Linear PSU Transient Recovery Response

[tombi]

Hello,

I've been following along with Peter Oakes' YouTube series on designing a bench PSU. I started doing my own design and am using LTSpice to model the behaviour of a MOSFET based regulator.

I am compensating the error amplifier to stop oscillation and now struggling with transient recovery.

For a general purpose bench supply, what would be sensible numbers for transient recovery?

My plan is to build a supply based around a 120VA transformer to supply 30V @ 4A or 15V @ 8A (or probably a bit less current - overkill for my needs). I've not done any current limiting circuit or pre-regulator. At the moment I am just working on the voltage regulator.

Under simulation when I set the output to 30V and attach a pulsed current source drawing 4A with 1us rise time and 200ms period it overshoots by 300mV and takes 10us to recover.

At 1A and 5V the overshoot is 150mV and about 10us.

Is this as good as I can expect or should I keep fiddling? Are these overshoots acceptable? Is 1us too short a rise time for the load?

I haven't modeled lead inductance or anything else yet.

Tom

[T3sl4co1l]

That seems reasonable, in the broad scheme of things.

It's always a question of need.  Do you need a precision supply that's rock steady at all voltages and frequencies?  Probably not, because the frequency requirement will be blown by the inductance of half a meter of hookup cables.

A good baseline is, will it work for logic supplies?  Meaning, can it do 5.0V or 3.3V +/-5%?  Holding that tolerance even under transient conditions?  Sounds like it can, or it could with some modest additional filtering.

Speaking of which -- do check to make sure it's stable with a capacitive load, or an LC load.  Resistors (or current sinks) are boring!

I have a bench supply that's built with large darlington transistors, and no active current limit; at the end of a 2m cable, I measured over 40A peak current capacity, before the test pulsing circuit terminated the pulse.  It takes some 20us for the current to rise to that point.  During which time, the load voltage is 100% less than nominal!   Clearly, such use is only practical with bypass capacitors at the point of load, and that's true, by necessity, no matter what's inside the power supply.  So, don't worry too much about extreme bandwidth or stability.  If anything, take the opportunity to reduce bandwidth or increase loop gain, to achieve lower noise or greater DC stability.

Tim

[tombi]

Thanks!

Haven't checked the behaviour with an inductive/heavily capacitive load. Will do this.

Otherwise I think I'll leave it for now then.  I want to add voltage sensing from the front terminals etc which is likely to stuff up the response. Also the pre-regulator is likely to stuff up the transient recovery response too (I wasn't planning on using a switching pre-regulator). Also when I proto-type it the real circuit is likely worse.

This is a project for fun/learning so overkill is my usual MO.

Tom

[dom0]

a 120VA transformer to supply 30V @ 4A or 15V @ 8A

You cannot draw 30 V / 4 A DC (120 W) from a transformer rated at 120 VA. When using a bridge rectifier you can draw about 1/1.8 of the rated AC current as DC.


Things I feel important about a lab supply:
- Reliability
- Reliability
- Reliability
- Thermal reliability (I often see way to small heatsinks or very optimistic thermal calculatins like "assuming 25 °C ambient temperature..." ; do it right, Tambj = 50 °C for hobby use).
- A dedicated switch to turn the output off w/o turning the supply off
- Little, if any, power on / off glitching
- For many uses noise isn't really a concern if it is not excessive
- Regarding stability, it is, again, not really a concern for most usage scenarios, unless it drifts excessively (like more than a percent due to heating or over the course of a day)

[Pjotr]

For a general purpose bench supply, what would be sensible numbers for transient recovery?

There is no number for transient recovery. There is overshoot or undershoot. An ideal designed PSU has neither. Further on a good PSU is insensitive to complex loads. Best approach is to design its internal impedance purely ohmic. That means you have to design the control loop (feedback loop) in such a way that is matches the output capacitor and its ESR (or vice versa). When the loop gain drops down, the output becomes inductive. Precisely at that point the output capacitor has to jump in so they complement each other.

A good way to investigate this all is to simulate the whole thing in PSpice as you did with a pulsed load.  Also look at the output impedance with an AC analysis. It will not bring you on the real spot but a far end in the right direction.

[tombi]

a 120VA transformer to supply 30V @ 4A or 15V @ 8A

You cannot draw 30 V / 4 A DC (120 W) from a transformer rated at 120 VA. When using a bridge rectifier you can draw about 1/1.8 of the rated AC current as DC.

Ahh that I missed! Thank you. Hmm 1/1.8 * 4 = 2.22 is not quite enough. Really wanted at least 3. I was hoping to keep the transformer small.

My plan was to pack two (or maybe three) of these PSUs into a box and have a Raspberry Pi with an LCD screen controlling them. Each PSU module will have its own transformer, uController, voltage DAC, current DAC and current measurement ADC. Each module will be opto-isolated from the Raspberrry PI and be isolated from the others. I was planning on having a mode where you can connect two PSUs to run in dual-tracking mode (via some relay on the board).

The Raspberry Pi will need a UPS and probably its own supply. Lots to figure out yet.

Things I feel important about a lab supply:
- Reliability
- Reliability
- Reliability
- Thermal reliability (I often see way to small heatsinks or very optimistic thermal calculatins like "assuming 25 °C ambient temperature..." ; do it right, Tambj = 50 °C for hobby use).
- A dedicated switch to turn the output off w/o turning the supply off
- Little, if any, power on / off glitching
- For many uses noise isn't really a concern if it is not excessive
- Regarding stability, it is, again, not really a concern for most usage scenarios, unless it drifts excessively (like more than a percent due to heating or over the course of a day)

Yes I plan to use a pre-regulator to reduce the heat dissipated but haven't gotten that far. I was planning on using a voltage reference and a DAC to control the output voltage.

I was thinking a relay mounted behind the front terminals to switch the output off. This way I can put the voltage sense close to the terminals. The easy way is not to use a relay and just set the output to zero volts but I think this is not as good.

For the moment my simulations show the output rising smoothly without overshoot at turn on. I think this is because I use a voltage doubler to generate the voltage for the MOSFET gate and this takes time to charge up. I'll keep watching as I add features. I imagine the real circuit will behave quite differently to the simulation.

Thanks again,

Tom

[Liv]

For example, transient response for Agilent E3615A (shown in the service manual).

"Load Transient Response Time Definition : This is the time for the output voltage to return to within a specified band around its voltage following a change from full load to half load or half load to full load."