no he [Amir] did not correct his mistake, he leaves a bump on the frequency response at 20kHz.

I think he'll never admit his DAC measures poorly.

He'll never admit because, regarding ALL his answers : He doesn't seem to have any clue what is going on in his DACs... He is a marketing guy, period.

 

[THW]

“i’ll keep repeating the same points, that’ll surely show him”

He’s becoming a meme at this point

 

[sweetsounds]

1) It's called quantization noise for a reason.
We see, the infinite train of harmonics only appears with a very special condition and the level of the spikes scale inversely with level, 0dBFS giving the lowest levels.
But mind you, the DAC was fed with a 24bit dithered signal, not undithered 16bit.

[...] This means the ADC's filter actually becomes the DACs missing anti-imaging filter when using the same or very similar sample rates.

You nailed it.
I guess, the R2R ladder has only 16 effective Bits (already great, that you can get resistor matching to that level.
It also has probably several ladders in parallel to push it a little more.

I checked the 6moons review: it has 24bit, but uses a mixing of the top 14bits and low 14bits separately. If measurements are confirmed, that seems to yield around 16Bit of noise level (if the measurement setup is correct).

Amazing, how much effort is spent to save this old concept. It's like pimping a 1950 Chevy.

But I now understand the graph by Briant on his webpage: a pure 14 Bit ladder DAC would never resolve such a low level signal. Because the DAC switches to the lower 14 Bits there is still a clean signal.

It would be interesting to really verify a -60dB 1kHz signal, once the measurement setup works.

The DAC should perform better with higher data rate, as this pushes out HF noise.

 

[zalive]

I agree with you. Putting a FIR in a NOS DAC smacks of amateurism.

I'll try to clear it up inside of me so i can come out with a clear output as well
I don't think it's amateurism. Possibly it's lack of understanding fully the philosophy behind the idea of NOS with no or limited filtering.
What's here problematic, in my view, is the price.
The justification for the price is not least the cost of parts used inside or manufacturing.
The justification for this can only be R&D cost. But this cost is high solely because of low scale of sold DACs.
And of course there will not be many sold since the price is obstacle for this. So it's a circle.
I don't like and I don't support this principle, because it changes who are the end customers.
Hifi is luxury, but as such, even above the entry level segment it's still affordable to a large number of hifi consumers.
Price of 13.000 euros (especially or a design which is clearly not a state of the art, unique topology and costly parts inside which might justify the price) is above a realistic limit of any mass segment of audio customers.
When you can buy something not only similar, but probably better in any aspect (not just measuring) like Soekris, something is wrong.
So even if R2R NOS is your thing and you do think there's a signature sound and like it, you can have it for much less.
(not to mention what I use which is based on a finished board which retails at ebay for less than 70 USD but it's basically similar or same concept, aside of being based on a chip instead of a resistor ladder)

So though I feel the initial measurements were (either unintentionally or intentionally) picked to show this DAC in the worst light possible, and that the reputation damage being done is irrepairable, and that this wasn't fair to the manufacturer, I'm feeling neutral for two facts; first one is that I can't support prices like this for a product like this (not about engineering but rather about too big a difference between manufacturing cost and the retail price); and the second one, I don't think manufacturer's usual end customers will be bothered too much with measurements anyway. So no reason this will kick the manufacturer out of the game even with a current product. However, his customers are rich people, not just plain audiophiles, and there's always a status element present there. But it has little with the world of audiophiles looking for performance/price ratio. Those would not likely pay the price even if these measurements never existed.

I can defend a design concept, though. I've no problem with the NOS R2R concept

 

[KSTR]

This you cannot get with the OS delta-sigma...

Of course you can. An OS delta-sigma DAC chip like the AK4490 will put out exactly this when it's digital filter is set to NOS ;-)

 

[jparvio]

Of course you can. An OS delta-sigma DAC chip like the AK4490 will put out exactly this when it's digital filter is set to NOS ;-)

Oh, the dac that has it’s own sound. I used to have Teac’s 503 dac. It was ”nice” sounding dac just as they wanted it to be.

 

[jparvio]

https://www.google.fi/amp/s/www.abs...akm-special-round-table-discussion?format=amp

 

[graz_lag]

He'll never admit because, regarding ALL his answers : He doesn't seem to have any clue what is going on in his DACs... He is a marketing guy, period.

Just marketing guy or rather Used Car Salesman?

 

[zalive]

Of course you can. An OS delta-sigma DAC chip like the AK4490 will put out exactly this when it's digital filter is set to NOS ;-)

Any screenshots to support this?
So far I find only this at Archimago's blog, the impulse response in NOS mode (TEAC UD-501):

There is some pre and post ringing present, much less than with any output filter on, but still present.
Probably a result of delta-sigma sampling digital filters which are inevitable in delta-sigma.

 

[jparvio]

https://www.google.fi/amp/s/www.absolutesound.com.sg/blog-2/2017/9/7/why-did-teac-use-the-ak4490-dac-in-the-ud-503-teac-x-akm-special-round-table-discussion?format=amp

I wonder if the ABX’d?

Kato: In addition, we had AKM engineers come to this listening room many times, and we listened to the sound of the product under development. We then asked them to make changes to meet our needs, and AKM took our requests into consideration in the making of the final product. This is what we do day to day in our technology exchange.

 

[KSTR]

Any screenshots to support this?
So far I find only this at Archimago's blog, the impulse response in NOS mode (TEAC UD-501):

There is some pre and post ringing present, much less than with any output filter on, but still present.
Probably a result of delta-sigma sampling digital filters which are inevitable in delta-sigma.

Will post a o'scope screen shot when I'm at home, later. You could as well look up the scope screenshot in the RME's manual (page 77). No preringing, but a bit of postringing from the analog anti-imaging filters (on the chip and in the RME analog section) at extremely high frequencies (== irrelevant).

The pic you show is a waveform view from Adobe Audition (recorded with the ADC of the RME Adi-2 Pro), which displays waveforms with a linear phase reconstruction filter just like it would look like coming out of a normal lin-phase FIR-filtered DAC or being sampled by a similar ADC (the ADC anti-aliasing filter in the RME has linphase and minphase settings). This is NOT representative of the analog waveform.

 

[MC_RME]

Any screenshots to support this?
So far I find only this at Archimago's blog, the impulse response in NOS mode (TEAC UD-501):

There is some pre and post ringing present, much less than with any output filter on, but still present.
Probably a result of delta-sigma sampling digital filters which are inevitable in delta-sigma.

IMHO the pre and post ringing is caused by an internal upsampling option (SRC in FPGA) that is never totally bypassed, even if deactivated. The ADI's don't show any in NOS mode.

 

[MC_RME]

Will post a o'scope screen shot when I'm at home, later. You could as well look up the scope screenshot in the RME's manual (page 77). No preringing, but a bit of postringing from the analog anti-imaging filters (on the chip and in the RME analog section) at extremely high frequencies (== irrelevant).

The pic you show is a waveform view from Adobe Audition (recorded with the ADC of the RME Adi-2 Pro), which displays waveforms with a linear phase reconstruction filter just like it would look like coming out of a normal lin-phase FIR-filtered DAC or being sampled by a similar ADC (the ADC anti-aliasing filter in the RME has linphase and minphase settings). This is NOT representative of the analog waveform.

Unfortunately in this case it is. I have seen this strange behaviour myself on a 503.

 

[zalive]

IMHO the pre and post ringing is caused by an internal upsampling option (SRC in FPGA) that is never totally bypassed, even if deactivated. The ADI's don't show any in NOS mode.

Is there an audible benefit at ADI's, say, NOS filterless mode vs FIR sharp linear? And what about the other filter options?
Btw. do ADIs implement anti-sinc roll off compensation respective for NOS or filters which might need it?