Question about potentiometer wiper failure in bench power supplies

[ferdieCX]

Hi, there is something that has puzzled me several years.
It is very common to see "scholar book" series regulators like the one in the picture. If the wiper looses contact for a moment, the load will receive the full input voltage and, may be, get destroyed.
The strange thing is, that it seems that many commercial bench power supplies could have this kind of failure.
Even the traditional HP linear regulator gives an output voltage proportional to the resistance of a potentiometer whose wiper is connected to one of the ends.
I wonder why the designers don't care about that.
After destroying several ICs about 30 years ago because such a failure, I got a Philips power supply that gives 0 V if the wiper fails.

[bsfeechannel]

It is very common to see "scholar book" series regulators like the one in the picture. If the wiper looses contact for a moment, the load will receive the full input voltage and, may be, get destroyed.

Tell me about it.

[David Hess]

It is very common to see "scholar book" series regulators like the one in the picture. If the wiper looses contact for a moment, the load will receive the full input voltage and, may be, get destroyed.

Experienced engineers avoid using potentiometers that way for the reason you identify.

Even the traditional HP linear regulator gives an output voltage proportional to the resistance of a potentiometer whose wiper is connected to one of the ends.

If the wiper is connected to one end, then an open wiper cannot cause the resistance to be higher than the total resistance of the potentiometer.

Tell me about it.

See above.

[Fraser]

I had three Thurlby linear Bench power supplies in my lab. Supposedly Lab grade equipment. All suffered the same horrible failure. The track on the voltage setting potentiometer would get worn in specific areas that are most used, such as 5V, 9V and 12V. One of them was set to 9V one day and suddenly jumped to the full 15V output for a moment and destroyed the power supply circuits in an expensive Multimeter/DSO that I was working on. It was then that I investigated what had happened and found all three of the Thurlby power supplies to have ‘issues’ with the voltage setting potentiometer. Having sourced the schematic, it was clear that the design is flawed and any intermittent contact of the potentiometers wiper causes full p.d. to appear at the output terminals. All three power supplies are currently sat in a box in my garage in disgrace. I was going to fit new potentiometers but have now moved to microprocessor controlled power supplies with no potentiometer wear issues to concern me.

Fraser

[David Hess]

One of them was set to 9V one day and suddenly jumped to the full 15V output for a moment and destroyed the power supply circuits in an expensive Multimeter/DSO that I was working on.

On the ancient Tektronix TM500 series power supplies that I have, they included instructions for changing the trip point of the output crowbar circuit to protect against accidentally setting the output voltage too high.  Another feature they have is that when tracking is enabled, the individual controls can be used to limit the outputs while the tracking control is used to set the output voltage but only within those limits.

I have seen old laboratory power supplies which included a screw adjustable potentiometer for setting the trip voltage of the crowbar on the output; the Power Designs TP340 is an example that has this.

[Ian.M]

Its not always possible to choose a circuit topology that avoids loss of regulation if the wiper contact fails. e.g. if a linear response is required over a wide voltage adjustment range without the complexity of digital control

When a potentiometer is being used as a two terminal variable resistor, it is however possible to sense the loss of contact by monitoring the voltage on the remaining track terminal (which obviously must not be externally connected to the wiper).  While the wiper is making contact, it will be at the wiper potential.  If the wiper looses contact, it will jump to the potential of the other track end.  This can be detected and used either shut down, or drive the regulator towards minimum output voltage.

e.g. In the LM317 circuit bsfeechannel posted:

if you put the wiper at the ground end of R2 and connect a small signal NPN transistor across R2 with its base fed by the remaining track end, you can get the output voltage to drop to approximately 2V if the wiper looses contact, instead of rising to close to the input voltage.

If the potentiometer is being used as a true potentiometer, provided the load on the wiper is small enough, it is always possible to ensure the wiper potential goes to that of the preferred track end on loss of contact, by adding an external bias resistor between that track end and the wiper, of large enough value not to significantly affect the potentiometer's linearity.

[tunk]

Ian.M, are you thinking about something like this (taken from an ebay listing):

[GlennSprigg]

Yes, for critical circuits I ALWAYS tie the wiper to one end of the Pot!  And even at times added
another higher value resistor in parallel with the pot, in-case the main pot track goes high !
(Sorry 'tunk'.  Didn't mean to come in on top of you... )

[Ian.M]

@Tunk,
Yes, but that LM317 circuit is questionable in other areas.  e.g. if the regulator is fed from a rectifier with no other loads, D3 is redundant, but as it has a significant value capacitor between Adj and ground, it *should* have a normally reverse biassed protection diode across R1, to prevent regulator damage if the output is shorted. 

Also, with that decoupling capacitor on Adj, one has to consider whether the transistor will be transiently outside its SOA if wiper contact is lost.   Worst case, it could have up to 28V across it with 5.7mA of base drive, which, allowing for h_FE dropping with increasing collector current, could result in a transient peak dissipation in excess of 3W.

@GlennSprigg,
That's only effective if the circuit (and any load) can tolerate the pot going to max. resistance.  As others have stated their  experience above, if not, it can get expensive!   IMHO, in most applications, linking the wiper to one track end does little except strengthen the pot's PCB pads!

[Seekonk]

HP had some supplies where the series pot was connected to the output.  Any time the output transistor shorted there would be full voltage on the output.  Of course the first thing you would do was turn the pot to lower the voltage and then the pot would burn up.  Like seeing a fixed resistance on the high side with the pot pulling down the feedback. Most pot dailures are an open so voltage goes to lowest voltage.

[bsfeechannel]

e.g. In the LM317 circuit bsfeechannel posted:

if you put the wiper at the ground end of R2 and connect a small signal NPN transistor across R2 with its base fed by the remaining track end, you can get the output voltage to drop to approximately 2V if the wiper looses contact, instead of rising to close to the input voltage.

Neat.

@Tunk,
Yes, but that LM317 circuit is questionable in other areas.  e.g. if the regulator is fed from a rectifier with no other loads, D3 is redundant,

However, if you connect the output to some circuit whose voltage is higher than that of the rectifier your LM 317 is toast. For instance, your load has a huge filter cap that has not fully discharged from a previous operation. Your power supply is not turned on yet, therefore, the voltage at the input of the regulator is zero. I think it's advisable to leave D3 there.

Also, with that decoupling capacitor on Adj, one has to consider whether the transistor will be transiently outside its SOA if wiper contact is lost.   Worst case, it could have up to 28V across it with 5.7mA of base drive, which, allowing for h_FE dropping with increasing collector current, could result in a transient peak dissipation in excess of 3W.

Maybe a resistor in series with the capacitor?

@GlennSprigg,
That's only effective if the circuit (and any load) can tolerate the pot going to max. resistance.  As others have stated their  experience above, if not, it can get expensive!   IMHO, in most applications, linking the wiper to one track end does little except strengthen the pot's PCB pads!

Bingo.

[ferdieCX]

HP had some supplies where the series pot was connected to the output.  Any time the output transistor shorted there would be full voltage on the output.  Of course the first thing you would do was turn the pot to lower the voltage and then the pot would burn up.  Like seeing a fixed resistance on the high side with the pot pulling down the feedback. Most pot dailures are an open so voltage goes to lowest voltage.

In several HP supplies, the output voltage is a linear function of potentiometer resistance. If the pot fails open, the voltage goes to the maximum.

[Ian.M]

@bsfeechannel,

Protection Diodes
When external capacitors are used with any IC regulator it is sometimes necessary to add protection diodes to prevent the capacitors from discharging through low current points into the regulator.  Most 10 mF capacitors have low enough internal series resistance to deliver 20A spikes when shorted.  Although the surge is short, there is enough energy to damage parts of the IC.

When an output capacitor is connected to a regulator and the input is shorted, the output capacitor will discharge into the output of the regulator.  The discharge current depends on the value of the capacitor, the output voltage of the regulator, and the rate of decrease of VIN.  In the LM117, this discharge path is through a large junction that is able to sustain 15A surge with no problem.  This is not true of other types of positive regulators. For output capacitors of 25 mF or less, there is no need to use diodes.

So its already got an internal protection diode equivalent to D1 rated for a 15A surge.  If you reckon you still need additional protection, you should be using a beefy Schottky there, not a relatively wimpy 1N4007.

I wouldn't put a resistor in series with the Adj decoupling capacitor as that will reduce its effectiveness.  However adding a 47R emitter resistor to the protection transistor would guarantee a peak current under 600mA, for only a 0.27V increase in the output voltage with the protection circuit active.  A much lower emitter resistor, or even none at all may be practical with appropriate choice of protection transistor to have a high enough Ic surge rating and low enough h_FE at high Ic, so it can operate Ib limited.

@All,
Monitoring the current through a potentiometer used as a variable resistor is another method of adding open wiper protection, that unlike my simple protection transistor suggestion,doesn't rely on the integrity of the track.  If you've got a negative bias supply available, it could be as simple as grounding the regulator's feedback network to the virtual earth at the inverting input of an OPAMP configured as an inverting current to voltage converter, and using a window comparator to check that the OPAMP output remains in the expected range for the normal feedback network current, to derive a protection signal that could e.g. be used to gate an output switch or trigger a crowbar circuit.

[bsfeechannel]

@bsfeechannel,

Protection Diodes
When external capacitors are used with any IC regulator it is sometimes necessary to add protection diodes to prevent the capacitors from discharging through low current points into the regulator.  Most 10 mF capacitors have low enough internal series resistance to deliver 20A spikes when shorted.  Although the surge is short, there is enough energy to damage parts of the IC.

When an output capacitor is connected to a regulator and the input is shorted, the output capacitor will discharge into the output of the regulator.  The discharge current depends on the value of the capacitor, the output voltage of the regulator, and the rate of decrease of VIN.  In the LM117, this discharge path is through a large junction that is able to sustain 15A surge with no problem.  This is not true of other types of positive regulators. For output capacitors of 25 mF or less, there is no need to use diodes.

So its already got an internal protection diode equivalent to D1 rated for a 15A surge.  If you reckon you still need additional protection, you should be using a beefy Schottky there, not a relatively wimpy 1N4007.

The original manual suggests a little differently, I'm a afraid.

(See attached image. This doggone image attachment plugin is b0rken).

I wouldn't put a resistor in series with the Adj decoupling capacitor as that will reduce its effectiveness.  However adding a 47R emitter resistor to the protection transistor would guarantee a peak current under 600mA, for only a 0.27V increase in the output voltage with the protection circuit active.  A much lower emitter resistor, or even none at all may be practical with appropriate choice of protection transistor to have a high enough Ic surge rating and low enough h_FE at high Ic, so it can operate Ib limited.

Makes sense.

@All,
Monitoring the current through a potentiometer used as a variable resistor is another method of adding open wiper protection, that unlike my simple protection transistor suggestion,doesn't rely on the integrity of the track.  If you've got a negative bias supply available, it could be as simple as grounding the regulator's feedback network to the virtual earth at the inverting input of an OPAMP configured as an inverting current to voltage converter, and using a window comparator to check that the OPAMP output remains in the expected range for the normal feedback network current, to derive a protection signal that could e.g. be used to gate an output switch or trigger a crowbar circuit.

I'd like to see a schematic of that circuit.

[Bassman59]

I have seen old laboratory power supplies which included a screw adjustable potentiometer for setting the trip voltage of the crowbar on the output; the Power Designs TP340 is an example that has this.

I have an old Power Designs 6050C which has that crowbar set-point pot. It’s easily adjusted with a fingernail.

The output adjust pot doesn’t have much friction so it can be easily disturbed. I learned early on to just set the output then set the crowbar.

[moffy]

I made this comment about a PSU design with the same issue:

"there might be a problem with the design. If you lose the connection between the pot and the 'adj' pin the output of the regulator is going to go high. If a pot develops a dead spot or fails you could get the full input voltage across your test circuit which you are powering. It would be better to fail safe or low output voltage. The 'adj' pin needs, say a resistor permanently connected, to make the output voltage low in the case of a failure of the pot or connection."

In this discussion: Projects, Designs, and Technical Stuff / Re: Building a simple compact dual rail linear bench supply

It takes a bit of thought and effort to make things fail safe. The other issue with pots is that they can be panel mounted and use wires and connectors to the board, which adds an extra failure point.

[Conrad Hoffman]

This is the same problem many audio amplifiers have with the bias pot. Pot goes open, amplifier goes to full bias and the outputs overheat and fail. Good designs don't do that. BTW, wiper current should be kept to a minimum for most pots if you want reliability.

[pardo-bsso]

Hi, check here: http://www.imajeenyus.com/electronics/20160517_potentiometer_feedback/index.shtml

Besides going over the open wiper problem that article also explores different configurations and the response they give (voltage output vs. pot position).

[TheMG]

Maybe the manufacturers didn't really expect the equipment to be in service for so long that the pots start going bad?

Anyways, when I designed my own bench supply, the pot simply provides a reference voltage to a negative feedback loop comparator. If the wiper of the pot goes open the reference voltage goes to zero and consequently so does the output.

[ArthurDent]

A lot of the better (read 'expensive') bench power supplies have OVP, OCP, and even OPP settings to protect equipment you're testing. Here are photos of a couple. The GW Instek power supply is a modern supply with an encoder and no pot for adjustments.

[ferdieCX]

A lot of the better (read 'expensive') bench power supplies have OVP, OCP, and even OPP settings to protect equipment you're testing. Here are photos of a couple. The GW Instek power supply is a modern supply with an encoder and no pot for adjustments.

Several expensive bench power supplies that were built for many years, have this issue with the pot wiper failure.
I suppose that microporocessor controlled power supplies could also have their own problems, for example firmware issues.

[David Hess]

In several HP supplies, the output voltage is a linear function of potentiometer resistance. If the pot fails open, the voltage goes to the maximum.

That may be deliberate so that the power supply can support remote programming.

[duak]

Based on a sample space of five or so lab supplies with hazardous problems I've encountered, only one had a bad output voltage pot.  It's one I'd like to fix and I'll get back to it in a bit.

The most common fault was a shorted pass transistor.  The hp 6002A had a shorted pass transistor and someones misguided replacement of an op-amp.  The crowbar circuit was tripped at anything other than zero output so I got the unit for free.

A Kikusui auto ranging supply had a bad output bypass cap that could break into oscillation at certain load currents.  Having a supply set to 12 V to test an electronic speedometer assembly and have it superimpose a 30 Vp-p AC signal on it was exciting.  I thought I fried the speedo but it's a good thing that it had good filtering and clipping circuits on the power input.

I suppose one way to make the supply fail-safe is to use a dual ganged pot and some sort of wired-OR circuit to select the lower voltage.  Or the second section could set a variable limit that would disable the output or trip a crowbar if there was a discrepancy.

In an hp type floating supply the regulator sources a reference current that is converted to a variable command voltage by the pot.  It should be possible to disable the regulator or trip the crowbar if the reference current drops below a threshold, as when the pot opens or goes over a rough patch.  However, because these supplies can be put into various series and parallel configurations, this will likely cause trouble in some cases.

To get back to the supply with the bad pot, I have a 40 V Lambda unit with a twitchy output. On these supplies, the output capacitor and circuit is discharged by the pot when it is turned down.  The pot  is fried, cleaning no longer helps and there are no new replacements.  It's a dual coaxial wirewound with values something like 8K+500R.  I don't like 10 turn pots because it takes so long to wind from one end of the range to the other.  The hp supply above has a 10 turn pot but I prefer the Lambda for many tasks.  At one time, it was easier to get coaxial pots so I've been casting about for solutions.  Some of them are:

1.) just get a 10 K WW pot and put a resistor across it to restore the original voltage limit,  Years ago, I ran across a two speed planetary drive where small movements were at a 3:1 ratio but larger movements (> 90 degrees) were 1:1.
2.) look for a coaxial switch and pot combo.  Use the switch to select fixed resistors that are then trimmed by the pot.  At one time things like this were easier to get - I'm sort thinking about an old Tek or hp scope attenuator or timebase switch assembly.  I've even thought of making something with gears, but there isn't much room.
3.) design an electronic replacement with an encoder and either a DAC or EPOT.  The circuit would not have to simulate a resistance but only have to clamp the programming voltage into the floating regulator to an arbitary voltage determined by the encoder setting.  A coaxial push-pull switch could control the resolution or some sort of speed sensitive ballistic action.

Any thoughts?

[phil from seattle]

with the encoder approach you could implement a proportional response - slow means very small increments, fast means much larger. Seems like a big win. I have a cheapo Chinese PS that works surprisingly well but the last few mV or mA of adjust is twichy. Would love proportional adjust.

[Vovk_Z]

Described in first post problem is why I like to use a multy-turn potentiometers in power supplys to control the voltage - they "jump" too, but usually they don't totally loose a contact but only jump for about one turn (so it isn't very high in voltage).

If I not used them - I'll loose to much devices too.