I am not very interested in a prehistoric serial port, in fact I implement the USB port in my devices.
But if you don't know how an R2R DAC works we are wasting time.
Here you can find the schematic, it could help.
Implementing a true FIFO buffer with low phase noise clock on the Soekris DAM1021 DAC
Last edited:
DACs 20bit and over, running down to DC, pose very special problems... I would mention here the 1/f noise which needs to be at least 3 times lower than the DAC LSB (which is about 9uV for a +/-5V output) meaning that the overall (DAC + voltage reference + signal conditioning) low frequency noise needs to be under 3uVpp, not a trivial task.
Also, the temperature drift would be a big issue, everything in the gain and scaling circuitry would need the best bulk metal foil resistors with under 0.1ppm/C, and we all know how much they cost.
And I'm not even going into thermo electric effects, where even a copper/copper junction drifts 0.2ppm/C, while an copper/copper oxide interface can drift up to 1mV/C, which would be a true disaster. This means breathing around a 20bit DAC at DC would walk the measurements results all over the place.
Bottom line, testing audio DACs at DC is absurd.
Stop bitching and do your homework. You are chasing a ghost, and I believe it's part of your marketing plan for your new product(s).
Testing Sigma-Delta audio dacs at DC for maybe about 2 seconds at each DC offset is a standard way of checking for one type of noise modulation. A two second period is long enough to acquire useful FFT data. https://www.diyaudio.com/forums/dig...-reference-dac-8-channel-265.html#post6491643 ...There is some more discussion as the thread continues.
Last edited:
Nothing to do with what is discussed here and now. And to add insult to injury, nobody is able to explain or reproduce that fuzzy data, beyond being some ESS DNR marketing exercise.
I know you like to attack me everywhere, but you should be a little less shameless.
You accuse me of marketing (no one knows what) in a thread started for a commercial product, which I bought and therefore I am a customer.
As you have already done this several times in your Canadian friend's thread, where all the offered devices are commercial one.
You accuse me of not sharing my projects in the same thread where a schematic has never been published.
The matter is even more comical when you talk about solid engineering referring to someone who has designed and sell the reclocker of his reclocker (that's solid business).
You are measurements obsessed but you don't like mine.
So it still escapes me, what do you think you're getting?
Really?
And when they run in AC the very special problems you are referring magically disappear?
Please, explain how AC is generated in a R2R DAC.
Wrong question.
Otherwise, any lesson of DAC dynamics for you would be $200/hour billed in hour increments. Proceedings will be donated to a charity of my choice.
I don't like attacking you, but like so many others around, I don't like you. You are stepping on everybody's nerves and you are collecting the net results.
OK, this means you don't know.
You have run out of miracles.
The last was that of the 100MHz oscillator which then magically became 10MHz.
Last edited:
So do you. You don't, because you very well know there is no negative advertising.
You're wrong again, I could never put you in my ignore list because I go crazy for your comments, especially when you talk about solid engineering.
BTW measurements are not negative advertising.
Last edited:
Well, 14 bit audio has the problem of quantization noise.
The DAM1021 on the other hand only produces harmonic (correlated) distortion. This is something much more benign than actual "14 bit performance".
I wouldn't worry too much about distortion. My DAM1021s sound a lot better than every delta sigma DAC I've heard despite higher harmonic distortion at full scale. It's not something the human ear is especially sensitive to.
The DAM1021 on the other hand only produces harmonic (correlated) distortion. This is something much more benign than actual "14 bit performance".
I wouldn't worry too much about distortion. My DAM1021s sound a lot better than every delta sigma DAC I've heard despite higher harmonic distortion at full scale. It's not something the human ear is especially sensitive to.
Although in my taste the Naim CD3 with TDA1541A sounds much better than my DAM1021 (but it's just my subjective opinion) here I have merely measured the accuracy of the DAC .
It is very curious that this simple measurement has sparked the indignation of some members, especially those obsessed with measurements.
Moreover the measurement has been done with the purpose of getting better accuracy using DAC digital calibration.
It is even more curious that the most obsessed with measurements accuse me of doing this for marketing purposes as negative advertising, while the measurement has simply confirmed what I claimed 6 years ago, which was even quite obvious from a mathematical point of view
Reference DAC Module - Discrete R-2R Sign Magnitude 24 bit 384 Khz
Edit: 7 years ago
Reference DAC Module - Discrete R-2R Sign Magnitude 24 bit 384 Khz
Finally, I found a possible problem of asymmetry between the positive and negative parts of the ladder which IMHO should be investigated, but it looks like nobody was worried/interested.
Obviously it is me who did not understand what the spirit of DIY audio is.
Last edited:
I wonder if this asymmetry would go away with changes to the PSU decoupling. My DAM1021s sound much better with a couple additional caps..
I don't believe because the asymmetry is not an analog problem, it's a digital issue.
With the same magnitude and opposite sign different switches are on.
Have a vague recollection of a dac problem measured by KSTR in one of the old 'Blowtorch' threads. Seems like a dac was off in the negative direction by one LSB. IIRC, Daniel (aka DPh?) pointed out that there is no 'negative zero' in two's compliment (or to that effect), and that somehow accounted for the one LSB offset in one output polarity. Unfortunately though its been awhile and I don't have a link to the discussion.
Could this be an artifact of the filter then? Have you tried different filters?
True, there is no negative zero in two's compliment, so when the word is negative (MSB=1) the magnitude has to be corrected by one LSB at ladder level (1111 1111 1111 1111 1111 1111 which means -1 in two's compliment becomes 1111 1111 1111 1111 1111 1110 in sign magnitude architecture DAC).
But in case of the DAM1021 the difference is far greater than one LSB.