It has been eight years since I reviewed the Keysight U1461A, an insulation resistance testing digital multimeter with a rather sharp OLED display. Since then, it has become my only insulation testing device, keeping me safe when salvaging electronics and testing dubious cheap products from China.
But as time has passed, unfortunately, it seems the Achilles Heel of OLED displays are beginning to rear their head. While it has been acknowledged that OLED displays have limited lifetime and can suffer from fading (with some colours being more vulnerable than others), lately, some OLED displays have been failing outright instead. My display, however, has started to become problematic despite rather light usage (estimated at about 100-150 hours in total since receipt). However, such issues are not entirely beyond repair …
The Problem
The problem came upon suddenly. The last time I used the U1461A, it seems nothing was amiss. However, when I pulled out the U1461A from its case to test the insulation resistance of a rather cheap and compact GaN charger from China, I discovered the display was not doing well.
Upon start-up, it was already clear that some pixels have just died while others are somewhat dimmer than the others.
This seems to have resulted in a scatter of inconsistent pixels across the whole screen. I tried taking the DMM outside and left it in the afternoon sun, just in case it was something like humidity getting into a flexible connector (as would commonly occur to some LCDs), but there was no change. I’ve also tried leaving it powered up for a long time, in the hopes it would somehow self-resolve and simply be caused by a lack of use, but this was not the case either.
I was running the latest firmware available, so this is not likely a software issue either.
Another thing I’ve come to realise is that the three brightness steps of low, medium and high aren’t working correctly either.
Low brightness is barely readable in ordinary room lighting.
Medium brightness is somewhat readable indoors, but not at all outdoors. Finally, high brightness wasn’t working at all! It would just remain just as bright as the medium setting.
From looking around online, it seems that some other popular models of Keysight DMMs using OLED screens have also suffered failures, such as the U1253A/B, U1753A/AX and 34450A. These failures often include dimming screens to downright complete failure. While it is known that OLED displays may degrade in brightness as a function of operational hours (more severe for certain colours compared to others), the sudden failure is a new one for me. From the discussions I’ve seen online, it is posited that the glass solder seal may be problematic, thus allowing the organic LED material to become exposed to the atmosphere and perhaps oxidising. As a result, the screens also seem to have a shelf-life, even if they sit unused. But as I’m no OLED chemist, I have no idea whether this is the cause, or whether it is due to something more mundane (e.g. internal thin film connection fractures). For those meters, however, drop-in replacement modules are available online for purchase. But the U1461A seems to not have been popular enough to see such attention. How unfortunate …
In doing some research, it seems the U1461A and related DMMs have been discontinued by Keysight. The discontinuation date was 1st March 2022 and I haven’t seen any replacements at this time. Perhaps this is a sign that Keysight is pulling out of the insulation resistance testing DMM market (or there’s something new on the way). While there is apparently product support for five years, but seeing as this unit is not in warranty due to age (and there is no list of spare parts available for it online), this would clearly be a problem for me to resolve on my own.
It seems that at least one person has had their display replaced …
Teardown
In order to plan a fix, I had to get a peek under the covers. This way, I can check the obvious things to see if it’s just a simple problem (and if not, then I can figure out what actually needs to be replaced). I typically don’t like to open up measurement instruments, as every case opening risks disturbing the “volatile ppms” (in the words of Marco Reps). As a result, there is a potential that the calibration may drift due to something as simple as disturbing the stresses on the board or by introducing contaminants into the system.
But in this case, it was more of an “all or nothing” situation, so I decided to open it up. After all, four-and-a-half digits shouldn’t be something too hard to retain after surgery. Taking off the rear cover is pretty easy - I first removed the two user-serviceable doors for the battery and fuse, removed the rubber surround (although neither may need to actually be removed in practice). I then went around and removed the six larger Philips screws around the perimeter to allow the haves to be separated. The piezo beeper can be unplugged from the mainboard, allowing both case halves to separate. I then removed the fuse as it was obscuring a screw - one which has to be removed as part of the next stage of the teardown.
For those who like close-up images of boards, here’s the top section of the rear. The transformer at the top is likely used to generate the high insulation resistance test voltages. There is a shielded inductor near the unpopulated J4 that seems to have been manually siliconed to reduce acoustic emissions or increase physical shock resistance. Aside from that, there are a few ICs on the board, but I didn’t pay much attention to that aside from the Hycon HY3131 Analog Front End on which the DMM is based.
The lower section near the inputs has a lot more spacing and anti-tracking slots for improved isolation. There also appears to be a lot of safety features including MOVs and thermistors with protective heatshrink (perhaps for current limiting in case of transient impulses or faults). The function selector knob mechanism is particularly fiddly as it consists of circuits on both sides of the board (!!!). As a result, one has to be careful to remove the screw and this rear plate along with three tiny gold-plated spring contacts that aren’t held captive. To further disassemble requires unclipping the auxiliary input jacks by pulling back on the tabs and undoing the bolts for the banana jacks. It also requires taking another five screws out from the PCB. Of note is that the PCB silkscreen indicates this is U1461-26500 (which may be an assembly number) of which this is the fifth revision. There is a UF840L N-MOSFET in a D2PAK which seems to be quite chunky. Might be related to generating high voltages or the Earth bond test?
Oh, and behind this metal shield appears to be a precision resistor array that serves a key function in ensuring the accuracy of the meter. Don’t touch! Not like you can with the metal can around it anyway.
Now the board is out, the internals of the casing on the other side can be examined. It seems both front and rear have foil shielding to either reduce EMI emissions or improve EMI/EMC robustness in case of being used in a radio-harsh environment. The three springy contacts on this side of the function selector can be seen - while they slot into grooves in the selector knob, they actually sit freely, so just a slight bump and those contacts can go flying. Because of this, reassembly is going to be an absolute pain.
The PCB from the front - I’d say it’s interesting to see more protection in the form of fusible resistors. The IXTH3N120 High-Voltage N-MOSFET D2PAK on the back seems to be switching currents for discharging. I’m not really intending to analyse the circuit in any depth - the key reason for opening it up is merely to get at the OLED panel. This is on its own PCB daughterboard that is attached to the mainboard using four smaller Phillips head screws.
Looking underneath the OLED module, I suspect this is a NEC/Renesas D78F1168A 8-bit Microcontroller that runs the whole DMM. Next to it looks vaguely like an EEPROM which may hold the calibration data. The PCB is dated to Week 46 of 2013, which indicates just how long ago this unit was made. My best guess would be that Q23 is responsible for switching the waveform to generate the high voltage.
A little bit of a closer look at the input jack segments - some of those resistors are covered in heatshrink, perhaps to stop them from exploding in case of a drastic overload? There’s another MOV on this side too. There’s so much protection, it’s just night and day compared to the cheaper meters that I’ve seen. The assembly does leave a little to be desired - it seems the PCB is not exactly clean in the section where the mode selector rides on - thankfully the switch points are sufficiently “sharp” and have sufficient pressure to cut through the contamination, but ideally all residues should be cleaned off for the best performance and longevity.
We finally get to the culprit (I think) which is the OLED display board. From the front, we can see the OLED display, seemingly as clean and neat as the day it was made.
From the back, we can see this board is marked as U1461-26501, the second revision. This PCB is dated Week 44 of 2013, making it ever so slightly older than the mainboard. Unfortunately, looking up this code doesn’t give us any options for purchasing such a part as a spare, which is a shame.
But if we look closer, we can see the Solomon Systech logo on the flex, along with the part number "SSD1305T7".
There’s a Part for That!
Doing a bit of digging reveals the SSD1305 is a display controller that has been superseded. The part number on its own doesn't so much define a screen size or resolution, as display manufacturers have the job of integrating such controller dies into their display products.
But by chance, it seems this type of display was also used in some Voltcraft DMMs and Pure DAB radios and these have failed too. There seems to be one lone seller on Alibaba with such displays in stock - no more choice on colour though as another listing had amber but no stock :(.
The product isn't exactly going to be as simple as a drop-in replacement as might be available for the more popular Keysight OLED DMMs. In fact, while the green flex seems to be the same, it comes with an adapter tail FFC to suit other devices. Would it be possible to break it apart without fatally damaging the ribbon? Only one way to find out.
To give this a go was not going to be an inexpensive endeavour. By the time the taxes and shipping have been added onto a declining Australian Dollar conversion fee, I was looking at about AU$53 for the display. That would only be one too - if I managed to ruin it, it would be another AU$53 for another try. When we consider that an ordinary Keysight DMM runs for the AU$400-500 mark and the whole set was AU$1088 at the point of the RoadTest review, this is not an insignificant investment. But I took it as a bet on my own abilities, so I plonked down the cash.
After a week of waiting, a very light box arrived at my doorstep.
Inside, there was a neatly-packed piece of bubble wrap.
Taking this out of the box reveals an anti-static shielding bag …
… containing the display and the attached FFC tail as shown in the image on the listing. Item definitely “as described”, but I have no way to verify if it is any good prior to installation.
There are some codes (UG-2864ASWDT16 WP12080063-03) printed on the adapter tail.
I’m not sure exactly how the adapter tail is attached to the main flex. At first glance, I thought it was adhered, and that may be partially true. There was no evidence of soldering, so perhaps there’s some conductive epoxy magic or something else going on.
The back of the display is pure glass, and has a code on the rear too (PG-2864ASWD P1222027-16-A05). There are two grey patches - they’re underneath the glass and are not self-adhesive pads (as I thought they might have been at first glance).
Now I am faced with a dilemma - the meter is working but the display is dim and spotty. If I try to replace the display, then I really only have one shot at it. In all probability, I might damage the existing display such that going back is not a possibility. I have no way to know if the new display is functional before trying to put it in - and I might damage it in the process of putting it in. In short, if I proceed, there is a chance that I could be left with a DMM that has no display at all, rather than a partially functional display.
Seeing as I’ve already put the money into purchasing the display, I decided to proceed with the repair for the simple reason that the U1461A also has an infrared interface port that allows for remote control. As a result, even without a display, I could use the Bluetooth or USB interface peripheral to provide a remote display (albeit with some additional fuss and latency).
The Replacement Process
The first step is to take the adapter tail off the display. While the tail only attaches on the top side, I did think of the possibility of just cutting off the tail but leaving part of it attached to the wires, but it would have complicated soldering by removing the ability to directly apply heat to the joint. So, I just took a brute force approach and peeled the adapter tail off leaving this … I suspect I might have weakened the connection wires by doing this - they’re as fragile as the copper foil on a printed circuit board. It is good that the lines are not affected from the rear of the flex, the side that needs to be soldered for this application.
The next step involved separating the existing display from its supporting daughterboard. This daughterboard is an important part, as it not only adapts the ribbon connection to the pins used by the mainboard, it also hosts a switching converter that generates the necessary (higher-than-logic) voltage to drive the OLED matrix.
To remove the ribbon from the PCB without damaging the PCB, I decided to use the “flood the pins with solder” technique, with a fine pair of tweezers lifting as I went along. Taking the OLED glass from the PCB was a bit more tricky, as the OLED was double-sided taped to the daughterboard with thin tape. I had to start with dental floss to clear enough of it from a corner such that I could get a box-cutter knife in to cut more of the adhesive. Eventually, I was able to lever it enough to remove it - thankful that I didn’t break the glass, or else it would have become a messy and potentially painful operation.
As expected, it was not a harmless operation as the wires of the original display were mangled by the desoldering process. I might have done better with hot-air, but I didn’t want to risk overheating any pads or providing insufficient heat and tearing it off by accident.
To install the new display, I laid it flat with the ribbon aligned to the connector. I bathed the PCB’s solder pads with fresh solder, cleaned off the excess flux and then applied some additional liquid flux. Holding the ribbon in place, I quickly heated the two end pins (which appear to be NC) to maintain the alignment. Then, with a “blob” of solder, I dragged it across the window, soldering each pin down in sequence while spattering flux everywhere. By the time I reached the end, the remaining solder was taken away on the tip of the iron.
Things didn’t seem to be over-heated, but the foil contacts are very fragile, so stress is a big concern for me as it could result in the already stressed conductors being torn later. I did think as to whether I would support them with hot glue or other adhesive, but I decided against this, as it would make repair more difficult in the future should it prove necessary.
After cleaning off the most major flux spatters, finally it is time to fold the display around and mount it with fresh double-sided tape. Due to the stiffness of the ribbon, it applies quite a bit of stress to the joint. Thankfully it held and visually doesn’t look too much worse than the original condition.
Will this be enough? Will the patient survive? It's time to reassemble and test.
Result
After swiftly reassembling the device, I rotated the mode selector and …
… success, or sort of. The new display has square pixels instead of round ones, so the display looks sharper but more jagged. It’s like the difference between LCD and CRTs - it looks like someone turned off the anti-aliasing. Nevertheless, the display is much brighter - you can see that the multimeter’s frame is now barely visible now that the text is properly exposed. The larger light emitting area probably also helps.
But there is also one defect - the rows with more pixels lit are darker than the rows with fewer pixels lit creating this banding effect.
Running in actual measurements and the issue is less apparent. At least the random-dead-pixels are now gone and the brightness is serviceable.
But one issue remains - the brightness control still doesn’t work for high. I get low and medium … then nothing more (i.e. medium again). Looking back, I suspect this also means that there may be a failure in the supporting board in the switching converter that generates the OLED drive voltage. Perhaps it’s dipping under load, or not producing the right voltage in the first place. Was it responsible for the partial death of the first OLED display? I’m not sure. For now, I’d have to guess that it might be the switching converter chip, bad output capacitors (even though they appear to be tantalum, perhaps they’re leaky or have high ESR) or perhaps problematic feedback to the switching controller. But seeing that it’s improved, I don’t think I’ll take another risk to try and fix it - if I heat up the SMD components from the rear, the heat may well damage the OLED display.
Conclusion
I think it’s clear that the OLED technology is one with its own pros and cons. While I have had occasional difficulties with LCDs, that is very rare and far between as of late. But when it comes to OLEDs, it seems this one decided to go with a sudden partial failure. As a product that wasn’t that common, it was not possible to find a drop-in replacement (unlike for the U1253A/B, U1753A/AX and 34450A), and using an obsolete controller means that there are few options for direct replacement. I was lucky that at this time, there are still some components left on the market, but this is probably not going to be the case for long. Even then, it did take some fussing around with removing the fitted FFC tail for a Pure DAB radio to allow it to be soldered down to the daughterboard. Even then, the residual stress on the ribbon has me a little nervous.
That being said, the repair was a partial success. There are no more random dead pixels and the screen is now serviceably bright. Unfortunately, the high brightness option isn’t working and there is visible horizontal banding in the display correlated with the number of lit pixels, suggesting that there may be an issue with the OLED drive voltage. Whether this was the cause of the original display’s failure is unknown, but attempting to fix it now will probably cause stress and damage this OLED display, so I think I’ll just leave it as it is.
Should another repair become necessary though, by then, there may not be direct replacement parts. By then, I’ll have to go to the lengths of reverse-engineering the pin-out, finding a screen of a compatible size and similar protocol (many SSD screens have similar protocols but some commands will cause problems), whack in a microcontroller or something as a “go-between” to fix up the issues with the differences in protocol and spin my own PCB. This is quite a bit of effort and the investment may not be worth the time (except for the challenge of doing such a modification and the sense of satisfaction could be worthwhile. The issue is that it’s not designed to operate “in the open”, so sniffing what is happening will require some creative soldering of ribbon cables and closing up the case to make it work.
Seeing as the U1461A and its family are discontinued by Keysight and it is unclear whether there will be any follow-up replacement products in the Insulation Resistance Testing category, I’d say this meter is perhaps quite the special one for me and I’d love to keep it going for as long as I can.
Top Comments