Linear power supply project

[pqass]

See digikey here for the DMS-20PC-0 (200mV) version; just pick your color. 
See Figure 2 arrangement here with external R12/R11 or R18/R17 divider. 
You'll still need the inverters and wiring configuration as per my last post.

EDIT: If you're going to use the Murata panel meters, they state "Devices are fully calibrated at the factory...and never require calibration or adjustment."   This is a problem since the values or R12/R11 and R18/R17 are pretty loose and are relying on the panel meter pot to calibrate.  Since this isn't possible with the Murata meters, you'll have to make R12/R11 10:1 and R18/R17 100:1, exactly.  The only case where this isn't necessary is if R12=0R and R11=open/no populate.

EDIT2: Okay, you can still go loose with the R12/R11 and R18/R17,  if you add pots between the two resistors or between the 10Ks in the inverters.  See attached.  And the CREF4METER calibration pot isn't really necessary as you can tweak R20.  But I earlier forgot that the CREF4METER has to be divided by 2 (see parallel 10K).

If you form a tunnel (with two half-heatsinks; see your link), the fan can pull inside air through the tunnel out the back. The fan mounted between tunnel and inside back of your enclosure.

If you are only using one of those heatsink halves, you can cover the tops of the fins with an Al plate, with fan mounted in the center of the plate.  Then air can enter from either side, turn 90° through the center and out the back of the enclosure.

I don't think you'll need another fan as the bulk of the heat is from the series pass transistors.
You'll need openings on either side or top/bottom of the enclosure near the front for air intake.

[enut11]

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

@Therm Mr.
I have used this simple 2-transistor circuit to provide temperature based cooling in a power supply. Choice of active components is not critical.
Use a 10K thermistor at J9 and 12v fan at J10. Adjust RV5 to suit your fan and the temperature where it starts.
Thermistor is mounted close to a power transistor on the heat sink.
enut11

[enut11]

Meters: eBay item number: 355659784116

This compact 4 digit 10A panel meter has a 1mA resolution. Basic accuracy is 0.5% but I see an IR pot on the back which suggests it can be trimmed.
Shunts in these meters are normally in milliohms so have minimal impact on voltage regulation.
I have not looked but there is likely a matching 4d voltmeter.
Cost is ~$17 AUD
enut11

[Kleinstein]

There are many cheap meter modules available. The problem is that they in most cases have the measured voltage / current linked to it's supply. In this case they will likely need a separate transformer or similar. There may be a way to shift the signal with a difference amplifier and use a voltage reading module with the same supply as the regulator. One just has to find a suitable module.

Even if only a few 10 mOhm, one usually does not want an extra shunt outside the regulation loop. If anyway using a separate supply one could as well place it in the loop.

[Therm Mr.]

I have eliminated most of the warnings. But the power Outputs are actually inputs in the symbol library. I can find no Output symbol.

@pqass My Output neg gives a "not driven by any output power pins" warning.

Output+ "Both S+ and +VDC are attached to same items; S+ will be used in the netlist".

Also it does not like ground and S+ attached to same line.

Can I just ignore these?

Also A whole bunch of Warnings "symbol or pin not connected or wire end off grid". But they are connected they just not one continuous line they overlap on the ends without a connection dot.

Can I ignore these also?

Thanks

Therm

[Therm Mr.]

There are many cheap meter modules available. The problem is that they in most cases have the measured voltage / current linked to it's supply. In this case they will likely need a separate transformer or similar. There may be a way to shift the signal with a difference amplifier and use a voltage reading module with the same supply as the regulator. One just has to find a suitable module.

Even if only a few 10 mOhm, one usually does not want an extra shunt outside the regulation loop. If anyway using a separate supply one could as well place it in the loop.

I plan on the display supply having it's own transformer. Something like a small 24v. I am still weak in my knowledge of the display section and all its little details. But I'm learning.

Thanks
Therm

[Therm Mr.]

I have used this simple 2-transistor circuit to provide temperature based cooling in a power supply. Choice of active components is not critical.
Use a 10K thermistor at J9 and 12v fan at J10. Adjust RV5 to suit your fan and the temperature where it starts.
Thermistor is mounted close to a power transistor on the heat sink.
enut11

Thanks I like that

Therm

[Therm Mr.]

See digikey here for the DMS-20PC-0 (200mV) version; just pick your color. 
See Figure 2 arrangement here with external R12/R11 or R18/R17 divider. 
You'll still need the inverters and wiring configuration as per my last post.

I will look at that wiring soon and try to grasp it. I love having this thread as a reference to refer back to.

EDIT: If you're going to use the Murata panel meters, they state "Devices are fully calibrated at the factory...and never require calibration or adjustment."   This is a problem since the values or R12/R11 and R18/R17 are pretty loose and are relying on the panel meter pot to calibrate.  Since this isn't possible with the Murata meters, you'll have to make R12/R11 10:1 and R18/R17 100:1, exactly.  The only case where this isn't necessary is if R12=0R and R11=open/no populate.

EDIT2: Okay, you can still go loose with the R12/R11 and R18/R17,  if you add pots between the two resistors or between the 10Ks in the inverters.  See attached.  And the CREF4METER calibration pot isn't really necessary as you can tweak R20.  But I earlier forgot that the CREF4METER has to be divided by 2 (see parallel 10K).

This is going to take a bit for me to completely comprehend. But I will eventually get it it. Thanks for the education.

If you form a tunnel (with two half-heatsinks; see your link), the fan can pull inside air through the tunnel out the back. The fan mounted between tunnel and inside back of your enclosure.

If you are only using one of those heatsink halves, you can cover the tops of the fins with an Al plate, with fan mounted in the center of the plate.  Then air can enter from either side, turn 90° through the center and out the back of the enclosure.

I don't think you'll need another fan as the bulk of the heat is from the series pass transistors.
You'll need openings on either side or top/bottom of the enclosure near the front for air intake.

I intended to use two one for each Darlington. Better too cool than not cool enough. Also I will put some vents in the case. You can't push air out that you don't pull in. I will configure the fan so it is mounted at the back-plate of the case. Will the PCB layout portion of KiCad allow me to place those heatsinks without interfering with the trace?

Thanks

Therm

PS one last thing. The parts list lists resistor types PWI, PWN, MO and TF. I know TF as thin film and MO metal oxide. The others have me a bit perplexed. Can anyone offer any clarity on this.

Also C2 is listed as a 1000uf and a 10000uf. I am thinking the 10000 is the right choice am I correct?

[Kleinstein]

I have eliminated most of the warnings. But the power Outputs are actually inputs in the symbol library. I can find no Output symbol.

@pqass My Output neg gives a "not driven by any output power pins" warning.

Output+ "Both S+ and +VDC are attached to same items; S+ will be used in the netlist".

Also it does not like ground and S+ attached to same line.

Can I just ignore these?

Also A whole bunch of Warnings "symbol or pin not connected or wire end off grid". But they are connected they just not one continuous line they overlap on the ends without a connection dot.

Can I ignore these also?

Thanks

Therm

There should be only one name for +S, GND and +output.  If additional names are wanted use just text elements (e.g. for + output).

The wires of grid are a bit anoying. This may happen from diagonal movements. To clean up it is often easiest to remove the effecte wire and redraw.
The wire not connected part should be fixed if later a PCB is planed. Sometime just move the part to the side and back to where is was to get it connected.

Chances are that PCB mounted heat sinks are too small for the main regulator (the 2 darlingtons / MOSFETs). They can be OK for auxiliary regulators (e.g. +-12 V), the display +5 V and the rectifier.
The main heat sink(s) are usually too large and more a PCB mounted to the hea sink, using the whole back side of the PCB and the power parts moutned below, with the heat tap down and a mount hole in the PCB.

[pqass]

@pqass My Output neg gives a "not driven by any output power pins" warning.
Output+ "Both S+ and +VDC are attached to same items; S+ will be used in the netlist".

+VDC and -VDC symbols are from the "Power" library so they have their pin defined as "Power input". That's why there's a complaint about needing the "PWR_FLAG". ie. a complaint that a pin needs power but no source was identified.  If you really like the large + and - symbols that much, you can make a copy of them in your own library and change their Electrical Type to Passive.

You can ignore the "Both S+ and +VDC..." warning; it'll just pick one for a net name.  It's required to attach GND (from the op amp power supply) to the +OUTPUT post otherwise the circuit won't work.

I plan on the display supply having it's own transformer. Something like a small 24v. I am still weak in my knowledge of the display section and all its little details. But I'm learning.

But it isn't necessary.  You could just hang a 7805  from the same transformer as the 7812.
The DMS-20PC models require a +5VDC regulated power source; 24V to 5V is quite a drop for a 5V regulator.

Will the PCB layout portion of KiCad allow me to place those heatsinks without interfering with the trace?

It's called a keepout area.  After placing, adjust its Properties to also keepout copper fill and footprints.

PS one last thing. The parts list lists resistor types PWI, PWN, MO and TF. I know TF as thin film and MO metal oxide. The others have me a bit perplexed. Can anyone offer any clarity on this.

My guess:  PWI=power inductive, PWN=power non-inductive, MO=metal oxide, TF=thin film

Also C2 is listed as a 1000uf and a 10000uf. I am thinking the 10000 is the right choice am I correct?

It's 10,000uF.  I saw that error too.


Also,
I've updated the panel meter wiring schematic for 200mV full-scale panel meters. See attached.
Changes from the last version are:
- R17 increased to an E12 series value,
- R12/R11 values reflect monitoring for up to 3A,
- removed need for R20 and R40, as CREF can be used directly in the 200:1 inverting amp configuration,
- added calibration pots for CREF, CMETER, VMETER to adapt to 200mV full-scale panel meters that are factory calibrated; that don't have their own calibration pots,

[Kleinstein]

Dividing the signal for the current and voltage so far down is not ideal. This would make it sensitive to amplifier offset and drift, so that the LM324 is no longer acceptable. The linked display modulte from DK at least has a differential input, otherwise the ground shift form a common link for power and ground could be another problem.
The display inverter circuit could likely get away without the extra buffers at the input, so just 1 amplifier for voltage / current.

I would consider getting a cheap module from ebay (e.g. there are some with voltage (100V) and current combined), to use an external shunt / suitable range (e.g. 1 or 2 V for the voltage).

[pqass]

So today I decided to open my E3611A to find out what's on the display board. I just found the expected (pair of) TC7107CKW but also a TSC9491 1.22V voltage reference; and no other chips. Much to my surprise though, the 7107 was configured for 2V full-scale; not the 200mV that I'd assumed! I confirmed this by checking the voltage between 7107 V+ref and V-ref pins which showed 1.00V; not 100mV.

This makes sense given, in the E3610A, the 1:1 current divider R12=0R/R11=open and 10:1 voltage divider R18=100K/R17=900K.  With a 2V full-scale meter, 3A looks like "3.00", and 15V (max for the supply) looks like "15.00".  I'm assuming (uh oh) my E3611A display board is virtually identical to the E3610A or E3612A in all but which decimal digits are enabled.  My E3611A has a 100:1 voltage divider and thus has only 100mV voltmeter resolution.

Given what Kleinstein said about working with small voltages, we can avoid them by choosing 2V full-scale panel meters like the Murata DMS-20PC-1 (last digit changed).  However, if choosing a generic panel meter like the "5135" that I proposed, it likely won't be possible to change the reference voltage to 1V.  And the most you could do is modify it to 200mV full-scale (by removing the RA resistor). Actually, I'm not so sure anymore what the reference might be on some of these DC5V, 20V, 50V, etc panel meters. 

Furthermore, the set current (CREF R20 wiper) is negative vs actual current (CMETER) but the way the E361xA series power supplies skirts this is by dropping the "+1" left digit, which you won't be able to do with a purchased panel meter.  Also, I believe the only reason negative voltmeter readings aren't shown is due to the floating supply allowing IN-LO and IN-HI being tied to R18/R17 center point and +S/GND/+OUTPUT post, respectively.  If you drop the isolated display supply requirement, I believe an inverting amp is necessary since meter ground will be tied to +S/GND/+OUTPUT post.

Unfortunately, with a 2V full-scale panel meter, the CMETER divider would be R12=0R/R11=open/no populate, which means we won't have an opportunity to fit a calibration pot. As such R2 (0.1R) would have to be very accurate or at the least slightly higher, which would at least give us the opportunity to put a high-value resistor in parallel to adjust it down to exactly 0.1R.

The brain hurts from too many trade-offs to consider.

@Kleinstein
It's true that the LM324 unit A buffer isn't needed as CREF is already driven by opamp U4A, thus CREF can be tied to the first resistor in the inverting amp.  However, I believe VMETER (R18/R17 divider center) requires a buffer (unit C) into its inverting amp, otherwise the transformation won't be correct.

Even-though the LM324 (powered by a +/-12V supply) would be working with small voltages (<30mV), the op amp inputs never come close to the rails.  If there is (crossover) distortion close to 0V maybe another op amp like the TL074 would be more appropriate.


Attached you will find another wiring revision for 2V full-scale panel meters (vs for 200mV FS meters in my previous post).
Changes are:
- R12=0R/R11=open/no populate, the entire R2 Vdrop goes directly to the panel meter.
- Addition of a high-value resistor in parallel with R2 for the option to lower R2 to exactly 0.1R; can't increase it.
- CREF inverting amp is now at a modest 20:1; changed to 100K (left resistor).
- R18=100K/R17=1M is now at a modest 11:1; changed R18.
- VMETER inverting amp remains 1:1 except for the calibration pot raised to 2.2K.

[andrewtaylor]

There are many cheap meter modules available. The problem is that they in most cases have the measured voltage / current linked to it's supply. In this case they will likely need a separate transformer or similar. There may be a way to shift the signal with a difference amplifier and use a voltage reading module with the same supply as the regulator. One just has to find a suitable module.

Yes, that's correct.

One point not mentioned yet:
Most of the digital panel meter cause a lot of EMC noise.
I don't want this in my lab bench supplies, That's why I prefer analogue (moving coil DC) meters.

No noise, no EMC -- and no need for a seperate power supply to power the meter.

When I really need to read out the voltage with higher precision, I use external HP 34401 or Fluke 87 V, whatever is handy.

[Kleinstein]

For the EMI / supply interference it depends on the panel meters. Some can be rather bad, with a switched capacitor circuit to generate a -5 V supply. The other point is a multiplexed LED display with some (not with an ICL7107).


The problem with using the LM324 with small voltages it not gettting close to zero or cross over. The problem is the offset voltage ( ~ 5 mV ) and it's drift.  It gets better with a 2 V range, or if only 200 mV available a divider just at the input to the module. For the current one may have to make sure to have the shunt close to the output, not to have too much cabeling included with the shunt.
I don't think the current display would want the inverter. One could consider a difference amplifier to allow a proper connection to the shunt.

The voltage sense can use the inverter directly from the neg. side, without an extra voltage divider. So as an active divider, separate from the feed back the an regulator.

[Therm Mr.]

I am really wanting to move on to the panel meter but this rules checker is getting the better of me. I have been able to eliminate most errors but the warnings will not go away no matter what I do. It tells me there are no connections where there are connections. What am I doing wrong?  See Pic

Also S1 has an A and B and so does S2 and this seems to produce errors. Though I see no way to split the switches like the op-amps.

I am really enjoying the discussion on the panel meters. It seems like it may even be the hardest most complex part of the project.

@andrewtaylor I have a fluke 87 also. I have had it for decades great meter.

[pqass]

See attachment regarding these issues.
A: R41 connected to the wrong wire.  It should be connected to U5A pin 2.
B: "stray hair"; unconnected wire or symbol with no wires attached.   Remove it or terminate with the "no connection" flag (see icon bar on right).
C: "SW1" appears twice as a symbol reference. One should be SW1A and the other SW1B to signify there are two ganged switches.
D: PWR_FLAG is not a connector. You need a connector symbol here. But don't choose from the Power library (like the +VDC/-VDC from before) as they have Power Input electrical type and demand a PWR_FLAG on the wire.
E: Wire is connecting to the line graphic. It should be connected to the green wire below it, where R12/R11 meet.
F: Delete the wire from junction to VMETER and re-add it.  There may be a broken connection along the path.

Also, not identified with a letter is T2 pin 2, loose junction and/or wire.  Delete whatever is there and re-Add.

[pqass]

The problem with using the LM324 with small voltages it not gettting close to zero or cross over. The problem is the offset voltage ( ~ 5 mV ) and it's drift.  It gets better with a 2 V range, or if only 200 mV available a divider just at the input to the module. For the current one may have to make sure to have the shunt close to the output, not to have too much cabeling included with the shunt.
I don't think the current display would want the inverter. One could consider a difference amplifier to allow a proper connection to the shunt.

The voltage sense can use the inverter directly from the neg. side, without an extra voltage divider. So as an active divider, separate from the feed back the an regulator.

But wouldn't any offset be compensated by the calibration pot (in the inverting amp)? 
With a 2V full-scale meter each right-most digit is a mV.  With a 200mV full-scale meter each right-most digit is a 100uV. 
So rather than the LM324, do you have any recommendations on an op amp with low (temperature induced) drift?

Actually, the CREF inverting amp is not strictly needed.  If the user is okay with seeing a negative current value when the CC Set button is pressed vs. when actual current is being displayed. The J1 connector has both + and - pins for panel meter IN-HI and IN-LO so they can be twisted and attached close to the source; shunt divider or R20 pot, if not using an inverting amp.

My reasoning for using an inverting amp on VMETER is because I don't think you can tie IN-HI to +S/GND/+OUTPUT post and IN-LO to the R18/R17 divider center if the panel meter power isn't floating. That is, can't do that if power meter GND supply is tied to the +S/GND/+OUTPUT post.  So if I'm wrong here, and the user is okay with seeing negative currents when CC Set button is pressed, then no inverting amps are needed at all.

The regulator feedback dividers are actually R15/R37 (for voltage) and R24/R23 (for current).  R18/R17 and R12/R11 are just for the panel meters.

[Kleinstein]

The pots are for the scale factor, not for the offset. A low offset variant of the LM358 is the LT1013. The classic low offset amplifier is the OP07.  If SMD (e.g. SO8) is not a problem, I would use the newer OPA202 / OPA2202.

Even for the voltage range (2 V or even 5 V with a divider) at the OP-amps output the offset of an LM324 could still be visible, though not much. One could add an extra trimmer for the offset to an inverting amplifier (works as a summing amplifier anyway).

The meter would likely be OK with the in+ tied to ground and in- as an input, at least for the 200 mV range. This at least works for the ICL7107 in 200 mV range and maybe also 2 V range with a -5 V supply.

I would consider getting one of the cheap ( ~ $8-1) voltmeter units from eaby and test if they can be changed to a 5 V or similar range so that it could also be used for the amps part with a simple extrnal amplifier. Chances are reasonable good, as there should be an input divider to set the 100 V range and the ADC itself is more like 1 V or 2 V.

[Therm Mr.]

See attachment regarding these issues.
A: R41 connected to the wrong wire.  It should be connected to U5A pin 2.
B: "stray hair"; unconnected wire or symbol with no wires attached.   Remove it or terminate with the "no connection" flag (see icon bar on right).
C: "SW1" appears twice as a symbol reference. One should be SW1A and the other SW1B to signify there are two ganged switches.
D: PWR_FLAG is not a connector. You need a connector symbol here. But don't choose from the Power library (like the +VDC/-VDC from before) as they have Power Input electrical type and demand a PWR_FLAG on the wire.
E: Wire is connecting to the line graphic. It should be connected to the green wire below it, where R12/R11 meet.
F: Delete the wire from junction to VMETER and re-add it.  There may be a broken connection along the path.

Also, not identified with a letter is T2 pin 2, loose junction and/or wire.  Delete whatever is there and re-Add.

Thanks for pointing all that out. I think I took care of most of it. SW2 has me completely confused. In CC set it shows as 1A and 2B? What am I missing here?

The Rules checker only sees three errors now the rest are warnings.

The errors are "SW1 not annotated" but it is. I even deleted it and put back in no change.

The plug "wall plug earth has been modified in library" error.

The "S+ and ground attached to same items" error by the CC set circuit. If I eliminate the ground I get an warning on U6. @Kleinstein how do I make them all the same as you suggested without causing a problem. Can I just make all s+ grounds?

I corrected lots of warnings by replacing the problem components. Most of the rest (66 of them) are "symbol pin or wire off connection grid' warnings. I redraw the wires and still get the same warnings. Will these warnings cause problems with assigning of footprints or the PCB editor?

I think I am getting real close to having a viable plan for the main board. Though 66 warnings makes me wonder. The display section looks like another worthy challenge.

Edit: add drawing

Therm

[pqass]

With regards to the "S+ and ground attached to same items...".
This is not really an error. It's just telling you which name it will use in the netlist.
In the ERC popup, if you select the error line, mouse right-click, then you can "Change severity to Warning ...".

[Therm Mr.]

With regards to the "S+ and ground attached to same items...".
This is not really an error. It's just telling you which name it will use in the netlist.
In the ERC popup, if you select the error line, mouse right-click, then you can "Change severity to Warning ...".

Thank you

I have posted the original drawing and circled SW1 in both locations. In the current section is shows as S1A and S2B Range is this a mistake? How does the S2 Range switch have anything to do with with this switch? When S1A opens the current Pot adjustment knob becomes active for current ajustment. Am I not comprehending something here?

[andrewtaylor]

You are mixing up some annotation.

S2 A and B change the range of the supply:
S2A lower left the transformer tapping, and S2B the max. possible current range.

S1A has nothing to do with it, because it is just for SET/ACTUAL current flowing.
So the pot belonging to S2B ist 0.2...100% of the current range.
the "100%" == full clockwise changes with the S2B setting.
100% is 2A in the low amperage (= higher voltage) and 3A (= 8V) range:


I hope now it more clear to you

[Therm Mr.]

@andrewtaylor Thank you.

@pqass I went back through to find where you had posted a sequence I should follow and re-read it. So the drawing I think is pretty much done. Its possible I still find errors but I think I'm good.

I realized I was out of sequence when I ran the electrical rules checker first. So I ran the annotation tool and it found the sw1 problem. Then I re-ran the rules checker and the problem was gone. I also made the ground error a warning as you suggested.

So now I am ready to run the footprint assignment tool. I still need to add the fan power and control circuit and any mods that may be needed for the display. Should I wait till this is done before starting to assign footprints? I am going to read back through the display discussion likely several times so I can understand it better before my next step forward.

I have been creating some parts list carts at DigiKey and Mouser. I may even need to add in Newark and Texas Instuments to get it all. I am ordering multiples of most everything figuring It won't hurt to have some spares for future experiments and projects.

@Kleinstein you mentioned some TL062 chips. I would like to order some of those as well as the LM358's I already have in my cart. I would like to be able to interchange like pqass suggested by using sockets. My question is there are many different suffixes on them any particular suffix I should choose? Also TL431LP's also have many suffixes I am guessing the TO-92 package does the suffix matter?

Thanks

Therm

[Kleinstein]

The suffixes with the TL062 are mainly for the case and temperature range. For a DIY unit the temperature range should not matter that much. I mentioned it, as a near 1:1 replcement for old obsolete LF442, not because it is an expecially good performance, more a cheap easy to get part.

The Tl431 versions have different TC limits. For the first test likely not an issue. For the PCB the details don't matter - the pinout is usually the same. It is a bit odd that Kicad has often so many essentially identical variants.

[Therm Mr.]

The suffixes with the TL062 are mainly for the case and temperature range. For a DIY unit the temperature range should not matter that much. I mentioned it, as a near 1:1 replcement for old obsolete LF442, not because it is an expecially good performance, more a cheap easy to get part.

The Tl431 versions have different TC limits. For the first test likely not an issue. For the PCB the details don't matter - the pinout is usually the same. It is a bit odd that Kicad has often so many essentially identical variants.

Thank you

Therm