Вставка
- Опубліковано 18 кві 2024
- In this episode Shahriar repairs a faulty Agilent 3458A. These highly desirable multimeters have continued to climb in price in the past few years. The last time The Signal Path has repaired one of these units was back in 2018.
This instrument generates several errors both during power up as well as Auto Cal procedure. The error points to a failure in the OHMS measurement mode. Detailed instrument block diagram and schematics are examined. Measurements also show incorrect DC current sourcing during OHMS measurements and the associated circuits are closely analyzed. The problem is traced to a quad-comparator IC. Replacement corrects all problems during ACAL. The battery backup of the SRAM ICs are also replaced to eliminate non-volatile errors. The IC of the comparator chip is also extracted and examined under the microscope.
The repaired instrument performance is verified by using voltage, current and resistance standards.
Please consider supporting The Signal Path:
www.TheSignalPath.com/Donate
/ thesignalpath
Other links:
www.TheSignalPath.com
/ thesignalpath - Наука та технологія
It's fun to see this circuit board and schematic again after so many years. I worked down the hallway and knew the entire design team. It's a testament to the strength of the original design that this instrument, with a few updates due to the age of the original parts, is still in production under the Keysight brand name. I believe this might have been their first voltmeter with a multi-slope + flash conversion architecture, and they knocked it out of the park.
So cool seeing people from HP in the comments. Same with IMSAI guy, love hearing the inside baseball from what I see as a golden age of test equipment
Is Scott Stever still alive?
I'd say the meter itself is excellent, however the standard ref used is rather average. I knew a guy in Poland who made a modified ref for it (build upon a few special versions of the Adr1000's if I remember correctly), along with some changes to the original circuitry, achieving stability way below 2ppm, practically better than the fluke dmm.
@@user-ui2fr2fq1w Does this guy in Poland have a project page or video of this modification?
@@user-ui2fr2fq1w: I don't know about Scott; his LinkedIn page is still up. Scott managed the 34970 DAQ project, one of the highest-volume instruments Loveland ever produced. If memory serves, the 3458 project was managed by Larry Desjardin.
Fluke 233 always look so bizzare in video :). Sockets in 3458A analog boards make me anxious, hah. Also good idea to run ADC A3 drift test to see if this 3458A needs new A3 or not. I'll send you the automated script for it as well. For measurements above 1kOhm you'd best use OCOMP ON with DELAY 0.1, otherwise some offset can be present if 3458A was adjusted before using OCOMP ON function.
P.S. I am cooking a beginner's guide about 3458As, so stay tuned for that. There were multiple revisions of various boards over the 38 years of 3458A project life, but boards are compatible between each other and still follow same analog design as was originally done by HP guru's in late 1980s.
Yeah, the meter always does my head in.
I'm convinced Fluke did this purely to mess with everyone's heads
Loved it. I don't think I've seen a failed comparator IC before this. It's good to see a wonderful instrument like this returned to service.
When I diagnose an instrument like this, I usually check if the PSU or something else simple is bad. It's easy to check. If the title starts with TNP you know it's that 😄
Thanks for the video as always Shahriar.
Shahriar makes the repair look easy...it's not. You need to know how to analyze the circuitry without guessing...thanks for another informative lesson.
Shahriar, the repair was done quite well.
Anyhow, before replacing the batteries, one should always download the complete content of the calibration RAM via GBIB, even if there is no officially known method of writing it back for this assembly. At least, one should read the 10V and 40kOhm calibration constants, CAL? 1,1, CAL? 2,1, so to have the chance to iteratively reproduce the original basic calibration.
The CAL 0 calibration can be performed by anybody, and SCAL might be omitted, if high frequency ACV measurements are not needed.
I miss DIP´s.. I Appreciate your working through the troubleshooting logic.
Thanks a lot. I always enjoy a good repair with your detailed explanations. And there is always something new to learn beside the troubleshooting.
When I saw the CD4094 I had to think of a glitch my college and I where hunting on one of our signal generator cards. In the end it was an 74HCT137 which had a single output pin glitching from time to time. I would never have guessed an 74 chip to cause those glitches but I was lucky and had a true wizard by my side... he is the one who found the faulty chip but was similar surprised that it really was the culprit after putting in a new one.
I still have the culprit on my desk and I'm still waiting to have some time to try and get the pin glitching on my breadboard setup.
I enjoyed the analysis of the silicon thanks. Some dimensions of pads and features would be interesting for these older devices.
Sure. Next time. :)
Wonderful thinking way for details and repair, love to follow it.
Oooh. I didn't realise there are schematics in the wild for the 3458A. That's great :)
Agreed the broken test gear prices are ludicrous now, so many items are now well out of my budget compared to being affordable just a few years ago, it is making it hard to buy equipment just to make videos about fixing it now.
Marcus Lorton got one about ten years ago for the "learning experience"...it has option 002 on it and it had less than 50 hours of use. He said he paid a bit less than usual for one with option 002.
Batterie backed-up factory calibration is evil...
Unfortunately it’s just good business sense 😂
I suspect that most highly integrated ICs change their specifications due to aging due to ion migration. These are often only small deviations, but they add up over time so that even the calibration is no longer able to compensate for them.
Hope it isn't the slope ADC that died, Marco Reps had a repair of one of these and they're not cheap, but seem to always break.
Or we only hear about the broken ones that were rejected and ended up on ePay ;) Observer bias, but indeed its not too uncommon for old units, mostly from HP era.
In a way it’s a shame it’s a newer digital board, the older type with the Dallas non-vol rams allow swapping them out for Fram ICs so no more batteries! Not sure you can do the same with this board.
It is still SRAM, just surface mount with pluggable battery, should still be able to adapt FRAM to it you would think...
did you have to wash the board with IPA after soldering those sockets?
The fluke. 233 is absolutely insane. Looks like someone took a hacksaw to a 173 and called it a day
Are there any differences with the keysight versions of the 3458As?
The digital board changed over the years. The black 3458a has some differences. Take a look at the 3458a video from Marco Reps
@@ofgjf nice will check it out. thanks.
I need to send my meter in to do a video on it cause these videos are to easy of a repair
What’s the problem with yours?
That was cool
Brilliant Error localization logic Sharimar ! I have a HP3478A that shows correct output currents but negative resistances at low ohms ranges.
Is there any other possibility than wrong calibration constants ?
Video 6:47 says that when the 3458A measures 1Gohm resistance, it will use 500nA. But 1G ohms x 500nA = 500V.
Is there something wrong with the manual?
All HPAK bench DMMs use an internal 10MOhm resistor in parallel with the DUT for ranges 100MOhm, 1GOhm, and calculate the DUT value from this parallel circuit.
Therefore, these 10MOhm x 500nA = 5V will never be exceeded.
You can find a detailed description for this method in the HP34401A Service Manual.
@@DrFrank-xj9bc Thanks a lot. I found the instructions in the service manual. The manual points out that when the ohmic current source output voltage exceeds the normal value, it will become non-linear, so a 10M resistor is required in parallel across DUT.
You've got to wonder if there's a modern equivalent to that LP365N...
Would it still be possible for you to fix this device if there werent any documentation like circuit diagram etc.?
It would have been significantly more difficult...
24:35 National Semiconductor logo in bottom left
Didn't the guy on CuriousMarc just throw the chip in a frying pan to soften the plastic and just snap the plastic off, ie, no acid.
You can put the chip in front of a hot air gun at 500 C for about 20 seconds and then break it with some pliers. But you should only do that if you have multiple parts to try it on, there's a good chance you'll break the die.
Do they still use the coax capacitor?
I think so.
You should be able to calibrate at least 2-3 more digits of accuracy.
Calibration is not the issue, I need more NLPC and guarded cabled to get the stability required.
No way that comparator is six metal layers. This is an old school chip, mostly hand layout probably (not sure if rubylith or CAD based, could be either). You can see that the actual comparator tiles are the same layout but the IO pad ring around them is custom routed (not copies of the same structure).
Six masks I could believe, though. It looks like (starting from the topmost text = lowest layer) three grayish layers that are probably related to doping or isolation, one contact/via cut mask, one metal layer, and finally cuts for the bond pads.
There might also be a 7th mask, looking just above and right of the first six near the NatSemi logo there seems to be a barely readable A as well. Maybe that was something done early on in the wafer processing before an etch process or implantation step or something? I'm a lot more familiar with modern planar CMOS than ancient BJT based tech made in a foundry that's probably older than me, so not sure of the exact flow for making them.
I meant layers, not metal layers. Doesn't look like it has more than 2, maybe 3 metal BEOL.
By my eye it seems to have 1 layer of metal interconnects. And then rely on doped silicon for any jumpers.
However, I somewhat doubt that this is an old chip. It has a ©mark on it, instead of the usual encircled "M" used back in the day. (Likely made in the 90's or early 00's)
And that modern nature makes me think it only has 1 mask for each layer. Since copy paste in CAD software is a thing, it is most likely not hand laid.
The diagonal lines also seem thicker than the vertical/horizontal lines (indicating that the edges are defined on a grid), hand laid masks usually keep constant width. This is most notable on the thinnest traces. (Also, I don't see why a hand laid mask would use so many differing line widths, have sharp pointy corners, and even the rounded corners seems odd for a hand laid mask.)
So "one contact/via cut mask, one metal layer, and finally cuts for the bond pads." sounds reasonable. And the three layers before is as stated, likely just doping/etching steps. And yes, there seems to be an etched 7th A at the top there. A bit offset to the right however...
good evening friend, I saw a video on your channel that you handle rf radios like erisson... I work in a telecommunications company and I repair those boards electronically... I have a fault in an ericsson rru 01b5 that does not raise the pa in the manager I get ( dependency failed) and it doesn't give tx or rx....could you please help me please since my work depends on that
Lol, a real repair this time (and not a faulty connection or something similar) nice one!
Out of the last 12 repair videos, only one has been a faulty connection.
Why does it have so many digits when it is so inaccurate? Is it at least precise, so you can compare tiny differences? Can it be calibrated to only have errors in the least significant digit?
I need more NLPC and guarded cabled to get the stability required.
*Abstract*
This video documents the repair process of an Agilent 3458A multimeter exhibiting several errors during power-up and auto-calibration. The host, Shahriar, systematically diagnoses the issue by analyzing the instrument's block diagram, schematics, and conducting various measurements. He identifies a faulty quad-comparator IC as the root cause of the problem, specifically impacting the ohms measurement mode and current sourcing. After replacing the IC and the SRAM backup batteries, Shahriar verifies the repaired instrument's performance using voltage, current, and resistance standards, confirming successful restoration of functionality.
*Summary*
*Introduction and Initial Diagnosis*
* 0:07: The video introduces the repair of an Agilent 3458A multimeter, highlighting its value and recent price increase.
* 0:30: The multimeter displays errors during power-up, prompting an auto-calibration (ACAL) procedure.
* 1:02: ACAL fails, indicating a hardware failure specifically related to the ohms measurement mode and internal overload.
*Circuit Analysis and Troubleshooting*
* 1:30: Additional errors reveal a depleted battery backup for the SRAM, requiring replacement.
* 3:37: The multimeter's block diagram is examined, focusing on the ohms measurement and auto-calibration sections.
* 4:20: Shahriar explains the principle of ohms measurement using precision current sourcing and voltage measurement.
* 5:27: The schematic for the ohms measurement ranging circuit is analyzed, highlighting the different current ranges and switching mechanism.
* 7:39: Observations point towards potential issues with the current sourcing and switching circuitry for the higher ohms ranges (1 mA and 10 mA).
*Identifying the Faulty Component*
* 12:09: Measurements confirm incorrect DC current sourcing during ohms measurements, deviating significantly from expected values.
* 14:01: Testing of the burr-brown op-amp (u300) reveals it to be functioning correctly.
* 15:10: Analysis of the FET gate signals confirms that the switching mechanism works for lower ranges but fails for the 1 mA and 10 mA ranges.
* 17:19: The investigation focuses on the comparator IC (u303) responsible for the faulty ranges, suspecting it as the primary culprit.
* 19:05: Measurements on the comparator IC reveal abnormal input voltage levels, further suggesting its malfunction.
*Repair and Verification*
* 19:53: Sockets are installed to facilitate easy replacement and swapping of the suspected comparator ICs.
* 20:26: Swapping the comparator ICs confirms a difference in behavior, solidifying the diagnosis.
* 21:01: Replacement of both comparator ICs with new ones restores proper current sourcing in the ohms measurement mode.
* 21:38: The multimeter successfully completes the ACAL procedure without errors, indicating successful repair.
* 22:17: The SRAM backup batteries are replaced to eliminate non-volatile memory errors and ensure long-term data retention.
*Performance Verification and Conclusion*
* 25:52: The repaired multimeter's performance is verified using precision resistance, voltage, and current standards.
* 26:01: The ohms measurement accuracy is confirmed across various ranges.
* 27:25: The voltage measurement accuracy is verified using a Fluke 754 process calibrator.
* 28:20: The current measurement accuracy is validated, demonstrating the overall success of the repair.
* 29:06: The video concludes with Shahriar expressing gratitude to his supporters and hinting at future repair projects.
i used gemini 1.5 pro
Token count
13,871 / 1,048,576
Moose!