Biased Bridge Rectifier as protection (Fluke 540B)

[essele]

Hi,

I'm trying to better understand the protection circuit from the Fluke 540B (mainly because I've got one that trips instantly for most of the time, then after a bit of probing is fine, then goes wrong again) ... so I'm trying to understand it properly and have a quick question about a bit of this circuit (I'm sure more will follow)...



So the above schematic is the first part of the protection circuit ... I have an AC input voltage to simulate an incoming voltage to the circuit, I've added a 1R resistor to include a bit of supply impedance, if you then focus on the bridge rectifier (rather than anything else for now.)

Unless I'm missing something the principle here is that the diode's will conduct once we hit a suitable voltage so that it clamps the "input" to plus or minus two voltage drops ... I need to go back and measure the actual voltage drop since I can't find a datasheet for the parts, so lets assume 0.7v each, therefore at 1.4v either way it will be clamped.

So my question is ... why is it biased with the +/-6.75v supply?  Simulation in falstad shows exactly the same behaviour whether the +/- 6.75v is connected or not.

Does it help with response time or something like that?

Thanks,

Lee.

[syau]

Hi,

I'm trying to better understand the protection circuit from the Fluke 540B (mainly because I've got one that trips instantly for most of the time, then after a bit of probing is fine, then goes wrong again) ... so I'm trying to understand it properly and have a quick question about a bit of this circuit (I'm sure more will follow)...

(Attachment Link)

So the above schematic is the first part of the protection circuit ... I have an AC input voltage to simulate an incoming voltage to the circuit, I've added a 1R resistor to include a bit of supply impedance, if you then focus on the bridge rectifier (rather than anything else for now.)

Unless I'm missing something the principle here is that the diode's will conduct once we hit a suitable voltage so that it clamps the "input" to plus or minus two voltage drops ... I need to go back and measure the actual voltage drop since I can't find a datasheet for the parts, so lets assume 0.7v each, therefore at 1.4v either way it will be clamped.

So my question is ... why is it biased with the +/-6.75v supply?  Simulation in falstad shows exactly the same behaviour whether the +/- 6.75v is connected or not.

Does it help with response time or something like that?

Thanks,

Lee.

A frequent trip is usually related the the leakage around the transistor Q303 & Q304, you can test this by disconnecting the 100 ohm resistor just before the 2u2F capacitor. If your 540B still trip then it is related to the leakage on those 2 transistor (their socket or the pcb trace)

[essele]

Thanks syau.

I've pulled the 100ohm resistor and it still trips ... there was some flux on the board, so I've given it a good clean, and had a quick go at the sockets with IPA but it doesn't seem to have made any difference.

I was starting to look at a differential between GND2 and GND3 which I was struggling to understand (hence the above question as a starting point) ... but leakage makes more sense.

The sockets are absolutely awful, very loose ... the transistors already have their legs doubled over to make them stay in the sockets which isn't the case on my other unit.

I did try the transistors from the other unit, but thinking back I don't think it was a viable test as I had to hold them in (because of the sockets) and therefore I think that would have tripped it anyway. I will try that again.

Any ideas what else I can do? Replace the sockets? The transistors don't seem to be available anywhere?

Thanks,

Lee.

[amspire]

The transistors become too leaky with age. The old transistors just cannot be used. I just replaced them with some modern Silicon transistors.

No need to replace the sockets - they are fine.

I used a MPSA56 for the PNP transistor (Q303) and a 2N3904 for the NPN (Q304). Both were in TO92 packages with centre base leads, so they plug into the sockets fine.

Just re-calibrate R314 to trip at 0.7V to 0.75V as in the manual.

The trip is rock solid with modern low transistors. No spurious triggers.

I fixed mine years ago. Just tested it a moment ago and it trips at 0.75V.

Richard

[essele]

Hi amspire,

Thanks for the suggestion ... you had my hopes up for a bit, but I've just tried a 2N3906 and 2N3904 and it's still exactly the same. I've also retried the other known good transistors from another unit and the behaviour is the same.

The strange thing here is that twice now it's all sorted itself out for a period of time while I've been probing and then eventually stopped working again.

Going back to the schematic, if I assume the transistors are good (I'm going to leave the 2N390x's in) and with the 100R removed, then the only place I think I need to go back to is the difference between the chassis connection and GND2 and I'm sure I was seeing strange things with CR305.

UPDATE: just retested, CR305 when the unit is on, but mode is off drop reverse bias is 0.245v ... if I switch to search it goes to 1.36v and the same thing with Q304 removed to stop the trip shows 0.9v, but both times this is then dropping. I've already replaced both of the capacitors, CR305 has been horribly bodged in although it tests ok with a normal diode test (I really need a curve tracer ... next project!) I think I'll replace CR303 - CR306 anyway, in fact a lot of the diodes look iffy ... someone has obviously had a go at them in the past.

The whole ground setup with this baffles me ... I don't really understand how an incoming signal referenced to GND3 can then be used referenced to GND2 (GND3 seems to be tied to the chassis) why am I then seeing such a differential between the two GNDs.

CR305 seems to ensure that the chassis can never be more than a drop less that GND2, but it can clearly be quite a bit above.



Lee.

[amspire]

The 2N3096 is not a good choice for the PNP. You want a PNP transistor with a absolutely lousy gain at low current. If you look at the datasheet for the MPSA56, you will see the gain is dropping like a rock below 10mA and they don't even bother to continue the graph below 3mA collector current. The 2N3096 on the other hand has plenty of gain still at 100uA.

Go through your collection of transistors and look for the highest voltage, highest current transistors for the PNP, and pick the one with the worst gain.

For the 2N3094, you need low collector current with the emitter-base shorted with a 47K resistor. You need a nA meter to test that most people actually have - it is called a multimeter. A 10M input multimeter on the DC volts range displays 100nA per volt on the display, so 1nA = 10mV.

My 2N3904 has less then 0.1nA collector current at 16 deg C. It will increase a lot when it is hotter, but as long as you have a really low gain (at low current) PNP transistor, the goal is probably to have less then 1uA in the PNP collector at no 540B input. If your PNP transistor has a gain of 20 or less at 1uA, that means you need the 2N3094 collector leakage less then about 50nA. Less then 5nA and you should be fine.

To test, I just lifted the collector of the NPN transistor and put the multimeter on a 1V range between the NPN collector and base of the PNP transistor. Then just give the meter plenty of time to settle.

Richard.

[essele]

Thanks - I've done a bit of testing as you suggest and I was seeing around 70-100mV, so 7-10uA which from what you say below sounds reasonable.

I'm really not sure this is the problem ... I've just put my oscilloscope across CR306 and CR306 in series (effectively across r313) and I was seeing a 200mV jump when I switched to search. However it's now suddenly all working again (well the 100R is still out, but it's not tripping and it was before) ... if I leave it off for a while when I turn it back on there is a much greater jump (~1V) ... which then drops slowly, so there is something capacitive going on.

And obviously when it's working it's working properly, the trip works fine (well not at the moment because the 100R is out, but it did before) so that does call into question the leakage ... unless it's intermittent.

I going to have to leave it for a bit now, but once I get back I'm going to try to methodically test the sequence of events that fixes it ... I suspect it's something to do with the ground lead of the scope.

Lee.

[wn1fju]

I will relate my story with the 540B protection circuitry, for what it is worth.

I had a similar intermittent false triggering of the protection circuitry when I first got my 540B.  The protection board had a bit of corrosion from leaking
batteries, so I scrubbed the board with alcohol and a toothbrush.  The thing seemed a little better, but often would false trigger for days on end, and then
sometimes would work OK for a while.

I did all the normal replacement of the two transistors, diodes, resistors, etc. to no avail.  Finally, I disconnected the board (pain in the ass) and removed it.
I took out the two transistors and a few other parts and put an ohmmeter across various traces.  I was getting readings like 250K across two traces
that were totally isolated (no components attached).

At that point, I figured what the hell - I couldn't make it worse, so I scrubbed the board with steel wool and muriatic acid.  And I mean I really scrubbed
and scrubbed.  Rinsed the board off with water, let it dry, reinstalled the parts I had pulled, and put the board back into the 540B.

The protection circuitry has been operating flawlessly now for over 6 months.

[syau]

I managed to use 2n3904 (Q304) and 2n3906 (Q303) by connecting the base of Q303 to the collector of Q304 directly (ie not solder to the pcb) 

[syau]

I managed to use 2n3904 (Q304) and 2n3906 (Q303) by connecting the base of Q303 to the collector of Q304 directly (ie not solder to the pcb) 

Also forgot to tell that I found the 47K resistor drifted up to 55K, replaced with a modern 1% one.

[essele]

I got back to this today after finishing my second LiPo power supply.

So I've now tried a few different MPSA56's and 2N3904's ... still trips! I tried the approach with the connection directly, no joy. Checked R313, that's ok too.

This is really frustrating ... for now I've just pulled one of the transistors so I'm going to run with the protection disabled but I'd rather I didn't have to do that as I'm sure I'll fry my thermocouple at some point.

I really don't want to have to resort to removing and cleaning the board, sounds like an absolute nightmare ... I wonder whether I could use a small bit of veroboard with the transistors on ... may be worth an experiment.

[RoGeorge]

lets assume 0.7v each, therefore at 1.4v either way it will be clamped.

So my question is ... why is it biased with the +/-6.75v supply?  Simulation in falstad shows exactly the same behaviour whether the +/- 6.75v is connected or not.

Does it help with response time or something like that?

It's about impedance, or how much current will flow through the protection bridge, not the clamping time.

Only an ideal diode will suddenly open at 0.7V.  A real diode will conduct current in the region 0...0.7V too, just not as much as it would conduct for a voltage bigger than 0.7V.

For the protection bridge, indeed, the maximum clamp voltage will be the same:  +/-1.4V with or without the +/-6.75V bias.  The difference is in the current flowing through the (upper side) clamping diodes when the input voltage is in the range +/-0.7V:

- with +/-6.75V bias voltages on, the current will be (ideally) zero, so infinite impedance, so the protection bridge will not affect the instrument readings for input voltages in the range +/-0...0.7V.  In other words, the impedance of the instrument is not affected by the protection diodes when the input is in the +/-0.7V range.  For input voltages in the range +/-0.7...1.4V, the clamping diodes will start to conduct current, so the impedance won't be infinite any more, so the measurement will be affected.

- with bias voltages removed, then there will be some current through the bridge even for the +/-0.7V input voltages, thus a low impedance that disturb the measurements.

[essele]

Ok ... finally some progress (although I don't really understand it.)

I pulled both transistors and recreated it on some breadboard. I disconnected the arming-switches wire from the board so there can be no leakage from that on the board anywhere. The relevant other signals I've just jumpered into the transistor sockets.

It made absolutely no difference!

With a scope across C309 I could see a 1.4v spike every time I switched to "search" which clearly is enough to trigger the whole protection circuit ... if I removed the base connection to Q304 I could still see about a 0.3v spike, so that definitely looked like a leakage problem with Q303. I tried a number of different transistors here to absolutely no avail.

Almost in desperation I decided to eliminate the thermistor and 100R pot, so I took a 13.5v feed off the board into a 10k pot to gnd2, with the wiper going to Q304, then adjusted it to rougly 100mV on the emitter of Q304.

It all worked perfectly ... trips at 0.75v every time.

BUT .. I've just gone back to it to check something before posting this and it's all gone mad again ... now it trips every time if the emitter voltage is anything less than about 280mV but it's not consistent, once it trips it seems to be more sensitive ... I wonder if there is a temperature or capacitive thing going on.

Aaarrrggghhh.

[syau]

I am using the following parts

ON Semiconductor
2N3904BU
Digi-Key 2N3904FS-ND

2N3906BU
Digi-Key 2N3906FS-ND

Note: The original transistor socket is leaky 

[essele]

Thanks Syau ... I actually don't think this is a problem with the transistors, I think the pcb is the problem (amongst other things.)

I've now removed the pot and thermistor and they both look horrifically corroded (see picture below.)

I've decided to go with a 20k pot in place of the thermistor, and a 100R in place of the pot, so I still have some controllability ... it seemed a bit more stable this way around, not sure why ... I was slightly concerned about heating inside the pot when tripped, so this made more sense. Although I didn't really see any evidence of a problem or resistance change.

I've put the relay-trip wire back in, and with transistors in place on the board it still has the problem, if I jumper to my breadboard with just the transistors in it then it's absolutely fine.

I suspect the pcb is the problem as it's been exposed to leaking batteries and then a plethora of different cleaning materials. I'm going to remove the transistor sockets and give that area a good clean, then I may have a go at some air-wiring.

[amspire]

I would try leaving the collector of the NPN and the base of the PNP transistor out of the socket. Connect the two free leads directly to each other. This eliminates leakage in the sockets or the PCB.

[syau]

Air wiring the base of Q303 to the collector of Q304 directly solved my problem 

[essele]

I did try the base to collector wiring before and it didn’t work, however I’ve now removed the sockets and soldered the new transistors directly to the board, and with the new pot/resistor combination everything appears to work.

It feels like there were multiple issues but probably all caused by the battery leakage.

This has been a real pain - thanks for all the help, I definitely wouldn’t have persevered without the various ideas.