DAC output - does it go -5V -> 5V or 0v -> 5V?

Hi! :slight_smile:

I am looking at this DAC (not in particular but more as an example): http://www.ti.com/lit/ds/symlink/pcm1690-q1.pdf

It seems all the DAC is getting is +5V and 0V (and +3.3V for digital). It outputs differential outputs for each channel, e.g. DAC1+ and DAC1-.

Page 39 shows an interesting diagram:

How much are +Vs and -Vs meant to be? And if they are only used after the output, I guess the output of the DAC itself are between 0 and 5V?

That's confusing but it does NOT say Vcc2 can be negative and if you look at the "absolute maximum" ratings, Vcc1 and Vcc2 should be the same (within 0.1V).

It does have a differential output and it says you can get a differential peak-to-peak output of 1.6 x Vcc1 (8V). But it doesn't go negative relative to ground.

Balanced (differential) audio connections are common in professional applications but they are not biased so they actually go positive and negative.

Ok thanks DVDdoug :) That confirms my expectations.

Then output of DAC1+ and DAC1- gives you two signals with opposite phase going from 0V to 5V - correct?

If so then is my next question - how do you get these signals to be between -2.5V and +2.5V then - when you only have one single supply og +5V?

I have read about the subject, and always found some sorts of "workaround" where you create a virtual ground at +2.5V - which makes your signal oscillate around that. There are tons of professional audio equipment which only have a single power supply (f.x. USB powered audio interface), and their output is 0V centered, are they also using the "virtual ground method"?

ToneArt: How much are +Vs and -Vs meant to be? And if they are only used after the output, I guess the output of the DAC itself are between 0 and 5V?

-Vs and +Vs are nothing to do with the DAC, they go to the opamps doing the buffering and low-pass filtering and would conventionally be +15V and -15V.

The DAC itself is strictly 0V -- +5V range. Its output is 8V peak-to-peak, ie the differential outputs each have a 4V range. The opamps convert differential outputs to single ended bipolar.

The + and - outputs are centred around mid-rail (2.5V), and Vcom pin is provided to decouple the internal 2.5V inside the DAC chip. The signs refer to the signal polarity, not the DC voltages.

Thank you Mark, understood - the DAC gives signals from 0V to 5V.

Hence my next question (same as the one from my last reply): how do you get your -15V and +15V (or -2.5V and +2.5V - I do not need to get louder than 7dBu = 1.7V RMS) from one single 5V supply?

Then you can use rail-to-rail opamps and generate a 2.5V virtual ground for them. Implement the output stages like this and then ac-couple the outputs to remove the 2.5V DC offset. More specifically you use 0V for -Vs, +5V for +Vs and 2.5V for the "ground" in the output stages.

The AD8656 is a high spec 5V-only rail-to-rail dual opamp that can drive headphones directly - note you need true rail-to-rail opamps for this, many are a bit weak on the output side when driven close to the rails. The TI LME49721 is another possibility. Most of the new rail-to-rail audio opamps are SMT, note.

Thanks Mark sorry I just saw your answer now :)

So ok the 2.5V virtual ground is the way to go then. Fair enough.

My first idea when trying this was to send the 2.5V ground and the signals on the audio outputs, but based on what you're saying I should: * only use the virtual ground for the output stage * then add a capacitor to remove the offset before I send the signals further (on an XLR connector f.x) * so that the XLR are getting the normal 0V as ground and signals is between -2.5V and 2.5V (instead of using my virtual ground on the XLR)

Correct?

Yes, but you might do better to use +/-5V supply, and opamps that can work with 10V supply difference, they wouldn't need to be rail-to-rail and you'd avoid those large capacitors on the output.

For sure - but to get these +/-5V supply I would need my own power supply which will generate these from a 220V AC voltage, right?

Actually my idea is to have a simple +12V power supply (still simpler and cheaper than making my own double supply I am tinking), that way I can make a +6V virtual ground and have plenty enough headroom.

Why not use a -5V rail generator chip? (switched capacitor type)

Ok that sounds super interesting - thanks! Never heard about these before! I will check :) A question though: the analog output of the DAC will still be between 0V and +5V so before I send this to an opamp f.x. I should add a capacitor to bring it around 0V right?

Also I was thinking about how to get the 2.5V virtual ground - most examples I found on the net were by simpy using two resistors (as voltage divider), but that creates a really unstable voltage as any load on it will affect the voltage - but then I realized that I should use a proper voltage regulator at 2.5 instead!

ToneArt: Ok that sounds super interesting - thanks! Never heard about these before! I will check :) A question though: the analog output of the DAC will still be between 0V and +5V so before I send this to an opamp f.x. I should add a capacitor to bring it around 0V right?

No, use a "differential amplifier" opamp configuration with ground as the reference, this will convert differential to single-ended about the reference voltage. Your opamps need to be rail-to-rail and able to work on 10V total supply.

Also I was thinking about how to get the 2.5V virtual ground - most examples I found on the net were by simpy using two resistors (as voltage divider), but that creates a really unstable voltage as any load on it will affect the voltage - but then I realized that I should use a proper voltage regulator at 2.5 instead!

Virtual ground is a divider, with decoupling to true ground, into a buffer amplifier (often a unity-gain opamp).

Thank you Mark :)