DAC gallery

[Markw4]

Thorp,
Interesting project. I don't know if its the best way to go in the end, but dacs of the type are are building can sometimes even sound a little more dynamic than the recording really is. On the other hand, SD dacs can sound bit flat in terms of perceived dynamics. There are ways to compensate both types of dacs for perceptually more natural sound. The question for me which type can produce a better overall sounding result with compensation/correction in place. Ultimately, I suspect a compensated SD dac will do better. Time will tell I guess.

[TNT]

How does one know a DAC to "sound a little more dynamic than the recording really is"... whats your reference?

//

[Evenharmonics]

Originally Posted by TNT

whats your reference?

That's a good question because I am not aware of any peer reviewed published research on that subject.

[dddac]

Thorp, not sure if you have the equipment to do a linearity test up to 24 bits, but if you want I can run that for you on my Audio Precision equipment.

for example, below is the linearity of the DDDAC 1794 - as you will see it is almost impossible to reach true 24 bit linearity

This was actually a test I ran as some one mentioned the Vcom decoupling on the PCM1794 was not necessary. I proved him wrong as you can see ...

Of course R2R dacs with non-laser trimmed resistors for LSBs and wanting to do 24 bit is quit a challenge, so checking this is a MUST so you know if you reached your design goals

any way up to you !

[Thorp]

Thank you @dddac for your feedback.
Now I see your point: of course it is not a trully 24 bit dac! When I named that, I reffered to the "size" of R-2R ladder.
This is a learning project, actually my goal is to (start) learn about FPGAs & CPLDs. I don't even have a modern RPi (but it will arrive today, the courier just called me )
It is just a hobby and, yes, you are right, I don't have any expensive equipment, but I wish to!


On the other side, I am on your side: I also want to reach the best possible version of something I work to. Even if I believe that no one can hear any difference between "with 47uF" and "without 47uF", I still woluld do as you did! Perception can be subjective, but the numbers (math) don't lie, actually they are our common language.

[dddac]

Indeed Thorp, doing what you did is a great learning process and I am very interested in your results. My offer stands, if you at any point feel it would be nice to have a Linearity, THD, Noise, Frequency transfer - measurement - straight from digital source to analog output, let me know

[NickKUK]

Originally Posted by Thorp

Thank you. Yes, I have planed to make the master version (RPi slave) of this DAC. I will use the same PCB, with some modification for signal isolator.
It is funny, because I thought initially that I have to reverse it 180 degrees only, but after a ordered PCB, I realised that even if data pins are simetrical, VCC and GND pins of the IC are not!!!

Next-next step is FPGA based project, to deal also with DSD(dop) signals from RPi

Not yet. Actually both (left + right) 5-bit R-2R ladders for INL calibration are inactive at this moment. I will use variable voltage on opamp non-inverting input to deal with INL. See please the picture below.

It depends on the clock rate on the output side but FPGA/CPLD clocks may give you problems. A lot of the R2R dacs use a shift register, load serially, then use the clock against the SR to enable the pulse for output. That way all the pulses are better aligned. Looking at the traces on the PCB, they're of varying lengths too which may cause problems are higher rates.

Btw you can get digital potentiometers that can be programmed via SPI etc, if you have a calibration step manually then use a digital to fine tune where human fingers can't.

Love the DAC name btw

[Thorp]

Hi, NickKUK, thank you for your feedback.
I agree, but unfortunately, it is hard to draw so many matched length wires on an RPi size PCB. Some FPGAs are already on the way: EP4CE22 in 16x16, 1 mm pitch balls - that means 17x17 mm sq. footprint, and EP4CE40 in 22x22 , 0.8 mm pitch balls - that means 19x19 mm sq. footprint. They are huge for such as small PCB size! More than that, I want to use them in combination with AK4137 - these ones already arrived -, so, it is hard to believe that I will be able to put all of them on the same RPi size PCB.
At least because of the isolators, some hi-q IC flip-flops are mandatory, I am still looking for them. Anyway, at this moment, this project is in my mind only.


I have finally managed today to compile my first functional kernel modules for Volumio! I've been working on this for a few weeks and I had during this time some moments when I wondered how Linux-based distributions actually work! But, I am happy now and I really believe that Linux is a fascinating world.