SA900C

[ppppenguin]

If the mains transformers are toroids and you can get at the centre hole you could add a few turns to the secondary, wired in inverse series to reduce the voltage. Would have to be rated for full secondary current but I doubt there would be many turns needed.

[Mark Hennessy]

I suggested just that. I reckon the main windings are just about OK as they are sagging under load to give the rated power and no more, it seems. I raised a concern about long-term sine-wave testing, but for music, they'll be fine. The windings for the 15V rails are 24V rather than 15V, so these could be usefully reduced this way, but it might be better to simply add heat sinks to the regulators (depends on the mechanics - haven't seen a photo of that bit).

[Diabolical Artificer]

Here's two video's I made that will give a better idea - https://www.youtube.com/watch?v=pRKDfStE_E8 and - https://www.youtube.com/watch?v=XxumrA4Wrd0&t=61s

I'll get it on the bench today and test it to see what is happening to the rails etc.

Andy.

[ppppenguin]

I suggested just that. 

Sorry, missed that.

[Mark Hennessy]

No worries! I believe the modern expression would be "TL;DR", apparently meaning "Too long; didn't read"

Andy,

That's a lovely workshop! I've very jealous of all that natural light and the views. My workshop is in my cellar, with no natural light and herds of wild elephants above me. Or at least, that's what it sounds like when the kids run up and down the hallway over me (really must fit a carpet one day!).

I've got a much clearer picture of what you've been dealing with now. Certainly not a easy job, but it's getting there. Very impressed with the paint job

Having seen those 15V regulators, I'd have no worries about fitting some kind of heat sink between the PCB and the IC itself. Something like this, perhaps? https://www.sparkfun.com/products/121 - or if there is space, this might be better? https://www.rapidonline.com/to220-heat-s...ting-29625 - these are the sorts of things that I salvage from old gear, as they're always useful. Alternatively, you could DIY something from a scrap of aluminium or copper sheet? Doesn't have to be pretty - only needs to have surface area. Might be wise to insulate the IC from it if there's a chance of something coming into contact with it...

That distortion meter might be better than you think - you quoted 0.1%, but that's the full-scale reading, meaning it ought to measure down to 0.01% with ease. It does rely on the source sine wave being clean, but that's easy to check. Frankly, if it's below 0.1%, it's likely well below audibility for most programme material.

I couldn't get a good freeze frame on the model number, but from a Google image search, I reckon it's the TF2331? Not the A version, which has a red knob at the top-left?

Looking at the manual, it looks like a nice bit of gear. The notch filter claims 80dB depth, which is 0.01%. The rectifier is mean rather than "true" RMS, but that's fine for most practical purposes.

Using distortion meters can be confusing. Setting the 100% scale reading before switching in the notch filter is critical, and how you do that varies from instrument to instrument. And picking the test level is fun. If it's too quiet, the noise of the amplifier can dominate the reading. Too high, and clipping might take place. Some trial and error might be needed. It helps to listen to the output - post notch-filter. On your instrument, there's a set of 4mm binding posts (bottom right) for that. Feed those to a small powered speaker (old PC speakers are ideal for this) - it can't drive a speaker directly. With practice, you can identify the type of distortion from what you hear - with 1kHz as the test tone, you might hear mostly 2kHz from a valve amp, or 3kHz from a solid-state amplifier - just depends. You can also judge the ratio of noise to harmonics - if there's too much noise, then increasing the signal level leaving the amp (and re-calibrating the 100% reading) might help to separate the harmonics from the noise floor. Similarly, if there's lots of 1kHz, that shows that the limiting factor is the depth of the noise floor, not the distortion of the source oscillator or amplifier. Viewing that same signal on a 'scope is also worthwhile. Mains hum jumps out at you if you lock the timebase to "line".

I use an ancient Lindos LA1 for most things. Like the Marconi, it's totally manual. It has a very similar performance (lowest range is 0.1% full-scale). I use it all the time for vintage radio work - it's easily good enough for that. The distortion circuit is fairly straightforward, so if you're thinking of building your own, you might get some ideas from the schematics: http://www.lindos.co.uk/la1-manual.pdf

Measuring distortion is something I do a lot of - both at home and at work - so could say more if it helps. We have a few old-but-nice boxes at work. Newer gear (or soundcard+free software on your PC) gives more detailed analysis. Certainly, plenty to explore

Hope this helps,

Mark

[Mark Hennessy]

Oh, just remembered the comment about the gear needing to be 10 times better than the DUT (device under test)...

That can be a good rule of thumb for some situations, but it's not the case for measuring distortion.

If the "floor" of your distortion setup is 0.01%, and you measure 0.01%, you can be confident that the DUT is better than 0.01%, but you can't be more precise than that...

If, however, you measure 0.03%, then 0.03% is going to be the answer, give or take.

Given the nature of THD+N measurements, extreme precision is pointless anyway

Cheers,

Mark

[Diabolical Artificer]

Hi Mark, my workshop is a bit small, but as you say it's light. It's nice in Summer as I can look up and watch buzzards or red kites. Bloody cold in winter though, but I just put the big Tek 585 scope on - toasty. I don't have kids upstairs, but have pidgeon's wearing lead boot's. They scare the crap out of me. I'm sure they hang round the corner and wait till i have my hands inside some lethal EHT device,  then they jump on the roof with a BANG!! giving me a near heart attack.

Having taken some measurements yesterday of the PSU, I'm getting 180v no signal, which drops to 170v at about 170w. So that should be ok. Thanks for the advice re tfmr, I wasn't looking forward to taking it apart or adding turns etc.

I've put some HS's on the 78/79 regulators, see pics. I had to use a low profile one on one PSU as a standing up type interfered with the fan.. I canabalise PCB's for everything from caps to heatsink's so have a reasonable stash and rarely have to buy them. I did buy some massive jobbies out of a welder that I was going to use for a DC load. After having looked at your dummy load I've used one for mine, see http://golbornevintageradio.co.uk/forum/...p?tid=6049.

I have the amp on test at present, only one channel though as one is overheating. I started this channel off at 180w, it's dropped to 165w ( dropped another volt RMS whilst writting - 35v) ; been testing for over an hour. 1khz @1.5v RMS in, into 8 ohm load. I briefly cranked it up to full volume and got 300w, same in/out etc. I keep getting different OP power, this may be down to some of my test gear being off calibration.

The channel that is overheating is a puzzle. I had it on test for about three hours yesterday at about 100w OP, it was only towards the end that it started overheating. Today I checked  the quiescent voltage across TP1, it was still 300mV, so changed it to 270mV, it got hotter quicker. This is the channel that blew the mains tfmr. I rewound the old tfmr and got only 50w OP, I think it's in that first video. While testing all transistors, my probe slipped and took out a pair of OP Q's. Fitted a pair of pulls, tested it and got nearer 250w, weird.

I'll do some more checking on the faulty channel, but wonder if it's the "new" Q's I've fitted having a higher hfe, than the old ones. the pair that blew were OP Q's, I put the drivers in their place and used the "new" Q's as drivers. I did check to see if there was any HF instability on the OP, but could find nowt.

I think your right about the Marconi TF2331, the manual as I read it is a bit vague and doesn't give a thd "range". I got the bit about test gear needing to be 10 times better than the DUT thd figure from an article in ETI I think. I really need to read up a lot more on the subject and have a few PDF's of "testing amps" books. I have that much to try and read up on though, it's finding the time.

Thanks again for all your advice Mark, Andy.

[Diabolical Artificer]

Did more testing yesterday but only on one channel as the other is overheating after about 20 minutes, so there's a problem there, bummer.

The good channel still kept dropping in output power, lost about 50w. The regulators after 3 hours were cold to the touch.

The right channel is a puzzler, but I have the left to compare it to, so I'll take some voltage measurements today. Somewhere a transistor is turning on too hard probably, this is where I usually start getting stuck after eliminating obvious culprits and I'll be honest my enthusiasm starts waning, I've been at this amp for weeks. Still, onwards and upwards.

Andy.

[Diabolical Artificer]

Took voltage readings of both channels and found a disparity on Q3 (BD235). On the good channel we get E 1.7v C 0v B 1.1v so all looks kushti there however I would have expected some voltage on the collector. I'll take the reading again today to make that's right. Any the faulty channel reads E 1.7v C 1.7v B 1.1v. The faulty channel also has slightly lower DC EG good channel = Q1 (MJE340) E 85.4v C 1.7v B84.1v compared to faulty Q1 etc E81v C1.7v B 80.4v which I think is down to the faulty channel pulling more current.

Anyway, I could be chasing a red herring but after checking all components around Q1A - Q5 the only thing I found was that the resistance on the good channel from R25's non wiper "terminal" to Q4 and 5's connected emmiter is 66k ohms but on the dud it is a few meg. this doesn't make sense, I would have thought a lower resistance would have caused more current to flow. The situation isn't helped by the schematic, I can't find any R24, it isn't listed in the parts list, it's not clear whether R24 ( which doesn't exist) is connected to C22. There is no direct track on the PCB from R25 to the emitters of Q4/5. I need to investigate more.

Andy.

[Mark Hennessy]

For the time being, let's stop doing the power tests. They're not needed, and don't do anything other than waste electricity. We know that the original designer has got the thermal design right - so once the actual problem has been fixed, the behaviour at high power will take care of itself.

It seems like we have a problem with quiescent stability. One channel is taking too much power even when the amp is silent. If you leave the amp powered but doing nothing, you should find that one channel runs much warmer than the other. Is that right?

So, Q3 is the Vbe Multiplier than sets the quiescent current. It has to maintain a constant voltage between its collector and emitter. Actually, that's a slight lie - the voltage actually depends on the temperature of the heat sink (quite deliberately). All of that you probably know already

What voltage should be between the collector and emitter? We have drivers and pre-drivers to consider, so that's 6 BE junctions in total, so about 3.6V between C and E of Q3. In practice, it might be slightly less.

However, your readings don't make sense. It is impossible for B to be less than E. We need to double-check...



R25 appears to be providing an additional feedback path between the output and the first of the transistors. I'm not sure it'll affect the quiescent current, but either way, it sounds like it has failed high.

Trouble is, the schematic has notes that refer to VR1, 2 and 3, but the presets aren't labelled as such. So my best guesses:

a: Adjust VR3 for symmetrical output clipping

I reckon this one is R12A

b: Adjust VR1 for 0.56V across R35 (TP1)

That's clearly R22, and as we established earlier, that should be something closer to 0.3V. The 900B schematic says 0.37V.

c: Adjust RV2 and check that pin 6 of IC2 is clipping at the same time as the output at 8 ohms

This must be R25.

Looking again, the transistorised part of the amplifier has a voltage gain that is determined by the ratio of (R24+R25)/R11. The gain of the whole amplifier is fixed by the global NFB (R9 and R6), so the preset shouldn't alter the overall gain, but it decides the gain structure of the amplifier. With that preset failing high, the output from the op-amp will be lower in level than intended, and that could lead to all sorts of interesting problems. So, definitely replace those presets (sounds like both might be wrong, but check the markings on the part itself) and re-align them according to the instructions.

I see what you mean about R24 being missing. It's possible it was added to later models to prevent a "twiddler" winding it down to minimum.