Review and Measurements of Totaldac d1-six DAC

[jparvio]

Do you know what are 13.500 € ? Six months of salary for an average Frenchman?

I am sorry but I fail to see the relevance here - at least to anything I´ve said.

 

[AndrovichIV]

Hmmmm. http://archimago.blogspot.com/2019/02/musings-measurements-on-why-2496.html

There's an extra step of DA in this experiment compared to Archimago, which makes this experiment invalid to conclude anything useful

 

[daftcombo]

I read it as 88kHz Input with 96kHz sampling, so would be again too close.
Graph 3 is 96kHz sampling and FIR-on at 44.1kHz Input.

Between graph 1, 2 only the sampling frequency changed and the height of side-bands (increasing with higher sampling rate) and noise changes (lower with higher sampling rate). So the result is measurement dependend. Doesn't say the sampling frequency of the first graphs, amirm's default last time was 48kHz sampling.

I hear you. What would you suggest? Sampling rate of 96 kHz, 44.1 kHz input, 20 Hz-20 kHz band, FIR-off?

We need to know the truth once and for all.

 

[Nightlord]

Well, it may be just me... but the shape of the DAC itself was enough to put me off... someone with a technical mindset would not build something like that...

 

[daftcombo]

I am sorry but I fail to see the relevance here - at least to anything I´ve said.

Sorry, maybe there wasn't. I thought about the money, all of a sudden.

 

[levimax]

There's an extra step of DA in this experiment compared to Archimago, which makes this experiment invalid to conclude anything useful

Does one AD conversion make an experiment "invalid" if the AD conversion can be shown to be "transparent" or very close to it and if any potential variation can be quantified to be significantly less than what is being recorded /measured? Is "bit-perfect" really a requirement to "prove" something? My understand is there is some "experimental error" in every experiment .... but I am not a scientist so would welcome being educated.

 

[jparvio]

people buy expensive DAC's that measure badly because they believe the PR of the company behind them, and believe that NOS is better, etc.
Also people fall in love with how the product looks. The reality is that the vast majority of people who own this DAC would be unable to differentiate it from a 70$ Behringer unit in a matched, ABX test.

So glad You found a connection - this really closes the book. So by following this analogue we must stop PR departments. Surely this is then a more profound problem and also concerns cars, food etc. One must be careful not to step into all the traps those clever bastards design. I´m glad I managed for Years without a telly. All them advertisements must mess them little brains.

Many here have already claimed that totaldac, the dac of this topic is uuuuugly as hell with it´s ****** remote and poor Renault-quality metal case. Not to mention the plasticky see-through front mask. Yikes. So looks cannot be the reason, for buying one, right?

Behringer is not a ****** product at all. For Years (5 to be exact) I tolerated 24/96 as my Sanders 10c speakers crossover. Funny thing is, at the same time I enjoyed my hefty vinyl stack, even if the signal had to go through ad/da. All the lovely snaps, cracles and pops were as crisp as ever before or after. It let also all the differences through. I mean the (not ABX´d) changes of riaa, dac etc. Yup, or so I lie to myself. Well, the noise level was at times a little disappointingly harsh, but music signal nicely masked it anyway.

Well nowadays I use mostly tube equipment from top to bottom. I guess not for the fidelity´s sake but for my own euphoria. There it is. Works like a charm, over 30 tubes glow like candles. OTL´s just have it all. You know, music flows like it should. I am sure they would not measure even as well as Amirs not-so-well measuring high end transistor amps.

So I am a lost soul here and in this context You can judge me and my opinions. But note that I do not defend bad measurements or ****** products. I just promote the freedom of choice and this is the wrong place for it.

 

[Haskil]

I see that some member like to play.

Someone attracted my attention on totaldac power supplies.

View attachment 30406

It's not a power supply, but a toroidal transformer...

 

[gvl]

44 pages and no end in sight... Folks, let's save some gun powder for the upcoming Airist R2R review.

 

[jparvio]

Please drop the ridiculous strawman arguments - of course "measurements do not explain it all" - as has been admitted over and over and over again - even by members here who are not only deeply involved in science, but also heavily immersed in the arts, especially music.

Psychology and psychoacoustics demonstrate quite nicely - as has been confirmed independently over and over and over again - that many subjectively determined claims of hearing sonic differences - quite often disappear completely if the listener is not aware of the identity of which piece of two pieces of hardware that he is listening to in a calibrated audio system.

Why do you refuse to verify (prove?) your claims? What are you afraid of?

(No need to answer with a flailing of irrational excuses, because there are many here, often with only a rudimentary knowledge of psychology and psychoacoustics, who are aware of the extraordinary powers of the human ego to suppress information that challenges their beliefs - and you demonstrate repeatedly here that your beliefs are very, very strong, and easily suppress any tendency towards logic and rationality.)

Quite often disappear completely? Like quite often people here do ABX and quite often there are no differences?

I rest my case.

 

[sweetsounds]

The interest/fun would be, as it was in archimago's test 1) is the difference audible 2) which one is preferred.

That test was really eye-opening: while I apparently could identify the best measuring device by ear, I still think I could have been just lucky at my n=1 personal level, since the differences were so small. In casual or standard listening scenarios, I am 99% sure I wouldn't have noticed a silent switch from the cheap motherboard to the reference device.

I really want to underline your post that archimago's test was also eye-opening to me. Electronic influences are very, very small.

It is a blind pair comparison. No pressure to be right on the answer, just say, what you like more. The closest you can get to field trials in audio.
And yes, it could mean that people prefer slightly imperfect.

The other test, which is quite good is "same/different". ABX is hard and stressful.

I'd recommend to also train listeners, like you train wine drinkers.
https://www.pooraudiophile.com/2014/08/blind-audio-testing.html

 

[pkane]

The skeptics will say that you ruined the sound of R2R by recording them with a normal ADC. And then playing them with ordinary DACs. So who are we trying to convince with such a test?

A recording using a fairly transparent ADC of the same track would enable not just A/B listening comparisons but also further analysis using software, such as DeltaWave or Audio DiffMaker among others. Gearslutz has a long-running thread where such recordings are posted from various DACs and ADCs and it’s very useful for comparisons. It’s amazing just how much information can be derived from comparing the original digital file to such a DA/AD encoded one.

I know it’s quite a bit of extra work for you, but it would be really awesome if you could make such recordings with all the DUTs that pass through your hands and share them with the community. Pretty please?

Think of it as an audio DNA recording and preservation project

 

[sweetsounds]

I hear you. What would you suggest? Sampling rate of 96 kHz, 44.1 kHz input, 20 Hz-20 kHz band, FIR-off?

We need to know the truth once and for all.

I am not a signal expert, but what sounds Logical:

Disconnect earth from signal ground (http://www.totaldac.com/use_note.htm)
FIR off. It was described before, that a NOS design will have a 3dB roll-off at 20kHz, accept it!
Be prepared to measure quantization noise, if you don't dither the last bit.

Bandwidth 20kHz (and if you like to see the two aliases of the missing reconstruction filters, try 50kHz. They will be there as on the initial review, not sure if the "infininte harmonics" will still be there.)

I think, the AP 2522 can't go higher than 192kHz.

Test A: 44.1kHz of input, 192kHz of sampling. 1kHz tone and intermodulation spectrum e.g. 1kHz + 10kHz to see the side-bands.
Test B: 96kHz of input, 192kHz of sampling. 1kHz tone and intermodulation spectrum 1 kHz + 10kHz.

This should give you the frequency domain behavior.
Reminder: if you believe, that CD (16Bit) sounds the same as HiRes music, than in general you should ignore noise below -90dB.

I expect
- noise level around -130dB, (Topping D50S: -150dB)
- jitter (the widening of the peak) at -105dB (D50S: -140dB),
- power supply noise spuriae at -110dB (D50S: -120dB).
- quantization noise at 97dB above 22kHz [= 16Bit * 6dB + 4.8db - crest, crest-factor is 3dB for Sinus, 12dB for music]

I might be wrong, but looking at the measurement data ...
Amirm's "infinite harmonics" might be simply the quantization noise of an undithered 16bit Sinus @ 97dB. No surprise, not a design error!
Amirm's "very high ultrasonics" nothing else than the aliases of a 44.1kHz signal without a reconstruction/HF filter. Not a design error per se!
Amirm's "ugly noise floor" is nothing else than a sampling artefact.

No idea on the audio side bands in his newer graphs. Let's see, if that is an issue.

So if above values will be observed, this is a perfectly designed NOS R2R DAC, folks! This is what is on a CD in terms of data.
And no, I don't think it is smart to design DACs without oversampling and reconstruction filters and rather live with quantization noise.

The other DACs aren't better (but smarter), they use filters (oversampling, reconstruction filters) to smoothen the result.
This looks better in measurements, but is a fitted result.

Again: I don't think it is smart to design a DAC like this! But to get this linearity by hand-picked Vishay resistors in a ladder is amazing craftsmanship. Why doing this in the age of sigma-delta? I don't know.

And again apologizing in advance, I am not a signal expert and might be wrong.

 

[AndrovichIV]

So glad You found a connection - this really closes the book. So by following this analogue we must stop PR departments. Surely this is then a more profound problem and also concerns cars, food etc. One must be careful not to step into all the traps those clever bastards design. I´m glad I managed for Years without a telly. All them advertisements must mess them little brains.

Many here have already claimed that totaldac, the dac of this topic is uuuuugly as hell with it´s ****** remote and poor Renault-quality metal case. Not to mention the plasticky see-through front mask. Yikes. So looks cannot be the reason, for buying one, right?

Behringer is not a ****** product at all. For Years (5 to be exact) I tolerated 24/96 as my Sanders 10c speakers crossover. Funny thing is, at the same time I enjoyed my hefty vinyl stack, even if the signal had to go through ad/da. All the lovely snaps, cracles and pops were as crisp as ever before or after. It let also all the differences through. I mean the (not ABX´d) changes of riaa, dac etc. Yup, or so I lie to myself. Well, the noise level was at times a little disappointingly harsh, but music signal nicely masked it anyway.

Well nowadays I use mostly tube equipment from top to bottom. I guess not for the fidelity´s sake but for my own euphoria. There it is. Works like a charm, over 30 tubes glow like candles. OTL´s just have it all. You know, music flows like it should. I am sure they would not measure even as well as Amirs not-so-well measuring high end transistor amps.

So I am a lost soul here and in this context You can judge me and my opinions. But note that I do not defend bad measurements or ****** products. I just promote the freedom of choice and this is the wrong place for it.

Well I don't think people are going to go after you for using tubes. I think the people who get criticized here are the people who claim that tubes are closer to the source than SS. Or the people who claim that a worse measuring DAC in general is closer to the source than one that has better measures. I think even Amirm has a tube amp

 

[AndrovichIV]

Does one AD conversion make an experiment "invalid" if the AD conversion can be shown to be "transparent" or very close to it and if any potential variation can be quantified to be significantly less than what is being recorded /measured? Is "bit-perfect" really a requirement to "prove" something? My understand is there is some "experimental error" in every experiment .... but I am not a scientist so would welcome being educated.

I would agree with you. But the people who you want to convince would claim that it's invalidated because of this step.

The way I see it, is that since even the worst measuring DACs have errors that are very small (compared to say loudspeakers), introducing a tiny error through AD and then a second conversion to digital by the DAC that you're using might decrease the signal to noise ratio of your experiment significantly

 

[Blumlein 88]

I am not a signal expert, but what sounds Logical:

Disconnect earth from signal ground (http://www.totaldac.com/use_note.htm)
FIR off. It was described before, that a NOS design will have a 3dB roll-off at 20kHz, accept it!
Be prepared to measure quantization noise, if you don't dither the last bit.

Bandwidth 20kHz (and if you like to see the two aliases of the missing reconstruction filters, try 50kHz. They will be there as on the initial review, not sure if the "infininte harmonics" will still be there.)

I think, the AP 2522 can't go higher than 192kHz.

Test A: 44.1kHz of input, 192kHz of sampling. 1kHz tone and intermodulation spectrum e.g. 1kHz + 10kHz to see the side-bands.
Test B: 96kHz of input, 192kHz of sampling. 1kHz tone and intermodulation spectrum 1 kHz + 10kHz.

This should give you the frequency domain behavior.
Reminder: if you believe, that CD (16Bit) sounds the same as HiRes music, than in general you should ignore noise below -90dB.

I expect
- noise level around -130dB, (Topping D50S: -150dB)
- jitter (the widening of the peak) at -105dB (D50S: -140dB),
- power supply noise spuriae at -110dB (D50S: -120dB).
- quantization noise at 97dB above 22kHz [= 16Bit * 6dB + 4.8db - crest, crest-factor is 3dB for Sinus, 12dB for music]

I might be wrong, but looking at the measurement data ...
Amirm's "infinite harmonics" might be simply the quantization noise of an undithered 16bit Sinus @ 97dB. No surprise, not a design error!
Amirm's "very high ultrasonics" nothing else than the aliases of a 44.1kHz signal without a reconstruction/HF filter. Not a design error per se!
Amirm's "ugly noise floor" is nothing else than a sampling artefact.

No idea on the audio side bands in his newer graphs. Let's see, if that is an issue.

So if above values will be observed, this is a perfectly designed NOS R2R DAC, folks! This is what is on a CD in terms of data.
And no, I don't think it is smart to design DACs without oversampling and reconstruction filters and rather live with quantization noise.

The other DACs aren't better (but smarter), they use filters (oversampling, reconstruction filters) to smoothen the result.
This looks better in measurements, but is a fitted result.

Again: I don't think it is smart to design a DAC like this! But to get this linearity by hand-picked Vishay resistors in a ladder is amazing craftsmanship. Why doing this in the age of sigma-delta? I don't know.

And again apologizing in advance, I am not a signal expert and might be wrong.

Seems like just buying a Soekris is a better solution if you dig this type design. I've been ignoring this trainwreck of a thread.

Design a broken DAC. Declare its brokedness is not a design error. Ask for special testing parameters that don't make it look so broke. I mean for $13 large why not a little leeway.

 

[sweetsounds]

Design a broken DAC. Declare its brokedness is not a design error. Ask for special testing parameters that don't make it look so broke. I mean for $13 large why not a little leeway.

I tend to agree.

Well, let's look at what the TotalDAC does: you send him a 44kHz bit-stream of 16Bit and it outputs an analog step function with 44kHz and 16Bits (see amrim's measurement), exactly as demanded. My 12 year old daughter would probably draw the same curve when I give her the data.

The other DACs guess, that "he probably wants me to output a 1 KHz sinus" and output a smooth analog 1kHz sinus.
My 16 year old wouldn't bother anymore to plot the data but just draw a curve.

So in some sense the non-filtering DAC is the most truthful to the digital data, the other is trying to guess what the original analog signal was by making assumptions.

There is one argument for not putting filters, namely the introduction of phase errors and voltage ripples/ringing in the time domain by these filters. Blind tests and calculations suggest, that these phase errors/ringing are not audible. These errors can't be seen in the frequency plots used here.

To see the benefit of the NOS, amirm would need to do time domain measurements.

 

[THW]

forgive me but I really don’t understand the obsession behind “preserving the original signals”.

if the output waveform is very close to the original waveform then why does it even matter?

 

[amirm]

So in some sense the non-filtering DAC is the most truthful to the digital data, the other is trying to guess what the original analog signal was by making assumptions.

There is no "guess." The filter has no intelligence. The system with a filter is following the sampling theorem that says if you band limit input and output, you get your original signal. No guesses involved.

A stepped output has infinite high-frequency components that were not in the original data. Reproducing it such is wrong.

 

[gvl]

To see the benefit of the NOS, amirm would need to do time domain measurements.

He did some. It wasn't pretty. Especially the sine waveform.