Keysight 34465A reliability

[KeithFisk]

I have a Keysight 34465A which has just died with a power supply related fault on the front panel control board. What should be 3.3V is only 2V so all that happens on power up is the fan runs and LCD backlight comes on. Before I offer up bodily parts to the local NZ agent in exchange for repair of this I'm interested in the experiences of others with the reliability of these DMMs.
I'm considering abandoning it and replacing with something else. The nearest equivalent 6.5 digit DMM of another brand that I can find is the Keithley DM6500. I originally chose the Keysight simply based on more prior experience (40+ years) with HP instruments than any other brand.

Thanks in advance for any thoughts
Keith

[HighVoltage]

From my own experience and from what I read, the 34465A is a very reliable instrument.

It would be interesting to find out, what is really broken on yours.

[jonpaul]

Bonjour. Sounds like failed PSU reg or perhaps shorted/overloaded bus.

1/ Recieved used 3365A 1 yr old last Aug, worked 12 hrs. Next turn on black screen. Not screen saver/off. Intermittent, banging case could temporarily fix.

2/ Contacted Keysight service USA, returned to factory under warranty,

3/ two days after, a brand  new unit received, with new CAL cert

4/  running perfect since Aug 2021.

Recommend to contact local Keysight service.

Bon chance


Jon PS:  Could use your dead  if we can figure how to ship to USA!!

[tautech]

The nearest equivalent 6.5 digit DMM of another brand that I can find is the Keithley DM6500. I originally chose the Keysight simply based on more prior experience (40+ years) with HP instruments than any other brand.

We have SDM3065X in stock for around $1200 NZ if you need a meter in a hurry.

[TheSteve]

I've seen a few of the 3446x series fail with a bad main processor. They typically get hot and drag down the power supply voltage. Generally they are quite reliable though.

[bdunham7]

I've seen a few of the 3446x series fail with a bad main processor. They typically get hot and drag down the power supply voltage. Generally they are quite reliable though.

That sounds like it might be the OP's issue?  Unless you are a full-time HPAK DMM repair specialist or work at Keysight, if you've seen two or more units with this issue I'd say that makes it a pattern failure of sorts.  Is there practical fix for this?  I'm guessing not...

[KeithFisk]

Yes, bad processor.  Service agent has reported back that it was sitting at over 100 deg C. Just waiting to hear back about availability of board before making the repair or buy something else decision.

Thanks for the replies
Keith

[SteveyG]

We've had 5 out of 12 purchased in 2019 fail the same way at work.

[KeithFisk]

Mine was purchased in early 2018 so out of warranty but probably had less than 100 hours powered on time. Perhaps if I had used it more it might have failed while still under warranty. The quoted cost of the repair including a recalibration is 60% of the cost or purchasing a new 34465A, 75% of the cost of a new Keithley DMM6500 and 130% of the cost of the Hantek 6 1/2 digit meter that you recently reviewed on your YouTube channel.  (All NZ price comparisons.) There is obviously a significant quality issue and once out of warranty it appears that Keysight don't give a toss. 

[kcbrown]

We've had 5 out of 12 purchased in 2019 fail the same way at work.

5 out of 12?!?

That means this is a consistent manufacturing issue that is worthy of a recall.  If Keysight doesn't perform such a recall and doesn't fix failed units for no charge (either of those actions would do), then they'll have basically lost any substantial justification for being regarded an A brand.

[KeithFisk]

We've had 5 out of 12 purchased in 2019 fail the same way at work.

Did Keysight provide an explanation for the fault and more importantly what they have done to stop it happening again?

[HighVoltage]

Mine was purchased in early 2018 so out of warranty

A few years ago, all warranties for all 344xxA DMM were extended because of the high voltage rating debakel.
So, there is a good possibility, that your DMM is still under warranty.
Or at least, I would ask Keysight to cover it under warranty and see what they have to say.

 

[HighVoltage]

We've had 5 out of 12 purchased in 2019 fail the same way at work.

They were all 34465A models?
That is ridicules!

[SteveyG]

All 34465As, but that's only because that's all we buy. They all got fixed, only notes were that the front panel PCB had failed if I remember correctly. The Keysight service van comes once a month for service an calibrations so it just all was handled through them with all the other equipment. Not had any issues since though.

[floobydust]

Likely not related to OP's fault, but something to point out (and I do hope I'm wrong here):
TI/Stellaris LM3S1D21 ARM processor is discontinued 2016 and withdrawn from the market due to the flash corruption issues. The parts are a lemon, known about in 2014.
"Flash corruption or device failure may occur at power on"
"Flash memory endurance cycle specification is 100 cycles"
"Flash memory corruption may occur when device is unpowered and stored for several months - Due to the storage oxide thickness and trap-assisted electron tunneling, there are more leaked cell values than originally expected in the Flash design within several months after programming. As a result, the ECC logic is not able to repair all of the errors in the Flash memory. Data derived from customer returns predicts that devices that are left unpowered at room temperature storage for 6 months can result in failure rates of 2000-3000 DPPM per year."
Ref: TEMPEST/INFERNO LM3S Errata Document (Literature Number: SPMZ861)

Bench multimeter product line 34461A, 34465A, 34470A use this MCU not as the main front panel processor but I think it's the Inguard processor.

If this is a real problem, silicon revisions need to be confirmed it's all 130nm parts, Keysight should be offering free extended warranty. It would be a massive debacle.

Front Panel processor ST SPEAr320S-2 is obsolete as well, but I did not dig into its issues.

[nctnico]

Likely not related to OP's fault, but something to point out (and I do hope I'm wrong here):
TI/Stellaris LM3S1D21 ARM processor is discontinued 2016 and withdrawn from the market due to the flash corruption issues. The parts are a lemon, known about in 2014.
"Flash corruption or device failure may occur at power on"
"Flash memory endurance cycle specification is 100 cycles"
"Flash memory corruption may occur when device is unpowered and stored for several months - Due to the storage oxide thickness and trap-assisted electron tunneling, there are more leaked cell values than originally expected in the Flash design within several months after programming. As a result, the ECC logic is not able to repair all of the errors in the Flash memory. Data derived from customer returns predicts that devices that are left unpowered at room temperature storage for 6 months can result in failure rates of 2000-3000 DPPM per year."
Ref: TEMPEST/INFERNO LM3S Errata Document (Literature Number: SPMZ861)

Bench multimeter product line 34461A, 34465A, 34470A use this MCU not as the main front panel processor but I think it's the Inguard processor.

If this is a real problem, silicon revisions need to be confirmed it's all 130nm parts, Keysight should be offering free extended warranty. It would be a massive debacle.

I'm afraid you are not wrong. It is a good find though so I appreciate you taking the trouble for digging this up. There is nothing to be found on TI's website about these microcontrollers nowadays. As if they never existed!

My 34461A also has the LM3S1D21 microcontroller revision A2 which is affected by the flash corruption problem. A way around it, is to leave the device on for at least 24 hours to give the internal flash controller the chance to fix the bits that have gotten corrupted. However, that system isn't failsafe as well and can actually corrupt data which was good. On top of that it is hard to tell whether a corrupted flash is signalled at all and how this affects measurements. It could be that the software continuous on with bad data affecting the measurement results.

All in all these units have a ticking time bomb inside them. 

[HighVoltage]

All in all these units have a ticking time bomb inside them. 

This is really bad news, if true. I have many of the 34461A in the field with customers.
Do we know when Keysight switched away from the LM3S1D21 microcontroller to a different one?

[Kleinstein]

I have a teardown photo (should be somewhere here in the forum) from an 34461, that shows a TM4C129 for the inbound µC. So chances are they have different versions. With the chip getting obsolete they kind of need to have a replacement. Given the problems I don't think KS would still use any NOS parts from an EOL buy.

There should be a flash checksum test as part of the POS. So flash corruption would likely be catched after power on. Ideally one would even get a specific error code for a checksum failure.
Chances are that the code for the actual measurements is only a small part of the code and a bad bit is more likely to cause a crash than a wrong result.

[hpw]

Likely not related to OP's fault, but something to point out (and I do hope I'm wrong here):
TI/Stellaris LM3S1D21 ARM processor is discontinued 2016 and withdrawn from the market due to the flash corruption issues. The parts are a lemon, known about in 2014.
"Flash corruption or device failure may occur at power on"
"Flash memory endurance cycle specification is 100 cycles"
"Flash memory corruption may occur when device is unpowered and stored for several months - Due to the storage oxide thickness and trap-assisted electron tunneling, there are more leaked cell values than originally expected in the Flash design within several months after programming. As a result, the ECC logic is not able to repair all of the errors in the Flash memory. Data derived from customer returns predicts that devices that are left unpowered at room temperature storage for 6 months can result in failure rates of 2000-3000 DPPM per year."
Ref: TEMPEST/INFERNO LM3S Errata Document (Literature Number: SPMZ861)

Bench multimeter product line 34461A, 34465A, 34470A use this MCU not as the main front panel processor but I think it's the Inguard processor.

If this is a real problem, silicon revisions need to be confirmed it's all 130nm parts, Keysight should be offering free extended warranty. It would be a massive debacle.

Front Panel processor ST SPEAr320S-2 is obsolete as well, but I did not dig into its issues.

This rings me a bell 

Does this means also, that the Wave Generator Trueform models 335xx & 336xx uses equal processors?

Hp

[TheSteve]

Likely not related to OP's fault, but something to point out (and I do hope I'm wrong here):
TI/Stellaris LM3S1D21 ARM processor is discontinued 2016 and withdrawn from the market due to the flash corruption issues. The parts are a lemon, known about in 2014.
"Flash corruption or device failure may occur at power on"
"Flash memory endurance cycle specification is 100 cycles"
"Flash memory corruption may occur when device is unpowered and stored for several months - Due to the storage oxide thickness and trap-assisted electron tunneling, there are more leaked cell values than originally expected in the Flash design within several months after programming. As a result, the ECC logic is not able to repair all of the errors in the Flash memory. Data derived from customer returns predicts that devices that are left unpowered at room temperature storage for 6 months can result in failure rates of 2000-3000 DPPM per year."
Ref: TEMPEST/INFERNO LM3S Errata Document (Literature Number: SPMZ861)

Bench multimeter product line 34461A, 34465A, 34470A use this MCU not as the main front panel processor but I think it's the Inguard processor.

If this is a real problem, silicon revisions need to be confirmed it's all 130nm parts, Keysight should be offering free extended warranty. It would be a massive debacle.

Front Panel processor ST SPEAr320S-2 is obsolete as well, but I did not dig into its issues.

This rings me a bell 

Does this means also, that the Wave Generator Trueform models 335xx & 336xx uses equal processors?

Hp

From memory they use a different processor part #.

[giovannirat]

If not already mentioned in the forum, there has been issued a service note for the 34465A regarding loose front panel (date 02/02/2022 )

[HighVoltage]

If not already mentioned in the forum, there has been issued a service note for the 34465A regarding loose front panel (date 02/02/2022 )

That is a weird service note.
The mounting of the front panel has not really changed in years.
If they have not made the mounting holes very large, I can not visualize how the front panel should come loose.

But then on the other hand, these 4 screws should probably have been installed since the early 34401A

[skander36]

Sad ... I just pull the trigger for Keysight 34465A, choosing between it and Fluke 8846A. Now it is too late to withdraw the order.
I just hope that the black case is not the only revision for the KS 34465A.





P.S. For those interested TME has now in stock Fluke 8846A at a good price for Europe. Though it is little outdated as display.

[nctnico]

For bench use I'd skip the Fluke 8846A and get the DMM6500 instead. Having some graphing / logging facilities on a bench DMM can be very handy at some times.

[skander36]

I have Keithley DMM 6500 from more than a year. It is better than 34356A in most respects and with 400 E cheaper!
I am buiyng KS for a "second opinion"    as Siglent SDM 3065X is already out of calibration after 3 years.

[Smith]

I have a 34465A at work, just died after maybe 40 hours. The transformer is blown (input side is open), I'm still waiting for a replacement from China. Keysight won't sell the transformer to me. This was one from only 2 we have.

I had multiple Keithleys (199, 485, 2000, 6485, 6517) and never had real issues, except for some dirty contacts, blown resistors (my bad). One 6485 at work had troubles with the EEPROM, but it had at least 10 years of almost daily duty and a 199 had lost it's EEPROM after it was written about 30 years ago. I will definitely be looking for a Keithley to replace the 34465A.

[alm]

I have a 34465A at work, just died after maybe 40 hours. The transformer is blown (input side is open), I'm still waiting for a replacement from China. Keysight won't sell the transformer to me. This was one from only 2 we have.

So it was only used for maybe 40 hours yet it's already out of warranty? Could be an open thermal fuse. But test equipment companies (Keysight, but also Keithley / Tektronix) certainly aren't as helpful with service information and spare parts as they used to be.

I had multiple Keithleys (199, 485, 2000, 6485, 6517) and never had real issues, except for some dirty contacts, blown resistors (my bad). One 6485 at work had troubles with the EEPROM, but it had at least 10 years of almost daily duty and a 199 had lost it's EEPROM after it was written about 30 years ago. I will definitely be looking for a Keithley to replace the 34465A.

And I've had numerous HP/Agilent instruments, none of which has had any transformer trouble. I do have a Keithley 6485 with an open primary in its transformer. Tektronix are ignoring my requests to buy a spare, so it's just sitting there. None of this gives any useful data about the reliability of the Keysight 34465A or the Keithley DMM 6500, however. Just anecdotes.

[HighVoltage]

Sad ... I just pull the trigger for Keysight 34465A, choosing between it and Fluke 8846A. Now it is too late to withdraw the order.
I just hope that the black case is not the only revision for the KS 34465A.

P.S. For those interested TME has now in stock Fluke 8846A at a good price for Europe. Though it is little outdated as display.

I had a 8846A and I am happy it is gone!
It is such old school instrument with a crappy display, completely outdated and too expensive!

The 34465A is a much better choice!

If one is buying a 34465A in 2022, I would think that all problems presented in this thread have been eliminated.
 

[AVGresponding]

I have Keithley DMM 6500 from more than a year. It is better than 34356A in most respects and with 400 E cheaper!
I am buiyng KS for a "second opinion"    as Siglent SDM 3065X is already out of calibration after 3 years.

Are you saying the Siglent has drifted out of its calibration limits and requires adjustment?

[bdunham7]

For bench use I'd skip the Fluke 8846A and get the DMM6500 instead. Having some graphing / logging facilities on a bench DMM can be very handy at some times.

 

This seems to be an oft-repeated discussion.  IMO any discussion of these three somewhat different meters should include their differences.  They're all 6.5 digit bench meters, but that's all they really have in common.  I don't recommend the 8846A to people in general because it isn't the best choice for many uses, but for the OP person wanting a 'second opinion' meter, it probably would, in fact, be the best by far.  The DMM6500 is simply a different class of instrument and has a lot of features--digitizing, the scanner card option, etc--that probably make it quite attractive for actual engineering use.  If I was going to buy another meter, I probably would go with it just for the scanner card function. 

The 8846A is a quick and efficient service bench meter, it goes from dormant (standby) to almost any range in 2 seconds and 3 button pushes or less.  Maybe 5 button pushes if you want dual display or something special like Hi-Z.  It also has 1000VAC, 1G-ohm and 10V diode ranges and is rated CAT I/1000V and CAT II/600V, unlike any competitors.  It also has actual and specified accuracy (if you read carefully) that significantly exceeds the other contestants.  So if those are the things that are important to you, the 8846A should not be overlooked.  I personally don't mind the display and logging can be done with Flukeview Forms (which is admittedly not great), LabView or HKJ's excellent Test Controller.  YMMV.

As for the current discussion, as an individual or very small business, I'd be very reluctant to invest in a 34465A given the recent issues with the model and the company.  It seems that unless you cycle them out before the warranty ends or you have an ongoing paid service contract with them and are a big enough fish to merit their attention, you may be left with scrap metal for your investment.  You might be better off with a decent HP 34401A and a PC for logging.  But again, YMMV--I have a high tolerance for things that are a bit odd or cumbersome to learn and use but a very low tolerance for things that are unreliable, don't last or can't be fixed.

[Smith]

So it was only used for maybe 40 hours yet it's already out of warranty? Could be an open thermal fuse. But test equipment companies (Keysight, but also Keithley / Tektronix) certainly aren't as helpful with service information and spare parts as they used to be.

It was sold about 2 or 3 years ago. I've done a lot of software for the last years, and haven't used it much.

And I've had numerous HP/Agilent instruments, none of which has had any transformer trouble. I do have a Keithley 6485 with an open primary in its transformer. Tektronix are ignoring my requests to buy a spare, so it's just sitting there. None of this gives any useful data about the reliability of the Keysight 34465A or the Keithley DMM 6500, however. Just anecdotes.

I never had any trouble getting Keithley parts. I've ordered displays, display assemblies, connectors, resistors (which I had blown), housing parts and such. Never had any transformer fail on me though. At this moment I have 7 Keithleys on my personal bench, so fingers crossed.

I'm almost sure it's the thermal fuse, but it would be quite some destructive work disassembling the whole transformer. I would if it would be my personal unit. Some transformers have the fuse accessible from the outside, this one was epoxied in the transformer. Took about 140 dollar for a new unit from China via Ebay.

[wraper]

It was sold about 2 or 3 years ago. I've done a lot of software for the last years, and haven't used it much.

AFAIK it has 3 year warranty. Warranty status checked on the Keysight website does not matter since purchase date in a proof of sale actually matters.

I have a 34465A at work, just died after maybe 40 hours. The transformer is blown (input side is open), I'm still waiting for a replacement from China. Keysight won't sell the transformer to me.

I don't think that a usual transformer is up to the task for such precision meter. It will probably inject too much leakage current into the DUT. I would take the outer insulation off and see if a thermal fuse can be reached.

[Smith]

I don't think that a usual transformer is up to the task for such precision meter. It will probably inject too much leakage current into the DUT. I would take the outer insulation off and see if a thermal fuse can be reached.

The transformer had the same number as the original, not just a compatible transformer. It was also sold as a 34465 / 34461 type transformer, it looked the same too, so I guess it should work. The transformer was epoxied stiff. It would take some serious work the get to the thermal fuse, with a reasonable probability of doing more damage. I did this type of repair to other transformers without any trouble, but not this one.

[tautech]

Siglent SDM 3065X is already out of calibration after 3 years.

Which ranges ?
Are you aware of the SDM Cal site Defpom made ?
http://www.thedefpom.com/siglent_sdm_calibration.php

[skander36]

Siglent SDM 3065X is already out of calibration after 3 years.

Which ranges ?
Are you aware of the SDM Cal site Defpom made ?
http://www.thedefpom.com/siglent_sdm_calibration.php

Yes I am aware. But this method is useless if you don't have a calibrated source like his Datron.
I know that after one year any DMM need calibration, but the ideea was that SDM3065X left me with the sensation of not being in the same league as Keythely,Fluke or KS. This is why I  headed to 34465A. But this topic make me anxious about ...

[tautech]

Siglent SDM 3065X is already out of calibration after 3 years.

Which ranges ?
Are you aware of the SDM Cal site Defpom made ?
http://www.thedefpom.com/siglent_sdm_calibration.php

Yes I am aware. But this method is useless if you don't have a calibrated source like his Datron.
I know that after one year any DMM need calibration, but the ideea was that SDM3065X left me with the sensation of not being in the same league as Keythely,Fluke or KS. This is why I  headed to 34465A. But this topic make me anxious about ...

Yes Siglent doesn't use aged references so they can drift outta spec. IIRC some owner here mapped the drift and found it almost exactly matched the LM399 datasheet drift spec.
You should contact Siglent in Germany as I believe they have a universal calibrator that can bring yours back to factory accuracy in a flash. Actually I will check on this as I know the US branch has one.

[skander36]

I will PM you to not continue to be off-topic.
Sorry, @OP.

[Fgrir]

When I bought a 34465A last summer, I got the 5 year extended warranty for an extra $82.  At that time similar 5 year coverage for a DMM6500 was quoted at $357.  I have a hard time believing Keysight is really experiencing massive failure rates on a product while selling extended coverage at such a low price.

[floobydust]

Why should customers suffer over a product defect?
It takes a lot of time getting the RMA, customs docs, shipping info, then there's down-time and the repair costs sending in the brick for repair, assuming it's not declared obsolete.

This guy's 34461A died after 6 months in storage, sounds just like the FLASH aging/corruption issue, a 2015 (at 4 yrs old) https://www.eevblog.com/forum/repair/keysight-34461a-179388/
At some point the affected models, years and serial numbers will have to be known. Your extended warranty, is it "to" or "additional" 5 years to the standard 3 years?

The "professional use only" shitstorm (almost 33,000 views) I wonder if it's to elude consumer protection laws which would apply here?
"They are not designed or tested for personal, domestic, or household use." https://www.eevblog.com/forum/testgear/keysight-officially-lost-the-plot-dont-buy-if-youre-a-hobbyist/

[TaylorD93]

If not already mentioned in the forum, there has been issued a service note for the 34465A regarding loose front panel (date 02/02/2022 )

That is a weird service note.
The mounting of the front panel has not really changed in years.
If they have not made the mounting holes very large, I can not visualize how the front panel should come loose.

But then on the other hand, these 4 screws should probably have been installed since the early 34401A

Ive had a 34461A come back from Calibration at Keysight, opened the bag to find the front panel loose on one side.

could see the plastic could have just be squeezed a bit and pushed back in, but we sent it back to make sure they were aware of it. But that was back end of 2021. Not seen any since, but most of ours are still in their first year of calibration. so will see when we send them out for the next rotation.

[Fgrir]

Your extended warranty, is it "to" or "additional" 5 years to the standard 3 years?

It extends the standard 3 year warranty coverage to 5 years, so I am paying $41/yr for the extra years of coverage. In my experience with other instruments, Keysight is happy to keep renewing this at similar pricing as long as the equipment is still supported by them. I've read all the threads about Keysight reliability over the last few years since I also own a MSOX3024A that was vulnerable to flash corruption. That unit was bought in 2014 and still running fine, but every time I read about the flash corruption issues I have to turn it on to see if it was still working. Yep, still boots.

To be honest, I am using all this professionally and a $1600 meter is not a big investment for me, but I'm not sure I would have gone with another Keysight without the cheap extended coverage. I still believe that pricing wouldn't be possible if their reliability was really as dismal as it seems based on the stories of people who have had bad experiences.

[bdunham7]

I still believe that pricing wouldn't be possible if their reliability was really as dismal as it seems based on the stories of people who have had bad experiences.

Suppose the annual failure rate is 1 in 20.  I would call that bad and that would be enough to generate howls of complaints, but if everyone had a contract, Keysight would have collected $820 for each failure.  That should cover it, I would think.  The incremental cost for an OEM to do a repair should be much less than retail.

[Kleinstein]

I still believe that pricing wouldn't be possible if their reliability was really as dismal as it seems based on the stories of people who have had bad experiences.

Suppose the annual failure rate is 1 in 20.  I would call that bad and that would be enough to generate howls of complaints, but if everyone had a contract, Keysight would have collected $820 for each failure.  That should cover it, I would think.  The incremental cost for an OEM to do a repair should be much less than retail.

That is especially true if they have a series fault on would only have the change the front part.
I would not expect the price for the waranty extension tp be calculated the normal way - it is more like an crude estimate upfront and rather hard to change a lot later on. Better pay a little for a few more repairs instead of unhappy customers.

[skander36]

New units come with screws mounted. The holes were left for rack mount kits, but with screws in place the face is rock solid. Without the screws it feel a little loosy.
Also they seem that changed the TI MCU (TM4C1292NCPDTI3 - G4), and what that seem to be an electric switch with a jumper. (see pics)
And they used for 34465A boards marked 34460.

[nctnico]

New units come with screws mounted. The holes were left for rack mount kits, but with screws in place the face is rock solid. Without the screws it feel a little loosy.
Also they seem that changed the TI MCU (TM4C1292NCPDTI3 - G4), and what that seem to be an electric switch with a jumper. (see pics)

If you put the datasheets / pinout of the TM4C1292NCPDTI3 next to the LM3S1D21, you'll spot many similarities. I think TI renamed the former Luminary devices and hopefully fixed their issues in the process.

[Bassman59]

New units come with screws mounted. The holes were left for rack mount kits, but with screws in place the face is rock solid. Without the screws it feel a little loosy.
Also they seem that changed the TI MCU (TM4C1292NCPDTI3 - G4), and what that seem to be an electric switch with a jumper. (see pics)

If you put the datasheets / pinout of the TM4C1292NCPDTI3 next to the LM3S1D21, you'll spot many similarities. I think TI renamed the former Luminary devices and hopefully fixed their issues in the process.

One major change between LM3S and TM4C is that the former is a Cortex-M3 and the latter is an M4.

[nctnico]

That is just the off-the-shelve CPU core. What matters is all that is around it.

[skander36]

They also came back with the kapton tape on top of the big caps ... 

[Keysight DanielBogdanoff]

I still believe that pricing wouldn't be possible if their reliability was really as dismal as it seems based on the stories of people who have had bad experiences.

Suppose the annual failure rate is 1 in 20.  I would call that bad and that would be enough to generate howls of complaints, but if everyone had a contract, Keysight would have collected $820 for each failure.  That should cover it, I would think.  The incremental cost for an OEM to do a repair should be much less than retail.

I can promise you the AFR is much better than that

[skander36]

I can promise you the AFR is much better than that

What does AFR stand for?

[alm]

annualized failure rate

[jonpaul]

Dear Daniel,

in case you are a Keysight afficanado or professional, can you please inform me...

I have a  new 34465a from Keysight in  August 2021,


serial  MY60016XXX

 FW A.03.02-03

Might it  have  the serious potential issues with the MCU, storage etc mentioned and is likely to fail eg if left un powered for  long periods?

Many thanks


Jon

[Traceless]

New units come with screws mounted. The holes were left for rack mount kits, but with screws in place the face is rock solid. Without the screws it feel a little loosy.
Also they seem that changed the TI MCU (TM4C1292NCPDTI3 - G4), and what that seem to be an electric switch with a jumper. (see pics)
And they used for 34465A boards marked 34460.

Thanks for the mini-tear down skander36, do you know the manufacturing date of your unit?

@KeithFisk: I'm sorry to hear that your unit failed do quickly I'd be curious to hear how the story ended, did Keysight offer any assistance despite the device being out of warranty. Did you repair or replace the device?

[skander36]

I wasn't able to find production date, but calibration was done in Nov.2021, so it is very likely to be in second half of 2021.
After finding that analog board MCU and also left side MCU from digital front panel have been improved, along with tightened of the front panel, now I am confident that major causes for fails that was reported have been addressed.

[Traceless]

I wasn't able to find production date, but calibration was done in Nov.2021, so it is very likely to be in second half of 2021.

I think your estimate is correct. AFAIK Keysight has a policy that devices that have been stored >6 months need to be sent back to KS (at least if you bought from an official KS partner). Thanks for taking the time to report back its good to know that devices manufactured >=Q3/2021 now use different hardware and thus may not be affected by the issue.

[KeithFisk]

@KeithFisk: I'm sorry to hear that your unit failed do quickly I'd be curious to hear how the story ended, did Keysight offer any assistance despite the device being out of warranty. Did you repair or replace the device?

@Traceless: I have asked the service agent to go ahead wit the repair but have no expectation of any concession or discount on parts and labour from Keysight.

[skander36]

AFAIK Keysight has a policy that devices that have been stored >6 months need to be sent back to KS (at least if you bought from an official KS partner).

Then there is a chance to be older than I have estimate. Calibration count is 75. I didn't know what that mean.
For me seem to be assembled unde the pressure of component crisis. (Keysight anounce 20 weeks delay for 34465A in Europe.)
I said that because it use 34460 board with a bridge under the main board. Maybe this is a solution for switching to rear terminals as the 34460 is not equiped with rear terminals. I am not sure though, maybe is just an improvement.

[HighVoltage]

That does not look like Keysight quality!
Not even cleaning the solder spots?

[skander36]

I don't know what to say. It is assembled on Malaysia on the same fab.
Those spots are where the mini doughter board is connected on the main board, maybe a rework?

[HighVoltage]

That definitely is a rework and looks like it was done by hand.

What is the part number and revision printed on the main PCB?

[HighVoltage]

I have just opened up my 34465A and took a few pictures.
This 34465A is in perfect condition and has worked well for many years already.

No extra PCB and no hand soldering anywhere.
The main PCB number is: 34465-26501 REV 004 KEYSIGHT MSS-B


Maybe you got a prototype of a new generation?

[skander36]

duno ... maybe @Keysight DanielBogdanoff  know 
The boards looks identical except components from pic attached (with red) that was moved onto the daughter board. Does not seem to be a prototype.
Also I am not sure if they changed Flash chip. Actual one that is used is Winbond as can be seen(attached).
Can you compare it with one from your DMM?

[TheSteve]

I would guess that the design was changed so that a single main board could be used with all versions, 34460 to 34470.

[Traceless]

@Traceless: I have asked the service agent to go ahead wit the repair but have no expectation of any concession or discount on parts and labour from Keysight.

Thanks for reporting back. If you have your device back and don't mind sharing I'd also love to hear what they charged you. Based on SteveyG's experience the problem seems to be no isolated case, just in case anyone else here runs into the same problem knowing the exact repair cost in advance would be interesting.

[Smith]

Small update, I now know why my transformer from the 34465A was fried. There was a dead short on the frontpanel, one of four 10uF 1206 caps was shorted. After I replaced it I luckily had the mains meter on it. The frigging thing took 120W, instead of the normal 11-12W from the socket. Even while it was turned off. Never had anything fry a transformer because of something so stupid. But there is a solution, I've ordered a Keithley DMM6500 to replace this crap.

[alm]

Glad you caught it before another transformer died!

I'd expect a primary side fuse that blows if the instrument is using 10x more current than expected. Is there none, or is it slower than the thermal fuse at this current level?

[tooki]

That definitely is a rework and looks like it was done by hand.

Nothing about that says “rework”. It’s a later PCB version where they decided, for whatever reason, to move components to a daughterboard. Since that adds cost (both for the extra PCB, but also the labor required to connect the daughterboard), they’ll have done it for some significant reason.

Soldered by hand? Probably, but not necessarily: it could also have been robotic soldering. Either way, it’s done with a soldering iron rather than wave or reflow soldering. (Yes, robotic soldering with a soldering iron is a thing.)

[AVGresponding]

But apparently cleaning up flux residue is not...   

[tooki]

It’s extremely common to find flux residue around hand-soldered (or robot iron soldered) components on boards that were otherwise wave or reflow soldered. It’s common on connectors and on assembly that isn’t just a single board.

This is a totally normal thing, yes, in commercial products.

[AVGresponding]

I know, but that still doesn't make it ok

[wraper]

Yep, often it makes more sense to use hand or robot soldering in production. For example when TH soldered on top side of PCB as in this case, or some other case where wave soldering is difficult to do or not justifiable from economic standpoint. For example if you just need a few contacts soldered, hand soldering is faster and cheaper.

I know, but that still doesn't make it ok

Why it's not OK? There is nothing wrong with this. As of flux, if it's not some low leakage circuit where flux may cover unwanted gaps between the traces, other than cosmetics there is no reason to clean it. Also cleaning it is no so straightforward as it may seem. Simply cleaning it with IPA will spread the flux all around and probably under SMT components where tiny leakage current actually matters. And getting rid of so there are no visible signs of its presence will take order of magnitude more time than soldering that connection.

[AVGresponding]

It's a high-end meter, not a home-made blinky; I'd expect better.

[wraper]

It's a high-end meter, not a home-made blinky; I'd expect better.

You'd be surprised. But as devices become more expensive and produced in lower numbers, the amount of bodges goes up. In this case it's not even a bodge.

[AVGresponding]

Yes, I understand; it's more practical to create a mod-board than to re-spin the main one, usually.

[skander36]

But apparently cleaning up flux residue is not...   

Keysight is not known for making perfectly clean PCB's.
This are the PCB from DSOX2002 - flux on the bottom of the board. Maybe they keep the boards as the laundry dries on the wire

[Smith]

Glad you caught it before another transformer died!

I'd expect a primary side fuse that blows if the instrument is using 10x more current than expected. Is there none, or is it slower than the thermal fuse at this current level?

There is a primary fuse, not exactly sure what is is rated at. I guess it just overheated until the thermal fuse went out. Probably for multiple months, until it died entirely.

About the solder flux residue. I don't have a big problem with it unless it is at a high speed path, of a low level measurement path. This tends to give problems in the long run, particularly with higher levels of moisture in the air.

I see lots of flux residues in higher end measurement devices, especially with user selected options from the factory, big input/output connectors, and bodges. Actually, it f you cant see it, its not potentially dangerous for you or the device, is it really a problem?

[Kleinstein]

For the flux residue it depends on the type of flux. Some flux is corrosive and must be removed. Other flux is OK even in higher humidity conditions and works as a reasonable protection and may be preferred over a "clean" surface.  With some parts fited (e.g. switches, cables, relays) is may be very hard to get a good clean. A poor attempt on cleaning, so that the flux is no longer visible can do more harm than good.

The main reason for the primary fuse is to prevent a fire in case the transformer fails, not to protect the transformer from overload. It would be nice to have also some protection for the transformer, but this is tricky with quite some tolerances in the fuse tripping points and possible inrush current spikes.

[tooki]

Yes, I understand; it's more practical to create a mod-board than to re-spin the main one, usually.

Do you understand, though?? This isn't a mod board as such. A mod board is commonly understood to mean a daughtercard made to implement a bodge to correct an error on the main PCB, rather than making a new main PCB.

As I explained above, early versions of the board had this part of the circuit on the main PCB. Then on this later version, the main board was respun to move that part of the circuit onto a daughtercard.

This isn't a cost-saving measure, as it's more expensive in every way: the added engineering cost, a respun main PCB, a new daughtercard, and the extra labor to solder the daughtercard onto the main PCB.

[tooki]

I know, but that still doesn't make it ok

As wraper said, cleaning in situ could get flux residue into places it shouldn't be. And with the two boards soldered together, cleaning becomes much more difficult.

So you'd be adding a cleaning step with no advantage, but plenty of risk.

I guess you don't open many devices. If you did, you'd have noticed the flux residue found in all sorts of equipment, from consumer electronics to IT to test gear to appliances.

[Kleinstein]

The part in the red rectangle looks a lot like part of the overvoltae protection. A good guess on the parts are 2 back to back depletion mode MOSFETs and that 2 units in series. This can be used as sereis element in the input protection, though is may fail over time from high energy ESD or similar events with very high voltage (e.g. > 2000 V). A separate board could ease repairs and maybe give the option to offer different versions with even better protection or meeting other regulatory limits (e.g. military not comftable with FETs used for protection). So having this somewhat vulnerable part on a seprate daughterboard is a very reasonable thing.

The KS meters did have some problems with the maximum votlage specs and this extra board could be reaction to this, though the rumors were more pointing to the relay voltage rating as the cause.

[skander36]

It is very possible that this section to be involved in overvoltage protection. I don't understand what is the role of J110 jumper.
I want to ask the owners of the KS34465A  what is happened when touch the red wire when the DMM is on DCV mode. A relay is clicking every time when I touch the red wire when is on DCV. None of my DMM (nor Keythely nor Siglent) behave like this.

[AVGresponding]

Yes, I understand; it's more practical to create a mod-board than to re-spin the main one, usually.

Do you understand, though?? This isn't a mod board as such. A mod board is commonly understood to mean a daughtercard made to implement a bodge to correct an error on the main PCB, rather than making a new main PCB.

As I explained above, early versions of the board had this part of the circuit on the main PCB. Then on this later version, the main board was respun to move that part of the circuit onto a daughtercard.

This isn't a cost-saving measure, as it's more expensive in every way: the added engineering cost, a respun main PCB, a new daughtercard, and the extra labor to solder the daughtercard onto the main PCB.

Even by your own definition, it clearly is a mod board, as the new main board is (what appears to be) not a respin but a new "universal" design, intended to be used in multiple models, but which lacks the onboard capability for these necessary components in a 34465.

I know, but that still doesn't make it ok

As wraper said, cleaning in situ could get flux residue into places it shouldn't be. And with the two boards soldered together, cleaning becomes much more difficult.

So you'd be adding a cleaning step with no advantage, but plenty of risk.

I guess you don't open many devices. If you did, you'd have noticed the flux residue found in all sorts of equipment, from consumer electronics to IT to test gear to appliances.

Please leave ad hominem attacks where they belong; on the editing room floor (inside your head).

Difficult is not impossible, and the risk is a balance between existing potential for harm, post-cleaning potential for harm, and the additional costs of either having unwanted warranty claims (something we can only speculate about as Keysight will certainly not release such commercially sensitive information), or additional costs of cleaning plus potential unwanted warranty cost as a result of more harm being done than good.

Yes, I understand very well, thankyou, nevertheless it's cosmetically unattractive, and in an instrument where I expect seriously anal levels of cleanliness, it does not sit well.

[tooki]

Yes, I understand; it's more practical to create a mod-board than to re-spin the main one, usually.

Do you understand, though?? This isn't a mod board as such. A mod board is commonly understood to mean a daughtercard made to implement a bodge to correct an error on the main PCB, rather than making a new main PCB.

As I explained above, early versions of the board had this part of the circuit on the main PCB. Then on this later version, the main board was respun to move that part of the circuit onto a daughtercard.

This isn't a cost-saving measure, as it's more expensive in every way: the added engineering cost, a respun main PCB, a new daughtercard, and the extra labor to solder the daughtercard onto the main PCB.

Even by your own definition, it clearly is a mod board

Absolutely not.

A mod board is a patch made applied after the fact to an existing part to correct an error discovered after manufacturing, so that you don’t have to trash the existing part. They DID redo the existing part, and this board was designed to be there from the beginning of this redesign.

as the new main board is (what appears to be) not a respin but a new "universal" design, intended to be used in multiple models, but which lacks the onboard capability for these necessary components in a 34465.[/color][/size][/b]

Since we don’t know for sure why those parts were moved off the main board, we certainly can’t conclude that it’s because of making it a universal design. (I frankly doubt that it has to do with reusing the board in multiple models: it’d almost certainly be cheaper to make different boards than to have the added labor to solder on a daughterboard. Remember that that’s just one person’s random guess, not even the result of circuit analysis.)

Also, by what logic is a new board design not a board re-spin? 

I know, but that still doesn't make it ok

As wraper said, cleaning in situ could get flux residue into places it shouldn't be. And with the two boards soldered together, cleaning becomes much more difficult.

So you'd be adding a cleaning step with no advantage, but plenty of risk.

I guess you don't open many devices. If you did, you'd have noticed the flux residue found in all sorts of equipment, from consumer electronics to IT to test gear to appliances.

Please leave ad hominem attacks where they belong; on the editing room floor (inside your head).

Difficult is not impossible, and the risk is a balance between existing potential for harm, post-cleaning potential for harm, and the additional costs of either having unwanted warranty claims (something we can only speculate about as Keysight will certainly not release such commercially sensitive information), or additional costs of cleaning plus potential unwanted warranty cost as a result of more harm being done than good.

Yes, I understand very well, thankyou, nevertheless it's cosmetically unattractive, and in an instrument where I expect seriously anal levels of cleanliness, it does not sit well.

An observation isn’t an ad hominem attack.

I said that I guess you don’t open many devices, and that still remains a reasonable assumption, since your surprise at a bit of flux residue in a place where it clearly does no harm (because if it did do harm there, they’d remove it!) would have been tempered by seeing it in device after device. (Either that, or you have opened lots of devices but just haven’t been observant.)

[Kleinstein]

The added extra board looks like it not only contains some parts, but also acts as a bridge to bring one of the signals from one side to the others. It very much looks like the extra PCB was planed togehter with the updated main board.

A reason for the new board could have been the problem they had with the high voltage rating, though the rumors were that this was from the relay ratings.
With the input signals routing needs quite some space for isolation and they may have just run out of board space in that corner and did not like a full redesign. The protection part on a separate board to simplify repairs may be an extra bonus.

The 34460, 34461, 34465 and 34470 are not that different. A combined board would make sense and should not take up more space. It should be mainly be a few parts missing with the 2 low end versions and maybe 1 or 2 bridges in the amps part to replace a relay or shunt.

[AVGresponding]

Absolutely not.

A mod board is a patch made applied after the fact to an existing part to correct an error discovered after manufacturing, so that you don’t have to trash the existing part. They DID redo the existing part, and this board was designed to be there from the beginning of this redesign.

You don't think it's an error to design a board that needs to be changed to move some parts to a daughter-board? There is a design error here, whether in the original that required a revision, or in the revision that could not be made to work without the extra expense.

Since we don’t know for sure why those parts were moved off the main board, we certainly can’t conclude that it’s because of making it a universal design. (I frankly doubt that it has to do with reusing the board in multiple models: it’d almost certainly be cheaper to make different boards than to have the added labor to solder on a daughterboard. Remember that that’s just one person’s random guess, not even the result of circuit analysis.)

Also, by what logic is a new board design not a board re-spin? 

Because it is not a new board intended only for the 34465, it is intended (by fairly obvious implication) to be used at least in the entire 3446X family, and as some have speculated in the 34470 as well.

An observation isn’t an ad hominem attack.

I said that I guess you don’t open many devices, and that still remains a reasonable assumption, since your surprise at a bit of flux residue in a place where it clearly does no harm (because if it did do harm there, they’d remove it!) would have been tempered by seeing it in device after device. (Either that, or you have opened lots of devices but just haven’t been observant.)

It isn't an observation, it's a subjective opinion expressed in a way to attempt to belittle the subject; you imply I am less experienced than you and therefore my opinion is of less value than yours. And you just did it again.


You also make an unwarranted assumption; that if the flux residue could be harmful, they would remove it. I explained in my last post why this is not necessarily so, and I can assure you I have the relevant experience to make the guesses that I did as to why they did not clean up afterwards.

[tooki]

Absolutely not.

A mod board is a patch made applied after the fact to an existing part to correct an error discovered after manufacturing, so that you don’t have to trash the existing part. They DID redo the existing part, and this board was designed to be there from the beginning of this redesign.

You don't think it's an error to design a board that needs to be changed to move some parts to a daughter-board? There is a design error here, whether in the original that required a revision, or in the revision that could not be made to work without the extra expense.

A mod board is a board bodged onto the ORIGINAL board, not to the new version! That’s the difference. If they’re designed to go together — as the new version and its daughterboard are — then it’s not a mod board!!

Again, to spell this out for you since you’re either not getting it, or are being deliberately obtuse: a mod is something designed AFTER a given version is manufactured, to be bodged onto already-manufactured boards. A mod board is, by definition, one for which the board it’s being attached to was not designed to accept.

Of course the original design had an error, that’s why they had to design a new board. But since the daughterboard was part of the new board design, and not an after-the-fact modification to the new board design, it’s not a mod board.

Since we don’t know for sure why those parts were moved off the main board, we certainly can’t conclude that it’s because of making it a universal design. (I frankly doubt that it has to do with reusing the board in multiple models: it’d almost certainly be cheaper to make different boards than to have the added labor to solder on a daughterboard. Remember that that’s just one person’s random guess, not even the result of circuit analysis.)

Also, by what logic is a new board design not a board re-spin? 

Because it is not a new board intended only for the 34465, it is intended (by fairly obvious implication) to be used at least in the entire 3446X family, and as some have speculated in the 34470 as well.

We don’t have any evidence for or against this claim. It could be right, but it could be wrong. So you shouldn’t treat it as gospel at this stage.

You also make an unwarranted assumption; that if the flux residue could be harmful, they would remove it. I explained in my last post why this is not necessarily so, and I can assure you I have the relevant experience to make the guesses that I did as to why they did not clean up afterwards.[/color][/size][/b]

Your explanation was so difficult to parse that it’s hard to understand exactly what you were saying.

But I stand by my claim that if it were harmful they’d remove it: if it harmed performance, then the performance would be suffering, and it’s not. And since they planned to leave it in place, they wouldn’t use a flux type that causes corrosion if left in place.

[nctnico]

Absolutely not.

A mod board is a patch made applied after the fact to an existing part to correct an error discovered after manufacturing, so that you don’t have to trash the existing part. They DID redo the existing part, and this board was designed to be there from the beginning of this redesign.

You don't think it's an error to design a board that needs to be changed to move some parts to a daughter-board? There is a design error here, whether in the original that required a revision, or in the revision that could not be made to work without the extra expense.

A mod board is a board bodged onto the ORIGINAL board, not to the new version! That’s the difference. If they’re designed to go together — as the new version and its daughterboard are — then it’s not a mod board!!

Again, to spell this out for you since you’re either not getting it, or are being deliberately obtuse: a mod is something designed AFTER a given version is manufactured, to be bodged onto already-manufactured boards. A mod board is, by definition, one for which the board it’s being attached to was not designed to accept.

I think you are both right/wrong. As Kleinstein already noted: Keysight did change the board layout but for some reason choose not to redo the entire layout but use an extra board instead. Likely Keysight didn't want to re-qualify / re-test a new board layout given the high performance they want to get out of the design.

[tszaboo]

I had 1 out of 4 fail with random reboots, in the last 4-ish years. We sent it back they replaced it without question.
I didn't put too much though into it back then.

[HighVoltage]

A reason for the new board could have been the problem they had with the high voltage rating, though the rumors were that this was from the relay ratings.

The high voltage rating debacle later turned out to be not a problem at all.
So, I think it is very unlikely that this was the reason.

[salvagedcircuitry]

Likely not related to OP's fault, but something to point out (and I do hope I'm wrong here):
TI/Stellaris LM3S1D21 ARM processor is discontinued 2016 and withdrawn from the market due to the flash corruption issues. The parts are a lemon, known about in 2014.
"Flash corruption or device failure may occur at power on"
"Flash memory endurance cycle specification is 100 cycles"
"Flash memory corruption may occur when device is unpowered and stored for several months - Due to the storage oxide thickness and trap-assisted electron tunneling, there are more leaked cell values than originally expected in the Flash design within several months after programming. As a result, the ECC logic is not able to repair all of the errors in the Flash memory. Data derived from customer returns predicts that devices that are left unpowered at room temperature storage for 6 months can result in failure rates of 2000-3000 DPPM per year."
Ref: TEMPEST/INFERNO LM3S Errata Document (Literature Number: SPMZ861)

Bench multimeter product line 34461A, 34465A, 34470A use this MCU not as the main front panel processor but I think it's the Inguard processor.

If this is a real problem, silicon revisions need to be confirmed it's all 130nm parts, Keysight should be offering free extended warranty. It would be a massive debacle.

I'm afraid you are not wrong. It is a good find though so I appreciate you taking the trouble for digging this up. There is nothing to be found on TI's website about these microcontrollers nowadays. As if they never existed!

My 34461A also has the LM3S1D21 microcontroller revision A2 which is affected by the flash corruption problem. A way around it, is to leave the device on for at least 24 hours to give the internal flash controller the chance to fix the bits that have gotten corrupted. However, that system isn't failsafe as well and can actually corrupt data which was good. On top of that it is hard to tell whether a corrupted flash is signalled at all and how this affects measurements. It could be that the software continuous on with bad data affecting the measurement results.

All in all these units have a ticking time bomb inside them. 

I just picked up a 34465a on the cheap and noticed that mine has a LM3S1D21 as well. How can I tell the revision of my LM3S1D21?
From the ti doc, it looks like my chip was produced September of 2015.
My unit has 2.14 firmware on it, but I have a feeling there is no firmware to fix internal MCU chip nand corruption :/
Thanks!

[floobydust]

Here is the TI EOL document PCN 20140630000 for the LM3S Tempest/Inferno defective MCU product line. Last shipments were through to end of 2015. We don't how much longer Keysight continued to produce H/W using the MCU. See end of doc but TI pegged the failure rate at 0.3% keeping down the costs of warranty credits Keysight surely received. There are no pin-compatible replacements.  Snippet of PCN last page:

Code: [Select]


July 1, 2014  Re: Acknowledgement and Limited Warranty Letter for Stellaris Tempest and Firestorm LM3S MCU Devices

Dear Ladies and Gentlemen:
Texas Instruments Incorporated (“TI”) has identified potential flash memory corruption risks with respect to the Stellaris Tempest and Firestorm LM3S part numbers listed in PCN 20140630000. You are receiving this Acknowledgement and Limited Warranty Letter (“Letter”) because TI records indicate you have purchased or sampled one or more of the LM3S Products affected. TI has made a decision to End of Life (EOL) these LM3S Products that were acquired from Luminary Micro. The Erratum (SPMZ861) for these LM3S Products is available at [url=http://www.ti.com]www.ti.com[/url].

TI did not undertake this EOL decision lightly. TI has undertaken significant diagnostic efforts and has replicated quality and performance issues, including flash corruption, that our customers have reported. TI has not been able to find any system level or manufacturing workarounds that completely prevent the flash corruption issues. Customer reported worst case fail rates for these LM3S Products, while using the entire flash in the largest 512kB configuration, is 0.3%. These issues may continue to manifest themselves as failures in customer designs over time, to a greater or lesser degree, depending on a customer’s system design, proportion of flash memory used, and storage times following a customer’s programming of the flash. Please see the Erratum for more details.

Importantly, we want to assure you that these reported flash corruption issues in the LM3S Products (which were designed using Luminary Micro methodologies) are not present in TI’s next generation TM4 products, which have been designed and produced in accordance with TI’s standard design and production procedures and methodologies. For the reasons explained herein, notwithstanding any provision to the contrary in any Purchase Order or any other purchasing or acknowledgement document, for shipments after September 1, 2014, TI’s sole and total warranty is a limited warranty. This limited warranty of credit or replacement is capped at .3% of the revenue for LM3S Products shipped after September 1, 2014.
When possible, TI will replace LM3S Products prior to the end of the EOL period based on then existing lead times. After completion of the deliveries at the end of the EOL period, this warranty will be limited to credit only. Except for this limited warranty, TI makes no warranty, express, implied or statutory, including any implied warranty of merchantability or fitness for a specific purpose, for these LM3S Products.
TI will not support full customer return analysis and any related corrective actions (8D reporting, cycle time metrics, etc.) for these LM3S Products, because of these known issues and uncertainty on how, or to what extent, these issues will impact actual customer systems. However, TI may, in its sole discretion, perform customer return analysis to verify warranty claims based on flash memory failures.
TI requires this Acknowledgement and Limited Warranty Letter for LM3S Products to be signed by you by September 1, 2014 to continue shipments after September 1, 2014.
If you desire further shipments of these LM3S Products, please have an authorized person sign below to indicate your agreement and acceptance of this Letter, and return a signed copy of this Letter to TI (at the e-mail address below ),or your authorized TI distributor, if you are buying from a distributor. This Letter will become effective upon your signature.
TI will accept cancellations of orders for LM3S Products accepted before July 1, 2014, if those cancellations are made by September 1 2014. Also, TI will accept returns of LM3S Products in customer inventory based on Return Materials Requests received by TI by September 1, 2014, that are undamaged and in original packing.
These LM3S Products should not be used for any safety-critical application in which a failure of the device could result in an injury to persons or property.
FOR CLARITY, PLEASE NOTE THAT ANY ORDERS REMAINING IN PLACE AFTER SEPTEMBER 1, 2014, WILL IMMEDIATELY BECOME NON-CANCELABLE AND NON-RETURNABLE. ANY INVENTORY REMAINING IN PLACE AFTER SEPTEMBER 1, 2014, WILL IMMEDIATELY BECOME NON-RETURNABLE, EXCEPT FOR THE LIMITED WARRANTY ABOVE.
Please do not hesitate to contact your TI representative or your distributor if you have questions.
Very truly yours,
Texas Instruments Incorporated


[bdunham7]

So anyone that shipped a product with it after mid-2014 was well aware that this defect existed?   

In some areas of the world, that would effectively extend your warranty to almost infinity for a consumer product.  Hmmmm.

[floobydust]

Apparently it's not a consumer product "... for professional and industrial use. They are not designed or tested for personal, domestic, or household use."
Who knows, legal might be eluding recalls and warranty extension this way, rights consumers would have.

[wraper]

I just picked up a 34465a on the cheap and noticed that mine has a LM3S1D21 as well. How can I tell the revision of my LM3S1D21?
From the ti doc, it looks like my chip was produced September of 2015.
My unit has 2.14 firmware on it, but I have a feeling there is no firmware to fix internal MCU chip nand corruption :/
Thanks!

Chip revision does not matter. It's garbage by design, not some bug which could be fixed. Read pages 24-25 https://www.ti.com/lit/er/spmz861/spmz861.pdf?ts=1651046449928&ref_url=https%253A%252F%252Fwww.google.com%252F
Shortly Flash is trash by design and can corrupt data when left unpowered for a few months, but ECC is trash too and can read the data incorrectly and then re-program incorrect data  .

[salvagedcircuitry]

Chip revision does not matter. It's garbage by design, not some bug which could be fixed. Read pages 24-25 https://www.ti.com/lit/er/spmz861/spmz861.pdf?ts=1651046449928&ref_url=https%253A%252F%252Fwww.google.com%252F
Shortly Flash is trash by design and can corrupt data when left unpowered for a few months, but ECC is trash too and can read the data incorrectly and then re-program incorrect data  .

Boy, it looks like someone goofed. Sure makes the TI acquisition of Luminary Micro look like a million dollar pile of bricks! 

[floobydust]

TI desperately needed to leapfrog into ARM core or get left in the dust. But then taking 5 years to fully realize the 3rd party FLASH IP used in the MCU's is total garbage? I've only seem Atmel's SAM7 was a lemon randomly erasing its FLASH (had to give up and end that product) so maybe it's early Atmel tech Luminary bought. But LM3S FLASH just rots and ages even when sitting there doing nothing, so reprogramming the FLASH is not always possible or a fix.
Then, replacement parts are not pin compatible, so basically you chuck the main board into the garbage making for a very costly repair. It's a total turd. TI got suckered with that acquisition and Keysight mixing both ST and TI MCU's in a product not the brightest move.

[wraper]

TI desperately needed to leapfrog into ARM core or get left in the dust. But then taking 5 years to fully realize the 3rd party FLASH IP used in the MCU's is total garbage?

FWIW TI made ARM based SoC (mainly for phones) for at least a decade at the time of acquisition (2009).

[wraper]

Also it's not just flash that is garbage. Looking on errata it's easier to count things that are working properly, rather than those which are broken in some way.
For example following cannot possibly be overlooked. You need to make an informed decision to release broken garbage.

LM3GPIO#05 PB1 has permanent internal pull-up resistance
Device(s) Affected: Stellaris Tempest-class Rev C5 and Firestorm-class Rev A2
Description: Regardless of its configuration (GPIO or alternate digital function), PB1 has a maximum
internal pull-up resistance of 800 Ω that turns on when the voltage on the pin is
approximately 1.2 V. Due to this internal resistance, up to 3 mA of current may be
sourced during the transition from 1.2 V to 3.3 V.
Workaround(s): When this pin is configured as an input, the external circuit must drive with an
impedance less than or equal to 300 Ω to provide enough drive strength to over-drive the
internal pull-up and achieve the necessary VIL voltage level. Ensure that the driver can
sink the temporary current. In addition, do not use PB1 in open-drain mode.
If this pin is configured as an output, be aware that if the output was driven High and a
non-POR reset occurs, the output may be driven High after reset unless it has a 300-Ω
resistor on it. Once the pin is configured as an output, the pin drives the programmed
level.

[skander36]

Sorry for asking again, all 34465A DMM's make a relay click sound when central (red) wire is touched? It is a normal behaviour?
Thank you!

[wraper]

Sorry for asking again, all 34465A DMM's make a relay click sound when central (red) wire is touched? It is a normal behaviour?
Thank you!

Does it click if you disable autoranging? Also is it in 10Mohm impedance mode?

[skander36]

Yeah ... it is from autoranging 
When disabled no more clicking ...
Thank you very much @wrapper!

[nctnico]

Likely not related to OP's fault, but something to point out (and I do hope I'm wrong here):
TI/Stellaris LM3S1D21 ARM processor is discontinued 2016 and withdrawn from the market due to the flash corruption issues. The parts are a lemon, known about in 2014.
"Flash corruption or device failure may occur at power on"
"Flash memory endurance cycle specification is 100 cycles"
"Flash memory corruption may occur when device is unpowered and stored for several months - Due to the storage oxide thickness and trap-assisted electron tunneling, there are more leaked cell values than originally expected in the Flash design within several months after programming. As a result, the ECC logic is not able to repair all of the errors in the Flash memory. Data derived from customer returns predicts that devices that are left unpowered at room temperature storage for 6 months can result in failure rates of 2000-3000 DPPM per year."
Ref: TEMPEST/INFERNO LM3S Errata Document (Literature Number: SPMZ861)

Bench multimeter product line 34461A, 34465A, 34470A use this MCU not as the main front panel processor but I think it's the Inguard processor.

If this is a real problem, silicon revisions need to be confirmed it's all 130nm parts, Keysight should be offering free extended warranty. It would be a massive debacle.

I'm afraid you are not wrong. It is a good find though so I appreciate you taking the trouble for digging this up. There is nothing to be found on TI's website about these microcontrollers nowadays. As if they never existed!

My 34461A also has the LM3S1D21 microcontroller revision A2 which is affected by the flash corruption problem. A way around it, is to leave the device on for at least 24 hours to give the internal flash controller the chance to fix the bits that have gotten corrupted. However, that system isn't failsafe as well and can actually corrupt data which was good. On top of that it is hard to tell whether a corrupted flash is signalled at all and how this affects measurements. It could be that the software continuous on with bad data affecting the measurement results.

All in all these units have a ticking time bomb inside them. 

In the meantime I had a rather in depth email exchange with Keysight support (without needing to pay or jump through hoops). It looks like Keysight has found the issue with the LM3S1D21 by themselves and reported back to TI. According to Keysight they did some kind of check on the part to see if it is affected before assembly so units in the field shouldn't be affected. In case there is a problem, it shouldn't result in bad/wrong readings. The latter is my main concern.

If my unit where affected, it would have shown by now as it is sitting switched off for periods lasting over a year.

[wraper]

In the meantime I had a rather in depth email exchange with Keysight support (without needing to pay or jump through hoops). It looks like Keysight has found the issue with the LM3S1D21 by themselves and reported back to TI. According to Keysight they did some kind of check on the part to see if it is affected before assembly so units in the field shouldn't be affected. In case there is a problem, it shouldn't result in bad/wrong readings. The latter is my main concern.

If my unit where affected, it would have shown by now as it is sitting switched off for periods lasting over a year.

It's not like that. If you read the errata, you'd know that sitting off for years does not necessarily cause the NAND corruption. It just increases its chance. Staying always ON does not help with corruption either since of garbage ECC circuit which may cause corruption of good data by itself. The flash corruption rates given in ERRATA because of both issues are similar. It's not like there are good chips and bad chips, there is a crappy design.

[autorepair78]

Is there anyone on the forum who had a problem with the processor in the 34XXX series and after replacing everything was back to normal? I am asking because I have the opportunity to buy a 34461A which, when turned on, only illuminates the LCD and nothing else works. When opened, you can see that the SPEAR320-S2 is very hot.

[CraigD73]

It's now 2024 has this problem been resolved with current production KS3446A DMMs?

[jonpaul]

Bought a few yres ago, screen had intermitten blank fail.

Keysight replaced with a new uint, calibration cert IN WTY.

Apart from that , 100% perfect.

Jon

[nctnico]

It's now 2024 has this problem been resolved with current production KS3446A DMMs?

From my email exchange with Keysight, it looks like they are using a different microcontroller as the microcontroller they used initially has been discontinued.

[voltsandjolts]

As I understand it, the self-destruct feature was mainly due to:

(A) Luminary MCU on the main board self-corrupting flash memory (Note 1)
(B) STM SPEAR320 processor on the front panel failing for some reason, at least some get very hot to touch
(C) NAND FLASH on the front panel getting corrupted, sometimes recoverable by interaction over the serial port.

Keysight is the undisputed champion of FLASH corruption, perhaps not always their fault but certainly early firmware made a poor job of recovering from FLASH corruption. e.g. depending entirely on one copy of the NAND ONFI parameters, ignoring backup copies when primary failed

My hope (I say staring at my DMM) is that these problems have been fixed in newer dark colour meters. Certainly the main board and front panel processors have changed, we have to wait and see if there are reports of NAND corruption on these newer meters. So far, so good on my bench

Note 1: TI EOL document PCN 20140630000 for the LM3S Tempest/Inferno defective MCU product line gives estimated failure rate of 0.3%

[HighVoltage]

I had a fair share of NAND corrupted Keysight scopes.

But even with so many Keysight DMM 34461A, 34465A and 34470A I never had a problem with those.
Maybe the problem is larger, if these instruments are not used on a regular basis?

[coromonadalix]

are they still running an WinCe thingy intheses,  WinCe was notorious to wear flash memories because of too many read and write, you had to do some optimisations

many systems uncompress the wince boot file into memory, i had problems with systems who booted from some sd cards, it took a while to get a good brand


not bashing,  but i'll thrust more an 34410a or 34411a   way before thoses models

[mawyatt]

We have 3 KS34465As, one tan and two dark cased, nary a problem with any!!

Best,

[voltsandjolts]

Well, they certainly do fail, my 34461A (2018 light colour) is currently whitescreen mode, fail to boot. Also, a quick search of the forum showed up these failures:

analogRF 34461A fail to boot

dc101 34461A fail to boot

kanzler 34460A fail to boot

晓看风云  fail to boot 34461A

jonpaul 34465A intermittant boot / black screen

300DT fail to boot 34461A hot SPEAR

maxtorque 34461A fail to boot

elcal fail to boot

autorepair78 34461A fail to boot hot SPEAR

yuhar fail to boot hot SPEAR

NandBlog random reboots

Smith fried transformer due to shorted cap

KeithFisk hot SPEAR

TheSteve seen several failed SPEAR

SteveyG "We've had 5 out of 12 purchased in 2019 fail the same way at work." (34465A hot SPEAR) 

skskelwl white screen fail to boot hot SPEAR

TERRA Operative fail to boot

Charos fail to boot hot SPEAR

Fixed_Until_Broken fail to boot 34461A hot SPEAR

rolkinas fail to boot 34465A hot SPEAR

kkayser fail to boot 34461A

Of course, if you're not a company Keysight will say you have no warranty, even though you can buy from Farnell and no one will mention that. Perhaps in that case Farnell would offer one year warranty, I don't know. I hope the dark colour units are going to prove more reliable. My advice to private individual purchasers would be only buy dark units, avoid older models on eBay.

OTOH I have a Keithley 2000 I purchased non-working from eBay. £15 for some new op-amps and JFETs fixed it. I can source most parts in that meter, including the eprom firmware and am very confident I could fix it if anything happened to it in the next 20 years. Keysight 3446x/70, hmm, not so much.

[voltsandjolts]

Seems that Farnell UK have a new promotion on 34461A at £978 + VAT, cheaper than the 60A.

https://uk.farnell.com/keysight-technologies/34461a-promo/digital-multimeter-6-5-digit-1kv/dp/4294711?st=34461a

My quote for a new front panel CCA for my whitescreen 61A will likely be near half that new price. Hmm.

[coromonadalix]