Negative to positive power supply design

[ini_kamoze]

Hey, so first a quick intro for my first time here. I came to seek your help for the design of a specific power supply. I have my background in electronics and some years of experience on the job but limited when it comes to low noise and taking heat dissipation into account. The things I've been making previously were more on a budget and did not have the need. This time around I'm trying to make a voltage supply to set the voltages for an out of the box high speed modulator (40Gbps). I want to have a flexible robust design which I can still use later on for other modulators (probably higher speeds).

These would be the requirements for the design:
- 1 adjustable output, range from -20V to 20V (pref linear slope)
- able to adjust output minimum and maximum range (eg. -2 to 8V)
- current draw around 500mA minimum constant load or higher
- stable in time (days, weeks constant operation)
- low noise
- low heat dissipation
- portable and compact

Optional:
- able to cascade the supply chips for added flexibility to up the current in later designs

This would be used three times on the same PCB for three different set voltages.

----- So:
My first thought would be to go with linear regulators for the noise with a virtual ground half way the range. But then with a less elegant big heatsink for the linear regulators so the heat spread in the enclosure does not make the modulators drift over time.

As this would be preferably something compact and portable my preference would be to go for a switching solution. I started looking into the products of linear technology, put an additional stage to make a virtual ground and additional filtering to cancel out the switching noise. This would be better for the heat dissipated but would decent filtering be enough to block the switching?
The modulators I'm using are fully enclosed so I'm not afraid of crosstalk between the tracks/components. I'm more worried for how clean a signal I can get from a switching supply to feed the modulator.

Maybe you already figured, I'd looking for something robust and reliable over simple
I'm open to any suggestions.

Thanks in advance,
Ini

[macboy]

If you really need something that can cross zero like that, then maybe consider using a power opamp as an output device. While power opamps that are marketted for industrial purposes are usually expensive, many devices made for use as audio amplifiers could do the job and are always inexpensive. e.g. LM1875, LM3875, LM3886, TDA7294/5/6.  A power opamp also has the benefit of providing a true four-quadrant output (it can source or sink current while the voltage is either positive or negative or even zero). All basic linear regulators are single quadrant. They will either have a positive output voltage and will source current, or have a negative output voltage and will sink current. For this reason, using a simple linear regulator with a virtual ground halfway through the range very likely can't work. Do you see why?

Besides the power opamp output device, all you need is a bipolar voltage reference and a means to control it. That could be as simple as a pair of TL431 (or similar) and a pot.  The opamp will simply amplify the voltage reference by its gain and that's that.

You will need to be much more specific with your requirements to get specific answers. These:
- stable in time (days, weeks constant operation)
- low noise
- low heat dissipation
- portable and compact
are all essentially meaningless.

[David Hess]

- 1 adjustable output, range from -20V to 20V (pref linear slope)
- able to adjust output minimum and maximum range (eg. -2 to 8V)

My first thought would be to go with linear regulators for the noise with a virtual ground half way the range. But then with a less elegant big heatsink for the linear regulators so the heat spread in the enclosure does not make the modulators drift over time.

This would be used three times on the same PCB for three different set voltages.

Is this three duplicated dual output tracking supplies or three duplicated single output supplies which can vary between a negative and positive voltage?  Do they need to be programmable?

- stable in time (days, weeks constant operation)
- low noise

Stability and low frequency noise mostly depend on the reference and error amplifier.

- low noise
- low heat dissipation

Low dissipation requires a switching regulator which is not a problem for low noise if care is used.  Switching regulators have special considerations if they are to be used over a large ouput (or input) voltage range.

- portable and compact

Portable like battery powered or portable like movable without a fork lift?

Optional - able to cascade the supply chips for added flexibility to up the current in later designs

So parallelable for increased output current?

As this would be preferably something compact and portable my preference would be to go for a switching solution. I started looking into the products of linear technology, put an additional stage to make a virtual ground and additional filtering to cancel out the switching noise. This would be better for the heat dissipated but would decent filtering be enough to block the switching?

I am not a fan of using a linear regulator to remove switching noise because line rejection is poor at the frequencies where switching noise dominates.  Passive filtering is usually sufficient and if precision is required, linear regulation should be done at the load.

Maybe you already figured, I'd looking for something robust and reliable over simple
I'm open to any suggestions.

Precision and low noise usually mean an external error amplifier and reference.

Is this suppose to be an off-line supply?  If not, what is the input voltage range?  A switching regulator can be used with a line transformer for high efficiency.

[ini_kamoze]

All basic linear regulators are single quadrant.

Thanks for pointing this put. I really did not took the sinking in mind. I'll have a look for some opamp based designs and go from there.

Is this three duplicated dual output tracking supplies or three duplicated single output supplies which can vary between a negative and positive voltage?  Do they need to be programmable?

Three times one output that can vary between negative and positive. Programmable would be nice, I did not consider it but it would be handy.

Portable like battery powered or portable like movable without a fork lift?

Movable on the workbench, can even be with an external AC/DC supply. Battery powered is not needed.

So parallelable for increased output current?

Yes, but this would really be an extra if possible.

Passive filtering is usually sufficient and if precision is required, linear regulation should be done at the load.
Precision and low noise usually mean an external error amplifier and reference.

Ok, will take this in mind.

Is this suppose to be an off-line supply?  If not, what is the input voltage range?  A switching regulator can be used with a line transformer for high efficiency.

I think this is one of my least concerns. Transformer or AC/DC converter (external/internal) will be the one that is best/easiest.

Thanks for all the input, this gives me already quite some food for thought.

Concerning these point:

- stable in time (days, weeks constant operation)
- low noise
- low heat dissipation
- portable and compact
are all essentially meaningless.

I don't have any specifics to meet. For now I'll just start with the basics from the design, keeping your input in mind. I'll check how good/bad things are when I have an output.
From the moment I have an output it will be easy to get some figures.

Thanks again

[David Hess]

Three times one output that can vary between negative and positive. Programmable would be nice, I did not consider it but it would be handy.

Ok, so like macboy says, that means a 2 quadrant output.  Linear audio power amplifiers with their class-B or class-AB output stages are a good example of this.  In practice this means using the same reference and error amplifier while mirroring the output stage so it can push and pull.

In theory an integrated power amplifier IC like the examples macboy gave could do this with an external error amplifier for precision but see below for perhaps a better idea.

So parallelable for increased output current?

Yes, but this would really be an extra if possible.

This could be tough but it is certainly doable.

- low heat dissipation
- portable and compact

These two sort of go together.  20 volts at 0.5 to 1 amp implies a minimum 10 to 20 watt continuous power dissipation for a linear design which is pretty reasonable for a bench power supply.  My Tektronix PS503As operate in that range but they have dual tracking outputs instead of a single 2 quadrant output.  But lower power dissipation means either a switching supply or some way to step the input voltage during rectification which some linear bench supplies use to keep power dissipation down.  Some big linear audio power amplifiers do the same thing.

Combining the requirement for lower power dissipation with the requirement for two quadrant operation brings up an interesting possibility.  Audio power amplifiers whether linear or switching have two quadrant outputs and there are a lot of inexpensive high efficiency switching class-D audio power amplifier ICs out there.  Add an LC output stage to them for filtering and an external linear error amplifier and they could be used as the power output stage of a 2 quadrant switching regulator.  Noise from the switching would then become the primary problem and the LC output filtering would complicate frequency compensation and load response.

Something like a TDA8922CJ should be suitable.  There are some synchronous switching regulator controllers which might perform better but they would be a lot more complicated to use.  If noise requirements cannot be met, then a linear class-AB output stage would be needed as well with the TDA8922CJ acting as a preregulator.  Layout and grounding would be critical to keep noise down.  The alternatives to the TDA8922CJ operating as a preregulator to keep the power dissipation down are more expensive and will take up more space

This application sounds a lot like something for ATE (automatic test equipment).  2 quadrant power drivers which can control both voltage and current are common with them although not usually at this power level.  If you can get away with it, I would consider separate 2 quadrant power supply designs with one being low current and completely linear for precision low noise and the other being high current high efficiency and not as quite low noise with a switcher or switcher followed by a linear as described.

[ini_kamoze]

I've spend most of my time today looking around with all the feedback in mind. To be honest, a lot of things I've found is making my head spin a giving me doubt about my electronics knowledge. Spend too many years with arduinos instead of the real deal .
On the other hand I think I found a few examples to go on.

This example from LT I tried to simulate in LTSpice - schematic on p33 (But I did not manage to get something out of it)
http://cds.linear.com/docs/en/datasheet/3091fa.pdf

This design from ti looks quite promising - schematic on p14
http://www.ti.com/lit/ds/symlink/opa541.pdf

I think I will try to build a schematic similar to the one from ti.

[David Hess]

Well, I had a detailed reply which the forum software promptly ate and I have no intention of typing it all back in.  Go check out Linear Technology magazine of August 2002 for another idea.

Good luck.