New Null Detector for old Fluke 335A/D

[Fred_47]

First time post ing a question to this forum.

About a year ago I purchased a Fluke 335D voltage standard & null meter. After a short period of replacing capacitors and not much else it was working just fine. Except for the null detector which was unsteady and wouldn't hold a zero setting on the lower voltage scales.

I eventually found that one of the neon lamps in the optical chopper was flashing occasionally. A little experimentation showed that adding a capacitor across it would settle things down nicely. After being energized for about 6 months it became impossible to set zero. This made it difficult to measure 0 volts. Who would have thought that measuring 0 would be so difficult. The zero setting pot simply didn't have enough range.

When I opened the beast, I found that the neon for the series photoresistor was weak and didn't illuminate fully. This was the one with the paralleled cap. Removing the cap didn't help. So, I used the 335D to test neon lamps from my meager stock til I found one that matched the turn on voltage and had similar length of electrode illuminated as the remaining neon on the shunt photoresistor. I couldn't match the color. That worked well for about another 6 months.

I'm now unable to zero the meter on the low voltage scales, the same as before. I haven't looked inside, but I expect that the photo chopper is the problem.

I've looked around and found a fix that uses LEDs but that adds load to the already marginal seeming ±15V supply. I think I also saw one that used photoMOSFETs but that also requires loading the 15V supply. I'm seeking something similar to the MOSFET chopper used in the Keithely 155. I hope to make as few changes as possible to the stock null detector board. No changes to anything not on the null detector amp board.

Breadboarding a ckt using stuff from my stock I came up with the test ckt below. I haven't worked up the courage to try it on the null detector card in the 335D yet. The ckt is on a solderless BB and works well except for the gate transient being coupled thru to the source and probably the drain (I haven't looked yet). I was able to add capacitance (0.02uF) too get the peak down to about .1V but any more seems to degrade the signal. The scope traces are Ch 1 on the gates & ch2 at the output from a x -100 amplifier (actually maybe times -20).

MOSFETS with the substrate brought out as a 4th terminal are not readily available. I suspect that this is what saves the Keithley chopper from switching transients.

Thoughts?

[RandallMcRee]

Well I have a non-working 335D at your disposal, if it will help.

My thoughts in no particular order:

Use a modern chopper, ala Conrad Hoffman's mini-metrology lab (modernize to use an OPA189).

Perhaps you can replace the neon with a modern optocoupler like the H11F1M

You can get mosfets with a substrate lead from LS (Linear Systems) e.g. http://www.linearsystems.com/lsdata/datasheets/SD5000%205001%205400%205401%20QUAD%20N-CHANNEL%20LATERAL%20DMOS%20SWITCH%20ZENER%20PROTECTED.pdf

I have a couple of SD5000, I think, if you need one at cost (they are expensive).

[Kleinstein]

If it is possible to add a new (lower turns) winding to the transformer, the load for driving LEDs could be relatively low - likely less than for the original neons.  Modern high efficiency LEDs don't need that much current - with a reduced distance to the LDRs one may not even need more current, just less voltage.

What are the points that make the 15 V supply weak ? One could increase R9 a little to use less current there.

Building a new null-meter based around a chopper OP is also feasible, especially if absolute lowest bias is not a requirement. The original circuit is not really low noise. So it would be more like low bias (though higher noise) AZ OPs to consider, e.g. MAX4238 or LTC2050 or ICL7650. If needed one could consider adding bias current canceling (e.g. via photodiode) and than use the AD8628.
Choosing between low bias / current noise and low voltage noise would be relatively simple.

[Fred_47]

Well I have a non-working 335D at your disposal, if it will help.

My thoughts in no particular order:

Use a modern chopper, ala Conrad Hoffman's mini-metrology lab (modernize to use an OPA189).

Perhaps you can replace the neon with a modern optocoupler like the H11F1M

You can get mosfets with a substrate lead from LS (Linear Systems) e.g. http://www.linearsystems.com/lsdata/datasheets/SD5000%205001%205400%205401%20QUAD%20N-CHANNEL%20LATERAL%20DMOS%20SWITCH%20ZENER%20PROTECTED.pdf

I have a couple of SD5000, I think, if you need one at cost (they are expensive).

I know that the existing ckt works if the choppers work so I'm starting out looking to replace the minimum amount of circutry as inexpensivly as possible.

One of the replacements that I looked at hinted at problems with the ±15V supply so I'm also trying to minimize changes there. Measurements that I took when I replaced DS1, showed that the PS on the detector board was +12.8, -16.8. The + supply was 15% low & the - supply was 2.5% more than spec'ed. The - ps is probably OK. My notes say that C3, 4 were previously replaced and have acceptable ESR & capacitance. I will check them when I open the beast.

The output waveform of the transformer low voltage winding is a not too bad square wave. (Newfile4.png)

I have a few shots of the neon ckt waveforms. I measured R2 (31.25kΩ) and used it as a shunt to get an idea of the neon currents. Newfile 9.png has Ch 1 at the node R2 to CR3, 4 and Ch 2 is at R2 to C1. Math waveform calculates (CH2-CH1)/31.25e+3. The RMS current is about 1mA. (I'm not sure how good the RMS calc is for that waveform.) The neon voltages (+ diode drop) are about 75V.

The output from the HV winding is also a square wave (Newfile 12.png) about 225Vpp. Ch 1 is on pin 1 of the card and Ch 2 is on the C1, R2 node. Math calculates the voltage across C1.

I figure that I can add about 6mA to the low voltage winding for every 1mA I eliminate from the high voltage winding without changing the primary current. I don't know much else about the transformer except that it's sealed, probably potted and about as high on the unobtainium scale as possible if I 'break' it.

The neon current is about 1mA so my 'budget' for the low voltage winding is 6mA give or take depending on what happens to the voltage.

I'll look into the H11F3M which doesn't have problems with charge injection. If the PS will tolerate 3 to 5mA it's probably the way to go.

Also, thanks for the offer of parts. Do you know what is wrong with your unit?

[Fred_47]

If it is possible to add a new (lower turns) winding to the transformer, the load for driving LEDs could be relatively low - likely less than for the original neons.  Modern high efficiency LEDs don't need that much current - with a reduced distance to the LDRs one may not even need more current, just less voltage.

What are the points that make the 15 V supply weak ? One could increase R9 a little to use less current there.

Building a new null-meter based around a chopper OP is also feasible, especially if absolute lowest bias is not a requirement. The original circuit is not really low noise. So it would be more like low bias (though higher noise) AZ OPs to consider, e.g. MAX4238 or LTC2050 or ICL7650. If needed one could consider adding bias current canceling (e.g. via photodiode) and than use the AD8628.
Choosing between low bias / current noise and low voltage noise would be relatively simple.

Changes to the transformer are pretty much out of the question. As I wrote above, it is sealed, probably potted and unobtainium if I break it. My goal it to make as few changes as possible. The ckt works well if the chopper works.

High frequency noise is (probably) not a problem because the output is an analog meter. Drift, however, is a major concern. Also effects on the measured circuit are a concern. Fluke put special effort into controlling leakage current. All of the parts in the input conditioning string on the board are on teflon standoffs. I don't know if it was to protect the rest of the card from the input or to protect the input from the rest of the card. Given that the input is normally 1mV FS or less, I suspect the latter.

As to what makes the supply weak, that is a very good question that I haven't looked into. According to the transformer ratios (and ignoring the loading which makes the whole calc 'how ya doin') the low voltage should be about 1/6th of the lamp ckt voltage using the ratios on the drawing. One measurement of the lamp ckt was 104.5V giving about 17.5V for the low voltage. Actual measurement was 14.5V. At the same time, the DC voltages were +11.79 (about 20% low) and -16.13V (about 9% greater than spec). The PS caps were checked about a year ago and found to be OK.

I started with the idea of replacing everything between from R4 on the left to C15 on the right with a chopper amp but opted to try for a simple fix first. I'm also trying to keep the beast as stock as possible.

[Kleinstein]

The Neons and a possible LED replacements are isolated from the rest of the circuit. So one could even power LEDs from the primary side - so the extra current would not need to go through the transformer at all. I think the glass rods should give enough insulation.

Already 6 mA should be plenty of current to power LEDs. High efficiency orange LEDs could be pretty good. One could even just compare the visibly intensity to the neons:  if the LED with some 3 mA is brighter as the neon it should be OK.

[Fred_47]

The Neons and a possible LED replacements are isolated from the rest of the circuit. So one could even power LEDs from the primary side - so the extra current would not need to go through the transformer at all. I think the glass rods should give enough insulation.

Already 6 mA should be plenty of current to power LEDs. High efficiency orange LEDs could be pretty good. One could even just compare the visibly intensity to the neons:  if the LED with some 3 mA is brighter as the neon it should be OK.

I'll have to trace the circuitry to see if this is possible but I know that the common is shared on the detector board ground plane and I suspect that I would  also have to make changes off the detector board which I want to avoid. If possible it is probably a good solution.

I suspect the the length of the light pipes is mainly to keep the radiated noise from the neons away from the amplifier input.

I suspect that the LED could be dimmer as long as both photo resistors are illuminated the same, i.e. , produce the same resistance when on.

[Kleinstein]

The is no need to have the same intensity for both LDRs. The copper does not need exactly same resistance, though it can help a little if there is an input bias from the amplifier.

So I see no need for a large change to the board, except removing the neons and wire the LEDs with a resistor. One may even wire the LED directly in place of the neons:  The diodes may have to change polarity (or be shorted) as the 20 V winding could have different polarity. The series cap for the neons would go out for a bodge wire to the 20 V winding. The resistor may need to be a little smaller.

Powering the LEDs from the primary would need2  extra wires to the transformer, that is not on the board. There is no need for ground connection for the LEDs (just the 2 LEDs anti-parallel and a series resistor). So the LED part would be floating. This could add some capacitive coupled common mode background.