-
Hello, I'm repairing a Keithley 3330 LCZ meter. I found that the output voltage is only 1/3 of the configured output voltage. It outputs 336mV when it should output 1V and it outputs 16.84mV when it should output 50mV. Both cases only about 33.6% of the configured voltage is output.
I have the service manual but there is no schematics so difficult to troubleshoot without taking the PCB out. There is a block diagram with an output buffer amplifier, which could be the culprit.
If anyone has the schematics I would greatly appreciate to get a copy!
Thanks.
/John. -
Hi to my knowledge i never saw an complete service manual, try to ask Keithley ??
There is a thread here with calibration procedures and software ...
https://www.eevblog.com/forum/metrology/keithley-3330/
And this web page further down as a story for the 3330 lcr may give you some small hints ??
https://dabbledoo.weebly.com/impedance-measurements.html -
Even the brothers like the 3321 3322 have no schematics in them
-
Hello, thanks for the info. I have removed the main PCB and trying to make sense of it all. I will attempt to replace the output driver transistor to see whether that increases the output voltage. Reverse-engineering without schematics is maddening but i may have to go down that route....
-
After a lot of probing and testing I have found a bad output transistor: Q301 PNP 2SB1414 has a 44 ohm short between emitter and collector, regardless of polarity applied. It is used in a complementary pair with Q302 NPN 2SD2134, which tested ok. I have not yet figured out the exact circuit configuration of these transistors as I have no schematics.
So, it seems I should either find the exact same part (2SB1414) or replace both with equivalent parts that are designed to be in pairs.
I suppose a next step could be to deep-dive into the datasheet and try to find equivalent transistors on Digikey.
I found some sellers of 2SB1414 in China but i'm not sure I can trust that i'm getting a properly functioning part.
https://www.ic12.com/product_info.php?products_id=320
Has anyone bought from IC12.com? Is it a trustworthy vendor?
Thanks,
/John. -
Hi, it's my 3330 on the Dabbledoo web pages!
I did not bother finding the exact output transistor models, but used BD139 and BD140 instead.
Interestingly, the output transistors in my unit were not 2SB1414 and 2SD2134, but 2SA1428 and 2SC3668.
At the same time, I replaced the output driver's op-amp U302 with an NE5532. I added a socket for the U302, just to make it easier to replace the U302 again, if needed.
In the input circuit I removed all four protection diodes, of which at least one was blown, and added two new ones (there seems to be two times two diodes in parallel), just common 1N4148. The input op-amp U401 I had to replace with an NE5534, and I added a socket to ease future repairs, if needed. I also replaced the Q410 and Q411 (I cannot recall exactly why) with common BC547 and BC557. Mind you that the legs of the BC547 and BC557 have to be twisted in order to match the pin-out of the 2SC1815 and 2SA1015 they replace. Check out the enclosed pictures :-)
-
Thanks Electrole; very much appreciated. Based on the application (very low frequency and low test voltage) the parts you proposed should be just fine. However, as i'm still not sure of the exact circuit configuration, I prefer to use parts as close to the original as possible. I found the op-amps on Digikey (AD826, AD817, both expensive) and similar transistors (150V) in TO-220. But I also ordered the exact parts you used as very low price. I already have those 1N4148 diodes in my lab. My diodes test good, as far as I can tell in-circuit.
It seems to me that the designers of the 3330 have used parts with way higher specs than actually needed. The transistor rating of 150V may perhaps be useful to withstand connection of charged up high-voltage capacitors. However, I don't see why they have used op-amps (AD826) with 50 MHz bandwidth and 350V/us slew rate when the maximum test signal is only 100 KHz sine wave. It almost seems that they used whatever parts they had on hand. That may explain why your unit had different output transistors than mine. I also found that the my unit never had the Q410, Q411 transistors assembled. I have the footprint on the PCB but they were never soldered in. Any idea what those transistors are used for?
My unit has serial # 2254.
# PRODUCT DETAILS QUANTITY AVAILABILITY UNIT PRICE EXTENDED PRICE
1 497-6657-5-ND BD140 TRANS PNP 80V 1.5A SOT-32 2 Immediate 0.57000 $1.14
2 497-6656-5-ND BD139 TRANS NPN 80V 1.5A SOT-32 2 Immediate 0.60000 $1.20
3 AD826ANZ-ND AD826ANZ IC OPAMP VFB 2 CIRCUIT 8DIP 2 Immediate 9.00000 $18.00
4 AD817ANZ-ND AD817ANZ IC OPAMP GP 1 CIRCUIT 8DIP 1 Immediate 6.31000 $6.31
5 296-7042-5-ND NE5534AP IC OPAMP GP 1 CIRCUIT 8DIP 1 Immediate 1.10000 $1.10
6 296-1410-5-ND NE5532P IC OPAMP GP 2 CIRCUIT 8DIP 2 Immediate 0.56000 $1.12
7 KSA940TU-ND KSA940TU TRANS PNP 150V 1.5A TO-220 2 Immediate 0.68000 $1.36
8 KSC2073TU-ND KSC2073TU TRANS NPN 150V 1.5A TO-220 2 Immediate 0.65000 $1.30
p.s. Tack sa mycket for hjalpen. Jag ar Svensk men bor i Kalifornien sedan 2000. Jag jobbade i Kopenhamn runt 1999. Hav det godt. -
Hej John!
There's something interesting going on with the selection of transistors and op-amps in the 3330! The op-amps in my 3330 before I replaced them were: U302 (part of the output buffer) = NJM5532D from JRC (replaced with NE5532), and U401 (the input current converter) = NJM5534D from JRC (replaced with NE5534).
Furthermore, the NJM5532D is a dual op-amp, and the two op-amps you mention are single op-amps. Which op-amp is AD846 and which is AD817 in your unit? Could you perhaps take a close-up shot of the areas around the output buffer and the input stage?
The AD846 is a current feedback amplifier, and one may not just replace a voltage feedback amplifier with a current feedback amplifier without looking into the details of the schematics. The NE5534 could likely be replaced with an AD846 if there's no capacitive feedback. If the AD817 is used for the output buffer, then Keithley/NE must have made some changes around the buffer's circuitry.
The serial number of my 3330 is N01150, so this could suggest that Keithley/NF replaced the buffer and input op-amps later, and even redesigned the circuitry, perhaps as an effort to improve the performance of the 3330 at the highest frequencies.
The fact that the Q410 and Q411 transistors were never mounted in your unit is further evidence of design changes. I have not reverse-engineered the input stage, but my first guess would be that Q410 and Q411 are used as output transistors to lower the output impedance and/or to increase current drive capability, and thereby improving the transimpedance conversion. It's likely that an AD846 does a better job alone as its open-loop transimpedance is quite high. Also, the additional phase lag of a pair of output transistors could be an issue.
In your case, I would try to use the same AD846 and AD817 op-amps as replacements. They are fairly wideband devices, and replacing them with other models could potentially spell trouble, or loss in performance.
If there are other owners of a Keithley 3330 out there, perhaps you could check which op-amps are used for the output and input stages in your unit?
With regards to the output transistors of the output buffer stage, I do not think the 150 V rating was any requirement. I would not connect the 3330 to a charged capacitor with even a mere fraction of that voltage. However, the specified typical transistion frequency fT is 200 MHz for both the 2SD2134 and the 2SB1414, which is higher than the 100 MHz for the 2SC3668 and the 2SA1428. Perhaps Keithley/NF found out that transistors with a higher fT were required when using another op-amp for the output stage? The BD139 and the BD140 have a specified ft of 190 MHz and 160 MHz typical, at least when specified by NXP.
BR, Ole -
Hi Ole,
Thanks for your response. Note that my op-amps are AD826, AD817 (dual and single) so likely compatible with the 5532 and 5534, respectively. The specs between yours and mine are quite different (mine are more high-end) but, since lacking schematics, i'm not sure its relevant. I have attached a high-resolution image of my board.
Since i'm waiting for the replacement parts, I'm thinking of de-soldering my output and input op-amps so I can wire them up on a breadboard to see whether they perform properly. I don't like to replace parts randomly on a hunch and I'll learn more by properly proving whether the parts are good/bad. This one is a hobby project, after all, so I don't mind going slow...
/John.
-
Oh, why did I read "AD826" as "AD846"?
It all makes more sense with the AD826!
Thanks for posting the picture!! It definitely shows that BOM changes were made, and that components were removed in the transimpedance input amplifier circuitry.
For comparison I have enclosed two additional pictures, one of the input stage and one of the output stage of my repaired 3330.
Good luck with your 3330...!
BR, Ole
-
I have made some progress. With the output transistors replaced, I'm getting the correct 50mV and 1V RMS output voltages. I now got stuck in the calibration procedure. The "Calibrating Frequency Response of voltage monitor" stage ran for 1 min 55 sec and then the "Check multimeter's operation and Check if signal cable is connected" error message is displayed by the cal program.
While it is running i can see that the cal program is correctly controlling both the 3330 and the DMM via GPIB. The DMM shows voltages around 1.000V and the 3330 goes into "remote" mode and also changes output frequency programmatically. The 3330 shows a B display voltage slightly lower than 1V (around 980mV). As the pictures show, i'm using an 8846A but with the 8842A command set enabled. There are no errors in the 8846 error log.
From here (https://dabbledoo.weebly.com/impedance-measurements.html) I figured out that I needed to connect the HCUR and HPOT together. Without that, the multimeter calibration phase fails after just a few seconds. I'm wondering whether the calibration manual omits something else that is important?
I would appreciate if anyone that has managed to run the complete calibration would clue me in whether i'm doing something wrong
The failure occurs exactly when the B display changes voltage measurements from some 980mV to 44.4mV. So, perhaps, the 1V range calibration passed but for some reason the 50mV range fails off the bat? Note that the LEVEL LED is still set to 1V in the picture.
Thanks,
/John.
p.s. I put the scope on the output and the cal program steps through a number of test frequencies (40Hz, 100Hz, 1KHz, 5KHz, 15KHz, 20KHz) with the 1V RMS output amplitude. Then a relay clicks, the B display changes from 980mV to 44.4mV and the cal program displays the error message. The oscilloscope still shows a 2.96Vpp, 20KHz sine wave and the DMM shows 0.997V RMS. I'm not sure what the significance of the relay click and the 44.4mV B display has...
-
Hi,
Did you follow the instructions and connect the shell of the HCUR BNC to the Ground terminal?
See page 7 of the attached document.
Also look at page 8.
Regards,
Jay_Diddy_B
pa390a(Model3321_22_30).pdf (717.27 kB - downloaded 114 times.)
-
Hi John,
If the 44.4 mV on display B represents the level as perceived by the 3330 then it appears a bit too low. However, I cannot recall how display B behaved when I did the calibration.
I have quickly measured the levels from my 3330 (in cold condition) with an 8842A, just for comparison:
120 Hz: 0.99825 V / 49.827 mV
1 kHz: 1.00138 V / 49.985 mV
10 kHz: 1.00127 V / 49.981 mV
100 kHz: 0.99702 V / 49.527 mV
BTW: The test fixture I made (should be equal to the 'Special Jig') is an 'Open' plus a connector for the level measurement.
The output for the level measurement simply connects to the HCUR and HPOT connection. I have enclosed a picture with the lid off.
The fixture also connects the LCUR and LPOT terminals together. That is at least one thing which differs from the picture of the 3330 you have posted.
Another idea is that the 8842A emulation by the 8846A is perhaps not 100% true to the original behaviour?? Do you know someone who has an 8842A?
Mvh. Ole
-
Thanks Jay & Ole. My connections are correct but I still get the same error. I figured out that I could log the last (or possibly all) GPIB commands sent during the calibration with the APPMON utility. I have attached the log. I now have to learn something about GPIB commands to decode all this. I briefly considered buying an 8842A but I had to pay $400 USD for one with AC voltage measurement capability. Too much since I already have several DMMs. I have a hunch that the 8842A emulation in my 8846A should be correct but I will have to see what the GPIB trace tells me...
Thanks,
/John.
-
Hi John,
I managed to get as far as decoding a fair amount of the Keithley 33xx GPIB commands used during calibration, and may (untested) have the abilty to dump and restore the calibartion on my 3322. See below for the commands.
https://www.eevblog.com/forum/metrology/keithley-3330/msg2376504/#msg2376504
That should help you work out the comands in your log file. If you are able to get a full calibration working, I would live to see the full GPIB log - I can then probably get some more modern tooling built to do calibrations.
Cheers,
Ash. -
I decided to go through a calibration of my 3330 this afternoon, but my old NI GPIB board does not seem to support GPIB analysis, so there's no GPIB traffic I can report
I noticed the levels on display B were lower than measured by the 8842A, but as far as I could tell not as low as 44.4 mV.
Mvh. Ole -
Thank you both.
I finally got my unit calibrated! But i'm not sure what made the B display suddenly display the correct 1V voltage, as configured in the 100 KHz mode. I suspect relay K401.
This morning, experimentation showed that the 1V mode output was correctly 1V RMS in the 100 KHz mode BUT the ADC thought it was only some tens of mV, as displayed on the B display. I noticed that relay K401 near the ADC was activated when switching from the 10KHz to 100KHz range. So, I wanted to probe the input to the ADC (U422/AK9201A-VP, pin 19). While probing I observed that no input signal to the ADC was available while in the 100 KHz mode. However, after switching the relay some 10's of times, the signal suddenly appeared and the B display displayed correct voltage. So, I'm suspecting that the K401 relay (TQ2-12V) had dirty contacts. Or some part behaved differently when heated up. Or some cold solder joint behaved differently after the PCB warmed up. I'm not entirely sure but from past experience, bad relay contacts on old instruments is quite common and can have similar effects.
I found a link to a schematic containing the ADC here: https://www.eevblog.com/forum/testgear/$50-for-schematic-of-hp-4338b-milliohmmeter!/
-Note: For a completely different product but I learned about the ADC pin-out.
With the 100KHz/1V mode working, the CAL program finished without further problems. I quickly tested some inductors and capacitors and the 3330 is working properly. The power on self test now passes okay. I have saved the calibration constants for later use, if needed.
I will check how the unit behaves tomorrow morning after having been being turned off for a long time. I might also order a couple of new relays since there is also one (K301) between the HCUR and HPOT connectors. They are only $3.38 each. If the issue comes back I can at least rule out bad relay contacts. https://www.digikey.com/product-detail/en/panasonic-electric-works/TQ2-12V/255-1002-5-ND/251774.
I also noticed that the 1V output in 100KHz mode is slightly lower than the other frequency modes. Both display B and my external DMM agrees with regards to RMS voltage.
120Hz: 998mV, 1KHz: 1.000V, 10KHz: 995mV, 100KHz: 981mV.
I also attached a full view of the PCB.
One question I have: Will it matter whether I calibrate using a Kelvin clip cable?
open: Leave alligator tips unconnected.
short: Short together alligator tips.
Resistor measurements: Connect between alligator tips.
I then also don't have to create any special jig and put my reference resistors in boxes with BNC connectors.
I plan to use the instruments with these alligator Kelvin clips so i figure the calibration would be more accurate if calibrated with the same cable?
https://www.ebay.com/itm/Kelvin-LCR-Test-Tester-Probe-Lead-Clip-Cable-With-4-BNC-For-LCR-Meter-Terminal/253809323050
/John.
-
Fantastic development! Well done!
Relays are one of my hate/love objects: They definitely have their advantages, but they may also be a recurrent source of trouble.
With regards to the calibration and open/short compensation of the 3330 using a set of Kelvin clips: Yes, that should be fine, and it may even be an advantage if you plan to explicity use the very same clips for your measurements. However, the open compensation should ideally be done with the same distance between the tips as you intend for the measurement. The capacitance between the tips obviously varies with the distance and can approach a couple of pF when the tips are close. The pictures I have enclosed show the capacitance at different distances between the tips of my Kelvin probe set. The open compensation was done with the tips about 20 cm apart. In order to get a low measurement uncertainty with a set of Kelvin probes when measuring small ceramic capacitors, for instance, one has to carry out the open compensation carefully.
Again, congratulations with a working 3330 !
Mvh. Ole
-
Thanks Ole for the cal tips.
I did another calibration since the 'refdata' file in my original test calibration was not exactly set up properly for my Vishay S Series film resistors, which i'm using for my calibration. The prior calibration was just quick and dirty to ensure it is actually working. This one I took a bit more care in performing.
I then played around some more with the 3330, checking against my DE-5000 LCR meter. The values are very close when measuring inductors, resistors, capacitors. Mostly, within 1%. I am however sometimes seeing a slight difference (like 1.4 ohm for 3330 vs 1.46 ohm for DE-5000) in ESR film cap (100nF) measurements but I suspect it is because the DE-5000 is not outputting exactly the same frequency as the 3330 (the DE-5000 is slightly high). It turns out that the ESR varies widely with frequencies on both meters, something I have not thought of before. The 3330 also outputs 1V vs the DE-5000's 500mV. Perhaps that effects the readings, too. I will have to read in on the details.
So, i'm getting close to be able to play around with some in-depth physics experiments of passive components, which is why I bought the unit in the first place.
(Next repair project: EIP 578 - 26.5 GHz microwave counter)
Mvh, best regards,
John