0-10V DAC or 0-5V DAC and opamp for gain - Meeting specs close to supply rails.

[petemate]

Hi guys, I'd like to remote control a PSU via its "analog programming voltage", which are two 0-10V signals that controls voltage and current. I plan to make a small PCB with an STM32F103 and power everything through USB. The PSU has a 10V output that can be used as supply/"reference" for the DAC and I would like to use this voltage to produce the 0-10V signals to the PSU, since that would mean that i won't have to include a 5V-to-10V boost converter on the PCB. This also means that the analog part of the design will be a unipolar supply.

I will need a 16 bit DAC to produce a ~1mV resolution to the PSU(its a 60V PSU) and it is hard to find a 2-channel 16 bit DAC that allows 0-10V operation from a 10V input voltage. Its much easier to find a 5V-compatible 2-channel 16 bit DAC, but then I would have to gain up the signal. Not a problem in itself, even though linearity, offset voltage, etc does have an influence. This can however compensated for via feedback. The real problem is the unipolar supply, because i need to keep the 1mV resolution when i'm close to 0V and 10V. Even a nice rail-to-rail opamp will only go a few millivolts within its rails typically, with worst-case being 10s of millivolts. In addition to this, the DACs also suffer from this problem of not being able to put out "true 0mV", as they also have some offset or zero error. some DACs do have a siwtchable "pull-down" resistor that could be enabled to get true 0V, but its kind of a hack that I'd like to avoid.

is there really no way around implementing a real bipolar supply to fix this problem?

[David Hess]

is there really no way around implementing a real bipolar supply to fix this problem?

That covers it; somehow you need to implement supply voltages less than and greater than 0 and 10 volts.

[magic]

What does the input impedance of the PSU look like?

What about some simple single-ended class A output stage controlled by a precision opamp? Say, 10kΩ to ground and a common source p-MOS with RDS(on)<1Ω to make 9.999V possible?

[OM222O]

Easiest solution would be a charge pump. If you have a 5V supply, you can create a -5v rail and by doubling the 10v reference, you can get 20v. It will not be symmetric, but should cover the voltage range you need. For the oscillator, easiest solution is using a hex NOT gate (preferably cmos) and either making an RC oscillator with one of the gates or feeding a single PWM to the first one. If the input impedance was low and the pump couldn't sustain the voltage, you can increase the current capability using external fets for switching.

[mawyatt]

If you don't want to create a + supply >10 volts and a - supply < 0 volts, which is the best approach IMO, then consider a discrete CMOS inverter as an output buffer for a rail to rail op-amp. Compensate the loop since you've added additional gain, and the phase is reversed because of the inverter. Some small footprint SOT23 FETs have very low Rdson, under 0.5 ohms, so the output will pull to 10 volts and ground with under 0.5 ohms effective output impedance.

This approach will cause heavy current draw from the 10 volt source for the mid output voltage ranges tho, so this needs to be considered. Also consider selecting FETS with higher threshold voltages and low Rdson. We use this discrete CMOS inverter approach to switch between rail outputs with very accurate output levels (used in precision voltage divider).

Best,

[magic]

Output impedance of the 10V source is another consideration. If it's some TL431 rubbish with 1Ω impedance then simply drawing 1mA of current will reduce regulation range by 1mV (edit: unless the control input is ratiometric )

How much effort would it be to actually bias a push-pull CMOS stage properly, even at mid-rails? All those micropower CMOS opamps gotta be doing it somehow.

[petemate]

Hi guys, thanks for all your answers. To follow up on your questions:

What does the input impedance of the PSU look like?

What about some simple single-ended class A output stage controlled by a precision opamp? Say, 10kΩ to ground and a common source p-MOS with RDS(on)<1Ω to make 9.999V possible?

The input-impedance looks to be 50kOhm. So there won't be any significant load on the output of whatever ends up generating the signal, meaning that it could work very close to the rail.

A common-source would solve the issue in one direction only, as far as I see it. Am I missing something?

Output impedance of the 10V source is another consideration. If it's some TL431 rubbish with 1Ω impedance then simply drawing 1mA of current will reduce regulation range by 1mV (edit: unless the control input is ratiometric )

How much effort would it be to actually bias a push-pull CMOS stage properly, even at mid-rails? All those micropower CMOS opamps gotta be doing it somehow.

As far as I can trace out, pulling a few milliamps should be possible. I don't have any specs for the 10V reference, but it goes all over the PCB and is powering other stuff, so I belive that I should be able to pull a few milliamps. (I am unable to tear down the PSU completely. Its one of those japanese high-end ones with no attention to ease of assembly. A ton of different screws and wire harnesses going all over.

I doubt that I'd be able to do a better job than a rail-to-rail opamp, but maybe I should just design the PCB as a first-shot and see what kind of performance I can actually get with a single-supply rail-to-rail opamp. It may be enough. The 10V input is equal to 60V full scale output, so 5mV should be ~30mV output from the PSU. That should be low enough to work with.

[magic]

A common-source would solve the issue in one direction only, as far as I see it. Am I missing something?

The idea was: resistor pulls down (almost to zero, limited only by off-state drain leakage current and/or leakage currents from the PSU), FET pulls up (limited by RDS(on) and pulldown resistance).