[buz]

I am wondering about the following: if you wanted the most neutral DAC possible, it would have to have as little in the signal path after the DAC as possible.

Starting from that premise, I started wondering if is there an USB DAC that basically pushes the analog out from the DAC directly to a line out to be picked up by an amp? No buffer, no opamps, no caps no nothing in between? Could that even work?

 

[Currawong]

The problem is that half (?) the DACs out there are current output, followed by the problem with the signal being very weak. That's why DACs have an output stage.

 

[audiofil]

I think this is an oversimplified view about DAC output stages.

3 things that must be pointed out:
1. Most DACs output current, not voltage. This requires a proper I/V (current to voltage) conversion stage.
2. It must be capable to drive whatever it's connected at (amp, headphones, etc) without distortion or linearity issues. Given the variety of input impedance alone (cans from 32ohm to amps of 100.000 ohm) this is hardly an easy task. Hence the need for a buffer stage.
3. Most DACs require an analogue low pass filter to cut out high frequency signal (over 20Kohm).

Now, output caps is another matter. They're installed to combat any DC that may creep into the output and sometimes to act as a low pass filters in conjunction with resistors.
The caps may be excluded if the design employs DC coupling.

The way I see it a well designed discrete output stage running in class A cannot be bettered.

 

[nikongod]

Quote:

Originally Posted by Currawong /img/forum/go_quote.gif The problem is that half (?) the DACs out there are current output, followed by the problem with the signal being very weak. That's why DACs have an output stage.

Quote:

Originally Posted by audiofil /img/forum/go_quote.gif I think this is an oversimplified view about DAC output stages. 3 things that must be pointed out: 1. Most DACs output current, not voltage. This requires a proper I/V (current to voltage) conversion stage.

I hate to be a nit-picker guys, but almost ALLL modern DACS are current output from the actual DAC. The ones that are V-out chips just do the I-V conversion in the chip, with an opamp.

I probably agree with you more than not:
Quote:

The way I see it a well designed discrete output stage running in class A cannot be bettered.

I agree.
Take a current output chip, run it through a nice external output stage, and then off it goes. Very nice!
There are tons of DACs that do this.

The problem of USB becomes smaller if you just get a USB to SPDIF converter, or perhaps TOSLINK for galvanic isolation.

 

[Currawong]

audiofil and nikongod: Thanks for filling in the details. I'd forgotten about most of that stuff.

 

[Jon L]

Quote:

Originally Posted by buz /img/forum/go_quote.gif No buffer, no opamps, no caps no nothing in between? Could that even work?

Even without an active output stage, the output caps are still necessary. Here's a USB-to-I2S DAC without an active output stage, but this requires paralleling 8 DAC chips per channel to get enough output voltage. The quality of the output caps are very important for overall sound signature in this case, thus the huge polystyrene and teflon caps.

 

[rosgr63]

Jon nice DAC! Which one is it?

 

[mikechai]

Looks like the once famous Storm Digital nos dac for me

 

[audiofil]

Quote:

Originally Posted by Jon L /img/forum/go_quote.gif Even without an active output stage, the output caps are still necessary

This is not accurate Jon.

In direct coupling designs output caps are redundant. Most high end manufacturers employ this technology for its intrinsic high linearity and low distortion.

Output caps are required in capacitive coupling designs to keep DC out (such as the one you've posted).
This is a serious compromise, also highly dependent on the quality of the caps. Other than keeping DC out of the signal there is little else to be gained here. Linearity, distortion (THD and phase) are considerably higher than direct coupled designs.
This also explains why output caps have gained such a bad reputation.

 

[buz]

So what is the challenge in building direct coupling designs (there has to be one, otherwise everyone would do it if its clearly superior)?

 

[nikongod]

Quote:

Originally Posted by audiofil /img/forum/go_quote.gif Linearity, distortion (THD and phase) are considerably higher than direct coupled designs. This also explains why output caps have gained such a bad reputation.

I dont think that the faults to linearity and whatnot are a result of the cap, but a result of trying to make something work in a way it was never meant to work.

The cap is not the problem, nor are cap coupled designs as a whole. The problem is that a cap is a fast and quick bandaged solution to what is often a more complicated problem and too often gets lumped in as the default "if it has a cap its crap" solution, when it can also represent a VERY elegant and solidly designed solution.

Put another way:
Is the problem that this circuit has output caps or that it does IV with a resistor?

I have seen articles saying (I have never seen actual tests, so take with grain of salt) that a cap has less distortion than a SS voltage shifter.

 

[audiofil]

Of course caps by themselves are not the problem in hand.
My post refers to caps when used for coupling (not in general), because they act as filters in conjunction with resistors. That's the problem I was addressing, as they're not the best solution for it.

Quote:

So what is the challenge in building direct coupling designs (there has to be one, otherwise everyone would do it if its clearly superior)?

DC coupling often employs the use of opamps. And in this form it is widely spread. Opamps are a convenient solution for being cheap (most of them anyway), flexible and take out very little space.
Of course this is not a cure-all-diseases-save-the-planet solution. The quality of opamps used and that of the design will dictate how well it gets the job done.
Cheap solutions (parts and design) will do a cheap job.

A better approach, favored by the "gods" of high end audio, is the use of discrete circuits instead of opamps.

The clear downside of this approach is cost! Well designed discrete circuits are expensive.
Also:
DC coupling circuits are inefficient and need beefier power supplies.
Complex design work is required for biasing problems, in/output impedance matching between stages, keeping DC offset low etc