Fluke 8050a True RMS U32

[tango17]

I found some TI LM3046s that should work for the CA3046. The CA3140 is available from Digikey.

[tango17]

I will admit that one of the reasons I started with the 3046 was it was easy to access. Easier than any of the other chips. I didn't have to remove the u32 circuit board from the main PCB in order to desolder the 3046. It's sort of like the drunk who searched for his lost wallet under the streetlamp even though the wallet was lost elsewhere simply because the light was better under the streetlamp. None the less the 3046 has now been removed and I'm pleased to report that two of the five transistors are definitely bad.

The bad ones were the transistor closest to the diode pair near the input and the transistor closest to the output.
Now I need to put together an order from digikey for all of the parts I need. I hate paying shipping cost for a 50 cent part. I might as well order the other parts as well just in case they need to be replaced also.

[ModemHead]

None the less the 3046 has now been removed and I'm pleased to report that two of the five transistors are definitely bad.

OK I'm a little late with this, but I'll post it for completeness.

Meter input remains at 1.00Vrms, at 200Hz.  The 8050A display reads 1.0007.  Scope settings remain the same.  I found my non-battery option 8050A and switched to it so the probing is easier.

Pin 1 of the 473777 is the module output at 1.069 Vdc.  The output of the other opamp on this chip is -1.805 Vdc.  All the inputs are very near 0V, with no AC signals present.  The -5V supply is -4.98V and the +6V supply is at 6.24V.

The transistor array pins have interesting signals which I've captured.  I'll call it U3.  Pins 6, 8, 9, and 11 are at 0 Vdc.  Pins 2 and 9 are both at -0.657 Vdc.  Pins 4, 5, and 7 are all the same signal.

I think it's safe to say the circuit does not follow the block diagram exactly. The input goes into the LF351 where it looks like the log(Vi) is done.  The CA3140E apparently then does the absolute value.  There's still an antilog and summation happening somewhere, but I don't know where exactly.

Hopefully the dead transistors will be your only problem...

[tango17]

Success! Just so I'd know if I needed to buy more chips, I cut off the legs of the two bad transistors, hot melt glued a couple of spare NPN transistors with similar specs on top of the 3046 and bent the legs to match the SOIC pins I'd cut off.  I soldered in the whole contraption and I'm now getting output that's looking pretty good.   I need to check calibration and make sure that it works on all ranges, but it was displaying values close to what I expected with the few tests I had time to make.

Thanks for the help, everyone, particularly the "beyond the call" effort from Modemhead with all that recent work. The information that there was no signal at the input of the 3140 helped me decide the problem was probably in the feedback  loop.

I still want to check signals around the "leaky diode" problem area to see if that's been fixed, and I'll compare my signals to those Modemhead just posted.

For the record, the bad transistors were on pins 6-8 and 12-14. I'm going to replace the chip anyway.

[ModemHead]

Excellent!  Great job on the diagnosis and repair.  Pleased to contribute a bit, sometime in the future someone will google "8050A RMS converter" and find some good info here.

Regarding the silver solder, the only thing I know about that is the ceramic terminal strips in old Tek scopes are supposed to be soldered with silver-bearing solder, because regular 60/40 can weaken the bonds between the metal contacts and the ceramic material.  Not sure if that applies here.

[tango17]

My report of success was somewhat premature. I had the scope on the 2 volt/div scale when I thought it was on the 1V/div scale. Everything looked more or less correct, but in reality, the repaired 8050a was reading about 50% low on all AC scales.  By chance, the numbers seemed to match up.  I'll be using Modemhead's numbers to track down what's going on.  At least I've got something other than zero on the display.

[tango17]

I took some readings from 1 volt to 200 VAC on another meter and the 8050a, then X-Y plotted them in Excel. They lie accurately on a line, so at least I know the error is just a linear scaling error.

[wiss]

You must get really bad matching-errors from the two hacked-in transistors!
These circuits are annoyingly temperature dependant. That's why no one try to build one with discreet transistors.
Even when you get a 3046-replacement I would guess that adjustments will be needed.

[ModemHead]

Even when you get a 3046-replacement I would guess that adjustments will be needed.

For those who haven't seen the schematic, there is a configurable voltage divider and trim pot on the output of this module. Hopefully when the matched array is installed the output will fall within the expected output range.

[tango17]

Yes, there are 4 resistors for the DC calibration, and two more resistors and two trim caps for AC calibration, plus one adjustable resistor on the U32 TRMS board itself.  I've been looking at the calibration procedures, but I decided not to touch anything until the transistor array is installed. I ordered three of them today. With luck, one of them will bring it within calibration range.

Again, I'd like to express my thanks for the help.  I'll let you know how it goes.

[tango17]

The 3046 came in yesterday and I removed the hacked transistors and bad CA3046.  The new chip was an LM3046 in an SOIC-14 package with 0.05" pitch  The old chip was a surface mount DIP package with 0.1" pitch, so the first step was to solder on 14 lead extensions and reform the SOIC to DIP shape. There's probably an adapter sold that would make this job easier, but I couldnt find it.

This morning I was ready to install. After some tricky soldering, with the U32 hybrid board still installed, I turned it on and ran ACV tests. The results were all within 1 percent of my best meter from 2 millivolts to 146 volts at 60 Hz. That meter (3.5 digits and purchased in the 1980s) only has 1.2 percent accuracy plus or minus 10 counts over the test range, so that looked pretty good to me. I'm not going to touch any of the 8050a AC calibration until I get access to a better meter and test equipment.

I guess that constitutes success!

Thanks for the help, guys!

[ModemHead]

Nice thorough job on the repairs, I hope you get years of service from your "new" 8050A.  Mine stays on 24/7, it's always ready for a quick measure or for monitoring something for hours with no battery-saver timeouts.

[wiss]

 

I would be surprised if the cal was dependant on the parameters of the 3046, that would mean lots of thermal drift.

[harrimansat]

RMS board with the same ic

[gearhead]

RMS board with the same ic

WHAT UNIT (TRUE RMS BOARD) IS THIS SCHEMATIC FOR?

[Swainster]

I didn't post the schematic however from the drawing number it looks to be from an advantest/takeda rikken multimeter.

It helped me out with understanding the converter on the F8050A as it's a pretty similar design. In my case my AC problems were caused by a faulty LF351 on the main board, plus a bad CA3046 on the hybrid. Working on the hybrid is tricky - probably best to use a heat gun or hot plate. I managed to pull off one pad but luckily I could pick up the circuit at a nearby test point.

[harrimansat]

TAKEDARIKEN TR6843