10 amp psu regulator

[JAY X]

Hi!

I was asked recently to design a linear psu for a recording studio. They used smps +/-15v 7.5amps per rail. These supplies are noisy despite LC filters used. I started looking at the technical documents section, and found the soundcraft psu CPS450, but it is rated at 3.5amps/ rail aprox.

I started searching for high current linear regulators, and just found available at reasonable price  LM338 5Amp regulators.
In the datasheet, its specified a peak current of 7,5 amps.  In the application notes of the datasheet there is schematic of a 10 amp regulator, with an lm107 as comparator. (it can be replaced for LM393 i think).

The schematic shown is just a building block for the full psu, that i still have to draw. I just want to have your opinions on this building block, as this is not the typical paraleling of voltage regulators....which is a tricky technique. ¿is it a good building block?

¡Thank you for your advise!

Jay x

 

[abbey road d enfer]

with an lm107 as comparator. (it can be replaced for LM393 i think).

I don't think so; LM107 is an opamp, and its inputs can be operated up to Vcc, which is the case in this circuit. There are certainly more modern opamps that can be used, but you must be aware of this requirement.

The schematic shown is just a building block for the full psu, that i still have to draw. I just want to have your opinions on this building block, as this is not the typical paraleling of voltage regulators....which is a tricky technique. ¿is it a good building block?

Although I have never used this technique, it solves elegantly the issue of load sharing.
Other commonly used methods resort to inserting load sharing resistors at the outputs. It has the disadvantage of increasing the actual resistance of the supply, but it's usually a non-issue, considering many mixers use power distribution resistors on the channels (typically 10-47 ohms).
If it happened to be the case, I would use multiple paralleled regs with load sharing resistors; it's too simple to pass.

¡Thank you for your advise!

Jay x
[/quote]

 

[Rob Flinn]

Try an LT1083 ( https://www.analog.com/media/en/technical-documentation/data-sheets/108345fh.pdf )

It will do 7.5A.  However all of these regs have a dimishing current capability the more volts you drop across them.  The data sheet is you friend.

 

[moamps]

I would choose CPS275 with one more high power pass transistor added per rail. 8)
http://www.lacompagnie-events.com/pdf/cps250_275_user_guide.pdf

 

[JohnRoberts]

Be careful about taking every application note as optimal design advice...

The example with LM107 may be designed by a FAE (field application engineer) trying to pimp 107 op amps.

Another stage of LC output filter might do the trick. Not rocket science.

Properly designed electronics should be relatively tolerant of PS noise that will get filtered several times before the rubber meets the road.

JR

 

[abbey road d enfer]

Be careful about taking every application note as optimal design advice...

+1

Another stage of LC output filter might do the trick. Not rocket science.

The role of the LM107 there is not to improve regulation, it's governing load-sharing between the two regs. It's a brilliant idea, as long as you don't try to put in practice IMO
You know probably too well that getting two regulation loops working together in theory often results in them fighting against each other in practice.

 

[moamps]

The additional problem with LM338 recently is that the TO3 version is obsolete, and I would never design the regulator where the dissipation on a TO220 package is greater than 10W on the long run.

 

[JAY X]

Hi All!

After reading the posts, i decided to build  two "normal" 5 amp LM338 psu as for the datasheet, and put their outputs in paralell.
I expect the current output will be about 3-4 amps maximum per psu. The heatsink will be quite large, but there are some models to choose and try... No fans because it is for a studio control room... ;D

Jay x

 

[abbey road d enfer]

Hi All!

After reading the posts, i decided to build  two "normal" 5 amp LM338 psu as for the datasheet, and put their outputs in paralell.
I expect the current output will be about 3-4 amps maximum per psu. The heatsink will be quite large, but there are some models to choose and try... No fans because it is for a studio control room... ;D

Jay x

I would put 2x 0.1 ohm resistors for load-sharing.

 

[JAY X]

Hi,

Attached is a section of the psu. Not finished yet.

+/-16v dc out.
5 amps per rail.
LM338T TO220


Snubber rectifier filter: 47R ( watts?)+47nf 100v

Filter caps: CRC FILTER:  10000uf 35v / 2.2R  ¿55watts? +10000uf35v

Voltage setting resistors: 1k8 / 100 R half watt.

Load sharing resistors: 0.1R  2,5watts

What i'm worried about is the wattage of the resistors. I used the I^2 x R formula.

Jay x

P.s. Q1 and Q2 are MPSA 42/92 transistors for a mutual shutdown circuit.

 

[abbey road d enfer]

Filter caps: CRC FILTER:  10000uf 35v / 2.2R  ¿55watts? +10000uf35v

No need for a resistor between caps. Just wastes heat. For noise performance, it is important to use hierarchical ground. That means ground does not loop back, leaving the reference points of C5/6 where they are, i.e. between the rectifier and the regulator's reference (junction C7/R3). No supplementary connection between C5/6 and 0V.

Load sharing resistors: 0.1R  2,5watts

I would put 5W at least.

 

[dogears]

No need for a resistor between caps. Just wastes heat. For noise performance, it is important to use hierarchical ground. That means ground does not loop back, leaving the reference points of C5/6 where they are, i.e. between the rectifier and the regulator's reference (junction C7/R3). No supplementary connection between C5/6 and 0V.

Hi Abbey, can you expand on this a bit?

I've often used the scheme showed in his schematic. The high and low AC sides feed the rectifier, but 0V on the PSU is tied to 0V AC is tied to earth (safety) ground in one place. Audio 0V ref is derived from the PSU 0V ref.  Even when I've done SMPS I'ved tied their 0VDC back to the AC 0V at one point.

Is there a better way to do it?

 

[abbey road d enfer]

Hi Abbey, can you expand on this a bit?

I've often used the scheme showed in his schematic. The high and low AC sides feed the rectifier, but 0V on the PSU is tied to 0V AC is tied to earth (safety) ground in one place. Audio 0V ref is derived from the PSU 0V ref.  Even when I've done SMPS I'ved tied their 0VDC back to the AC 0V at one point.

Is there a better way to do it?

For best performance, chassis ground should be connected to 0V reference; in particular the xfmr center tap should not be redirected to ground. Neither should the 0V side of smoothing caps.
For compliance with safety rules, the conductors carrying this dirty 0V should be of same size as mains wires (at least), in case there is a direct discharge from a leaky xfmr.

 

[moamps]

.... i decided to build  two "normal" 5 amp LM338 psu as for the datasheet, and put their outputs in paralell.

That's  not a good idea, IMO, but if you insist on it,  I will suggest inserting high power schottky diodes in series with R7 and R14.

 

[benb]

If you feel you must roll your own, you may want to read the "sample chapter" here which happens to be on power supplies.
http://theartofelectronics.com

I looked for linear open frame on digikey and mouser - digikey didn't have anything close, but Mouser has 15V at 9 amp and 15 amp available at big-bucks prices.  I'd get two of the 9 amp (you said you needed 7.5 amp, right?), wire the outputs in series and be done with it. I'd feel a lot better with a COTS solution than a one-off something I put together myself.

 

[dogears]

For best performance, chassis ground should be connected to 0V reference; in particular the xfmr center tap should not be redirected to ground. Neither should the 0V side of smoothing caps.
For compliance with safety rules, the conductors carrying this dirty 0V should be of same size as mains wires (at least), in case there is a direct discharge from a leaky xfmr.

So for a bipolar supply, is this correct?

Transformer center tap, 0V side of pre-regulator smoothing caps, and regulator reference are all tied together at a node we'll call 'dirty 0V'.
Audio ground is established after the regulator (his C8 and C9 0V).
Dirty 0V and audio 0V meet back at the chassis ground, which is also safety to earth.

 

[abbey road d enfer]

So for a bipolar supply, is this correct?

Transformer center tap, 0V side of pre-regulator smoothing caps, and regulator reference are all tied together at a node we'll call 'dirty 0V'.
Audio ground is established after the regulator (his C8 and C9 0V).

Correct.

Dirty 0V and audio 0V meet back at the chassis ground, which is also safety to earth.

No. You leave dirty ground alone. That's 0V you tie to safety ground.

 

[JAY X]

Hi!

Ok, i will implement all recomendations.  About the rectifier snubber...¿is it necessary? ¿what resistor wattage?

jay x

 

[abbey road d enfer]

About the rectifier snubber...¿is it necessary?

I don't know. It depends on the xfmr and rectifier. You must experiment.

 

[dogears]

Correct.
No. You leave dirty ground alone. That's 0V you tie to safety ground.

Hm. So the dirty 0V which is also regulator reference is only related to audio 0v by the regulator?

This also means dirty 0V is floating, but audio and chassis are tied to earth. Is that correct?