EEVblog Electronics Community Forum

Electronics => Projects, Designs, and Technical Stuff => Topic started by: Recyclojunk64 on December 11, 2016, 11:17:28 am

Title: Modifying a server power supply into an adjustable bench supply
Post by: Recyclojunk64 on December 11, 2016, 11:17:28 am
I am trying to modify a switchmode server power supply for adjustable output, from 0 to 42V, by tapping into the feedback loop and adjusting the reference voltage with a 10 turn pot.

I have previously done this with a few ATX supplies using TL494 controller chips, and it usually worked quite well. Except that none of them have survived the torture i put them through (see profile pic). So this time I hope to have a somewhat more robust and powerfull supply that will handle any electronic circuit I throw at it. In the future I also plan on adding a linear tracking output stage to it for better regulation and contstant-current/current-limiting, but I won't worry about that for now.

BTW, the 42V output is rated at 66A, which is about 2.8kW, so I think it should be suitable for most of the stuff I'm going to use it for. Plus 12V, 3.3V and 5V rails adding up to about 300W (from a separate transformer). It also uses a 15A C19 input connector (at 240V).


Now on to modifying it. Firstly I dissabled the over/undervoltage protection. The controller chips is on the primary side, using an opamp and optocoupler (in linear mode) for voltage feedback. The problem though is that the circuit is only adjusting the on-time of the transistors. This is a half-bridge topology, and when I measured the duty cycle I found that I could modify it from 48% down to 33% by decreasing the reference voltage on the opamp, but I soon realised that I was also increasing the frequency from 76kHz to nearly 500kHz! Not so good for the drive transistors. The controller is a CXA8038P http://www.datasheet4u.com/pdf/CXA8038P-pdf/500903 (http://www.datasheet4u.com/pdf/CXA8038P-pdf/500903) which has two pins for timing resistors, one for On Time and the other for Dead Time control, with the optocoupler connected to the On Time pin. And two external totem poles for driving the main transistors.

So I'm asking for advice on what the best thing to do here might be. Perhaps someone knows of a simmilar smps controller chip that has fixed frequency but variable duty cycle that I could deadbug into the circuit somehow. As I'd like to keep the overcurrent and other protections. Or an easy method of adjusting both the On and Off time at the same time maybe? Or maybe I should use a low side controller and use two optocouplers to get the drive signals to the mains side totem poles? And also anything I might want too look out for, I have some idea of what I'm doing but i'm no expert.

More pics to come
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: bktemp on December 11, 2016, 12:16:53 pm
LLC resonant power supplies have a quite limited output voltage range. Adjusting the output voltage over a wide range is impossible.
Using PWM does not work, because the transformers are specially designed for resonant operation (large leakage inductance).
If you want an adjustable output voltage using this SMPS, you need to add a buck converter at its output.
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: Recyclojunk64 on December 11, 2016, 03:02:26 pm
Thanks for the reply. Not really what I was hoping to hear, I think if I was to have a buck converter on the output that will probably be just as large as the power supply itself for 3kW. In which case I might as well make it an isolated buck converter and run it straight from rectified mains.

But is there any chance I might be able to convert this supply to a different topology? Is there another topology that is driven with a half bridge and does ZVS, that I could adjust the duty cycle to adjust the output? Or maybe something that doesn't use ZVS but will still output over 2KW using only the three half-bridges of 20N50F mosfets, perhaps I could replace them with faster ones so that they dissipate less heat when switching? And either replace the transformers or modify them to have less leakage inductance if necessary? I have lots of different cores and litz wire. I would quite like to make it adjustable as it's pretty much useless to me if I can't adjust it over a wide range.
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: bktemp on December 11, 2016, 03:50:47 pm
One big adavantage of LLC resonant converters is the absence of an inductor at the output, because it is integrated into the transformer.
So if you change the topology, you most likely need to add an inductor after the output rectifier.
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: Siwastaja on December 11, 2016, 03:55:03 pm
Quote
I think if I was to have a buck converter on the output that will probably be just as large as the power supply itself for 3kW

Nope - a non-isolating, simple, 42V input buck converter is likely to be 10x smaller, 10x cheaper, and a lot simpler to design than a mains-powered isolated supply - and it's easy to achieve good efficiency.

You can also add proper current limit to the buck stage to make it a real lab supply.
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: Recyclojunk64 on December 11, 2016, 05:17:03 pm
Hmm, well in that case I think I'll probably go with an external buck converter then. There isn't really room for any additional output inductors in the server supply anyway, I was thinking the output inductors were rather small (they are in parallel, and have two turns), now I know why.

Tomorrow I'll raid my parts boxes and see what large powdered iron torroids and mosfets I can find and hopefully I'll be able to build something that isn't too large. And maybe I'll have a linear tracking stage on the output as well eventually.

I guess the other advantage of having an external buck converter instead is that I can use it to supply more than 66A at lower voltages, though I'll have to draw the line somewhere of course when designing it. Maybe 200A max (when below 14V), or maybe more. And perhaps I might make it a buck-boost converter (non-inverting 4-switch type) for when I need more than 42V
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: bktemp on December 11, 2016, 05:48:10 pm
At that power levels I would use a multiphase synchronous buck converter: Because they are phase shifted, the ripple at both input and output is much lower.
It also reduces the current in each stepdown, making the inductor design much easier.
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: flyrod on December 01, 2018, 04:29:59 pm
Hello all.  I brought this thread back up because I'm interested in doing the same thing.  I want a nice bench supply with enough power to charge batteries, run motors, etc. but one that can do this is at least a few hundred dollars.  Server power supplies have good components and are much more affordable, so I think it would be a fun and useful project to modify one or use the parts to build a nice bench supply.  Has anyone had much luck doing this?

At that power levels I would use a multiphase synchronous buck converter:

Are there any good sources or plans for one of these?  There are some server or telecom supplies that put out 48-50v, so as a start I could just stick one of these on it, right:?

https://www.ebay.com/itm/273509350150 (https://www.ebay.com/itm/273509350150)

(https://i.ebayimg.com/images/g/EaYAAOSwX3FbbEdE/s-l500.jpg)

This would get me adjustable voltage and constant current from the way I read the listing.

Thanks for your input!

Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: 001 on December 01, 2018, 08:22:11 pm
Hello all.  I brought this thread back up because I'm interested in doing the same thing.  I want a nice bench supply with enough power to charge batteries, run motors, etc. but one that can do this is at least a few hundred dollars.  Server power supplies have good components and are much more affordable, so I think it would be a fun and useful project to modify one or use the parts to build a nice bench supply.  Has anyone had much luck doing this?

At that power levels I would use a multiphase synchronous buck converter:

Are there any good sources or plans for one of these?  There are some server or telecom supplies that put out 48-50v, so as a start I could just stick one of these on it, right:?

https://www.ebay.com/itm/273509350150 (https://www.ebay.com/itm/273509350150)

(https://i.ebayimg.com/images/g/EaYAAOSwX3FbbEdE/s-l500.jpg)

This would get me adjustable voltage and constant current from the way I read the listing.

Thanks for your input!

You can try to get up to 50 volts and up to 200 ampers from dirty cheap china welding inverter
Title: Re: Modifying a server power supply into an adjustable bench supply
Post by: flyrod on December 04, 2018, 12:07:51 am
You can try to get up to 50 volts and up to 200 ampers from dirty cheap china welding inverter

The point is to get adjustable voltage and current for use as a bench supply.  It seems that a name brand server supply would have better components than a chineze bench supply.  I saw a 50v 40A HP blade server supply on fleabay for $20 shipped.  But maybe I should just buy a cheap bench supply.  I was looking at this one:

(http://tekpower.us/media/catalog/product/cache/1/image/363x/040ec09b1e35df139433887a97daa66f/a/z/azgq_2256.jpg)
from here:

http://tekpower.us/power-supply/switching-powersupply/tp3030e.html (http://tekpower.us/power-supply/switching-powersupply/tp3030e.html)

I found someone talking about it here:

https://www.eevblog.com/forum/reviews/teardown-tekpower-tp3030e-30v30a-adjustable-switching-power-supply/ (https://www.eevblog.com/forum/reviews/teardown-tekpower-tp3030e-30v30a-adjustable-switching-power-supply/)

Does anyone have suggestions?