Fullscreen HTML5Med Res (??MB)HTML4

SERVICEMAN'S LOG Well-designed thoughtlessness A recurring theme for these columns as of recent times is the prevalence of designers who, whether intentional or not, put a lot of thought into not thinking when designing devices. People of a certain age may recall an English sitcom named “The Fall and Rise of Reginald Perrin”. I’m not talking about the insipid recent remake, but the original show, which aired way back in the mid-70s. The premise of the show was the hum-drum life of an ordinary, middleclass, middle-management worker and his eventual descent into mid-life crisis. He wanted more, and ended up reinventing himself. The show was satire, and an indictment of then-British society (and her colonies). It took every opportunity to skewer the class system, the unions, the nationalising and de-nationalising of various industries and much more. One of several running gags involved the trains, where because they always ran late, Reginald was always late for work. In the first series, he was always 11 minutes late. In series two, he was always 17 minutes late and in series three, 22 minutes late. He always offered a different excuse for his lack of punctuality, and these excuses were increasingly outlandish, such as: “seasonal manpower shortages, Clapham Junction”, or: “Seventeen minutes late, water seeping through the cables at Effingham Junction.” Without spoiling it for anyone who wants to watch the show (I recommend it, though it isn’t everybody’s cup of tea), part of the storyline involved a shop named Grot. A sly dig at rampant consumerism, Grot’s stock was made up of items that were purposely designed to be bad or useless, such as salt and pepper shakers with no holes in them, non-stick glue, elastic tow-ropes and square rugby balls. I mention this, admittedly in a rather long-winded lead-up, because recently siliconchip.com.au I’ve been working on some items that could have come from this shop! Over the years I’ve regularly called out what I see as lousy design, and I’ll keep doing it, because it often seems the person who designed the machine, appliance or manufacturing method has no idea of how the appliance, machine or manufacturing method will actually be used in real life. Examples include the light on my vacuum cleaner that shines up the wall, rather than on the floor, or the lawnmower that doesn’t cut grass short enough and has handles and levers that Australia’s electronics magazine Dave Thompson Items Covered This Month • • • Misanthropic designers producing ill-considered designs HP8595 spectrum analyser repair 3A USB charger repair *Dave Thompson runs PC Anytime in Christchurch, NZ. Website: www.pcanytime.co.nz Email: dave<at>pcanytime.co.nz protrude wider than the cutting track of the mower, making it difficult to mow right up against a wall. Another example is the pickup selector switch on my Fender Telecaster; it is almost impossible to actuate when in the bridge pickup position because it is July 2020  61 almost hard-up against the tone knob – a design flaw that has persisted since the late 50s. (For those pedants who are preparing to flame me, like many I flipped my Tele’s tone-plate around for easier access, swapping the volume and tone pots). And don’t get me started on car engineering! Some cars have the battery under the driver’s seat, so you have to remove the seat to replace it. One gets the impression that after the initial prototype rolled out, the engineers discovered that they had forgotten to include a battery and so they had to scramble to find a place to put it! Many of the cars I drove as a youngster had steering-wheels that obstructed the view of the instruments, and in one car I test-drove, I had to sit at an awkward angle because the steering wheel and pedals weren’t in line. And the number of times I have needed triple-jointed limbs or specially-made tools just to be able to access nuts and bolts to disassemble machinery to get to faulty parts… It’s a miracle that any of these designs get put into production with these quirks. Indeed, there are web pages and YouTube channels dedicated to this subject: stair-wells heading into brick walls, inward-opening toilet doors with notches cut into them to fit around the bowl or basin, water pipes right next to electrical outlets; the list goes on. Don’t get me wrong, these follies are always good for a laugh, but usually, it isn’t the people who have to deal with them that are doing the laughing! would have cost more than a new one is worth, so it made sense to send the faulty board (swapping the board fixed the welder, so we know it is at fault). The usual suspects are capacitors, semiconductors or simply solder joints gone bad, but working to resolve any of these problems becomes a major mission due to the varnish coating. One saving grace is that the customer supplied a circuit diagram for the board, and while components on the PCB were clearly marked, it always helps to have a circuit diagram for troubleshooting. When I encounter a coating like this, the first thing I do is see if I can soften it using solvents. None of the solvents I have touched it. Next is cautiouslyapplied heat. While a proper heat gun is ideal, I tend to use my desoldering heat gun more these days, as it is easier to control and aim. I found an unpopulated corner of the board and judiciously applied heat to the area to see if I could make a dent (so to speak) in the varnish. I couldn’t. It didn’t even get softer; it just got hotter! It turns out, though, that I could melt it with my soldering iron, so that’s what I did. Messy, but effective. That partially solved one problem: getting to the soldered connections on the bottom of the board. But I still had to deal with the component side. Desoldering the leads below was one thing; extracting the components was another. The first thing I did was to check the soldering on the bottom of the board. Though the varnish was thick, it was Enter the culprit Now we get to the meat of the matter. The other day, I received a faulty PCB to fix, and the entire thing was covered with a very thick layer of varnish, top and bottom, despite being a single-sided board. Every component is well-embedded into this coating, making parts incredibly difficult to desolder, let alone extract. What genius thought this would be a good idea? Obviously, it is designed to be replaced rather than repaired, and I’ve made my opinion on that subject well known. It must be evident to the manufacturer that end-users would want to repair faulty boards; they aren’t a cheap replacement part, but a multi-hundreddollar investment. This one is from a heavy-duty welder, which no longer held its output. Shipping the welder 62 Silicon Chip Australia’s electronics magazine siliconchip.com.au mostly clear, so a visual inspection was possible. Like many such boards, there are multiple, well-tinned heavy tracks for the likes of Earth returns and power supply paths. Welders generally boast serious current-handling capabilities, so the boards have to be up to supplying that current without sagging and compromising the welds. Where there is resistance, there is heat, and as these boards would heat up and cool down regularly, they will expand and contract. This makes any physical connection a potential weakness. As the resistance of a bad joint increases, so does current and heat, and the cycle continues until something eventually gives. In many ways, it is better for the serviceman to get a board that has utterly failed; at least the faults (or the consequences of the faults) are patently obvious. Intermittent or partial faults make things more difficult, as does not having the ability to bench-test the board. Several other satellite boards drive this particular controller, and without those, I can’t test the board at normal operating levels. I can test each component though, and the overall physical integrity of the board, which is the process I had to use. There were some very large, heaped solder joints which looked a bit dodgy – this is typical of every high-current power supply board. But I saw no obvious faults like overheated tracks, discolouration of the board or any other visual clues to explain the failure. I burned through the varnish over several of the more dry-looking joints and cranked up my heavy-duty soldering iron to reflow them, but overall the board looked well-made, and the joints generally were physically sound. So I moved on to check the components. Component checking Six large capacitors dominate the landscape of the board. I could measure them in-circuit using my Peak ESR tester, but I prefer to do my component testing off-board, just to be sure. At least these caps were relatively easy to remove because I could get some purchase onto them. I did have to first cut through the fillet of varnish around the base of each one using a razor blade; a task made more difficult by the proximity of other components. Still, I got them out, and though it was hard to tell visually whether the siliconchip.com.au goop coating the outside of some of them was leakage or runs in the varnish itself, it proved to be the varnish. All the caps measured very close to their stated values, and the ESRs also read very low. So I moved onto the handful of smaller electrolytic capacitors and a dozen or so high-voltage ceramic types; I removed them the same way, and they all tested fine. I also pulled a medium-sized transformer from the board and measured it for resistance; the figures I took from the primary and secondary corresponded roughly to the turns ratio supplied on the schematic (no other specs supplied). But I was more interested in shorts or open circuits, of which there were none. A megger check also proved there was no breakdown in the windings or insulation. The board has three 24V 10A relays mounted on it and these I tested by clearing the varnish from their terminals on the track side of the board and soldering test leads to their normallyopen contacts and coils. I downloaded the data sheets and used my bench supply to raise and lower the voltage, testing each relay’s current draw and operation and drop out voltages. I connected my multimeter’s buzzer across the terminals; while not technically a perfect indicator of the electrical condition of the contacts, my musical ear can detect even subtle variations in the frequency of the tone, which changes with resistance. Any deviation from the closed-circuit tone (with the meter leads shorted, for example) means there is resistance in the circuit. On these relays, the buzzer tone remained the same, indicating no significant changes in contact resistance. Again, this is not a definitive test for contact integrity, but adequate for my purpose. Semiconductors were my next target, and this board boasts many different types. Here I had to cut some corners; while the relay driver Mosfets and the large, paired diodes in the voltage multiplier section were relatively easy to remove and test, the smaller DO-35-sized zeners and regular diodes were in almost every case totally enclosed in varnish. They would likely be impossible to remove without damage. All I could do was clear the varnish underneath (there was a lot of varnishclearing going on!) and measure them Australia’s electronics magazine with my semiconductor tester. As this automatically detects and takes into account whether they are in-circuit or not, I had to trust it was doing its job correctly. There was nothing suspicious in my measurements with any of the diodes. All the zeners measured as-rated, and the other identical types all had very similar breakdown voltages, which in itself means nothing other than that no individual component stood out as a potential problem source. Measurements of the three IRFZ24 Mosfets did show some discrepancies, so even though not a smoking gun, I decided to replace them. They are as cheap as chips anyway (LOL!) and as the TO-220 packages stand proud of the board, they are easy (relatively!) to remove and refit. There is one small TO-92 type NPN transistor, and because I broke one of its legs removing it, I subbed in a BC549, one of the suggested alternatives in my transistor manual. There were also sundry components, such as a 4-pin DIP-style opto-isolator, which was buried in varnish and bridged a physical channel cut into the board. I could only resistance-test this device, and it appeared to pass. There are also several series-connected thermistors, used as inrush current limiters, and a chunky metal-oxide varistor (MOV) used for surge protection; these all tested fine. Having replaced the only parts I could find that could potentially be causing faults, all I could do now was to clean up where I’d been and re-coat the places I’d dug into the varnish with some standard polyurethane. I doubt the board really needs it for July 2020  63 electrical protection, given it’s all relatively low-voltage, and there would be no stray coronas developing on pointy solder joints. I’m assuming it is there in case metal dust or welding swarf might find their way into the cabinet and potentially short out something on the board. In the end, I did as much as I could without testing the board in the welder, then sent it back to be reinstalled and tested in-situ. Theoretically, checking joints and testing individual components should resolve any problems, but we all know there is more to it than that, and it will only be dumb luck if the board works when put back into the machine. More badly designed junk Another potential Grot shop candidate is a USB3 hub I worked on recently. It had a problem that’s common with many other modern devices. This hub was relatively new, but the socket inside had come away from the PCB, rendering it useless. The owner wondered whether it could be repaired, not because it is a particularly expensive device, but because it irks him (as it does me) to throw something away that isn’t that old or has had much use. The problem with this, and other devices, is that it is designed to be small and portable, but the cable that comes with it is very heavy and not overly flexible, so the thing will never sit where it is placed, and the stress and strain on the socket is very high. The new USBC connectors might solve these issues, but we shall see about that. I’m sure this same problem affects all of us; my phone, which is a few years old now, is starting to show signs of socket wear, mainly because many of the OTG cables available now are quite heavy gauge, and unless I am careful, I can put a lot of strain on the charging socket. Editor’s note: this is one of the benefits of wireless charging; while slow and inefficient, it doesn’t wear the USB connector! I also purchased a Raspberry Pi 4 a while ago, and this uses a USB-C connector for power and micro-HDMI for video output. Both these cables are so stiff that I just have the Pi sitting in midair, at whatever angle comes naturally with the cables plugged in. To do otherwise would probably rip the sockets off. Given that these sockets usually rely on only a few tiny solder pads for adhe64 Silicon Chip sion, it’s no wonder they come adrift, even with normal use. Re-attaching the USB socket to the hub wasn’t too taxing; the challenge was getting the thing apart without breaking the plastic clips they used to hold it together instead of screws (another Grot idea). I suggested the owner let the repaired hub hang naturally on the cable and hope it doesn’t break again. I wonder if someone has upset the designers of these devices, and they are exacting their revenge on society by designing shoddy, unserviceable products. If so, I wish they would take their frustrations out in some other manner, such as with a stress ball or a punching bag. Do us servicemen a favour, please! HP8595 spectrum analyser repair A.L.S., of Turramurra, NSW has been up to his usual hobby of buying cheap test instruments from internet sellers. And as is so often the case, they turned out to need a bit of TLC (by which we mean ‘serious repairs’) to get them back into full working condition... You can buy a second-hand HewlettPackard HP8595E spectrum analyser quite cheaply on the internet. These devices can analyse signals from 9kHz to 6.5GHz, but they are starting to age a bit as they were new in the late 1980s. Many now have little gremlins growing inside them. The HP85xx series was very popular 25 years ago, because these instruments are portable and easy to use. So there are thousands of them for sale, and many parts available on the internet. The one I bought was a real find because it included several options, in- cluding the HP-IB/parallel port interface for external control and printing. It was this option which convinced me to buy the instrument, so that I could keep records of various traces. Its specs are really impressive, and it analyses an incredible array of RF and modulated signals, including TV signals. It also has an FFT function to analyse harmonic distortion of AM/ FM audio signals. On receiving the device from the USA, I immediately tried out the print function and connected up my “Print Capture” device (parallel-to-USB module). This allows me to download screen grabs, and has worked tirelessly on all my test instruments with parallel ports. But it refused to work this time. Another way I can obtain screen grabs is via a GPIB-to-USB adaptor, but that also failed to work. To my horror, no matter how many combinations and permutations I tried, I could not get any screen grabs out of the device! Of course, you can photograph the trace on the screen easily, but the result is not as crisp and neat as a digital hardcopy, because the signal moves a bit and blurs. An obvious clue as to why this was not working was that the option “041” was not listed on the setup screen. But the other three options that were supposedly installed according to the seller were listed there. So my immediate thought was that the board was installed, but not connected correctly, so it was not being detected or used. The instrument is relatively easy to open up. I just had to unscrew four Allen-head bolts and four Philips-head screws. The cover then slides off. The HP-IB/parallel port interface, which is used for external control and printing, was a welcome inclusion with the spectrum analyser. Australia’s electronics magazine siliconchip.com.au The first thing I noticed after opening it up was that there were some missing screws, which suggested someone had previously been inside it. I feared that a dodgy board had been fitted just so that the instrument could be sold with more options. However, once I got a chance to inspect it, I found that the GPIB board looked pretty good. It was a bit dusty, and I noticed that the multi-pin DIL connector looked a tiny bit crooked. I disconnected everything and applied some contact cleaner to the plug and socket. It was then that I noticed a bent pin on the male header. I wasn’t sure if I had bent it during the disassembly, as the plug was very stiff with age when I disconnected it (I guess we all end up that way!). Anyway, after cleaning the board and connectors, I straightened the pin and plugged it back together. It snapped into place. On start-up, the option “041” appeared, and I was finally able to obtain beautiful hard copies via both parallel and GPIB. Sadly, though, that is not the end of the story! The plot thickens Some weeks later, I noticed that the analyser amplitude readings seemed a bit low. I connected its internal calibration signal up to the input and obtained a reading about 18dBm lower than expected. So I ran the “cal amplitude” routine. To do this, you need to connect a BNC patch cable from the calibration output directly to the input and then press the “cal amplitude” soft-key. The instrument should be warmed up for at least 30 minutes before doing this. It takes several minutes, and during this time, you hear plenty of relays clicking in and out. However, in my case, it stopped after a minute, and a message came up saying “Cal gain: Fail”. The service manual explains that this means the signal was too weak and outside the specified minimum level. Either the calibration signal was poor, or there was an internal problem with the analyser. The manual suggests checking the following parts of the circuit: A3 front end, A7 analogue interface, A9 third converter, A11 bandwidth filter, A12 amplitude control, A13 bandwidth filter and A14 log amplifier. That really narrows things down – not (consider that these assemblies total about half of the instrument)! siliconchip.com.au A beautiful ‘hardcopy’ finally emerged after fixing the improperly connected circuit board. Hoping it was just the calibration signal at fault, I hooked up a 1GHz generator but found that the reading was still 18dBm low, proving that the problem was with the measurement side of the instrument. I was hoping it wasn’t a front-end problem, because the attenuator is buried deep inside, whereas the other boards are merely plug-ins and changing them is an easy job, if time-consuming. I did some internet research and discovered a great three-part YouTube video about fixing an HP8590, which is a similar device but with a 1.5GHz maximum frequency. I highly recommend it if you enjoy repair stories. See: https://youtu.be/kV4BOf3Oqk8 This inspired me to check out the symptoms of my instrument, and I noticed that when I manually adjusted Australia’s electronics magazine the attenuation, I could get a correct reading when it was set to -20dBm. But the readings were all over the place at other attenuation settings. I also got obscure readings at different frequencies; precisely the same symptom as in part three of those YouTube videos. Unfortunately, this meant that the attenuator was the immediate suspect and so it would be a significant repair. Hunting around the internet, experts reported that 90% of problems with these instruments were the result of poor or damaged attenuators, so I immediately looked around for a secondhand or reconditioned attenuator. As luck would have it, I found somebody selling a brand new attenuator, all sealed up in its original HP box, so I made him an offer (which he didn’t July 2020  65 The analyser’s ‘front end’ which processes the input signal via the attenuator. The faulty attenuator (top right) was deep inside. It looks more like plumbing than electronics! refuse), and the part arrived in a few days from Italy. Now the fun and games began! The YouTube guy, who calls himself “FeedbackLoop” (siliconchip.com.au/link/ ab3c), did not go into details of how to extract the attenuator. I could not even 66 Silicon Chip find it after searching for some time! The diagram in the manual is somewhat simplified, and the assembly (labelled A3) is just shown as a dotted line. You cannot get to it from the side, so the whole aluminium assembly has to come out in one piece. Australia’s electronics magazine Everything looks pretty simple until you figure out how to extract it because it’s a bit like a Rubik’s cube. You have to start by loosening screws and then gently shaking things to discover how to extract the entire “box” containing the attenuator. For fear of boring readers, I won’t describe the exact procedure here. But if you find yourself in the same boat as I was, you may wish to write to Silicon Chip so your message can be passed on to me. I will then reply in excruciating detail. I think the money I spent to obtain a new attenuator just drove me on to replace the suspect one, despite the herculean task before me, because it would be such a waste to have a beautiful new part and never use it. Sort of like those blokes who buy an expensive car which then just sits in the garage, never being driven. What a waste! You will see in the picture here that the assembly looks somewhat like a UHT dairy plant and is more about plumbing than wiring, because of all the semi-solid cables which require disassembly. Caution is advised here, because they must not be bent. I had to work slowly and patiently to unthread some of the wiring harnesses between the semi-solid cables. Finally, I managed to extract the culprit and replace it with the brand new part, but it required the same amount of patience to re-assemble everything. Naturally, I made some mistakes and had to do it all over again when I realised that I couldn’t re-fit one of the retaining screws because an aluminium housing was blocking it. But finally, it was done, and I checked it all thoroughly in case something was amiss. Very cautiously, I switched it on, hoping there would be no nasty noises. Amazingly, it all started fine, and the measurements were almost exact to within 0.5dBm. The frequencies were also spot on! After celebrating for the required 30-minute warm-up, I performed a self-calibration, and the accuracy improved even more. I was really glad I purchased the brand new attenuator (at significant cost) because if it was a second-hand ‘dud’, it would have been a colossal waste of time and effort. Now I have a really precise and importantly, working instrument. I intend to protect the attenuator by using a DC-blocking device and an external attenuator, in case a DC voltage might siliconchip.com.au accidentally be applied. Any applied DC will destroy it, as will RF signals which exceed 1W or +30dbm. 3A USB charger repair B. P., of Dundathu, Qld has had the same simple fault fell multiple devices in his possession. Is he cursed, or is this a case of bad designs multiplying? You be the judge... When chatting with my mate via Skype on a Samsung Galaxy S 10.5 Tablet, I found that its battery would discharge even though the supplied 1A USB charger was plugged in. I ordered a 2A charger on eBay, but I found that it was also unable to keep the battery at 100%, so then I purchased a 3A charger. This one was finally able to keep the battery charged at 100% while using Skype. As I was quite happy with it, I decided to get a spare, so I ordered another identical one. The original 3A charger worked well for a couple of years, but recently I noticed that the battery was discharging even while it was plugged in. I felt the charger and it was cold, so it clearly wasn’t working, as it was usually quite warm when in use. Swapping it for the spare charger got me back in business. I decided to try to fix the failed unit. It appeared that the two halves of the case might be glued together, as is common with many chargers, so I clamped it lightly in the vice with padding, to see if it would crack open. It popped apart and I found that it wasn’t glued, but instead clipped together. This was siliconchip.com.au good news, as it would be much easier to reassemble it later. I had a close look at the circuit board and noticed blue corrosion build-up on two of the 1N4007 diode leads, but these diodes and the other diodes on the board tested OK. I then noticed a 1W 0.5W resistor marked “F1” on the circuit board, indicating that it was used as a fuse. This resistor was connected between one of the mains wires and the rest of the circuit and when I tested it, it was open circuit. I didn’t have any 1W 0.5W resistors in my parts bin, only 1W types, but I managed to salvage a similar resistor that tested OK from another dead charger. I reassembled the charger and tested it, and it worked just fine. However, after a few weeks of use, the charger failed again. I wasn’t surprised when I opened it up and found that the same resistor was open circuit. I decided to replace it with a 1W 1W type, and it has been working reliably ever since. Even though this 3A charger only cost me about $5, it was an easy fix which not only saved me $5 and the wait for a new one, but that was one less device going into landfill. I had a similar problem with the sensor light on our front verandah. Its ‘fuse’ resistor failed several times, so I ended up replacing it with two higher-rated resistors in series. That repair then lasted the life of the sensor, which eventually disintegrated due to UV deterioration of the plastic. I’ve also had mates bring me other USB devices which had stopped working, and I was able to fix those by, you guessed it, replacing a fusible resistor. So this is a very common configuration in devices where a low voltage is derived from the mains, and failures of this part are a common occurrence. It’s likely that the resistors are just barely rated for their use in this configuration, so it may be necessary to increase the rating of the resistor to compensate for the inadequacy of the original resistor to cope with higher ambient temperatures and high mains voltages. SC The 3A USB charger PCB shown outside of its housing. F1 is the 1W resistor shown sticking out at lower left. Australia’s electronics magazine July 2020  67