What is the tradeoff to run a DAC single ended?

[DougL]

I have a DAC based on the PCM1794.
I have a passive analog section using a transformer.
I was looking at the Zen I/V and wondering if I could use a single output and either:
A. Ground the unused input.
B. Place a small resistor to ground.
C. Place an "active load" on the pin and not use the output.

This was just a thought, it would avoid a balanced to Single Ended conversion, and allow a more standard volume control. The question is what do I give up for this?

Thanks in advance.

Doug

[SoNic_real_one]

You give up to 6dB of signal/noise ratio...

[Pano]

How are you doing a passive output stage? Doesn't the PCM1794 need an I/V stage at the output?
I could see maybe a transformer after that.

[DougL]

How are you doing a passive output stage?

Step up transformer and resistor. Started with a Edcor 1:2 and a 1k resistor, liked it so much I picked up the RAKK passive line stage, which is a Lundahl LL1674 and a 3K resistor.

[DougL]

You give up to 6dB of signal/noise ratio...

I think can live with that. I was afraid that the output had a lot of second harmonic that needed the other output to cancel, or some other show stopper.

Doug

[jcx]

transformers do diff to single ended

modern monolithic DACs use balanced output to reject large high frequency digital noise current in their substrate - its impossible to get decent S/N without differential output

balanced circuitry also gives even harmonic cancellation

the PCM1794 has large output DC current bias - your I/V has to balance this current with low Vos from gnd - otherwise you get nonlinearity as you start to have signal current shunted by internal protection diodes

the Zen I/V may not have low enough input Vos for the PCM1794, doesn't have DC current balancing so the circuit needs some modifications to be compatible with this chip

where ever you put it, your transformer must not have any net DC current in the windings - blocking C absolutely mandatory


any or all of these ideas will perform measurably worse in distortion than the datasheet 5534 I/V - and much better op amp I/V is possible

[DougL]

do you have a circuit or just playing with "audiophoolish" ideas?

I will leave you to judge that. I currently have a step up transformer and a resistor acting as my analog stage.

transformers do diff to single ended

Yes, I have heard that.

modern monolithic DACs use balanced output to reject large high frequency digital noise current in their substrate - its impossible to get decent S/N without differential output

Good to know.

balanced circuitry also gives even harmonic cancellation

Yes, but is it significant?

the Zen I/V may not have low enough input Vos for the PCM1794, doesn't have DC current balancing so the circuit needs some modifications to be compatible with this chip

What sort of modifications are you talking about?
Zin was quoted as 15 ohms, and the common source input should be self biasing, especially since the outputs are not DC coupled. It certainly might benifit from 2 pairs of FET in parallel.

where ever you put it, your transformer must not have any net DC current in the windings

Agreed.

blocking C absolutely mandatory

That's one way of accomplishing it.

any or all of these ideas will perform measurably worse in distortion than the datasheet 5534 I/V - and much better op amp I/V is possible

Measurements aren't everything. I should put together an op-amp I/V and compare for myself.

[jcx]

The Zen I/V does sim ridiculously well, IMD as low as -100dB – but who knows how good the jfet models are or how well any particular current out DAC performs with nonzero input impedance and offset

jfet parameter match dependance could be explored - I really expect the complete 2nd harmonic cancellation seen in sim with idendical models is impractical with real world device variation

There is offset of -124 mV at the I/V input with the nominal models and Nelson suggests a low Ohm pot V offset adjustment for mere mortals who don’t have hundreds of the fets to select the Idss matched pairs to get the 1 mV Vos he claims in his article, balanced I/V requires good matching of 4 fets

I show a sim with AC coupled xfmr (it could give selectable balanced or single ended output) to the degree that the caps match there is some psrr but filtered/regulated power is still a good idea

Cap physical size may be a challenge if you insist on film caps and low frequency response

Xfmr ratio, frequency response and Cload modification of the response are all details that need to be addressed

the current sources aproximate the PCM1794 +,- Iout, I show bias current cancellation R but not Vos adj

[Ken Newton]

Originally Posted by DougL

I think can live with that. I was afraid that the output had a lot of second harmonic that needed the other output to cancel, or some other show stopper.

Doug

Actually, you only give up about 3dB in SNR.

[DougL]

jcx,

Thanks for the simulation. Very interesting.

Doug