I want XLR out from AD1933 but getting very confused!

[alexanderbrevig]

Hi everyone!

I've been working on and off on a project and finally I have decided (I think.. heh) on which DAC to use. I landed on AD1933 with balanced/differential outputs.
My initial hope was that I could simply bring those outputs to an XLR connector. But after thinking a bit and getting some input from a friend I've started to worry about:

• Phantom power DC protection
• Low Pass filtering < 20kHz

Then I started to think about how to do these things on a differential signal.
Could I use a signal transformer to protect about DC, and simply do a small shunt C to common for the filter?
Should I go back to a single ended output version? Do the filtering and buffering there? Then decouple/bypass into a typical unbalanced to balanced configuration.

Do anyone know of a 'best practice' example of how to do I2S/DAC to XLR line out?
Any pointers much appreciated!

The dream (if it would work) is attached with crappy paint.

EDIT: added LTSpice simulation from 1 to 1MHz, -3db at around 25kHz  [and corrected the upload...]

[Cerebus]

Take a look at That Corporations balanced line driver ICs. Even if they are not to your taste, That Corporation have a boatload of useful design notes/ app notes on balanced input and outputs that will probably answer any question you might have on the subject.

Transformers are always a good option, but you have the usual trade-offs in price and quality. If this is a project that is likely to be 'on the road' then I'd heartily recommend a transformer solution. Transformer coupling can save you from nasties (like say a few hundred volts common mode) like nothing else can.

[dmills]

Second the recommendation for the THAT Corp app notes, all good stuff.

The usual way to skin this is to convert to single ended, lowpass to taste (For a sigma delta part I like ~60Khz or so, third or fifth order to avoid in band phase shifts, but it depends a bit on where the noise shaping starts to put meaningful energy), then convert back to differential and scale to get your full scale to somewhere in the +24dBu or so region at the connectors.

Phantom protection is usually diode clamps behind the output build out resistors, no big deal if the build out resistance is 50 ohms or so. 

Regards, Dan.

[alexanderbrevig]

Thanks guys, just scrolled through That application notes (recognize the name from my guitar fx playing and general audio interest googleing) but can't find anything about balanced/differential.
Is my THE DREAM schematic 100% crushed?

It will definitely be on the road.

My project is for semi professional bands.

[dmills]

Pin 1 goes to chassis, it goes directly to chassis, it does not have caps and resistors, and going via the internal 0V reference is right out!
See AES48 for the right way to do that.

Look at the massive phase shifts thru the audio band in your simulation, ~30 degrees of differential phase between near DC and 1KHz and it just gets worse from there.

Also, you will only see a few volts peak to peak with this, pro line level is up around at least +18 to +24dBu, so some gain after the DAC is indicated (easily combined with the filters).

Page 25 of the AD1933 has a suitable filter circuit and you just stuff that into a line driver chip, or if you are feeling cheap, go 'impedance balanced' (Which is perfectly valid balanced line for audio) which just involves an extra resistor.... TIs SLOA150A note is possibly interesting if going here.

Regards,  Dan. 

[alexanderbrevig]

As an update I've gone for AD1934 with single ended output, and a filter from their PDF into a THAT1646 line driver with design according to the schematic of a demo board they have.
I'll figure out what kind of PSU I need for that to hit approx line level out. I don't need it to hit it exactly as it will go into a mixing console preamp either way.

Thanks for your feedback and getting me to research more! Feel more confident now

[dmills]

Still not sure why your ground connection goes to the IO connectors (U4 and U7 I am guessing), it really shouldn't.

I would have stuck with the differential output ADC and a differential filter, easier to keep the single ended references very local that way, but whatever.

Personally I would have the output coupling electrolytic caps the other way around, just so they are better behaved in the presence of P48, and add some low value build out resistors and clamp diodes to protect against P48 accidents.   A common mode choke on each output (the small ferrite kind) and a hundred pF to chassis on each output is a good idea as it helps to keep mobile phones out of the audio doings.

Regards, Dan.

[alexanderbrevig]

Thanks dmills!

I'm out of pin on my current version so waiting for a new license for DipTrace. I'll make it configurable (earth/gnd) I see it done in many ways, so I'll just test what seems best for me.
TVS will be added, as well as a choke. I'll add pads for a filter cap to chassis as well. Thanks for the tips!

U4 and U7 are just some symbols I found. I'll make clear ones when I decide on which conn to use. Probably a black Neutrik with chassis conductivity and chassis pinout.

[alexanderbrevig]

Front end of this is evolving
Added your suggestions, and a GND lift SSR. Nor sure if the filter caps (100p) should maybe be mandatory, so not be part of GND lift. Ideas?

Thanks again!

[dmills]

Ground lift on an SSR??! WTF?
Never seen that before, it is almost certainly horrible, how does that relay behave up at cell phone frequencies?

Seriously tie pin 1 directly to chassis, and connect your PCB 0V to chassis at ONE point, job done.

Regards, Dan.