0 - 30 v 0 to 5-10 amp Lab Bench Power supply design and questions / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

0 - 30 v 0 to 5-10 amp Lab Bench Power supply design and questions

Move Thread

LAN_403

Inducktion

Fri Dec 14 2018, 06:14AM

Registered Member #3637 Joined: Fri Jan 21 2011, 11:07PM
Location: Buffalo, NY
Posts: 1068

Hi All, been a while since I posted here.

I've been brainstorming a method to create an easy lab bench power supply using some common, relatively "cheap" parts.

I bought a server 960 watt power supply for about $15 off of eBay, brand new. 80 plus platinum, a pretty damn good deal considering the amount of power it can push out.

So, my thought process goes something like this

I bought a buck-boost converter, one of these:

*

I'm planning on swapping out the fets and diodes with TO-220 package ones, and set each of them on separate heatsinks, for increased power handling. The original fets are * ones, which are already pretty decent ones. The new fets are * , which are pretty similar to the originals, except a slightly lower on state resistance, and they can handle higher voltage. The gate capacitance is marginally larger, but I don't think it should be a big issue.

The original schottky diode is a *, a pretty damn good one. I couldn't find anything necessarily "better", but still needed a TO-220 package. I chose this as a replacement *.

I'm also going to swap out the inductor with a better, higher power one. I have a few large "transformer" style ones from Coilcraft that should work.

For regulation, I'm planning on adding a linear regulator after the buck boost converter. Using an arduino, and a digital potentiometer system I'm thinking to track the buck/boost and keep it just above the linear regulator's dropout voltage, to add the smoothness of linear regulation but keep it efficient. For the linear regulator I've spent a bit of time researching what to use. There are a number of solutions out there but I think an LM723 fits the bill pretty well; by adding a pass transistor, one can achieve some pretty ridiculous power outputs. The pass transistor is gonna be *, from an NOS heatsink set I got from eBay. They're big transistors, with really low thermal resistance junction to case (0.584 C /W!!) , so I thought they'd be a good choice for this. Either PNP or NPN, but from what I understand a PNP transistor works better since it has a lower Vcesat ?

I'm hoping to make the voltage tracking all digital, using a keypad and the arduino to dial in voltages. The CC will be handled using discretes (differential op amp, current sense resistor) I think... or I may try to use digital control for that too, I'm unsure. I have parts for both, so either is an option.

A lot of this is still in the planning stages, but overall, a few questions:

1. Does this sound feasible? Is there anything specific I should try to watch out for, or keep in mind when designing/putting all of these parts together?

2. With the LM723, it has a built in zener reference, but it's 7 volts; to keep the dropout voltage more reasonable, would it be okay to use an external reference, like a TL431 with a voltage divider to get a 1.25 vref? Or is that a bad idea? I didn't find any information on it but I'm not sure if the error amp or something wouldn't like that.

3. I know that there's going to be issues with oscillations/loops. Is there a general guideline for how to approach and solve them? I've been reading up on how to design linear supplies, and found a lot of different information but wasn't sure if there are more general rules and stuff that people have learned from doing it themselves here.

4. Is 10 amps asking a lot for this? I have a nasty habit of slowly increasing a project's scope until it becomes unreasonable. Would it be better to shoot for a lower current rating? Or is this more something I just have to test and try out once it's built? The 2N568x transistor's SOA shows that with a Vce of under 6, it can do it's full 50 amp rating, so I would think that it'll be okay.

Please let me know your thoughts on this undertaking! I'm excited to try and build something like this least to say, since it'll be the first true electronics project I've taken on in a while.

Sulaiman

Fri Dec 14 2018, 07:47AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Thoughts;

. No matter what the available output of a bench supply is there are always situations where I want more voltage or current capability.

. The most common usage of bench power supplies is to provide common voltages such as 5, 12, 24 Vdc

. Variable voltage output is usually a low current requirement

. If you replace smd components with TO220 devices you will need to use as short as practical wires as lead inductances and capacitive couplings can cause instabilities,
even pcb layout can be problematic - I've read.

. I've not looked at the datasheets that you pointed to but ensure that switching speeds and capacitances are similar or better

. Add rotary encoders or similar for optional manual voltage and current settings, manual digital setting can be rather tedious.
coarse/fine or fast/slow control is useful.

Based on the above, I would;

. Provide good terminals for the 960W power supply as those fixed voltages are very useful

. use the buck-boost converter un-modified except for external control of settings and a fan for the heatsink.

. forget the linear regulator for a while,
the main value of a linear regulator is low noise,
the buck-boost converter ripple should be acceptable for most uses,
it will be surprisingly difficult to get a 'quiet' linear supply with all of the smps action nearby.

Leave space for
front panel rotary encoder/potentiometer, display(s), terminals
and the linear regulator pcb and heatsink

get the power supply working with your microcontroller

use the power supply and modify it later IF you find it inadequate.

P.S. an unexpectedly common use of power supplies is battery charging,
controlling and/or detecting end-of-charge is the challenge.
and
I would remove the heat spreader and heatsink, clean and re-assemble
because heat transfer is vital,
and assembly is often sub-optimal with cheap modules.

You will want to be able to display actual voltage and set voltage, same for current.

2Spoons

Fri Dec 14 2018, 09:50PM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

For regulation, I'm planning on adding a linear regulator after the buck boost converter.

Don't waste your time. For some reason people think of linear regulators as some kind of magic filter that will make all your switching noise go away. If you look at the datasheets, in particular line rejection vs frequency, you will find the large majority are completely useless above 10kHz. They just don't have the internal bandwidth to deal with high frequencies. You will do much better using an LC filter with a cutoff a decade or so below your switching frequency.

Inducktion

Sat Dec 15 2018, 02:38AM

Registered Member #3637 Joined: Fri Jan 21 2011, 11:07PM
Location: Buffalo, NY
Posts: 1068

2Spoons wrote ...

For regulation, I'm planning on adding a linear regulator after the buck boost converter.

I've seen a few people asking about this sort of thing, with a SMPS and a linear reg after, and there's even a paper on it too; Aren't linear power supplies also "faster", in that they can react to changing loads and transients better too? I'm genuinely curious as to why people would attempt this if it offers no benefits.

I know that they aren't perfect. I'll likely add an LC filter per your suggestion to help with switching frequency.

Sulaiman

Sat Dec 15 2018, 11:58AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

At work we sometimes get a particular brand of power supply to repair that uses an smps to provide about 2V Vce for the linear regulator transistor.
The layout and especially 0V wiring looks like it 'evolved'.

There used to be a model of linear power supply that had
on the same shaft as the the front panel voltage setting dial;
. a potentiometer to set the linear regulator output voltage
. a variac to provide via transformer-rectifiers-capacitors etc. just enough dc voltage for the linear regulator to operate properly.

Neither of these two approaches have become widespread,
presumably because the extra cost provided little benefit.

flyrod

Sun Dec 16 2018, 04:16AM

Registered Member #61905 Joined: Sun Nov 12 2017, 03:27AM
Location:
Posts: 23

Inducktion wrote ...Hi All, been a while since I posted here.I've been brainstorming a method to create an easy lab bench power supply using some common, relatively "cheap" parts.

I did something similar. I bought a 2000w server power supply that puts out 50v, then I stuck a couple of these cheapo chineze buck converters on it:

It seems to work well enough. It doesn't go all the way to zero volts, but it's way cheaper than a regular bench supply with similar performance.

You can get amazing deals on the server power supplies, and they use much better components than the chineze stuff on ebay or amazon. It seems like people would be doing more with them, like converting them to adjustable output, etc.

Inducktion

Thu Jan 10 2019, 09:22AM

Registered Member #3637 Joined: Fri Jan 21 2011, 11:07PM
Location: Buffalo, NY
Posts: 1068

I've redone the design of the project a little bit;

Instead of using a chinese built buck/boost converter with unknown characteristics, I'm designing my own LTC3780 based one. Should be a lot better since I have control over it.

My biggest question is, how do I decide the current sense resistor for it? It's a current-mode chip, but I don't entirely understand how one decides what value to use; should the peak current be some multiplier of the maximum current draw? Is there a limit to what it can handle?

Otherwise, I'm also considering still adding a linear stage after; I found a very interesting chip that utilizes a mosfet for linear regulation (and has a charge pump to provide the gate voltage for it) that I may try out. Not sure still.

the_anomaly

Thu Jan 10 2019, 09:01PM

Registered Member #19 Joined: Thu Feb 02 2006, 03:19PM
Location: Jacksonville, FL
Posts: 168

I only briefly looked at the datasheet but I saw choice of Rsense begins at the end of page 17. Just follow the equations after you calculate the inductor. I would expect the value you get to be somewhere between 0.005ohm and 0.1ohm. You will probably have to do a lot of iterations to get a standard off the shelf value. Same for the inductor.

Do your calculations assuming maximum current output, the converter should be able to handle everything below this. If you don't want any ripple, design for a higher max current than you intend to use.

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob