Linear power supply project

[pqass]

- need GND symbol attached to wire with TP6 for continuity to the transformer center-tap winding.
- U2, U6 GND pins need junction on TP6 wire.
- remove junction to TP6 wire between R32 and R33.  Sometimes Kicad puts in a junction when you cross wires and it goes unnoticed.
- WRT U5B, did you edit the symbol to change the pins?  Because you didn't need to do that. Remove the symbol and re-add original. Select symbol then right-click mouse > Symbol Unit to B and right-click mouse > Mirror Vertically (or "Y" key) to flip.
- WRT R11, it's actually not 1K but "NO POP" (not populated).  And the asymmetry with R12 being a jumper but called R12 is giving me ticks. You could make R12 a resistor symbol with "0R" value.
- "stray hairs" off of CR6 (anode) and R46 (bottom).
- JP1 is technically (from a DRC view) correct.  But it sure is odd (from a human view) for not using the open jumper symbol.  But yes, I agree to your "... I was thinking couldn't I just eliminate the JP1 wire, it does nothing now?" but then why have JP2 if there's no JP1?  My opinion is remove both and tie CR7 cathode to S+.

[Therm Mr.]

- need GND symbol attached to wire with TP6 for continuity to the transformer center-tap winding.
- U2, U6 GND pins need junction on TP6 wire.
- remove junction to TP6 wire between R32 and R33.  Sometimes Kicad puts in a junction when you cross wires and it goes unnoticed.
- WRT U5B, did you edit the symbol to change the pins?  Because you didn't need to do that. Remove the symbol and re-add original. Select symbol then right-click mouse > Symbol Unit to B and right-click mouse > Mirror Vertically (or "Y" key) to flip.
- WRT R11, it's actually not 1K but "NO POP" (not populated).  And the asymmetry with R12 being a jumper but called R12 is giving me ticks. You could make R12 a resistor symbol with "0R" value.
- "stray hairs" off of CR6 (anode) and R46 (bottom).
- JP1 is technically (from a DRC view) correct.  But it sure is odd (from a human view) for not using the open jumper symbol.  But yes, I agree to your "... I was thinking couldn't I just eliminate the JP1 wire, it does nothing now?" but then why have JP2 if there's no JP1?  My opinion is remove both and tie CR7 cathode to S+.

I have done these I will post the revision later. I can just eliminate R12 and replace with wire like the jumpers correct.

Also regarding U5A, I have been fighting those A's and B's ferociously. Now in only four sentences you have ended my struggle. Thanks.

[pqass]

I can just eliminate R12 and replace with wire like the jumpers correct.

Yes but no jumper necessary;  just a wire from SW1 pin 1 to left of R2. that's it. 
Remove R11 entirely too as it would be pointless.

Also regarding U5A, I have been fighting those A's and B's ferociously. Now in only four sentences you have ended my struggle. Thanks.

[Therm Mr.]

Getting closer.

[ArdWar]

Are you trying to match the original design as closely as possible? Otherwise it might be worthwhile to substitute the opamps with more modern readily available parts.
Compensation may require slight adjustment and can be a bit tricky however.

Maybe also recalculate the resistors so it won't need something rather exotic like 900k or 50k or 2k5 (why is that even?). Shifting E96 to E24 values etc.
Also that's quite beefy low value R46 bleeder there.

[pqass]

@Therm Mr.
You forgot R16=10K in the original HP diagram. It's between U5B pin 5 and U? (should be U1) pin 6.

Sorry, I've reconsidered my earlier advise (in reply#52) of removing R12 entirely. 
By having both R11 and R12 (both resistor symbols) across the R2 shunt, you have the option of populating them both (like in the E3611A) to pick-off a fraction of the shunt voltage drop so as to adapt to your chosen panel meter.
Or, populate R12=0ohm and not populate R11, if your meter can use the full shunt voltage drop as is.

You could change the values for U1 and U4 to:
U1:LF411 as in the original E3610A which you can still buy from Digikey. But even a 741 may work.
U4:LM358 since you can't buy the original LF442 from Digikey any more.
But if you install sockets, then you can change these later in testing.

@ArdWar
HP seems to like odd-ball resistor values; eg. R10=21.5R vs 22K.  Most aren't critical and can be nudged to an E12 series value.

@Therm Mr. will need to revisit each. He should understand their purpose and what a change in value would have on the circuit.
Non critical: R10, R33, R40, R35 and R36 (can be  >0.3ohm and <1ohm), R41, R46.
Likely need to change to adapt to panel meters: R18, R17, R12, R11    Joke: R27, R34    Likely: R27=0ohm, R34=no populate
Critical: R15 (with R37 decides the voltage range on the output), R23 and R24 1% (adapt CREF with R2=0.1ohm shunt voltage drop; in this case 300mV shunt Vdrop=3A when setting CREF=-6V), R22 (probably should match R23).

All the diodes except for CR3 and the bridges can be 1N4148; don't need 180V PIV or >1A diodes here.
CR3 can be a 3A diode like 1N540x where x=1 or greater.
CR9, CR13 bridges can be standard all-in-one 1A 100V bridges or be made from standard 1A diodes like 1N400x where x=2 or greater.
CR2 bridge should be 6A, 100V or better.

[Kleinstein]

The OP - amps have capacitors in the feedback and the compensation for the regulator is set by those capacitors and the resistors around the op-amps. Substituting another OP-amp type would not change much. If at all to slow an OP-amp could be an issue. A near 1:1 replacement for the LF442 would be a TL062. If it does not have to be DIP case one is free to use more modern types. The current regulator may want a lower drift type  (e.g. OPA202 or similar) as the voltage is relatively small.

The resistors R22-R24 should not be that critical. They are for setting the gain for the current regulation and usually the shunt is the weak part. If R22 is needed at all, depends on the OP-amp. With the LF442 or other very low bias types there is no need for R22. The quality / power rating for the shunt can be an issue: The burden voltage is usually a bit on the high side to alow reasonable low drift / noise also with a reduced current. This leads to relatively large self heating and related drift in the current regulation.

In the last plan CREF is still wrong. It should be from the pot wiper.
The meter switching part also still looks odd. Keep in mind that the voltage at Cset is quite a bit larger than the read back case. One would have R23=R24 to get the same scale factor.

[pqass]

In the last plan CREF is still wrong.

Not sure I see an issue here as the meter gets a fraction of CREF (R20 pot wiper).  CREF is meant to feed R24.

I don't know why HP chose CREFs range to be 0 to -6V only for it to feed R24=50K.
They could have instead chose the CREF range to be 0 to -600mV, then R24 would only need to be 5K (or 2 parallel 10Ks) but R23=2.5K would still be an odd duck.
Then R40 may be 100K and R20 would be dialed-in to read 3A on a (0-200mV FS) panel meter when CREF=-600mV. 
See attached for this option (simulation here).
Of course R33 and R28 would change by x10 but the R31,R32 trimmers and R19 10 turn pots could still be 10K.
With my alternate 0..-600mV CREF, R24/R23 could also be 10K/(2 parallel 10K); no odd values.

[Kleinstein]

Indeed, the Cref part looks OK. I somehow mixed up the pot and trimmer for the full scale.

[ArdWar]

R17 and R18 should be ratio, so the usual 1:9 pair of 20k:180k or 30k:270k should be fine. A bit higher quiescent current, but shouldn't be a problem.
R23 and R24 appears to be ratio too. 1:20 are plenty. 1k:20k, 1k1:22k, 1k2:24k, 1k5:30k, 1k8:36k, etc.
R22 looks like attempt to balance out bias current. Make it equal R23, or even just leave it out when using low bias opamp.
It seems that you can get away with 15k on R15.
Any resistor in series with potentiometer are obviously noncritical. R33 could be 3k3; R40 could be 270k (maybe, not sure which range R20 is usually in); R10 could be 22R; R6 just pick whichever low value resistor available tbf.
R27 and R34 seems unnecessary. Just pick random low value R27 and random high value R34 if you insist.
Just put 4k7 for R41.
While at it, let's see if you can get away with higher value for R46. 390R dissipates 1.5W at 24V output...

C7 having different value from C13 is also rather curious. Did they do that as an attempt at power sequencing?

If you're changing CREF scaling, might as well change 5VREF to 2V5 so it doesn't need two TL431. Of course you need to recalculate R15 R23 R24 for that.
Honestly this CREF and dual range business is making things way more complicated for not much benefit. Constant current control could be much simpler without CREF, just like how it's done in 3620 (and practically everyone else).

As for OPAMP, I'm not sure tbf. TL062/072/082 sounds adequate, but I'd like something with lower offset here.

[Kleinstein]

The 2 ranges for the current setting make some sense, kind of an alternative to a 10 turn pot. One quite often needs current limits well below the full capability of the supply.
There is no real need to overdo it with the resistor accuracy in the set part, as there is anyway the pot to set the values. It would only be the read back part that may want reasonable stable / accurate resistors. For the voltage regulator the offset of the OP-amp should not be that relevant. So a TL071 is OK there. It is mainly the current part that may want low offset / low drift, so that the reading of the set point has not to much offset from the final actual current limit.

[andrewtaylor]

That's correct with the pot choice.

most times it's a flvour of how one likes the (daily) use of the lab supply:

- 10 turn pot in easier to build in because you only have one hole in the frontpanle and save the space there.  Negative aspect for use is you have to turn and twiddle several turns to get over the (full) range -- sometimes one prefers to reduce the current quick when the magic smoke vents from the powered device.

- 2 pots is quick for course /fien seup - and as kleinstein mentioned, modern circuits often want a precise limit in the low current range for some tests.  A "one suits all" are dual pots that have a dual-shaft concentric (e.g. BOURNS
81A2A-B28-D15/D15L) -- but of course hard to get or quite expensive.

For my projects I prefer BOURNS 3680, most because the FAE here dropped me 100 pieces for free evaluation.


One thing for the lab supply to keep in mind: ON/OFF output switch to quick unpower the output.
Especially when you have a dual/muti voltage supply .-)

[ArdWar]

But the 2 range on 3610 is not for that. It is there to extend the I/V curve by allowing higher current when supplied with lower voltage.
The two ranges are 3A/8V and 2A/15V (or 0.85A/35V and 1.5A/20V), so you need multiturn for control regardless.

For coarse-fine control you can just put two pots with one or two order of magnitude difference in series. Especially here since all of the controls are in rheostat configuration. A bit ugly but shouldn't be a big problem for DIY.

[xavier60]

Getting closer.

If you will be testing the design on a breadboard, consider connecting R6 to the ORing node instead of directly to the output of the CV opamp. This idea suggested to me by Kleinstein greatly reduced the voltage overshoot in my PSU design whenever it transitioned from CC to CV mode.

[Therm Mr.]

Are you trying to match the original design as closely as possible?

Yes Am trying to replicate original design because its proven quality and works well. But there are a few components that will have to be substituted due to obsolescence.

Thanks for the advice

Therm

[Therm Mr.]

@Therm Mr.
You forgot R16=10K in the original HP diagram. It's between U5B pin 5 and U? (should be U1) pin 6.

Sorry, I've reconsidered my earlier advise (in reply#52) of removing R12 entirely. 
By having both R11 and R12 (both resistor symbols) across the R2 shunt, you have the option of populating them both (like in the E3611A) to pick-off a fraction of the shunt voltage drop so as to adapt to your chosen panel meter.
Or, populate R12=0ohm and not populate R11, if your meter can use the full shunt voltage drop as is.

You could change the values for U1 and U4 to:
U1:LF411 as in the original E3610A which you can still buy from Digikey. But even a 741 may work.
U4:LM358 since you can't buy the original LF442 from Digikey any more.
But if you install sockets, then you can change these later in testing.

@ArdWar
HP seems to like odd-ball resistor values; eg. R10=21.5R vs 22K.  Most aren't critical and can be nudged to an E12 series value.

@Therm Mr. will need to revisit each. He should understand their purpose and what a change in value would have on the circuit.
Non critical: R10, R33, R40, R35 and R36 (can be  >0.3ohm and <1ohm), R41, R46.
Likely need to change to adapt to panel meters: R18, R17, R12, R11    Joke: R27, R34    Likely: R27=0ohm, R34=no populate
Critical: R15 (with R37 decides the voltage range on the output), R23 and R24 1% (adapt CREF with R2=0.1ohm shunt voltage drop; in this case 300mV shunt Vdrop=3A when setting CREF=-6V), R22 (probably should match R23).

All the diodes except for CR3 and the bridges can be 1N4148; don't need 180V PIV or >1A diodes here.
CR3 can be a 3A diode like 1N540x where x=1 or greater.
CR9, CR13 bridges can be standard all-in-one 1A 100V bridges or be made from standard 1A diodes like 1N400x where x=2 or greater.
CR2 bridge should be 6A, 100V or better.

I don't know how that rascally R-16 escaped but thanks. I will undo the R12 change. The component advice will be helpful too. My schedule is been busy right now and my time limited but next week I will be back at it 100%. Choosing a display is something I need to do to move on.

Thanks
Therm

[Therm Mr.]

andrewtaylor That bourns is quite interesting. Thanks for posting.

xavier60 Yes I will be testing on breadboard. If the smoke is going to fly, I want it to happen when its easy to fix.

Kleinstein as always your help is appreciated.

Thanks everyone

Therm

[Therm Mr.]

I have made the change-backs you recommended pqass.

I do plan on dip sockets because it will absolutely prevent burning up those chips soldering and provide changeability. I am pretty sure I could solder the chips without a problem.  If a different footprint chip that is direct solder were better I am open to that.

Kleinstein why the TL062?

Regarding the ICL 7107 based panel meters is there a particular spec or version I am looking for? Like 0-15v 0-3a or is that not a factor? The expense is less of a concern to me than quality. I don't expect this project to be a bargain.

R19 calls for a 10k 5% 10-turn Cermet and R37 calls for a 50k 5% 10-turn WW. I found both relatively easily but there is a spec for linearity what is that?  Is that a factor I need to be concerned with.

As mentioned before in this thread (by pqass I think), the HP has a very large aluminum heatsink that is basically the whole back panel of the device. The Darlingtons were mounted on a small porches on each end of the heat-sink. I know there will be a lot of heat to reject. I plan on mounting the Darlingtons on small aluminum heat-sinks maybe even fan cooled if that's even possible (I don't know). At minimum I planned on a cased mounted fan to circulated air through the cabinet. Maybe even get fancy a use a thermistor circuit to control a fan. I would have to research that a bit. I don't want to have a heat issue. Is this over kill?

What do you guys think?

Thanks

Therm

[Kleinstein]

The TL062 is an easy to get part near identical to the LF442. From the performance it is not the best however. For better stability with the current I would consider a OPA2202, however like many newer chips it is only available in SMD and would need an adapter to use on a breadboard or in a DIP socket.  If it has to be DIP, maybe an RC4558 could be an option.

For the ICL7107 / TC7107 there should not be much difference between versions / manufacturers - they should all be OK. The display is 1999 counts. So a voltage range like 24 V or 30 V would not be ideal, as less of the range is used.  3 A is not ideal for the display, as the resolution would be 10 mA. Getting a 2nd range is however also a bit tricky. It is more effort (extra BCD-7 seg driver), but the ICL7135 version may be an option too.

For the trimmers the accuracy is not that important. One usually reads the voltage from the display and less from the scale at the pot. A highly linear 10 turn pot would mainly make sense if used with a dial indicator to read the set value.

One would definitely need a large heat sink, especially if no tap switching is used. For 3 A and 15 V range the raw DC voltage would be at some 20-25 V or a bit more. This would be some 60-75 W to dissipate worst case. Small heat sinks would need a larger fan. The safe way is a sizable heat sink and a fan as back-up.

[enut11]

Hi @Therm Mr.
You picked a difficult design to start with. The reason being, it is generic for a range of Agilent power supplies from 8vDC to 120vDC. The circuit diagram tries to cover all models so is difficult to follow for one specific model, eg E3610A, because a lot of the parts are redundant.

You may have this already but there is a table that lists the different components between the models.

I may be interested in building the E3610A version too so I will be following your progress.
enut11

[Therm Mr.]

Hi @Therm Mr.
You picked a difficult design to start with. The reason being, it is generic for a range of Agilent power supplies from 8vDC to 120vDC. The circuit diagram tries to cover all models so is difficult to follow for one specific model, eg E3610A, because a lot of the parts are redundant.

You may have this already but there is a table that lists the different components between the models.

I may be interested in building the E3610A version too so I will be following your progress.
enut11

Thanks for the sheet. I know the design is difficult that partly why I chose it.

[Therm Mr.]

I ran the rules checker due to my thinking I was in fairly good shape. I was wrong it found 13 errors and 77 warnings. some of the warnings were edits, like switches and the power plug. Also I had U5 LM393 and U4 warnings. I had used dual op-amp instead of just LM393 it was split between 393 and TLZ something. So I was getting pin errors and also unit C not placed error?

So Trying to fix this I remembered my A, B lesson from @pqass. I deleted all and re-inserted from the library. Then I found that I was not getting the 4 and 8 pins. And I discovered they were option C. So I added the pins with the C option to the correct half of the op-amp and thought I was good. re-ran the checker still getting errors. Different ones. They are "Duplicate items UC" on U4B and U4A "unplaced units" [UnitC]. U5A says the same unplaced units error.

Also many Warnings: Symbol pin or wire end off connection grid. I'll include a screen shot. There are other things I need to look at but I figured I would start here.


Thanks
Therm

[pqass]

So Trying to fix this I remembered my A, B lesson from @pqass. I deleted all and re-inserted from the library. Then I found that I was not getting the 4 and 8 pins. And I discovered they were option C. So I added the pins with the C option to the correct half of the op-amp and thought I was good. re-ran the checker still getting errors. Different ones. They are "Duplicate items UC" on U4B and U4A "unplaced units" [UnitC]. U5A says the same unplaced units error.

Make sure you've deleted all units first (A, B, C-power pair) from previous Adds.
Then Add back "LM358" under "Amplifier_Operational", placing all three (A, B, C) units.  Similarly for "LM393" under "Comparator".
Place power symbols on C units and change the reference to replace "?" with the appropriate number.
The C units don't have to be placed over-top an A or B unit.  Some people put them off to a corner with bypass caps.

Also, make sure you don't have any "?" remaining in a reference anywhere.  Use Cntrl-F to find "?".

Then run DRC.

[pqass]

Regarding the ICL 7107 based panel meters is there a particular spec or version I am looking for? Like 0-15v 0-3a or is that not a factor? The expense is less of a concern to me than quality. I don't expect this project to be a bargain.
...
As mentioned before in this thread (by pqass I think), the HP has a very large aluminum heatsink that is basically the whole back panel of the device. The Darlingtons were mounted on a small porches on each end of the heat-sink. I know there will be a lot of heat to reject. I plan on mounting the Darlingtons on small aluminum heat-sinks maybe even fan cooled if that's even possible (I don't know). At minimum I planned on a cased mounted fan to circulated air through the cabinet. Maybe even get fancy a use a thermistor circuit to control a fan. I would have to research that a bit. I don't want to have a heat issue. Is this over kill?

I'm searching Google Shopping or Amazon for "digital panel meter".  There are so many that aren't useful to your application. Like:
- 2 wire only; only good for 2.5VDC to 100VDC
- AC input
- ammeters with built-in shunt or AC current transformer (torroid)
- combination voltmeter and ammeter
- frequency meter, tachometer, counter

For those that have a power pair (5VDC) AND a sense wire (single or pair),  no information is provided on how to configure it or about power isolation with regards to the sense inputs.  Ideally, you need a meter with a separate power pair AND 0-200mV (full-scale) sense inputs (with IN-LO/-Vsense that can be tied to supply ground). Most people won't know how to configure the resistor divider for their application, hence, you'll see 5, 20, 50, 100, 200, 500V in the description.

You can find exactly what you need for name brand meters like the Murata DMS-20PC series but then you're looking at >$60!

As I'd said earlier, an ammeter is just a voltmeter that measures across a shunt so you can buy the same panel meter to do both power supply output voltage monitoring and ammeter service.  We don't want to add another shunt (for reasons see my reply#36) as we already have R2.

However, hope may not be lost if we take a bit of a gamble. 
If you search for "digital panel meter 5135" you'll see the attached. They are sold by many suppliers like this one.  According to the ICL7107 datasheet, Figures 13, 14, IN-LO/-Vsense can be connected to GND if it has an external reference (TL431 seen in another photo) and -5V supply is provided, both of which may be true in the photo.

To use this meter, you'd just need a 7805 similarly configured like the 7812 from the same transformer's unregulated positive in the op-amp supply; no need for another isolated 5V display supply.   DO NOT USE the +5VREF label to power the panel meters.

For the ammeter configuration up to 1.999A, R12=0R (a short), R11=open/no populate.  For one up to 19.99A, R12=10K, R11=100K.  Short the "2" or "20" option on the panel meter to enable the correct decimal point (0-1.999A vs 0-19.99A).  IN-HI of the meter is on the common of SW1A switch, IN-LO on [+S]/GND/+OUTPUT post.  HOWEVER, R20 wiper is a negative voltage!  To turn it positive, you'll need an inverting opamp.  See attached panel meter wiring schematic. 

The voltage monitor panel meter can also use the same op-amp power supply's new 7805 positive output and GND.  With IN-LO tied to [+S]/GND/+OUTPUT post and IN-HI tied to the R18=10K/R17=900K divider. HOWEVER, it too will report a negative voltage, thus, another inverting opamp will be needed. See attached panel meter wiring schematic.  Short the "20" option on the panel meter for 0-19.99V. 


WRT heatsinks...
CPU heatsink+fans can transfer lots of heat in a compact space.
Attach on back of enclosure or at least exhausting the fan out the back.

[Therm Mr.]

Make sure you've deleted all units first (A, B, C-power pair) from previous Adds.
Then Add back "LM358" under "Amplifier_Operational", placing all three (A, B, C) units.  Similarly for "LM393" under "Comparator".
Place power symbols on C units and change the reference to replace "?" with the appropriate number.
The C units don't have to be placed over-top an A or B unit.  Some people put them off to a corner with bypass caps.

Also, make sure you don't have any "?" remaining in a reference anywhere.  Use Cntrl-F to find "?".

Then run DRC.

Done and worked. So my mistake was deleting the reference to C. Thanks easy fix.

You can find exactly what you need for name brand meters like the Murata DMS-20PC series but then you're looking at >$60!

I saw these and was going to ask about them. I am ok with the $60 just knowing they are quality and will work no problem. Thanks for this too.

WRT heatsinks...
CPU heatsink+fans can transfer lots of heat in a compact space.
Attach on back of enclosure or at least exhausting the fan out the back.

Would this be enough with another fan venting the case? Or could I just configure this to vent out the case?
https://www.ohmite.com/assets/docs/sink_c40.pdf?r=false

pqass Thanks Again

Therm