High-end DAC designs?

[sidd0rs]

I'm interested in the experts' opinion on what makes a good DAC circuit. I understand this is a very open question, but I'm happy for open answers. I would appreciate it if someone could list the factors which make a particular DAC circuit better than another (assuming both use the same DAC chip)?

Are there any current examples of real high-end DAC design that I could examine?

Is there any reading material available for someone wishing to study and begin experimenting with DAC design?

My sincerest apologies if any of this has been asked before.

Cheers.

[psycoman]

You can't find any different in the digital circuit. They are all refer to the datasheet. Only the analog section you can design by yourself, you can use opamp, tube or transformer. All the information you can find in http://www.ti.com/audio

If you like traditional Multibit DAC, you can use SRC4392+DF1706+4*PCM1704.

If you like 1bit dac, you can use SRC4392+(PCM1792,PCM1794,DSD1794,...)


If you don't want to do it from the beginning. You may buy some kits such as http://www.kandkaudio.com/digitalaudio.html or the DAC from China. They are good quality and low price.

[poynton]

there are so many to choose from.........

try:-
Building the ultimate NOS DAC using TDA1541A

[sidd0rs]

OK, so am I to understand that the main principles of a good DAC design are:

- Choosing the best DAC chip
- Designing a good analogue stage
- Sample rate conversion.
- High quality power supply important?

DAC chip choice
What makes you suggest the Texas Instruments chips (PCM1792, PCM1794, DSD1794, etc.) rather than the Wolfson WM8740/8741? I notice that the TI chips have better SNR and THD than the Wolfson chips; is that all one should look for when choosing a DAC chip?

Analogue Stage
Assuming I wish to keep the analogue stage solid-state (for now), could you point me in the direction of any 'top-end' opamps that would be suitable? Would a suitable discrete analogue stage be a significant improvement over an opamp based section, or are the current opamp offerings just as good, these days?

Sample Rate Conversion
I have read up on the principles and understand the concepts. I am curious, is there any reason not to upsample?

Thank you for your patience.

[JC Fardo]

Originally posted by sidd0rs
I'm interested in the experts' opinion on what makes a good DAC circuit. I understand this is a very open question, but I'm happy for open answers. I would appreciate it if someone could list the factors which make a particular DAC circuit better than another (assuming both use the same DAC chip)?

Are there any current examples of real high-end DAC design that I could examine?

Is there any reading material available for someone wishing to study and begin experimenting with DAC design?

My sincerest apologies if any of this has been asked before.

Cheers.

It depends mostly on what you consider listening qualities. Pin point detailing soundstage or long run relaxing listening sections. Too much detailing can be anoying. You can only realise this with listening sections Start with simple minimalistcs designs like Audiosector NOS DAC or Monica2 kits (musical/non-fatigueing) them go further to over sampling multi chip designs 192/44Hz (detailing sound stage) if you´re not happy. Sorry for poor english and good luck

[psycoman]

Multibit DAC usually has warmer sound such as PCM1704, TDA1541. 1 bit DAC has clean sound and good dynamic. Analogue stage is very important. Only the high-end opamp can win tube stage or passive transformer. I recommend opa128sm, opa128lm, opa627sm, opa627bm. Opa128km/jm has very special sexy vocal but the high extension is not enough.

You also need a regulated power supply. The circuit is in my blog. http://drcylau.spaces.live.com/

For the choice of DAC, don’t rely on the spec. You should do the research on the high-end model to see which chipset they use but I can say you only have two choices. Multibit or 1 bit delta sigma.

I didn’t try Wolfson. First, I don’t know where to buy it. At least, I can’t find in digikey. Second, I did go to HIFI show and listened Cambridge Audio. They use Wolfson. I can say it absolutely not a high-end. I don’t trust Wolfson and Cirrus Logic.

Don't wate the time to compare Wolfson or TI. May be you compare PCM1792 or AD1955 for the 1 bit chip. They both appear in the high-end model. I believe TI is better.



SRC is important. It improves the sound a lot.

[regal]

In my experience the best DAC designs use a multibit DAC chip and a no feedback analog stage. For SS the Pass D1 analog stage is good, but a passive I/V +SE tube gain is even better. Also probably best to not fool with reclocking, it is rarely right and usually just causes problems.

[analog_sa]

Originally posted by psycoman


SRC is important. It improves the sound a lot.

Yes. Unless it completely ruins it

[audiofish]

Originally posted by sidd0rs
DAC chip choice
What makes you suggest the Texas Instruments chips (PCM1792, PCM1794, DSD1794, etc.) rather than the Wolfson WM8740/8741? I notice that the TI chips have better SNR and THD than the Wolfson chips; is that all one should look for when choosing a DAC chip?

Note that there is a subtle but important difference between the Wolfson and some of the TI parts: all of the Wolfson chips I have come across have a classic differential voltage output whereas the very high end TI chips have a, slightly unconventional, differential current output. This generally allows you some extra flexibility: there is less pressure on the designer to place the analogue section directly next to the DAC. This may account for the TI parts' better specifications.

Note also the price of the TI chips! TI's PCM1794 would cost me £12 whereas Wolfson's WM8740 only sets me back £3. (Typical external circuitry is very similar for both: three op-amps per channel and some assorted passives.)

I'm currently designing a DAC based on the WM8740 which I shall probably make within the next couple of months if you're interested.

Originally posted by sidd0rs
Analogue Stage
Assuming I wish to keep the analogue stage solid-state (for now), could you point me in the direction of any 'top-end' opamps that would be suitable? Would a suitable discrete analogue stage be a significant improvement over an opamp based section, or are the current opamp offerings just as good, these days?

Linear Tech and Analog Devices are your best bet for op-amps. Be warned about the price of these high end chips though, especially when using six of them for a design! Make sure the input noise is low, something below about 5nv / sqrt( Hz ). For what it's worth I'm using two quad op-amps from Analog for my design: AD8674s. For just the two I'm spending over £8.

You would be hard pressed to design a discrete output stage with better specs unless you really know what you're doing (which I, alas, do not!). The price for parts would also be a lot more for a discrete design.

[philpoole]

That's a good question, and open to all kinds of responses!

Another thing to consider is also your limitations. I don't mean that in a discouraging way, but you may not be able to build PCBs or use SMD devices. And you may well benefit from building things as modules, so you swap them in and out.

For instance. I decided to use the TDA1541A. Some people love it, at least as many probably hate it (I think you'll see that kind of distribution with most DACs). But, I chose it because I found some, because it was DIP and thus easier to work with, but also because it could work at NOS, and being multibit typically used lower frequencies on the board, so I could just about get away with using stripboard.

I did implement a simple NOS DAC, and I was chuffed with the results, but wanted to know what OS sounded like and compare. So added an SAA7220P/B digital filter module - which sounded awesome.
I now have a PMD100 which I hope to one day use to improve on my oversampling (which will probably lead me eventually to a more modern DAC, with more than 16-bits resolution, and not require so much glue logic with a PMD100).
I also have a small IV module, based on opamps biased into class A. I have tried and so far failed to implement a discrete IV stage (I will return to it). I found, for my DAC, that a dual Burr Brown op amp sounded the best for my setup, as opposed to the likes of LM4562 and LM6172.

Another major breakthrough happened when I paid some attention to regulation. I initially used 7805/7815/7915 regulators, and it worked. Replacing these with simply LM317/LM337 as a first regulator stage, followed by TL431 programmable zener diode based regulation drastically improved things. I'm sure there is further room for improvement there - there are lots of exotic regulation methods to choose from.
Again, having your box as a bunch of modules makes it easier to chop and change and improve things without having to rebuild everything.
Mind you, eventually, it would probably be best as a one off PCB, but that could wait while you experiment.

AND, don't forget about having a decent clock!

Cheers,
Phil