Repair of a 24V x 75Amp Mastervolt battery charger 

 I've repaired, or tried to repair, a few complicated items from boats and the latest is a Mastervolt (Mass 24/75) battery charger. Not the simple type with which car owners are familiar, but a big really expensive one for charging batteries on a boat. In this case there are twelve lorry-sized 12 volt batteries connected in series/parallel to supply lots of amps at 24 volts. I repaired this several months ago after a mains surge had blown up a couple of protection devices and it worked OK for a bit before two of the batteries got extremely hot and expired. This was almost certainly due to a temperature sensor (connected to the charger) not being fitted to the bank of batteries causing the pair nearest to the charger to expire from excess charge.

Oddly, the charger then failed to work and arrived back at the (ex) Low Cost Repair Centre (which is now only open for friends and family).

 

 Plugging the charger into a mains supply failed to power it up so I looked for any evidence of internal DC voltages. This is not so easy because the upper part of the circuit board and the components are conformally coated and this is a particularly thick coating making it a bit hit and miss when probing for voltages and measuring components. I did discover the high voltage supply derived from a large bridge rectifier was present. This is used to develop at least one set of low voltages which power the control circuits and these were missing. The problem then was to identify the chopper power supply parts, especially the chopper chip itself. The identity of some chips is easy enough as the codes show through the conformal coating but the majority are anonymous, requiring the coating to be removed without erasing the codes. Oddly I didn't find a chopper chip so I guess it's one that is difficult to get at so instead looked at components typically responsible for power supply failure.

I found several capacitors that measured either low in capacity and/or with a high ESR but replacing these one by one didn't fix the problem. To test the effect of each replacement I checked the voltage at the logic circuits near the on/off switch. Each time this read zero.

At the start of proceedings I'd noticed a larger than average resistor marked brown/grey and whitish but got an indeterminate reading across this (the effect of which I put down to large capacitors or residual voltage) but after swapping the capacitors I decided to look again and to check the voltage at each end of the resistor. I discovered one end was 320 volts and the other end zero volts which seemed wrong. I switched off and waited until the high voltage had decayed then examined the resistor again. I reckon the white band should have been yellow but probably heating had discoloured this and it was now white. This makes sense as there's usually a high value resistor associated with chopper supplies which triggers the thing into life. After being triggered the chopper circuit is self-powered but, if a capacitor fails, the resistor will carry the trigger current for a prolonged time. Typically a 180K resistor in this application fed from 320 volts will dissipate a bit over half a watt, but add to this some voltage stress (a parameter not widely understood) and reliability will be poor.

I removed the 180K resistor (which indeed measured open circuit) and fitted two 91K one watt resistors in series.

This time, once a pair of series-connected 12 volt batteries were in place, and with mains power applied, the set of LEDs around the on/off switch lit up. The red charge LED lit initially then extinguished correctly as a pair of orange LEDs lit once the batteries accepted charge.

This is one area of the circuit board carrying chopper power supply components. I fitted two new capacitors in this area. One can be seen at the back and the second left centre. These are 22uF x 63v replacing two smaller capacitors marked 22uF x 25v. Neither chip in this area is a chopper controller, the nearer being an LM339 quad op amp (coating removed) and the other an HEF4050 (hex buffers). The pair of 8-pin chips at the back are optical couplers.

 

 

 

 

Below is the adjacent area to that opposite and shows two new capacitors and the pair of 91K resistors. The near 100uF x 63v capacitor replaces the original 100uF x 25v which measured about 40uF and nearly 10 ohms ESR.

The new capacitor at the back is 10uF x 100v and replaces a 10uF x 25v which read open circuit.

Again, I found a hex buffer chip, a quad op amp plus a "555" timer (centre). At the rear are a couple of chips whose conformal coating masked their identities but one may well be the chopper control chip.

 

 

Above.. the duff parts. 

Below is a general view of the main circuit board. Access to components is difficult and those aluminium parts mostly used as heatsinks are riveted to the board adding to the complexity of fault finding. Looking at the back of the printed circuit board you can see factory modifications (those white gluey areas) suggesting power supply or controller design problems.

 

 

 
 

 Stability Control Board

Although I'm retired I still do the odd job when it sounds interesting especially when it comes to things to do with ships. Maybe it's a kind of throwback, because I'm not the first engineer in my family.Way back in 1910 my great grandfather James Murray retired. He'd been working for Cunard since he was a lad and steadily climbed the promotion ladder sailing on the Liverpool to New York run and ending with his last ship, the Mauretania which held the Blue Riband as the fastest ship across the Atlantic. My great grandad was Chief Engineer.

 

 

 

 Anyway, back to repairs... A chap from a local shipping line handed me a faulty printed circuit board made in Italy by a company which is still very much in business over there and, after some puzzling over it, contacted their design engineer about what I'd found.

It looked pristine or nearly so except I noticed a pair of resistors that looked slightly discoloured.. not too much but they were very small and not even as big as nearly all the others.. so why weren't they larger? Discoloured meant they'd been too hot so why hadn't they fitted bigger ones?

This is what I found... the board carries a Traco DC-DC converter marked to accept 24 volts input and plus/minus 15 volts out so I assumed the main supply to the board was 24 volts and sure enough there's a large 2200uF capacitor across the input to the DC-DC converter marked 50 volts.

Because the board is faulty I initially applied 5 volts across the capacitor with a low current limit and firstly monitored a 5 volt regulator also fed from the 24 supply. As I gradually increased the voltage the 5 volt supply stabilized and seemed OK at 12 volts. I then switched over to monitoring the plus 15 volts from the DC-DC converter and as I reached about 22 volts the 15 volts appeared. The neg 15 volts was also present and the supply current drain was relatively low. I'd hoped the DC-DC converter had been faulty as this would have been an easy fix, but it was fine.. in fact the whole board seemed to be OK so I turned on my thermal camera and looked for anything odd.

Pretty soon I found two transistors (marked Q1 and Q2) running fairly warm.. at more than 60C, and still rising, which isn't common for TO92 devices.

 

 Top left is Q2 marked PN2222A and top right Q1, a JC2907.

Left is the 5-volt regulator with 24 volts input and 5 volts out.

Over a long period the temperatures would increase somewhat and whilst regulators like this 7805 generally run quite warm it's unusual for a general purpose transistor to run hot.

 I checked their markings.. a PN2222A and a JC2907 (that latter not one I've come across before but in fact likely to be the equivalent of a PN2907, which is the complementary PNP equivalent of the NPN PN2222A). I looked at the board again and I noticed both transistors were fitted the opposite way round to the circuit board markings.

I tested both and initially the PN2222A gave odd results on my universal tester so I checked it on my Peak transistor tester and this found a parallel diode across base-emitter. The next day the same tester said it was two diodes with common cathodes. My universal tester oddly declared it to be an PNP transistor, so I tested it with my multimeter and this indicated it had what seemed to be either an extra diode across base-emitter or a bad resistive leak. I guess the latter because, in both directions the base-emitter reads 0.540 instead of about 0.680 with my diode tester.

Now I was puzzled because although there was a wired mod at the other end of the board there was no evidence of resoldering at Q1 and Q2 which might have explained why they were back-to-front. So, fitting these had been deliberate and presumably correct according to the makers documents so I traced the circuitry around them. The DC-DC converter spec revealed the pins at which the plus/minus 15 volts emerged.

At this point things began to get quite complicated because of the pinouts of the transistors. Some have their emitter/collector pins the opposite way round and tracing the circuit revealed that both the PNP and NPN must be pinned the same as each other to match the tracking. Try as I might I failed to discover the spec of the JC2907 but that wasn't critical because it actually tested as a good PNP transistor.

That's when I contacted the manufacturer.

 Above is the circuit board with the two transistors Q1 & Q2 removed and shown on the right (Q1=JC2907A, Q2=PN2222A)

Now here is my paraphrasing of the manufacturer's engineer's explanation.

The board was originally designed for Q1/Q2 having opposite pinouts to those fitted, but during manufacture, the original transistors became difficult to source, and I'm informed, two alternatives were substituted viz. the PN2222A and the PN2907A. These replacements have their emitter and collector pins reversed to those in the original design so were fitted back to front ie. not matching the board markings.

No problem.. the board would work OK, and so manufacturing would proceed, correctly working from modified production drawings to reflect the change, meaning the alternative components were fitted correctly, but here the JC2907A was substituted for the PN2907A.

To repair the board I needed to trace the circuitry to identify the pinouts of Q1 and Q2, and make sure I selected and fitted new parts that match the circuitry. During this puzzling exercise I found that the original Q1 device must have been a version of the 2907A but pinned the opposite way round. Just to confuse matters some manufacturers number the TO92 pins back-to-front and there were even identically marked transistors with reversed emitter/collector pinning.

To recap...the matching pair of devices fitted to this board are the JC2907A & PN2222A as shown and these are both fitted back-to-front.

What were the original transistors? I'm told they were the 2N2222A & PN2907A but this can't be so unless they were pinned differently? After much delving into transistor specs they could have been the P2N2222 & P2N2907 with or without the "A" (as shown opposite).

If you carefully study the drawing above right you'll see exceedingly confusing numbering differences between different transistor bases. 

My choice for new transistors is below with pin numbering as shown below left:-

Q1 = BC327-40 tested as 1(C)-2(B)-3(E) and PNP

Q2 = BC337 tested as 1(C)-2(B)-3(E) and NPN

 Once the new transistors had been fitted (above right), I powered the board from 24V and it drew 106mA. The voltage across each of the 22 ohm emitter resistors measured 490mV. This represents a transistor emitter current of 22mA. The emitters measured plus and minus 14.5 volts resulting in a dissipation of about 320mW each. Both transistors are rated at 340mW at 80C and their temperature now measured over 70C. As the NPN transistor was hotter than the PNP I believe that temperature was the reason for the board failure as an ambient rise over my workshop temperature, which was only about 12C, to over 20C would have meant the NPN transistor was operating outside its safe area with consequent poor reliability.

 

Although I've assumed the original design used TO92 devices it's possible that TO18 types could have been initially chosen. The reason I say this is that the 2N2907 is available in a metal case. The 2N2222 may also have been selected as some makers supply that device in a TO18 case. Both these can accommodate a 50% increase in dissipation and at a 50% higher case temperature.

Could the board failure have been due to a failure to appreciate the reduced rating of the TO92 plastic devices?

 

I wasn't happy once I'd re-checked the data sheets so removed the TO92 devices and fitted a pair of TO39 metal transistors viz. a 2N1711 NPN and a BC461 PNP.

The board again drew 106mA at 24 volts and a temperature check revealed the TO39 transistors were operating at around 35C, so well within their rating of over 800mW at 50C.

 
 
 

 A Marine Phone

  I received this odd-looking box purporting to be a telephone the other day with the request to repair it as spare parts are no longer available. In fact, besides the box, there were three identical circuit boards in newish condition and one with really bad water damage (below).

 

 

 I was also supplied with the picture below which shows the main equipment with which the phone connects.

 

 Three boards plus one in the box looked like this and all had the same fault.

 

 

 Below, looking very innocent, are the failed components. These are marked IR1J (Schottky diode = 10BQ100 made by International Rectifier) and 2597H M-5.0 (Buck switcher = LM2597HVM-5.0 made by National Semiconductors). Note that I've shown later data sheets not those available when the parts were made (ie.manufactured prior to the specific published datasheets). It's possible therefore that the failed parts in question do not have the exact specs to those in the datasheets. I'll mention this later. Click for Nat Semi chip markings.

Manufacturing dates for the two chips appear to be 2001 and 2003 respectively.

 

In all four cases of the newish-looking boards The Schottky diode was short-circuit and the switcher chip had a bad leak or short between at least two of its pins. The various capacitors all measured close to perfect. The circuit, parts and its (critical) layout follow fairly closely to those given in the manufacturer's data sheet but with the 24 volt Schottky diode shown as 1N5817 replaced with a 100 volt type.

 

 Interestingly, I've encountered this type of failure in the past when one particular type of a numeric indicator board for a lift had very similar faults, but I recall that in those examples the fault was most likely due to a bad capacitor. Nevertheless the Schottky rectifier and its associated voltage regulator chip were always bad. I never figured out the reason for their failure but I'm tempted here to try harder.

 My plan is to work out a way of repairing and testing without shutting out any chance of getting the phone to work (spares are no longer an option and the microprocessors are critical as they carry proprietary firmware). In several cases of the numeric indicators I've repaired in the past the incoming high voltage had punched through the 5 volt regulator and blown up all the chips.. typically an SN74HC14 SOIC.

The question is therefore.. are the three surface-mount chips on these boards OK? They're all complex types viz. an ATMEGA128L, PSB21373 and a MAX202. The first is not available as a programmed chip.

I applied a low voltage to the repairable boards and found that one had a very hot processor so I removed the three complex chips from that board and then the three corresponding chips from the scrap corroded board and fitted the three "new" complex chips to the repairable board. I removed the bad regulator and Schottky diode and measured the current to the board using an external 5 volt supply. This was much less than the previous current draw and the processor ran cool. This board requires a new regulator and diode. Call this Board "B".

Next I repaired a second board (still with its original complex chips) by fitting a new voltage regulator and diode. As the former is not readily available I used the regulator from the scrap board but fitted a better Schottky diode. One of the potential reasons for power supply failure was a weak Schottky diode (100V and 1A) so I found one with a better rating (100V and 2A) and fitted this. Call this Board "A".

 

 My repair philosophy is to fix two boards and test with the least risk of destroying the last working microprocessor.. These are "A" and "B" boards.

Board "A" has a new power supply and original complex chips. If this works then two of the remaining repairable boards should work once new power supply parts have been fitted. The fourth board "B" with salvaged complex chips should also work once new power supply parts have been fitted. The repair of these three will have to wait until new parts arrive from China. These are LM2597HVM-5.0 and STPS2150A (150V and 2A).

Good news... board "A" went off to the customer and it worked! With a bit more luck three repaired spare boards plus extra power supply chips, in the event of future failures, should be available by mid-May.

Below, repaired board "A" fitted into its case.
 

 

 Now some of the theory concerning the fault

One board can of course be discounted as the problem is water damage which has destroyed the carbon pads and made a mess of lots of solder joints. The other four including that fitted in the metal box all have much the same fault which is failure of the 5-volt regulator circuit. Examination of the circuit board reveals that the designers have essentially used the circuitry recommended by the chip manufacturer so in theory all should be well and the regulator will be reliable.. but clearly it's not.

The diode used is a 10BQ100 which is rated at 100V and 1A against the 1N5817 (24V and 1A) used with a supply of 12 volts.. This change is of course necessary because of our 48 volt supply. The regulator chip is the LM2597HVM-5.0 which is rated by one manufacturer as having a max input of 57 volts and another at 60 volts. The original Industry chip was an LM2597-5.0 (missing the "H" in the coding) rated at 40 volts so at first sight unless the chips have been wrongly coded with "HVM" they should work fine with the system voltage of 48 volts DC.

 The 48 volt supply to the circuit board is fed via a pair of "Line" wires (two of the black wires in the picture above). My first suspicion for the failure was reversal of the plus/minus wires (you can see in the above picture that this would be an easy mistake to make).

However, supply reversal, which would have destroyed the regulator and diode, isn't important because, on examining the circuitry (right), the designers had included a full wave bridge comprising four small surface-mounted diodes (D1-D4). This will guarantee the correct connection of the 48 volt supply to the regulator circuit. On all boards the diodes were found to be OK.

On the right you can see I've removed the bad 5-volt regulator U1 and the Schottky diode D8 together with L2 (for access) and R6 which became unsoldered with U1.

The smaller square black chip marked "GFX39" is the SMCJ48CA TVS diode.

Also note that blue capacitor C3 which will be mentioned below.

 

 What about the effect of an excessive supply voltage which might result from a fault in the main system equipment? In this case, to protect the circuitry the designers have included a TVS diode at the input to the bridge rectifier. This is an SMCJ48CA (the "CA" means the TVS will work on either plus or minus 48 volts). It has a stand-off rating of 48 volts so should be virtually invisible to a good DC supply voltage. Its critical parameter is that it will break down at a voltage of between (plus or minus) 53.3 to 61.3 volts. This means that it will protect a circuit from a voltage of anything higher than a figure between these two values. It will do this by shunting up to 1500 watts through itself with the effect that a fuse (if it's fitted) will blow before it gets destroyed. So given a TVS meeting the "max" figure, we should be able to guarantee that the regulator will never see anything greater than 61.3 volts. This figure might be higher if the worst case tolerance of the TVS isn't met. This raises the question.. do the numbers in the manufacturers spec include manufacturing tolerances and ageing and even if this is true will these figures be met over a long period of storage or use?

According to published specs the regulator chip will definitely work OK up to 57 volts and nominally up to 60 volts so this means that in a worst case scenario the chip might fail because the TVS limit is greater than 57 volts (ie. 61.3 volts). Of course for this to occur the system voltage would need to be high by say a nominal 13 volts making it 61 volts.

If we look further into the circuit board design there's one factor which may have been overlooked by the designers.

The full wave bridge used for polarity reversal protection has a prime feature of rectifying AC. Although the 48 volt system supply voltage is DC (see the picture of the main equipment above) the telephone "Line" circuit may be relatively long or unscreened and might induce interfering noise, which if powerful enough would be added to the voltage seen by the regulator and exceed its maximum rating. This noise would need to be outside the normal operating characteristics of the TVS diode which is fitted before the bridge (eg. a high frequency). In fact after studying the design of the line circuit I see a 470nF capacitor C3 (that blue component) is fitted across the output of the bridge rectifier. That capacitor will raise the composite DC plus AC noise to a peak voltage level whose amplitude will be determined by the load power. In fact load power is pretty low so it wouldn't take much induced noise to kill the regulator. Another factor to bear in mind is that the layout of the circuitry may well result in a high susceptibility to local VHF signals (ie. a nearby VHF transmitter or perhaps radar may raise the DC input to the regulator to something greater than 60 volts).

Further testing on a repaired board disclosed another interesting factor. The regulator chip works with power rather than voltage. This means that given an input voltage of something above the minimum required to supply 5 volts output (in fact circa 8 volts) it's the power input which results in the power output. The 5 volt drain was measured at about 80mA meaning a power requirement of 400mW. At the minimum nominal 8 volts input the current drain from the incoming supply will be 400mW divided by 8 volts or 50mA. In fact due to less than 100% efficiency I measured this at about 100mA. As the input voltage was increased to 30 volts I measured 20mA or 600mW. At the system voltage of 48 volts I assume the board would draw 600 divided by 48 volts or between 10 and 15mA. This makes the input impedance of the order of 5Kohm. This is high enough for the line circuit to be easily susceptible to RF pickup from the ships radar or VHF transmissions.

Remembering that the input bridge will convert incoming RF to a peak DC voltage added to our 48 volts it would only require say 10 volts RMS of RF pickup to damage the board ie 48+10 x root 2= 62 volts.

 I was able to source new regulator chips plus 2A x 150V Schottky diodes from China and once these had been fitted the boards worked OK with all drawing similar currents. Also, reducing the current limit to the repaired boards I found the input voltage varied cyclically which is indicative of microprocessor or similar switching activity. Once the current limit was raised the input voltage remained stable. If more failures are met I can redesign the regulator circuit.

 Return to Reception