Fullscreen HTML5Med Res (??MB)HTML4

A cure for a serious fault By LEO SIMPSON When readers began assembling kits for our Induction Motor Speed Controller, problems started to arise. Either it would not reliably drive induction motors or it failed – sometimes in spectacular fashion. This caused great consternation in the SILICON CHIP camp until we could finally be sure that we had a successful cure. In brief, the published circuit is fine – but the PCB needs a few mods to cure an interference problem. I t was probably too good to be true. After we published the Induction Motor Speed Controller in the April & May 2012 issues, there was very little feedback from readers and most of that involved queries about when kits would finally be available from Altronics and Jaycar. Some Controllers were assembled by keen readers who obtained the PCBs and programmed micro from SILICON CHIP but most people wanted a full kit, to avoid the hassles of obtaining all the components separately. For most readers, that is indeed the best and easiest approach since it is often cheaper, you get all the parts and you know the kitset suppliers will have already built their 80  Silicon Chip own prototypes to check that everything is OK. Once the Motor Speed Controller kits became available, a lot more were built and with so many out “in the wild”, eventually problems surfaced. This resulted in us receiving a number of unhappy emails! Some readers, not to be put off by circuit malfunctions, took matters into their own hands and did some extensive investigations, to see what the problems were. And in some cases those investigations led to even more spectacular failures and further grief. All of this was most frustrating for us since we had two faultless prototypes. Eventually, we obtained a defective Controller from Alsiliconchip.com.au It’s been a very popular project but not without a few problems. The mods described here should eliminate those problems, most of which were due to interference. tronics and indeed, it would not drive a reasonably high power load. The really frustrating part was that it would work if a motor was connected between the terminals labelled W & V on the PCB but not the W & U terminals! Even more frustrating, when the diagrams on pages 69 & 74 were being prepared, I had nominated terminals W & U as the ones to be used when connecting a single-phase motor. Paradoxically, if I had nominated the two left-most terminals, W & V, it probably would have taken a few more months before we would have known about the severity of the problem. In fact, many builders would probably have experienced no problems at all. It also happens that those terminals are wrongly labelled on the wiring diagram but that is something that we would have just put down to a drafting error. That arises because the chip manufacturers, ST Micro, have adopted a very different pin-out numbering convention to the norm. So what we showed as pin 1 on the PCB overlay diagram (May 2012, page 69) is actually pin 16. And so what we had labelled as output U on CON2 is actually W. Confused? Fortunately, none of this actually affects circuit operation. It seems that no matter how much proof-reading we do on each issue, errors can still be missed. The only consolation we have is that large commercial electronics manufacturers usually make a series of short siliconchip.com.au production runs to iron out errors in their new designs – and even then, they sometimes have recalls. Anyway, from the foregoing, it appeared that there was a problem with the W output on the PCB (pin 18 on the IGBT chip). We very carefully checked everything we could on the PCB: voltage checks without a load, waveforms, whatever. Nothing appeared amiss. At the same time, another reader reported that he found that the supply provided by zener diode ZD1 was too low at around 4V. This would cause faulty operation of the overvoltage comparator (IC2a) and could be fixed by increasing the bias current through ZD1. So maybe we could fix the interference susceptibility of IC2 by generally reducing the circuit impedance around it. We tried the effect of reducing the bias resistor to the zener diode to 560Ω and reducing the positive feedback resistors at pin 4 of IC2 by a factor of 10. The changes did not work. OK, so I decided to connect a radiator with several heat settings, across the W & U outputs (as originally labelled), together with a 100W incandescent lamp. The idea was that I could easily see whether the output was being varied while it drove a substantial resistive load. The Controller was set into pump mode. At the lowest setting, all appeared OK and encouraged, I switched the radiator over to the highest setting which was 1.8kW. This is nominally more than our published power rating of 1.5kW but I figured that it should handle this since December 2012  81 4004 4004 150k 150k WARNING! NE-2 NEON BR1 GBJ3508 (UNDER) CUT THIS TRACK TH1 SL32 10015 DANGEROUS VOLTAGES (COVERED) FUSE1 10A Neutral Earth Active CON3 FLT1 YF10T6 470F 470F 400V (UNDER) + U W PIN 18 IGBT module pin numbering does NOT follow conventional pin numbering. NOTE: + MOTOR V 16k 100nF 620k 1 10F 10F 100nF 10F 10F 100nF 470F 400V (UNDER) 4.7k 5W ZD1 470F LM317T ISOLATION BARRIER T2 6V+6V 5VA (UNDER) 100nF 100nF REG1 D5 D6 D7 D8 100nF 100nF + 4004 IC3 CON4 RAMP 100nF VR1 VR2 10k 100nF 10k SPEED 100nF 100nF 100nF dsPIC33FJ64MC802 10F 470F CON5 CON7 CON6 ZD2 BC337 Q1 PP Ext O/S Flt A A Rev Run Fault A 1: Detach PCB from heatsink and disconnect the thermistor wires. 2: Cut pin 18 of IGBT module very close to the underside of PCB then straighten out remaining pin horizontally (taking care not to break it!) so that it SEE: PIN 18 projects away from the module and cannot make contact with the PCB pads. 3 Remove barrier terminal strip CON2 and refit it to the PCB one terminal space to the left as shown above. A new 1.2mm hole will need to be drilled in the PCB to accept the left terminal lug. 4: Under the PCB, run a length of mains-insulated wire between the straightened pin 18 of the IGBT module and the lug at the left end of the terminal strip (dashed line with red sleeve, above).       This terminal will become the new “U” motor terminal. 5: On the top side of the PCB, run a length of mains-insulated wire between the PCB pad which originally connected to pin 18 of the IGBT module and the negative terminal of the centre rear 470µF/400V electrolytic capacitor (blue sleeved wire shown above). 6: On the top of the PCB, cut the copper track connecting to the negative pin of the 470µF electrolytic located just to the left of thermistor TH1. SEE: CUT THIS TRACK 7: Under the PCB run a length of mains-insulated wire between the negative pin of the 470µF electrolytic capacitor referred to in “6” above and the earthy (upper) pin of the 100nF MKT capacitor located just below the lower right corner of the IGBT module (dashed line with light blue sleeve shown above). 8: Re-attach the PCB to the heatsink as before and re-connect thermistor. MODIFICATION STEPS: W U CON2 2.2k 220nF X2 470F 400V (UNDER) 0.015 2W WARNING: ALL PARTS IN YELLOW AREA OPERATE AT LETHAL VOLTAGE & LETHAL VOLTAGES REMAIN FOR SOME TIME AFTER POWER IS REMOVED – SEE TEXT EARTH 4004 D1 D2 D3 D4 4004 4.7k 5W 4004 4.7k 5W 1.5nF IC1 STGIPS20K60 (UNDER) 4004 T1 6V+6V 5VA (UNDER) 47nF X2 47nF X2 8.2k 10k 1.1M 470 4004 100 620k 8.2k 8.2k 5.1V 8.2k OPTO2 100nF 4004 10 47k 1.5k HCPL2531 OPTO3 ISOLATION BARRIER OPTO1 4N35 HCPL2531 IC2 LM319 100 4.7k 100nF 10105121 +3.3V 15V 100 Vin S1– 4 680 100 GND 470 100 REV 100 RUN 100 D9 110 1.5k 180 100 1.5kW Induction Motor Speed Controller 4.7k Having turned it on, the output voltage ramped up for a few seconds, whereupon there was a loud bang. Not good! EST Bang #1 GND the continuous single phase current rating was 8.5A RMS. And anyway, if it did not like it, the current overload protection would cut in and no damage should be done. 100 82  Silicon Chip 100 + + ICSP + + Fig.1: the amended PCB component overlay for the Induction Motor Speed Controller using the original PCB (the circuit remains the same). As well as the mods detailed in this article, it also corrects U/V/W (motor connection) confusion. Upon opening up the case there was the typical burnt component smell and the lid had a burnt patch which just happened to coincide with the position of the 15-milliohm surface-mount current shunt on the PCB. Umm – where’s the current shunt? It wasn’t there – it had been completely vapourised! The siliconchip.com.au 75 Dumb and dumber! 65 Then I did something really stupid. I decided to do a live voltage check around the unit. The reasoning was that the main fuse had not blown. This should mean that there was nothing wrong with the power supply itself and all the rest of the circuit should be OK, even without the current shunt because it would not have any load connected. I duly put the lid on the box (a good idea, as it turned out) and resolved that I would power it up and then take off the lid to check that the neon indicator was on. This would indicate that the high voltage power supply was OK. So that’s what I did. 5 5 ALL DIMENSIONS IN MILLIMETRES 5 on-board 10A fuse was still intact though. Well, this wasn’t supposed to happen. What about all the protection features? I carefully checked the PCB for any other signs of damage and could find none. I then did as many continuity tests on various components such as the optos, comparator IC, the IGBT bridge etc. 200 45 60 85 105 130 ALL NINE HOLES ARE TAPPED M3 There was an enormous bang inside the box and all the computers and lights in the office went down as the main circuit breaker tripped. For a moment, I was too dumbfounded even to swear. But then I let loose: a long stream of invectives about a stupid and incompetent idiot, someone who should not be let within ten metres of mains operated equipment and so on and so forth. As Bugs Bunny was often heard to remark, “What a revoltin’ development!” Once I disassembled the PCB from the heatsink, it turned out that the IGBT bridge had large bits of plastic encapsulation blown off it. You could see the remains of tiny PCBs and surface-mounted components. Of course, this confirms that the IGBT bridge is not a single large monolithic chip but is made up of a number of separate tiny PCBs for logic, boost supplies and the IGBTs themselves. Other components that were likely to have been damaged were the three optocouplers and the comparator IC. These were all replaced from our component stock, various checks were done and then we powered the unit up with a load. Guess what? It all worked. We could not fault it. Any serviceman doing such a repair would be very happy. It’s all fixed. Beauty. Button it and and send it back to the customer. But we weren’t happy; not in the least. We still did not know what the original fault was. The now-repaired speed controller was returned to Altronics. 25 Bang #2... We then had some very useful feedback from reader Geoff Clulow, who had found problems with two units that behaved very similarly to the defective unit that we had destroyed and repaired. After a lot of investigation he determined that there was interference between the U output of the IGBT bridge, IC1, and the adjacent LM319 comparator, IC2. What was happening was the comparator was sending a false error code if a load was connected to the U terminal. His solution was to cut the track from pin 18 of IC1 to the respective terminal U on the 3-way output terminal block. He then ran a separate wire, well away from the LM319 siliconchip.com.au 5 What next? Fig.2: corrected heatsink diagram. The measurements were correct but one hole was out of position. dual comparator down under the filter block (FLT1) and terminated at the edge of the board. OK, so this confirmed that the problem involved the pin 18 output from the IGBT bridge and the comparator IC, although we now think that there is also interference coupling into the control circuitry inside IC1 from this track. This IGBT bridge is actually a “hybrid module” which December 2012  83 ADDED WIRE CUT THIS TRACK CON2 MOVED The top side of the board showing the three modifications required here – the black wire is an addition and CON2 is moved over one hole (new hole required). The third mod is the cut PCB track immediately to the right of the 470µF capacitor (top left of board. PIN 18 LIFTED, WIRE CONNECTED ADDED WIRES We’re only showing a section of the underside of the board here for clarity, so you can see exactly where the additional wires go. Note that PIN 18 of the IGBT module is also cut and bent up clear of the PCB. 84  Silicon Chip siliconchip.com.au contains a number of components including the six IGBTs, six normally reverse-biased power diodes and the driving and control circuitry. Tracks on the main PCB running close to this module, carrying high currents with fast voltage ramp times, could possibly interfere with the internal control circuitry. This presumably affects the operation in such a way as to bypass the module’s protection features and we think that is why the modules can blow despite having short-circuit and over-temperature protection. So why didn’t we spot this problem during the prototype stage? It appears that most modules work fine with the original design but some small percentage are “fussy” and pick up enough noise so that they do not operate correctly. To put the problem into perspective, we believe that more than 100 motor speed controllers have been built to date but only a handful of constructors have experienced problems. Regardless, it quickly became obvious that we needed to find a solution. As a result, we have a devised a procedure which incorporates the modifications suggested by Geoff Clulow. In essence, it involves isolating pin 18 of the IGBT bridge and connecting it direct to a pin of 3-way connector CON2 which itself is moved over to the left. At the same time, the now-disconected track from pin 18 is then connected via a wire to the negative terminal of the central 470µF 400V capacitor (shown in blue on top of the PCB). This earths the disconnected track. Also to correct an error we discovered in the star point earthing on the PCB, we have cut the track to the negative electrode of the 470µF capacitor near D4 and connected it instead to the negative pin of the 100nF capacitor near pin 16 of IC1. The modifications are shown in the diagram of Fig.1 and this includes the step-by-step instructions. The photos also show the modifications. IGBT module appearance While investigating this issue, we also discovered that the STGIPS20K60 module is made in two different factories. These modules differ slightly in appearance – see the photo above. Besides the laser-engraved labels, other differences include the shape of the isolation cut-outs between the pins and the finish of the plastic encapsulation. Both are genuine ST Micro parts and presumably their internal structure is the same. We believe either type can be subject to the failure mode described here. OK, let’s go through the steps. First: remove the PCB and heatsink assembly from the case. Detach the PCB from the heatsink. To do this, you need to remove the five screws for the mounting pills, the two screws for the IGBT bridge (IC2) and the one for the bridge rectifier (BR1), which attach these devices to the heatsink. You might also like to disconnect the thermistor (TH1) because too much flexing of its leads will break them. Second: cut pin 18 of the IGBT module very close to the underside of the PCB and then straighten it so that it projects out horizontally from the chip. Again, not too much flexing or you could break the lead off. Third: remove the barrier terminal strip CON2 from the PCB and refit it on the PCB on terminal space to the left, as shown in Fig.1. A new 1.2mm hole will need to be drilled in the PCB to accept the left-hand terminal lug. Naturally, that leaves one original hole vacant. siliconchip.com.au Here’s a close-up, not too far off life size, of the underside of both versions of the GIPS20K60 IGBT chips. The top one, with the square notches, is made in ST Micro’s Chinese factory as indicated by the “CHN” in the label. The one shown below is made by a subcontractor in Taiwan (“TWN”). With the mods detailed here, both should be quite OK. Fourth: under the PCB, run a length of mains-insulated wire from the straightened pin 18 on the IGBT bridge to the left-hand terminal lug on CON2. Solder the other terminals of CON2 to their respective (ie new) PCB pads. You could also place a short length of heatshrink tubing over the soldered connection to pin 18. Fifth: ideally you should also move the adjacent supporting pillar for the PCB so that it is not too close to the relocated terminal W on CON2. This will require another hole in the PCB and a drilled and tapped hole in the heatsink. Sixth: above the PCB, run a length of mains-insulated wire between the PCB pad originally connected to pin 18 of the IGBT module and the negative terminal of the centre rear 470µF 400V electrolytic capacitor (shown as a blue sleeved wire on Fig.1). This effectively grounds the now unused track and provides some shielding to the LM319 comparator IC. Seventh: on the top of the PCB, cut the copper track connecting to the negative pin of the 470µF electrolytic capacitor located just to the left of the thermistor (TH1). Then under the PCB, run a length of the wire between the negative pin of the now isolated negative pin of the just-mentioned 470µF capacitor to the earth (upper) pin of the 100nF MKT capacitor located just below the lower right corner of the IGBT bridge (dashed blue line with light blue sleeve shown in Fig.1). This corrects the error in the star-earthing on the PCB, as mentioned previously. Finally: you need to reattach the PCB to the heatsink and reassemble it into the case. Then run all the checks described in the original article. If you are assembling a kit with the original PCB (see overleaf), you would obviously not solder pin 18 of the IGBT bridge to the PCB but would bend it out horizontally and solder the red mains-insulated wire from it to the W output terminal on CON2. 3-phase wiring Finally, a note on the 250VAC-rated cable to be connected to the output connector CON2. Instead of supplying the kit with a surface-mount 3-pin chassis socket, Jaycar supply the kit with a short extension lead which is meant to be cut and stripped to provide an input lead with moulded 3-pin plug and an output lead with moulded in-line 3-pin socket. This is quite a valid approach if you are driving a standard single-phase induction motor. However, this is not appropriate if you are building the December 2012  85 Rev ZD2 CON6 CON5 RAMP CON4 SPEED U 8.2k 8.2k 8.2k 620k 620k IC1 STGIPS20K60 (UNDER) 16k 1.5nF 220nF X2 150k WARNING: ALL PARTS IN YELLOW AREA OPERATE AT LETHAL VOLTAGE & LETHAL VOLTAGES REMAIN FOR SOME TIME AFTER POWER IS REMOVED – SEE TEXT CON3 4004 4004 4004 4004 EARTH FLT1 YF10T6 D1 D2 D3 D4 (COVERED) FUSE1 10A NE-2 NEON 150k BR1 GBJ3508 (UNDER) TH1 SL32 10015 470F 400V (UNDER) WARNING! 47nF X2 470F Neutral Earth Active 47nF X2 + DANGEROUS VOLTAGES W CON2 0.015 2W V MOTOR 100nF 100nF VR1 VR2 10k 100nF 10k 100nF 100nF 8.2k 1 100nF 1.5k 10F BC337 Q1 100nF 100nF dsPIC33FJ64MC802 10F IC3 ZD1 OPTO2 2.2k A A A Run Fault CON7 HCPL2531 OPTO3 10F 100nF 10F 470F 100nF 100nF REG1 D5 D6 D7 D8 LM317T ISOLATION BARRIER T2 6V+6V 5VA (UNDER) 470F HCPL2531 10F 10k 1.1M OPTO1 4N35 100nF 5.1V 470 4004 T1 6V+6V 5VA (UNDER) How do you tell which board you have? The easiest way to ensure you have the new PCB is to look at the number on the silk-screen overlay. New boards will have the number 10105122; original boards will be numbered 10105121. SC IC2 LM319 4004 470F 400V (UNDER) 100nF 10 4004 470F 400V (UNDER) ISOLATION BARRIER 100 4004 + +3.3V 47k 100 4.7k 5W Vin 100 470 100 D9 + GND 100 4004 4.7k 5W RUN 100 S1– 4 4.7k 5W 4.7k 100 180 110 + 86  Silicon Chip REV 1.5k 15V 4.7k EST 680 10105122 10105122 100 + To accommodate the changes in this article without having to add extra wires, etc, we have revised the original PCB pattern and it is reproduced below with component overlay. To recap, if you build this project with the new PCB shown below, none of the changes we’ve detailed above will be needed – they’re all taken care of in the PCB pattern. However, there may be many original kits (with the old PCB) in the marketplace and/or in constructor’s hands but not yet built. Obviously, if you have the original board, the modifications will be required. (Alternatively, new PCBs can be obtained from the SILICON CHIP PartsShop – see page 104). GND 100nF 1.5kW Induction Motor Speed Controller Revised PCB pattern 100 + This cable is not difficult to obtain from electrical wholesalers and even large hardware stores. If an electrician does the installation, make sure it is done this way. 100 + ICSP PP Ext O/S Flt Controller to power a 3-phase motor. For a start, you cannot use a standard 3-pin mains socket for the job and you cannot use the green/yellow earth wire as one of the phase outputs. The earth wire must never be used as an active conductor, not even for a brief test. The correct cable is a 440V-rated, 4-wire flex, with 3-phase coloured conductors for the motor terminals and a green/green yellow earth conductor which must be connected to the motor frame (a terminal is usually provided in or adjacent to the motor cable entry box). Fig.3: component overlay for the revised PCB which does NOT require the modifications in this article. Simply follow this diagram. Check to make sure that you have the newer PCB by looking for the new board number (top centre of upper side of board). siliconchip.com.au