Cambridge Audio A3i repairs and mods

Thanks for the tip, spandrel, I'll check those points, though the fact that the heatsink gets too warm too quickly suggests to me that too much current, rather than too little, is being drawn.
As to the Towers, and other data books, it's a case of no-one is infallible. To me, it isn't surprising that the occasional error slips through the net. I have one rather old edition of Towers, several of the Mullard Data Books from years gone by, plus two 'dov' transistor data books, and I wouldn't be surprised if an error or two didn't appear in one or the other of them. Your experience highlights the virtue of 'check and double check, whilst your last comment shows up the difficulties inherent in repairing DC coupled transistor amplifiers. I sometimes feel the same way!!
 
I've just noticed your comment re; DC offset, spandrel, but surely a consequence of setting the bias as per the maker's instructions should, all other things being equal (O.K. they seldom are!!), result in the DC offset at the o/p stage center point being at the minimum value possible?
 
No. It's more likely that the offset would be worse with no quiescent current. It's the DC feedback which controls the offset. I've never measured DC offset while adjusting quiescent current so I've just tried it on a NAD 3020 which has a similar o/p arrangement. The offset started at 11mV with the correct quiescent current and jumped to 12mV without any. Some amps have active control of DC offset, referred to as DC servo. This NAD has an adjustment preset but the A3i relies on the matching of the long tailed pair input transistors. The one I just mended had an offset of 60mV because I didn't match the gain of these. I'll have to do that when/if I get it running again.
 
Thanks for the explanation. Earlier today I found that Q10 was completely open circuit. Not having any BC639s to hand, I looked in my semiconductors box and found som 2SC 1845s, so I fitted one as a replacement. I then ran the power amplifier for 40 minutes or so with no sign of any misbehaviour, so refitted the amp. PCB and re-connected the control and pre-amplifier boards, expecting all to be well, but, when I switched on again, with the phono input selected, the relay still tripped in and out as the volume control was adjusted. Admittedly I haven't yet checked and adjusted the bias settings, but I' still puzzled. I have replaced c1 (and C101) with new 10uf 63v caps, wired the correct way round-i.e. +ve terminals to the inputs, and also replaced C8 & C108. Still more problems to sort out!!!
 
I don't think C1 is relevant, it's only there to avoid DC coming in through the CD input. That's the only direct connection. Have you got an Avo or similar meter? It's easier to see a needle twitch than watching digits. See if there's a DC offset blip when the relay trips out. I'd set up the quiescent current though before doing anything.

If the DC offset isn't changing it's not that that's tripping the relay, so it must be the over current sensor or a fault in the protection circuit. As it's only protecting the speakers I'd ignore it for now until you're sure the amp's working ok. Use headphones connected through 100R resistors direct to o/p. If you have an oscilloscope check it's not oscillating at RF.
 
Hopefully I have at last found the cause of the faults. Being puzzled as to why there was up to +10v on the 'hot' side of the volume control, depending upon it's setting, I did some voltage checks on the control PCB only to discover that the -15v supply was missing at the anode of ZD502. With hindsight this also explained the crackling noises when operating the 'direct/tone/mute switch. The cause was a break in the track at the ZD502 end of R512. After having re-soldered R512 and two other nearby resistors which had dry solder joints, plus fitting a wire link between the end of R512 and the anode of ZD502, then re-assembling the amp. I test-ran it for maybe 20 minutes with no sign of any problems. No DC voltage where it shouldn't be and no crackling noises when operating function switches either. Not the easiest of amplifiers on which to work, but I'm glad I've finally solved the problems. Can't quite work out why lack of the -15v supply caused a positive DC voltage on the v/c, though!
 
The volume control is fed from IC501 which would skew full +ve with no -ve supply. Well done for solving that one, it's always worth poking about with a meter. I found an odd fault in an Arcam like that. Are you going to proceed with the upgrades? I'd recommend doing one channel first and checking it works after each change.
 
As it's not my amplifier, I'll have to ask the owner, but for now I'm just glad that it's working again. During my checks & tests, I noted that the rectifier diodes actually do have parallel capacitors, although none are shown in the diagram. It would be relatively easy to replace D210 & D211 with 10 ohm resistors, but I may leave the diodes in place. It would be much easier, as I'm sure you'll agree, to repair or modify this amplifier if panels could be removed from the bottom to expose the printed side of the circuit boards, as is the case with some other amps.
 
I think it's the LED and zener resistors that need to be higher power and lifted off the board, and the higher power VAS and drivers. The resistors instead of D210/11 are a performance improvement.

It would be much easier, as I'm sure you'll agree, to repair or modify this amplifier if panels could be removed from the bottom to expose the printed side.

You should try an Arcam.
 
I did once repair an Arcam, but I won't say 'Never Again!!' Almost feel that way about Cambridge amps. after battling with the A3i, though I have repaired Cambridge amplifierd before. I'll check out the LED & Zener resistors. Most are miniature 0.25w. types, but changing these components will involve dismantling the amp. again. Still, if it improves long-term reliability perhaps I should do so.
 
Following on from the above, there's no sign of overheating or stress in any of the resistors you mention, spandrel, so, as I've other things to do, I may put the A3i on soak test to check it's long-term reliability before spending any more time on it. Interstingly there's been no word from x-pro(Alex) the designer of the amp. about any of this. The pre-amp board, as you no doubt know, is marked as being by Michael Creek, designed presumably by him before he set up Creek Audio.
 
Last edited:
Something occurred to me which is nothing to do with the faults or the difficulty of working on the A3i, it's simply that I' curious why the legend on the rear panel tells the use only to connect to the lower set of speaker sockets when using two speakers, when the two sets of sockets are simply connected in parallel, so it makes no difference which set is used. No doubt if an owner so wished the connections could be separated, and A/B speaker switching added (not that I'm planning to do that).
 
The DC offset on the L.H. o/p, measured across R22 & R23 is approx 50mV, whereas that across the R.H. channel is only a few mV, and can be adjusted to the 24mV (12mV) across each 0R22 resistor specified. As the amp. is running O.K., subject to a longer test run, I'm inclined to leave things as they are as I can see no way of reducing the 50mV by very much, short of changing components. Any opinions/suggestions regarding this welcome.
 
I'm sorry if I've not explained things well but you seem to be confusing DC offset with quiescent current. The DC offset is the voltage across the speaker terminals and should be less than 0.1v, preferably zero. The only way to affect it is to make sure the input long tailed pair have the same gain, not easy to do.

The presets adjust the quiescent current (no signal) through the o/p transistors, measured by measuring the voltage across both emitter resistors. 50mV divided by 0.44 ohms is 114 mA which normally is much too high but not a problem for these chunky Sanken transistors. The recommended voltage is 24mV (between emitters) which is 55mA. Make sure they are both about 24mV.

x-pro does suggest upping the quiescent current to 80 or 90 milliamps (40mV between emitters) to reduce distortion, but that's your choice.
 
Spandrel, I apologize if I misunderstood your explanation, though I don't think so. I'm aware that the pre-sets adjust the quiescent current through the o/p transistors, but the approx. 50mV measured across the emitter resistors is essentially the same figure as that across the speaker terminals with a speaker or equivalent dummy load connected. I may, however increase the Iq to around 80mA, depending upon audible results before and after long-term 'soak' testing, though the first priority is obviously to reduce the Iq through the channel with the higher reading.
 
Possibly, because the true DC offset is measured across the speaker terminals, the negative of which is grounded, so I should have said between the junction of R22/r23 and earth. In fact the voltage drop across the two resistors is very small - I didn't not the exact figure.
 
Hi OXAUDIO,
I used to do warranty service for Creek in Canada. I'd follow any advice given by x-pro, but I wouldn't want the heat sinks to run too warm. Just saying, and I'm known to be a cautious technician.

If you have a 50 mV DC offset on one channel, I would be inclined to fix that. What it needs is a matched pair of input transistors. The tighter the match is, the lower your DC offset will be in general. This also lowers distortion. When installing these transistors, put them in thermal contact with each other (a little dab of heat sink grease) and a piece of heat shrink tubing over them to ensure they stay in contact. If you can do the same to the other channel, it's not a bad thing to do.

Matching transistors can be tricky. The two transistors have to be at the same temperature and your breath can throw them out - touching will make it impossible to find a match. What you might find helpful is to lay them out and leave them alone for 1/2 an hour or so. THen use tweezers to insert them into the test jacks on your tester. It's easier to do in a room that doesn't receive sunlight (at all), and blocking any heating vents for this period in time will help as well.

This advice generally works for all transistor differential pairs. Sticking random transistors in will not be an efficient use of your time, and any other matches you find can be used in other amplifiers.

No matter what paperwork says, 50 mV DC offset is my cut-off point for acceptable offset. The exception would be an amplifier that has high offset designed in. Normally cheaper units with simplified circuitry. You can figure this out by calculating the base current (approx.) and find the voltage drop through the base to common resistor. Then repeat for the feedback resistor that runs to common. If there is a blocking capacitor, then the voltage drop you calculate for the input will be the expected DC offset for the amplifier. Otherwise it would be the difference in voltage drops between the two sides of the diff pair. This is a very approximate, quick way to find out if you are wasting your efforts to reduce DC offset. I have seen amplifiers that had over 200 mV of designed in DC offset.

Probably too much information, but hopefully it would be useful to you in general.

-Chris
 
Thanks for the info anatech, you've convinced me to use matched gain transistors and strap them together. I haven't got an hFE tester. Are the simple designs on the net good enough? Would a dual transistor package be a better idea?

The collector current of the ltp transistors is just 0.3mA so the base current will be less than 3μA. The feedback resistor is 56K (the gain is 57!) so the voltage drop will be less than a millivolt. ltp have constant current load in tail and current mirror for collector loads. The next stage is a fet, so no current there.

OXAUDIO, the voltage across the two emitter resistors R22/23 is the same as measuring the voltage between the emitters. It's just more practical to use the resistor leads than to poke about under the board. This is the voltage you need to check to set the quiescent current. You shouldn't have speakers connected for any measurements.
 
Hi spandrel,
Don't you have a meter with an Hfe section? Most new digital meters throw that in there, the less expensive ones do at any rate. I don't know if Fluke or Keysight does, and they are my go-to brands for great meters.

You can make a diff pair setup to do this. I designed one and gave it to the forum. You'll find it near the end of one of the GFA-565 amplifier threads. A few members have designed PCBs for it that are a little different than mine, but they work fine.

-Chris