Modurn lab power supply ideas and advice

[oliver_the_potato]

I plan to build a lab power supply into a Tektronix TM500 system plug-in enclosure because such plugins are hard to come by in the UK and tend to be unreasonably pricey if you can find them at all.  Most of the available PSU modules are also looking a little dated now.  If this works out i'll publish the project with PCBs.

Requirements:
* capable of operating as a split power supply, with + and - rails.  this could either be in fixed dual tracking mode or as 2 separate supplies that can be seriesed together.  the latter is slightly more flexible with slightly more potential issues around current control.
* variable voltage and current adjustable down to '0V' and '0A'... ish.  A few mV,mA is acceptable
* im considering switching pre-regulation to improve efficiency as the supplies in the TM500 mainframes are a little anaemic.
* current and voltage readouts, to be provided by a digital display.  this should be the easy bit.
* +- about 20V max.  TM500 presents each (high power)bay 2 separate dedicated 1 amp 25VAC windings and one dedicated 4 amp 17VAC winding.

Options
* LT3081 based designs.  Linear even did a demo board https://www.analog.com/media/en/technical-documentation/user-guides/DC2132AF.PDF single channel but i think two could be seriesed up.   it's an expensive design though,  I don't mind expense if it buys me something, particularly if it buys simplicity, but here it seems over the top, £50 of semiconductors just for the two supplies alone.  the need for a minimum load on the LT3081 is annoying, though interestingly the datasheet just glosses over that with a 249Ohm resistor to ground.  at the max 20v that is only a couple of watts so maybe not a huge issue, but i'd have to calibrate the readout for that extra current draw.
* there is a power op-amp, https://www.ti.com/product/OPA549  which even has a current limit feature.  They are also expensive but very simple, potentiality.  Unfortunately hard to find in the UK.  getting a current reading could be annoying though as it doesn't use a shunt resistor, so i would have to add one, then fight with high side current sensing anyway... so..
* What about alternative power op-amps intended for audio?  for example the LM3876 in conjunction with a high-side current scene amplifier such as the INA296?  it should make for a simple topology, as an added bonus one could even feed in an arbitrary signal and use the psu as a power buffer.  I have an HP psu/amp combo like this and it is surprisingly useful.  On the down side i'm not sure how well a power op-amp intended for audio would like having it's input voltages changed on the fly, by a pre-regulator or tap switching to improve efficiency.  I am also not sure if there are hidden pitfalls in this strategy, particularly around the current limiting which can be hard to make stable.
* yes i suppose there are also legacy designs.  I don't like them.  they are hard.  they have an ungodly number of zenners, a mix of ancient integrated and discrete techniques.  yes i'm not ashamed to say i may have reached the limits of my expertise working with these designs and have failed with them brffore.  Go on, tell me i should just use an LM723, i dare you 

opinions and advice welcome.

[Kleinstein]

For a +- supply one would likely build to idetentical parts that could be stacked.

With the limited space a preregulator may be a good idea, though it adds complexity. The alternative would be several transformer taps and a fan to get rid off the heat. It is kind of a balance between heat (power rating for the given size) and noise.

For the regulator the standard way is to use the floating regulator scheme and OP-amps for the control and power transistors. An integrated solution is tricky, as the heat would effect the accuracy. So it is a good thing to have the heat  and precision parts separated. It looks a bit ancient, but it is still the best way to build a lab supply. The frequency response part (compensation) needs a bit care to get good stabilty also with a difficult load. Here a simulation can really help, if one does not copy a tested design. The supply for the regulator part can also power the read-out and control part.

[moffy]

You might wish to specify what your voltage and current requirements are, as any advice would need to take that into account.

[oliver_the_potato]

@moffy
Yes, good point i did omit that.  edited the original post, max about 20v, + and -

[jbb]

You mention a 17V ‘AC’ winding. Is that 50 Hz (or 60 Hz)? Of so, a classic approach is to deploy a phase controlled rectifier using Silicon Controlled Rectifiers (SCRs) aka thyristors as a pre-regulator.

(But a 17 V AC winding won’t reliably get you a 20 V DC output, I guess. Maybe it was sized for 15 V?)

[PCB.Wiz]

* +- about 20V max.  TM500 presents each (high power)bay 2 separate dedicated 1 amp 25VAC windings and one dedicated 4 amp 17VAC winding.

What current do you target ? > 1A ?

You can work backwards from the current and cooling.  The negative regulator is more restrictive these days.

There are circuits around for older LM337 that can go to 0V, like this  https://www.eleccircuit.com/0-60-volt-dc-variable-power-supply-using-lm317lm337/
The 337 is easy to cool, and relatively simple, but does not have stellar specs.

A more modern part like TPS7A33, comes in a harder to use and cool QFN20 package and is rated to just 1A, but has lower dropout and lower noise.

* What about alternative power op-amps intended for audio?  for example the LM3876 in conjunction with a high-side current scene amplifier such as the INA296?  it should make for a simple topology, as an added bonus one could even feed in an arbitrary signal and use the psu as a power buffer.  I have an HP psu/amp combo like this and it is surprisingly useful.  On the down side i'm not sure how well a power op-amp intended for audio would like having it's input voltages changed on the fly, by a pre-regulator or tap switching to improve efficiency.  I am also not sure if there are hidden pitfalls in this strategy, particularly around the current limiting which can be hard to make stable.

If you cannot find an existing LDO, then building one can be very like designing a power amp.
Fortunately, good opamps are easy to find, and so are precision voltage references.

There used to be split rail tracking regulator controllers, but they seem harder to find.*

or, you could make two identical positive floating supplies, your windings may allow for that ?
Getting those to track when floating, is more of a challenge. Maybe a part like IL300 can be used to achieve that ?

* Not split rail, but parts like MIC5158 control external power devices.

Addit:
There are also CV/CC controllers for power supplies, which have 2 OpAmps and a Vref  - see
https://www.diodes.com/products/power-management/ac-dc-converters/secondary-side-controllers-and-switchers/?

[jonpaul]

No need to reinvent wheels.

  get the PS501, etc schematic, manual and duplicate the fine  desing.

Both TM500 and TM5000 mainframnes and plugins are commn in France, EU.

In UK Try BVWS and amateur radio fles and boot sles.

Lots of PS501 on ebay, co reasonable proces

j

[David Hess]

Each plug-in slot also has access to an isolated PNP and NPN power transistor which is mounted to the frame of the mainframe for heat sinking.  These are sized to handle a 20 volt 1 amp power supply.

No need to reinvent wheels.

  get the PS501, etc schematic, manual and duplicate the fine  desing.

I would modernize the design of the Tektronix PS503.

[oliver_the_potato]

I was looking at the PS503 design today, it's not as bad as i anticipated.  A good fallback certainly.  I'm not so keen on the way it mirrors some things to accommodate the PNP/NPN pass transistors in the chassis, or the high side current sensing i'd have to work around to get something useable by a digital current meter.

I'm currently investigating the possibility of using a single LT3081 per channel, no pre regulator, use either diode+capacitor negative voltage generators off the AC, or a suitable tiny transformer to get enough load to regulate down to 0V (only need about 8mA)  the two identical channels could then be stacked for split supply or paralleled if more current was needed.  that would be an upgrade on PS503 capabilities, at the cost of potentially ~50W being dissipated in the plugin, worst case.

investigations continue!

[PCB.Wiz]

I'm currently investigating the possibility of using a single LT3081 per channel, no pre regulator, use either diode+capacitor negative voltage generators off the AC, or a suitable tiny transformer to get enough load to regulate down to 0V (only need about 8mA)  the two identical channels could then be stacked for split supply or paralleled if more current was needed. 

If you can tolerate a low 10's mV, you can also use a good LOW SAT transistor as a min load current sink, set for 8-10mA.
Very simple, and no negative rails.

[David Hess]

If you can tolerate a low 10's mV, you can also use a good LOW SAT transistor as a min load current sink, set for 8-10mA.
Very simple, and no negative rails.

The original Tektronix power supplies take advantage of the other side to include a pull-up or pull-down current to the supply output, and then also bias the error amplifier so that it can drive the output slightly negative or positive no matter what the offset of the original operational amplifier was, so both sides can cross zero output by a couple of millivolts.

They used 741s or 301As but with a -5 or +5 volt supply below ground so that their common mode input range includes ground.

[Marco]

The LT3081 also gets you an microsecond fast current limiting regulator that can operate without output capacitance.

[Kleinstein]

Without preregulation the LT3081 is very limited as it has a hard time to get rid of the heat.

Also keep in mind that the votlage regulators are made for a well defined load, while a lab supply should work with most sensible external loads. So a voltage regulator may oscillate in some cases or at least have a stronger ringing than a good lab supply.

When using a +- supply with a connom ground, one is kind forced to use high side current sensing. To allow for low side sensing one would need 2 seprate supplies, not a directly linked +- supply.

[oliver_the_potato]

I'm going to attempt to go with a switching pre-reg and deal with the switching noise it makes.  there is some president for this, both in the datasheet and someone over on diyaudio did a rather more complicated implementation here: https://www.diyaudio.com/community/threads/triple-outputs-160w-lab-psu-epsux3-version-2.263471/

Unsure what length im going to have to go to to get switching noise down to an acceptable enough limit, probably i'll at least end up isolating the switching part with ferrites everywhere.  Preliminary schematic excluding most of that filtering and a lot of quality of life additions that will come later attached

[PCB.Wiz]

You could split it into 2 separate boards, then you can experiment more with filtering and decoupling.
You have a current sink included, but the 1.2M is too large to sink 5mA min.
For a bench supply,  I'd also be a bit wary about sink current derived from the raw DC - use a LED or a zener or a proper VREF part, to drive the injection resistor keeps things clean at very low output voltages.
for current mirrors you usually pick matched parts, and where you want lowest sat, (closest to 0mV Vout) you could look at PHPT610035NK or NSS40301MDR2G parts.

[oliver_the_potato]

@PCB.wiz
All good points, thanks.  Multi board is the way to go i think, I'm designing a prototype/base board for the TM500 that will take multiple small plugin boards for just this purpose.  it will make it much cheaper and easier to iterate on this and future tek plugin ideas i have. 

[jbb]

I couldn't stop myself... and knocked up a simulation bench for a thyristor controlled rectifier.

Varying the firing angle of the thyristors from 'a bit before 90 degrees' (in this case 80 degrees) to 'nearly 180 degrees' (in this case 150 degrees) yields a smoothly variable DC link voltage (assuming a DC link capacitor and DC current load).

Pros:

• Switching noise should be roughly similar to the simple diode rectifier that you've got already
• Incremental power losses are minimal - about 0.4V extra drop in a thyristor vs a silicon diode

Cons:

• Power factor (and this RMS current draw from the Tx winding) is terrible
• I don't know if you can get controller chips for this job any more. They may have all gone obsolete.

[PCB.Wiz]

I couldn't stop myself... and knocked up a simulation bench for a thyristor controlled rectifier.

I'll post my 'I could not stop myself' variant then.   
I checked out a PFET heat spreader and dropper.

Has more loss than SCR's but works on both rise and fall of the AC signal, for 200Hz ripple, and can deliver a reduced current during the highest AC too.

Comes in between a linear dropper and a SCR, (much lower noise than SCR) tho you can tune the PFET model to be very like a SCR one, with no current between the flanks.
An advantage of the PFET is you can tune the gate drive force, to trade off a bit more switching loss for lower noise.

This simple variant uses a PNP gate driver to the PFET to roughly give a 'flat' M1 power during the conducting part of the AC cycle