U1241B multimeter woes

[PA0PBZ]

The real protection would usually come in the form of a resistor ladder after the PTC. It's weird that they don't have that in a meter of this class (or at least it's not visible in the image posted above).

Found it: Other side of the board, after the range switch, 5 x 2M.

[joeqsmith]

Joes reverse engineering (in his video) shows similar values.

The PTCs I measured were not even close to OP's 22 ohms.   I have never seen a PTC with this low a value used in a handheld meter.   

It's weird that there's no resistor ladder as part of the input protection.

There are two legs.  One is an input, the other an output.   While the input side has a very high input impedance and will commonly have some sort of high voltage resistor or several smaller parts, the output side is a low impedance path.

I have seen meters with nothing for protection on the input side.  They count on the high impedance to protect the meter.   The drive side is a little more tricky but it's not rocket science.  The protection schemes are fairly well known. 

Dave's vapormeter is a little odd because it could drive at such a high output voltage.   If you caught those early videos I made of the meter, you may remember me adding a TVS to that circuit along with changing to a CMOS part to try and increase the margins.  It seemed Dave had made some comments about the meter doing as good as the 87V which only ever fails on my test jigs and that the cert was good enough.  Later he posted about this circuit using a TVS.  It will be interesting to see the released version of that meter, but that's for another topic. 

Back to OP's meter.  As I mentioned, there will normally be a low voltage clamp on the backside of the switch.  There will be enough current that will cause the PTC to heat.   Even if the contacts shorted, all that should happen is the PTC would heat up.   But with a 22 ohm part, I wonder what the thermal time constant is....

[Fungus]

Joes reverse engineering (in his video) shows similar values.

The PTCs I measured were not even close to OP's 22 ohms.   I have never seen a PTC with this low a value used in a handheld meter.   

We were talking about "the left resistor".

the left resistor in OP's drawing is only 275 Ohm in my meter.

[Fungus]

The real protection would usually come in the form of a resistor ladder after the PTC.

Found it: Other side of the board, after the range switch, 5 x 2M.

Told you so! 

[PA0PBZ]

The PTCs I measured were not even close to OP's 22 ohms.   I have never seen a PTC with this low a value used in a handheld meter.   

The PTCs in my meter are around 800 Ohm.

[joeqsmith]

The PTCs I measured were not even close to OP's 22 ohms.   I have never seen a PTC with this low a value used in a handheld meter.   

The PTCs in my meter are around 800 Ohm.

I would buy that.  If OP measured them correctly, it will be very interesting to see how they are marked.  If they really are 22 ohms and marked this way, the next question is did the factory do this, or was the meter damaged at some point (I've cracked a few PTCs) and someone with exception intelligence wanted to save a few bucks and play technician?   Still it does not explain why it would arch across the switch in the first place, unless if the meter was previously damaged, maybe it had arced across before... 

I like a good murder mystery. 

[najrao]

I have rechecked the PTC's to find that 22 ohms is wrong: they show 680 and 710.  It was my fault entirely. The 'fusible' resistors are 1k on the Volts path and 100 ohms on the Ohms line. Markings are hidden below shrink sleeves. Keysight want all evidence retained, so I will not cut open.
The switch track clearance may have been a bit more than 0.8mm;  there was much charring and I did not use a  vernier.
There are no other protection components on the board.
Once again, on the Volts range, there is no issue other than the fact that the Ohms switch remains hot on one side, stressing the small clearance. The 2×5M resistors should definitely be enough to survive 1000V, even with a crest factor of 3.
On the Ohms range, current is limited essentially by the 100 ohm protection resistor and the PTC,  unless the measurement circuit is able to take up much of the applied voltage. If it clamps, there will be too much current and the PTC should heat up and go high, but not for a time. The current can reach several hundred milliamps initially and last seconds; can the measurement circuit hold its clamping action?
On the other hand,  nearly all the impressed voltage will reach the measurement circuit if there is no clamping. This seems almost impossible, given the track clearnce of 0.25mm.

[joeqsmith]

Thanks for rechecking the PTCs.    There for sure are other protection parts on the board.  The meter would not survive without them.  These will be on the backside of the switch you damaged.  You would need to trace it out.   This clamp will hold that side of the switch fairly low.  The remaining voltage will be across the PTC and resistor as you suggest (at a KV, basically all of it). 

Yes the second clamp I am referring to should be designed to handle the current for the length of the thermal time constant of the PTC.  Likewise, the resistor should handle the full voltage as well for that short time.  My guess, based on the pads being damaged the way they were is the parts making up that secondary clamp would be visibly damaged. I'm actually surprised that with the pads repaired that the meter would still read resistance, capacitance or diode check.  If the clamp failed short, the parts down stream may survive but the meter would not function.  This is the best case failure.  lol.   I've blown them open which never works out well for the parts down stream from the clamp that was supposed to protect them.  Normally this is after the PTC arcs over and I have seen a lot of piss ant PTCs come apart..

If you cut away the jacket, you may be able to find the datasheet for the PTCs which will have the curves for the time constant.  You could calculate if it was a problem easy enough.  It still would not answer why the switch had such a violent ending.

Of course, if the input side GDT fires it has nothing to do with the drive side.  So the switch sees the spike until the drive side tube fires. But these are transient conditions.  It sounds like you are claiming that you are not running anywhere near where the tubes would even turn on.    Really strange..   

It will be interesting to see if Keysight provides any further guess to what went wrong..   

[najrao]

Excellent, Joeqsmith. At last someone got to the points I made in the first place. No Cats at all!
Agree there has to be a low side clamp, but tracing what and where is well nigh impossible, there are no discretes as such energy clamp just has to be one. Applies to both sides of the board. But then one may be hidden behind the display, for example.
I am equally surprised that the failure event did not destroy the clamp or the measuring circuit ahead. Indeed the PTC and fusible resistor look like they never saw any duty. Even now, the Ohms, Diode and Cap work as nothing happened at all. Not calibration checked, no, but good for all purposes.
PTC's can heat up very fast once they get going in a maintained current situation, and simple thermal time constant calculation is all but absurd. Wild guess is best.
My repair solution is not all it should be. I have just wired a reed relay to points leading to the obliterated switch, and operate it by a Nd-Fe magnet from outside the case. The reed contacts are just 0.5 spaced, and it is easy to forget withdrawing the magnet before a voltage measurement. I would not recommend it.
Keysight appear to have gone slow after they made the replacement offer. But I will give the failed unit even if they don't replace. There are logistics difficulties in shipping.

It is possible that other models by Keysight have the same issues, would their owners check? Not really difficult, you hv to take out the board and examine visually.

[HKJ]

As far as I understand the input protection:

The MOV is there to protect the switch and high impedance path.

There are usual two or 3 path from the input: A 10Mohm, often a 1Mohm and a low impedance for driving ohm/capacity (Usual in the 0.5-2kohm range).

After the switch are low voltage clamps on all the paths, the low impedance clamp must handle fairly high current until the PTC heats up.

[najrao]

Thanks, HKJ and Joeqsmith. Your input drove me to check again, more carefully this time. I have to eat my own words again.
Bingo! There are TWO  discrete diodes, may be zeners, and one of them clamps positive voltage to unknown level to ground, negative to a diode drop. This diode has a marking "457A" as far as I can read by magnifying glass.  Some in situ testing shows clamping action at 3V or so for a modest 8mA through the Ohms input. A negative current of 9mA clamped at -- 0.73V, as expected. All voltages referenced to COM input terminal.
The other discrete diode  is marked K7, but looks identical. It's cathode ends up somewhere deep, so I cd not trace further.
How Keyswitch protects the Ohms range is now clear. It is unlikely that they have not engineered the circuit for the PTC heating up time.
This resolves the track spacing issue as well: the potential on the cold end of the PTC can never exceed the clamp level, positive or negative, and so the 0.25 or so is still good enough.
But it does not absolve them of the poor switch clearance, 0.8mm.
We have got somewhere after All! Thanks.

[joeqsmith]

I was going to point you to a high res picture that showed what appeared to be the secondary clamps.  Glad you took the time to find them.   I suspect the one is a TVS.   For me, we are no further along.  You have a better idea of how the circuit is designed but it still does not answer what happened to the contacts.  You got rid of the body before presenting the case and asking to help solve this murder. 

If I owned one of these meters, I would have no problems putting some DC into the input and seeing where it clamps and the dialing it back and letting it sit for some extended period.   Obviously if the switch breaks down we have a problem.   

Dave did a review not too long ago using a insulation tester on a cheap ANENG meter.  This meter has the input, through a fuse into a switch with the other side of the switch to a TVS that returns to the common.  In this case, those contacts would arc (I would need to check at what level).   But it is not a big surprise.  The pads I am sure had some damage from the arcs but that was a fairly low energy transient.  The were not destroyed as you suggest yours were.    When I test them, I use up to around 20 Joules.  The generator has a two ohm source impedance.  This has not been enough to cause the level of damage you describe.  At most, maybe some discrete components come apart.  It appears yours was hit with something much larger.  High res pictures of the damaged area may have helped answer that.   

I at times will run what I am calling a half cycle test on them.  It's nothing I keep data on.  This waveform simulates a half AC cycle of something a little more reasonable energy wise.  It also uses a two ohm source.  The generator can be applied to a quality meter and nothing will happen as the voltage is lower than what the meters are rated for.   However, it works with the higher voltage generator.  Basically, the idea was to simulate a transient on an AC line but still keep the energy to something I personally feel safe working around.   If the high voltage transient causes something inside the meter to breakdown, the low voltage generator then feeds the fire.   It's nothing like what could happen in real life but does give me some idea how a meter may behave.     Anyway, this setup can cause the sorts of damage you describe.  And it is very typical for the switch to provide spark gap and needed copper.   

So if you were looking at an AC bus and something like this came through things like this can happen.  This is where the CAT rating comes in.  It's a safety standard with as I understand it is there to help prevent damage to the operator.  From talking with meter  manufacture's some design their products to survive the transients called out by these standards, other do not.  I only ran the one Keysight meter.   It is no longer with us....   You may have felt you were looking at a 440 AC bus which the meter was rated for....  Again, hard to say.   

[najrao]

Thanks, Joeqsmith.
I continue to believe that the clamp is just a zener and not a high voltage TVS, because it has to clamp at very low voltage which is unusual for TVS. At anything over a few volts, it will need to dissipate, and survive after,  several watts. Remember the applied voltage may be any value below 1000V; at 100V, there is a current of just under 100mA,  and the PTC would take longer to cut this down. At some even lower voltage,  the current will get to be too small to heat the PTC,  yet be large enough to cook the zener, unless it clamps at just a few volts. The zener is 500mW at best and has no lead conducted cooling. More, it has done its job through the catastrophe when the switch arced over, and is none the worse for it after. I would not hesitate to test it all out, but Keysight want to investigate further in their own way and would need the evidence in tact.
Agree that we still don't know what caused the switch to arc over in the first place.
There is data on 1N4574,  but the marking is 457A. 4574 is a special temperature compensated reference zener at 6.4V.

[joeqsmith]

TVS's are not at all limited to high voltages.   Normally people don't measure the output resistance of their active power supplies, but even if they do the PTC is there to save you from yourself.  Check the curves for some PTCs and see how quick something like the one you have will switch.  Sure, the peak power can be quite high but that is not a steady state condition. 

The majority of the meters I have looked at will use two back to back SOT23 transistors or diodes to form the secondary high speed clamp.  Cheap, low leakage.  Off the top of my head, the BAV99, BAV199 and 3904s seem to be common.   They can handle a fair amount of current for a short time.   

I doubt very much in your case that the PTC did anything but act as a resistor.   A pulse large enough to arc over and cause this sort of damage, I would guess was all over before the PTC budged.  However, you are correct that for those of us, (I am talking about myself) who seem to have a strange desire to measure the resistance of their active power supplies just because we want to see how robust these meters are  (I'll give you this is not a typical case),  I really haven't seen much of a problem with the clamps or switches in the sub KV levels.  If you watch any of my videos, you will know it is not for a lack of trying.  The typical life of a meter in my hands is measured in days..     

To be clear, I am not suggesting that your meter doesn't have a problem but I really have no way to verify it without checking it first hand.  If I had one, I would gladly attempt to smoke it just for you. 

[najrao]

Thanks, Joeqsmith.
I continue to believe that the clamp is just a zener and not a high voltage TVS, because it has to clamp at very low voltage which is unusual for TVS. At anything over a few volts, it will need to dissipate, and survive after,  several watts. Remember the applied voltage may be any value below 1000V; at 100V, there is a current of just under 100mA,  and the PTC would take longer to cut this down. At some even lower voltage,  the current will get to be too small to heat the PTC,  yet be large enough to cook the zener, unless it clamps at just a few volts. The zener is 500mW at best and has no lead conducted cooling. More, it has done its job through the catastrophe when the switch arced over, and is none the worse for it after. I would not hesitate to test it all out, but Keysight want to investigate further in their own way and would need the evidence in tact.
Agree that we still don't know what caused the switch to arc over in the first place.
There is data on 1N4574,  but the marking is 457A. 4574 is a special temperature compensated reference zener at 6.4V.
Keyswitch hv just said that they will go ahead with the further investigation and are arranging to take it back. Let us wait.

[najrao]

True to their word, Keysight has sent me a replacement. Very generous of them.
In turn, I have given them the old unit to investigate. We need to wait and watch for any report from them.

I will get back here if/when they do send their findings.