Help understand and repair LM317 High voltage/ current circuit

[Bratster]

Hello, I have the attached circuit in a piece of custom test equipment I have been asked to repair.

It's supposed to supply 125VDC to some solenoid valves, rated at 35 watts each.

It was built 6 years ago, but I do not know how much of that time it was actually used.



Power is fed from a 120V isolation transformer. the DC out goes to some switches to connect to the valves.

It looks like it is basically a copy of the examples given for the LM317 regulator.

I have not connected a scope yet, will get to that tomorrow.


Here are the voltage readings:

Using incandescent lamps for the load.

No load:
AC in 138V, Rectified DC 176V, Regulated DC 126V.

25 Watt load:
AC in 129V, Rectified DC 153V, Regulated DC drops to 58V when lamp turned on, then rises to 103V.

40 Watt load:
AC in 129, Rectified DC 153V, Regulated DC drops to 10V when lamp turns on, then rises to 97V.


I don't fully understand how the circuit works, so I am not sure where to start looking. Nothing is visually bad on the board.

It seems like it is current limiting, but I don't know why. Supposedly it worked fine and could run 3 of the valves, so 105 Watts, 840mA.

I know the main transistor is there to handle the majority of the regulation, with zener D1 to keep the voltage difference the LM317 sees within its rating.

Any help would be appreciated.
Thanks

[Bratster]

Pics of the unit.

Sent from my Fi Moto x4 using Tapatalk

[floobydust]

The circuit is straight out of the LM317 datasheet, although that example is lower current 25mA version.
More here: High Voltage Adjustable Power Supplies National Semiconductor Linear Brief LB-47 Michael Maida March 1980 describes the 'floating' regulator a bit.

If the regulator drives inductive loads, it's missing a protection diode at its output, or across each solenoid, for long life.
Also missing is short-circuit protection. I think the power ratings are a bit overdone, the isolation transformer looks sized for ~50VA.

I would measure the LM317 voltages on the 3 pins when the power supply has collapsed (insufficient output current). That would give some idea what is weak.

[PKTKS]

That is a very good (or bad)  example how things taken to the limit
may fail.. or will fail  sooner or later.

LM317 and 78xx or 340  they all have a typical internal limited power rate
Roughly about 5Watt is safe  10W is overkill and a proper dissipator is required.

Even advertised supporting 1/2 Amp safely  if the power rate surges
the regulator fails (several reasons to happen that)

Floating the reference pin with a potentiometer is tempting.
However POTS DO FAIL and chances are if the contacts of that
POT fail the regulator was exposed to a differential higher than
the claimed 30V max

The POT should be carefully tested and cleaned and my guess that the
regulator failed due to over-voltage  spike

That arrange is on the limit and the POT is really a weak point

Paul

[commongrounder]

Th OPs circuit is fixed output at 125 volts, and has no pot in the circuit. That weakness can be dealt with by wiring the pot in rheostat mode, with the wiper tied to one end of resistive track.  That way if wiper contact is lost, it won’t go open circuit, just to the full resistance of the pot.

[duak]

Incandescant light bulbs are really the wrong things to use as test loads because they have a much lower resistance when cold.  Measure one and see.  I suspect the current limit is caused by the LM317 overheating and shutting down.

I think it's kind of working the way it should.  I would locate a 1K0, >=15 W power resistor or network of resistors to use as load to see what happens.

Cheers,

[floobydust]

Sometimes I get things like this to repair and I find the original design was never working.

I did a hack simulation without the LM329B and pass-transistor spice models.
It didn't regulate well, sagged under load. Found it to be the Darlington pass-transistor is starved with 100k base-drive. Lowering to 33k (1W under short-circuit) fixed that.
MJH11022 hFE can be low.

SOA is ~0.25A at 150VDC and because this supply does not have short-circuit protection, it really should because field-wiring always seems to get abused or a solenoid coil burns up and shorts.
I agree, best not to use light bulbs as a load because their inrush current is up to 10X. You can measure a lamp's cold resistance to verify. A fuse might enough. Or add a few ohms on the emitter for current-limiting.

The LM317 current limit is a few amps with low input-output differential and will not help.

[Bratster]

It wouldn't surprise me that this wasn't done right from the beginning, although it would have had to have worked "good enough".

There are MOVs on the outputs, V150LA2P.

I will get some resistors to simulate the valves, unfortunately I can't get any of those.
Thinking some 450ohm 50W and 400ohm 50W.

I am going to get new transistors and regulators while I am ordering anyway. and a new main CAP since who knows how long it was run next to the heatsink.

Will get some 33k 1W resistors to test in place of the 100K for base drive.

For adding short-circuit protection, I see the example shows a resistor between the transistor and regulator, but unless I missed it how do you calculate the value?


Checked the ripple:
Rectified DC:
No load, 0.2VAC
1K, 2.2VAC
25W, 3.5VAC

Regulated DC:
No load, 0VAC
1K, 0VAC
25W, 1.15VAC


I did happen to have a 1K 30W.
1K load, Rectified DC: 163, Regulated DC: 112

\Edit: correct measurements below

Measured the LM317:
no load:
Pin 1, adj, 124VDC
pin 2, out, 125.3VDC
pin 3, in ,  128.3VDC

1k load:
Pin 1, adj, 116.5VDC
pin 2, out, 117.5VDC
pin 3, in ,  120VDC

/edit

Now for the annoying part, there is definitely something being affected by temperature.

Depending on how long I run it with different loads the readings I get change.
The test with the 1K actually passes just fine after it warms up.

The LM317 is not getting warm, or at least that I can tell by finger.
The transistor on the other hand is uncomfortable to touch its heatsink.


Thanks

[duak]

A couple o' things:

- the LM317 should have +1.2 V between output and adj.  If I'm reading your measurements correctly, something is amiss.  LM317s have a minimum output current spec of 3.5 mA typ to 10 mA max.  This is usually satisfied with a 240 ohm resistor between output and adj.  This circuit may not meet that.  What voltage do you measure across D3?

- I would consider adding a 1N4005 rectifier across the 125VDC Out to protect the circuit from the compliance voltage generated by the inductor when the current is switched off.  I expect the voltage to go negative so the cathode (band) of the rectifier should be connected to the + output.

Cheers,

[Bratster]

A couple o' things:

- the LM317 should have +1.2 V between output and adj.  If I'm reading your measurements correctly, something is amiss.  LM317s have a minimum output current spec of 3.5 mA typ to 10 mA max.  This is usually satisfied with a 240 ohm resistor between output and adj.  This circuit may not meet that.  What voltage do you measure across D3?

- I would consider adding a 1N4005 rectifier across the 125VDC Out to protect the circuit from the compliance voltage generated by the inductor when the current is switched off.  I expect the voltage to go negative so the cathode (band) of the rectifier should be connected to the + output.

Cheers,

I redid my measurements above, before I was power cycling so I could move probe clips around.
New measurements are with power staying on and probing.

note that the loaded measurements were constantly changing, I tried to get as close time wise as I could.

Measured the LM317:
no load:
Pin 1, adj, 124VDC
pin 2, out, 125.3VDC
pin 3, in ,  128.3VDC

1k load:
Pin 1, adj, 116.5VDC
pin 2, out, 117.5VDC
pin 3, in ,  120VDC

Measured across D3:
 7.1VDC no load and 1K
6.85VDC 25W
3.8VDC 40W

[floobydust]

The power supply is seems undersized for what it's apparently supposed to drive. A 35W solenoid valve is huge, usually they'll get peak-and-hold drive because the heat dissipation is too much given the physical size of the coil 0.28A at 125VDC. I would check the solenoid valve's datasheets. You don't even need a voltage regulator for solenoids.

With three, that's almost 1A load. The pass-transistor does all the work and normal for it to heat up really hot. As temperature goes up I think the transistor's hFE droops even though VBE drops, which is what you are seeing.

I think changing from 100k->33k will fix things.
But next the power transformer will probably be the one sagging, it appears to be small and ends up doing some current limiting.

The LM317 voltages are fine, PSU output is low because the pass-transistor is sagging first. The LM317 never sees more than ~4V across it.

For a current-limit resistor, the problem is the power transistor can't take more than 0.25A into a short, so how do you get 1A output with that as the limit?

The MOV is not helpful because the V150LA2P clamping voltage is 216-264V and with an inductive load it's the -ve spike that kills semiconductors. I would add a 1N4007 across the PSU output. The PSU needs a fuse too.

[Bratster]

The power supply is seems undersized for what it's apparently supposed to drive. A 35W solenoid valve is huge, usually they'll get peak-and-hold drive because the heat dissipation is too much given the physical size of the coil 0.28A at 125VDC. I would check the solenoid valve's datasheets. You don't even need a voltage regulator for solenoids.

With three, that's almost 1A load. The pass-transistor does all the work and normal for it to heat up really hot. As temperature goes up I think the transistor's hFE droops even though VBE drops, which is what you are seeing.

I think changing from 100k->33k will fix things.
But next the power transformer will probably be the one sagging, it appears to be small and ends up doing some current limiting.

The LM317 voltages are fine, PSU output is low because the pass-transistor is sagging first. The LM317 never sees more than ~4V across it.

For a current-limit resistor, the problem is the power transistor can't take more than 0.25A into a short, so how do you get 1A output with that as the limit?

The MOV is not helpful because the V150LA2P clamping voltage is 216-264V and with an inductive load it's the -ve spike that kills semiconductors. I would add a 1N4007 across the PSU output. The PSU needs a fuse too.

Yep, its looking like this thing was doomed from when it was made.

It is for temporary testing of the valves, not the usual method of control.

Unfortunately I don't have access to specifics about the valve.

Will add the output spike diode.

Mains AC goes through a 1A fuse before the switch and transformer.

[duak]

Floobydust is right about Q1 causing the voltage sag.  HFE tends to go up at high temperatures at least in part to collector to base current leakage.  I think that decreasing the value of the 100K resistor will help but Q1 will also dissipate more power and could fail.  A power resistor could be added in series with the collector of the transistor to help handle the excess power and could be mounted on the case.

However, without the valve solenoid and heat sink specs, it's hard to say what will happen and what should be done.

Cheers,

[Bratster]

Ok, well suddenly they are now up to 4 of these "failing", everybody that has looked at it agrees it was only designed to run one output at a time, barely.

No one knows how they went 6 years and didn't notice.



I am going to be basically redoing the whole thing.

New transformer rated for 150VA, new case to fit that...

I am looking at using a Texas Instruments TL783 regulator. Their app notes have the exact circuit I need.

Unfortunately the specified darlington transistor, TIP150 is only available from Newark UK.

Going to use the one isolation transformer, rectifier and filter cap, followed by 3 of the below circuits, one for each output, will add the spike protection diode to each.

Overkill, but should take whatever they throw at it.

Am I missing something obvious that will bite me?

Thank you all for your input.

[floobydust]

I don't see a big difference between the LM317HV and TL783 circuits here. Both are a linear HV regulator so heat is a big concern.
The TL783 circuit 120V zener just means most of the heat will instead be dissipated by the TL783 now. The TIP150 drops voltage when things are over 120V input-output. It's also a small, obsolete  transistor compared to MJH11022.
The LM317 circuit was the opposite with the 6V zener, it let the MJH11022 dissipate all the heat in dropping the voltage down.
Check the thermal calcs, 1A output and 140VDC input is still 15W of heat and a large heatsink for that to rise say 30-40C.

I still question the need for a voltage regulator with solenoid loads. They can take +/-20% variation and a decent sized power transformer will "regulate" close to that. What's wrong with using a 90VAC transformer and bridge rectifier only. I find 125VDC is only used in railroad and utility substation applications.

Note also there is a liability due to the shock hazard with the high voltage going outside the product.
I would increase the safety of the output connectors and field wiring. Make them touch safe at least.
Banana plugs might not be insulated well enough for lay people to mess with them.

[Bratster]

I don't see a big difference between the LM317HV and TL783 circuits here. Both are a linear HV regulator so heat is a big concern.
The TL783 circuit 120V zener just means most of the heat will instead be dissipated by the TL783 now. The TIP150 drops voltage when things are over 120V input-output. It's also a small, obsolete  transistor compared to MJH11022.
The LM317 circuit was the opposite with the 6V zener, it let the MJH11022 dissipate all the heat in dropping the voltage down.
Check the thermal calcs, 1A output and 140VDC input is still 15W of heat and a large heatsink for that to rise say 30-40C.

I still question the need for a voltage regulator with solenoid loads. They can take +/-20% variation and a decent sized power transformer will "regulate" close to that. What's wrong with using a 90VAC transformer and bridge rectifier only. I find 125VDC is only used in railroad and utility substation applications.

Note also there is a liability due to the shock hazard with the high voltage going outside the product.
I would increase the safety of the output connectors and field wiring. Make them touch safe at least.
Banana plugs might not be insulated well enough for lay people to mess with them.

I was thinking of using 3 copies of that circuit, so each should only see the 280mA for its dedicated output.

I'm not set on regulation or that circuit, I would be quite happy to not need "active" circuitry.
But I do want to limit the output voltage pretty close to 125-128VDC.



I am not having good luck finding 90VAC transformers in the 120-150VA range. Found a 100VA with 90V tap, Hammond 169C, but that would be running at a slight overload (I think?) with all outputs on.
I would really like to use an isolation transformer as opposed to a standard one, although I don't know how major the differences are.

I am also concerned with the output voltage running too high, the above transformer is spec'd for 115VAC in, our mains voltage is 120Vac generally, 125Vac in some areas.
After rectification and filter cap thats 127VDC (which is fine), from 115VAC. 90V*1.414.
Unless I am screwing up my math(probably) 125Vac is 8% higher, so the DC would be 137Vdc. I am somewhat uncomfortable with that.


Although I don't particularly care if the voltage is lower, currently the "working" unit drops to 100VDC, and they didn't notice apparently.
I don't want to go that low, but 115-120VDC may work.

So perhaps a 80-85V transformer, which seems even harder to find.

Or perhaps a second transformer in a buck configuration to trim the voltage to where I want using a easy to get main transformer.
Efficiency is not a priority.

After thinking some more I am liking the idea of: isolation transformer, with an additional buck transformer and then bridge rectifier/ filter cap.

[Bratster]

Current train of thought in image below.

Thanks to Floobydust for reminding me to "keep it simple, stupid"

[james_s]

If it doesn't need regulation, you could make it even simpler by winding a custom transformer. I've done a few and was surprised by how easy it is, I like the cut-C cores from Alphacore, the cores and bobbins are inexpensive and easy to wind.

[floobydust]

I think it's better (more robust) approach to consider. The original circuit using LM329 for 100ppm drift seemed interesting but not really needed.

You can also switch the output voltage low (minus 36V tap), as sometimes undervolting solenoids is a way to test their mechanical pull strength.

I would add an RC snubber after the transformers (bridge input) like 33R 1W and 47nF to tame the leakage inductance.
A push-button circuit breaker on the primary, as this will probably get abused.
Also a bleeder resistor or LED or something to discharge the filter cap in under 10 seconds.

My tests show 150VA transformers have around 7% regulation and 30VA around 12%. The smaller 36V transformer will make the no-load voltage high.

[Bratster]

Waiting on the new 28V transformer to double check but here is where i'm at. (attached image)

Using the 36V buck transformer dropped the voltage lower than I would like, and the 18V center tap was Higher than I wanted.

Hopefully the 28V should be just right.

Added a 2K 25W resistor to bleed down the HV quickly and also keep the voltage from going too high due to no load.
Added snubber circuit.
Switched from fuse to circuit breaker.

[james_s]

When I need an unusual transformer I wind my own, cores and bobbins and such are available here http://www.bridgeportmagnetics.com/index.html

it's really not particularly hard, although in smaller sizes the required turn count gets pretty high, but the bigger the core the fewer turns are required.

[Armadillo]

Hi, change the 1uf capacitors.