Anatomy of an A-100 PSU3. Please help me check

[SPoT_D]

Hi everyone.

I'm looking into building a power supply and was looking at how a A-100 PSU3 works. I'm especially interested in the (adjustable) regulators part.
Now, I only have some pictures on the internet to go from, so I would like to know if anyone can help me check some things:
- is it indeed based on a LM317 and LM337?
- What is the value of the resistors underneath the heatsinks? (220R?)
- Can you actually connect the negative out of an IRM-30-15 to the positive out of the IRM-20-15 to get a +15V/-15V?



oh and does anyone have a KiCad 6.0 footprint for a IRM-30-15?

Thanks!

[guest]

from looking at pictures online, it looks like C7,8 are at the output stage, along with a C9,10 which are 0.1uF. and there is another set of 10uF caps between the ADJ pins and ground. its wierd they put a resistor under the heatsink. the LM317 has ok thermal stability, but maybe they wanted to up this? so the 220ohm might have a tempco to it. its hard to say what the right tempco would be, though, as its a nonlinear response, and depends upon load current. ive seen graphs where its goes up, and then comes back down. and it probably varies from part to part.

[SPoT_D]

from looking at pictures online, it looks like C7,8 are at the output stage, along with a C9,10 which are 0.1uF. and there is another set of 10uF caps between the ADJ pins and ground. its wierd they put a resistor under the heatsink. the LM317 has ok thermal stability, but maybe they wanted to up this? so the 220ohm might have a tempco to it. its hard to say what the right tempco would be, though, as its a nonlinear response, and depends upon load current. ive seen graphs where its goes up, and then comes back down. and it probably varies from part to part.

Ah yeah filtering the output does make a lot of sense. I'll update that picture later. I see those caps are next to the LEDs on the original. they are not in this schematic.
I'm not 100% sure about there being a resistor underneath the heatsink, but it looks that way judging by the solder points on the bottom of that board.
If these are indeed LM317 and LM337 they are what set the voltage output and should not be tempco's (R4&R5)

[dot matrix madness]

I assume the reason for having the resistors under the heat sinks was simply motivated by having them
close to the voltage regulator. If you look at the data sheets for the LM337 and LM317 it's obvious that
these aren't any tempcos.
The LM337 TO220 delivers up to 1.5 Ampere on the negative rail, thus that makes sense.
However for the positive rail that is specified at 2.0 Ampere, the LM317 TO220 would fall short.
Anyway you can email Dieter Doepfer directly (hardware@doepfer.de), he's very willing to provide infos
on such details.

[SPoT_D]

However for the positive rail that is specified at 2.0 Ampere, the LM317 TO220 would fall short.

ah, you are right.
Possibly a LM338 then? Those should have a 5A output

[dot matrix madness]

When you calibrate the output voltages on the plus and minus rail make sure you put some
load to ground. I reallized that just the current going through the indicating LEDs is not
enough for that purpose.
Can find infos on the LM338 right now, but instead of the adjustable voltage regulators
you can alternatvely use those that deliver fixed +/-12 Volts and leave out the adjusting
resistor networks: L7912CV TO220 (1.5 A) and LM1085 IT-12 (3.0 A).
Adding reverse protection diodes is highly recommend.
See for example Ken Stone's power supplies CGS14
and CGS66.

[noobyist]

I drew this footprint in kicad. Not tested yet...

[al2o3cr]

Screen Shot 2023-05-23 at 10.44.30 AM.png (28.94 KiB) Viewed 590 times

Big safety callout: putting a fuse in the "neutral" side of the AC input wiring is at minimum a bad idea, and at maximum literally illegal (depending on your country's electrical codes).

[SPoT_D]

Screen Shot 2023-05-23 at 10.44.30 AM.png

Big safety callout: putting a fuse in the "neutral" side of the AC input wiring is at minimum a bad idea, and at maximum literally illegal (depending on your country's electrical codes).

Thanks. I'll check the PCB again. Could have sworn that is how it was connected

[guest]

I assume the reason for having the resistors under the heat sinks was simply motivated by having them
close to the voltage regulator. If you look at the data sheets for the LM337 and LM317 it's obvious that
these aren't any tempcos.
The LM337 TO220 delivers up to 1.5 Ampere on the negative rail, thus that makes sense.
However for the positive rail that is specified at 2.0 Ampere, the LM317 TO220 would fall short.

thats the thing i dont get, is why you would want those resistors near the main sources of heat, if you werent going to use tempcos, especially considering the rest of the adjustment setting parts are not under the heatsink, so trace length isnt any more of an issue. just seems like an odd choice to bake your resistors when not necessary. also, the LM317 can provide 2.2A in its typical operating point, and thats just limited by thermal disipation anyways, so as long as you can get the heat off, you can probably go beyond that as well.

as for minimum current for regulation, the current through the adjustment network is 6mA, which should be enough (3.5mA typical), and the extra led current helps out here.

[dot matrix madness]

I assume the reason for having the resistors under the heat sinks was simply motivated by having them
close to the voltage regulator. If you look at the data sheets for the LM337 and LM317 it's obvious that
these aren't any tempcos.
The LM337 TO220 delivers up to 1.5 Ampere on the negative rail, thus that makes sense.
However for the positive rail that is specified at 2.0 Ampere, the LM317 TO220 would fall short.

thats the thing i dont get, is why you would want those resistors near the main sources of heat, if you werent going to use tempcos, especially considering the rest of the adjustment setting parts are not under the heatsink, so trace length isnt any more of an issue. just seems like an odd choice to bake your resistors when not necessary. also, the LM317 can provide 2.2A in its typical operating point, and thats just limited by thermal disipation anyways, so as long as you can get the heat off, you can probably go beyond that as well.

as for minimum current for regulation, the current through the adjustment network is 6mA, which should be enough (3.5mA typical), and the extra led current helps out here.

Maybe somebody with that PSU at hand can have a closer look at the actual components. But if you confront Dieter himself with such
expert knowledge I'm sure he'll explains you his concept. Don't hesitate to email him (see in my postings above).

[dot matrix madness]

The installation manual for the PSU3 reveals some more details on the components (if you can decipher the values):
The negative rail uses a LM337 and a 200R resistor under the heat sink, whereas the positive rails
comprises two resistors (220R and 10k) and a LM350 (rated at 3.0 A).
Still I guess that "missing space" on the PCB was the reason for putting the resistors there.

[SPoT_D]

The installation manual for the PSU3 reveals some more details on the components (if you can decipher the values):
The negative rail uses a LM337 and a 200R resistor under the heat sink, whereas the positive rails
comprises two resistors (220R and 10k) and a LM350 (rated at 3.0 A).
Still I guess that "missing space" on the PCB was the reason for putting the resistors there.

Thanks for that!
Now it gets mysterious.. any clue what that second resistor is for? judging from the PCB photo it looks to run parallel..
I don't see this suggested in the datasheet though.

[guest]

it also doesnt look populated in that photo. the LM350 has slightly worse thermal regulation, and that resistor runs in parallel with the 220ohm setting resistor, and seems to read 10k to me, which is a common value for an NTC thermistor.

[SPoT_D]

I drew this footprint in kicad. Not tested yet...

Thanks for this! I only just had time to check it. Will doublecheck by the time I order the actual part

[al2o3cr]

thats the thing i dont get, is why you would want those resistors near the main sources of heat, if you werent going to use tempcos, especially considering the rest of the adjustment setting parts are not under the heatsink, so trace length isnt any more of an issue. just seems like an odd choice to bake your resistors when not necessary.

From several of the LM350 datasheets - no idea which one is "the source", they all have similar phrasing.

The current set resistor connected between the adjustment terminal and the output
terminal (usually 240Ω) should be tied directly to the output (case) of the regulator rather than near the load. This
eliminates line drops from appearing effectively in series with the reference and degrading regulation.

With R2/R1 around 9 for a 12V output, that means that any connection resistance between the regulator and the resistor will effectively be 10x worse for regulation.

No idea what that extra 10k resistor is about, though.

[dot matrix madness]

it also doesnt look populated in that photo. the LM350 has slightly worse thermal regulation, and that resistor runs in parallel with the 220ohm setting resistor, and seems to read 10k to me, which is a common value for an NTC thermistor.

That's a good point. An additional common 10k resistor in parallel to the 220R makes no sense.
(with R1=220R, R2=1880R (1k8 plus the 200R trim poti), Iadj=50µA, and Vref=1.25V you arrive at Vout=12.023V)
Possibly they intended to use a 10k NTC thermistor, but then dropped it.
If I take the curves form the National Semiconductor data sheet then I read Iadj=55µA and Vref=1.247 at 75°C (ca. 175°F)
giving rise to Vout=12.007V. Thus thempature compensation is obviously not really needed.
But, as pointed out before, it's not a good idea to expose parts to unneccessary heat stress.
There are much worse design flaws around, for example in my (old) Audiolab amplifer that really bakes
the resistors with two much current to an extend that the PCB turns brown.

[d.simon]

I have a psu3 here...

I can't remove the heatsinks on this one without desoldering - the screw heads are underneath the heatsink (the regulators are folded backwards) and it appears those were added before they soldered the regulators.

I measured 219.5 ohms on the negative and about 220 on the positive. Not sure whatever other parallel resistance is involved, but that's the measurement at the resistor terminals.

If there's a 10k involved on the plus side...its funny that the negative side measures less. But either of them I'd say 220R would work fine (not that I know anything about psu design but 219.5R vs 220R seems insignificant.).

But honestly both sides have 5 through hole leads- 3 for the regulator and 2 for a single resistor. Plus I shined a flashlight in there and could only make out two resistors total.

On the plus side there are those two filled in vias - I thought those were test points, but on the top side I do see a silkscreen indicator of an unpopulated part. So those are probably pads and where an unpopulated part would go.

psu3 back01

psu3 back02

psu3 top01

[d.simon]

unpopulated part pictures...

psu3 unpopulated01

psu3 unpopulated02 closeup

[KSS]

unpopulated part is probably an option to use SMD instead of TH for the 220s.

[SPoT_D]

unpopulated part pictures...
WIN_20230525_15_07_07_Pro.jpg
WIN_20230525_15_06_58_Pro.jpg

Thank you for quite literally shining a light on that.
I've updated the schematic in the first post.

If you want to, could you quickly check to what pin on the IRM-30-15 (and other IRMs) the fuse is connected?

[dot matrix madness]

Thanks d.simon for these details. If there would be a 10k resistors in series with the 220R one, the total resistance would
be around 215R. The measured 219.5R are well with the 1% deviation of a 220R resistors.
My assumption is that the two 220R resistors were initially placed as close as possible to the voltage regulators
(as suggested by the datasheet) already in the earliest version of the PSU (A-100 NT12) and that part of the layout
wasn't changed for the following versions (A-100 PSU2, PSU3).
Doepfer doesn't seem to use SMD parts if not really necessary (probably to allow easier repair the old fashioned way).

[dot matrix madness]

If you want to, could you quickly check to what pin on the IRM-30-15 (and other IRMs) the fuse is connected?

The fuse should be connected to the AC/L terminals of all three bricks (as shown in the updated schematic).
Specs are 1.6A time lag for 115V, and 800mA time lag for 230V.

[SPoT_D]

The fuse should be connected to the AC/L terminals of all three bricks (as shown in the updated schematic).
Specs are 1.6A time lag for 115V, and 800mA time lag for 230V.

Yes. al2o3cr also pointed out that it should be.
However.. If I look at the PCB It looks to be connected to the N pins

[d.simon]

Yes. al2o3cr also pointed out that it should be.
However.. If I look at the PCB It looks to be connected to the N pins

Looking at my psu3 here - That appears to be correct. The fuse is connected inline with N.

This stuff gives me a headache. After looking at IEC connectors and the PSU3 installation pdf...I've convinced myself that neutral is fused.

I don't think that's good...but anyway blue is Neutral and Black is Line/hot correct?
You can see here blue is going through the fuse
as always use at your own risk!

EDIT - oh forgot one important thing...the IEC connector has its own fuse inline with hot.
I'm pretty sure I've been through this same existential PSU3 crisis before...but just forgot about it.

so...make sure to use something similar with an inline fuse. The connector in the diagram has 3 parts : a plug receptacle, a fuse, and a switch.

dieter psu3 wtf?

dieter.jpg (19.22 KiB) Viewed 179 times