Review and Measurements of Totaldac d1-six DAC

[amirm]

You think in your head like a standard measurement technician.
I think in my head like a R&D engineer (and optimiser by listening tests).
We won't agree, we don't make measurements with the same aim.

When you post something on your website, expectation is that you are following industry standards. If you are not, which in this case is the reality, then you need to explain the motivation and parameters that went into it. As it is, your measurement misled some of the members here, thinking it would show distortion products which it did not.

You see the consequences of posting measurements that are not representative of product performance. Someone like me shows up and you wind up having to explain things in a hurry. Best to get ahead of the train and provide proper explanation for your measurements. Again, I refer you to Benchmark manuals which extensively explain these things.

 

[amirm]

It is a key point, and there are other such key points in the Totaldac DAC making it difficult to measure, or leading to unexpected problems.

So you can't measure some aspects of your design? Exactly what aspect is that and how do you know you got it right if you can't measure it?

 

[KSTR]

Happy to hear that.

For example, I'm sure you implemented the zero-order-hold drop correction at HF correctly, based on the math. If one measures frequency response without a steep enough tracking band-pass or low-pass filter, the result will be an incorrect display of the HF response with too high levels for both the corrected and uncorrected case (from the image components), just what I think we see in Amir's measurement. But when measured with the Farina log-sweep plus convolution method which strictly sees the fundamental only, I assume it would be 100% correct.

 

[amirm]

Happy to hear that.
Amir discovered the strairs of a NOS DAC on a scope for the first time?

Many of us on the side of audio science have written at length trying to explain to audiophiles that digital audio does not have steps due to use of reconstruction filter. Then come vendors like you that validate their fear, potentially giving all digital audio a bad name, not just broken ones like yours.

So no, the problem is not me. It is you throwing out the book on digital audio principles.

 

[amirm]

If one measures frequency response without a steep enough tracking band-pass or low-pass filter, the result will be an incorrect display of the HF response with too high levels for both the corrected and uncorrected case (from the image components), just what I think we see in Amir's measurement.

No, the results are the opposite and the unit actually implements a filter to correct the droop:

The dark line is the standard output which as a drop in high frequencies, not a peak as you state.

The correction filter is in red but it overshoots, rather than provide a flat response.

I also don't know what test you are describing. The source here is NOT a wideband signal, Farina or otherwise. A digital generator produces one tone at a time, and measures what is produced in the DAC. That better be accurate or all of your audio will be wrong. There is NO bandpass filter involved in this test. You are confusing how open-loop room measurement systems work with how a test instrument works to generate a close-loop frequency response.

 

[totaldac]

Does your frequency response measurement takes the image frequency image into account? Oops you have forgotten that pont maybe.

NOS DACs are specific to measure...

 

[amirm]

Here again is the online manual:

Flat response is NOT achieved using the filter, making me wonder what measurement was ever used to verify this.

 

[totaldac]

Forget standard measurements NOS DACs (+ other specificities).

 

[totaldac]

NOS DACs are not your domain.

 

[luisma]

NOS DACs are not your domain.

If that's the case Brian, (I'm sorry I believe it is Vincent right?) prove him wrong, claims like that doesn't make your product or your reputation look good. I'm not a detractor of your DAC, it would help if you can provide some details on measurements, unless you don't care of course but you are here for a reason

 

[totaldac]

Here again is the online manual:

View attachment 30233

Flat response is NOT achieved using the filter, making me wonder what measurement was ever used to verify this.

You should know what a sin(x)/x curve should give at 20KHz with a 44.1KHz sampling frequency. This is not what you measure. Reconsider your process please.

 

[amirm]

Does your frequency response measurement takes the image frequency image into account?

Of course. Anything out of the DAC up to the limit of the one-half of the sampling rate is included in the meter in the analyzer. I thought you said you had an AP and know how it works? Now I have to explain the basics of frequency response measurements to you?

Remember, my measurements replicated your online manual that there is a treble droop. Why are you questioning it?

 

[totaldac]

unless you don't care of course but you are here for a reason

I don't care your measurements, I just don't want to let some of you say that it is a poor DAC design and with fake graphs on Totaldac web site.

 

[gvl]

Does your frequency response measurement takes the image frequency image into account? Oops you have forgotten that pont maybe.

NOS DACs are specific to measure...

All images lie above 20kHz, or Nyquist for that matter. Are you referring to their IMD products maybe? Looking at the awful IMD performance I can believe IMD products mess up the FR test. Is that supposed to be a good thing? Sigh.

 

[totaldac]

Of course. Anything out of the DAC up to the limit of the one-half of the sampling rate is included in the meter in the analyzer. I thought you said you had an AP and know how it works? Now I have to explain the basics of frequency response measurements to you?

Remember, my measurements replicated your online manual that there is a treble droop. Why are you questioning it?

Don't mix up the DAC under test sampling frequency and the analyser sampling frequency. An image frequency around 25KHz is taken by the analyser if it is set internally to 96KHz or 192KHz.

 

[FrantzM]

Care to

I don't care your measurements, I just don't want to let some of you say that it is a poor DAC design and with fake graphs on the web site.

Well.... Please come back with "real" measurements.. because based on those "fake" measurements, this is as you wrote it, a "poor DAC design". Note in passing. many DACs, when subjected to the same "fake" measurements, come out with flying colors ...
ad hominem won't cut it here. Come with proofs. solid ones.

 

[gvl]

Remember, my measurements replicated your online manual that there is a treble droop. Why are you questioning it?

You didn't measure enough of a droop for 44.1k sampling. I was trying to bring it to your attention several times. I think this is what Vincent is getting at.

 

[luisma]

I don't care your measurements, I just don't want to let some of you say that it is a poor DAC design and with fake graphs on the web site.

It is not "my" measurements and I'm not saying it is a poor DAC design

 

[totaldac]

Care to

Well.... Please come back with "real" measurements.. because based on those "fake" measurements, this is as you wrote it, a "poor DAC design". Note in passing. many DACs, when subjected to the same "fake" measurements, come out with flying colors ...
ad hominem won't cut it here. Come with proofs. solid ones.

I mean, fake graph on Totaldac web site, not on this web site.

 

[amirm]

You didn't measure enough of a droop for 44.1k sampling. I was trying to bring it to your attention several times. I think this is what Vincent is getting at.

The sampling rate is 44.1 kHz and the digital generator is programmed to go 22 kHz. I could go to 22.5 kHz if you want to be pedantic about it. Is this what you are asking?