Help ned with repair of Armstrong 621 Amplifier

[Sassenach]

I am repairing an Armstrong 612 and am struggling to understand a fault on the output stage on one of the channels. Here is the circuit diagram:


When I power up the channel, it draws excessive current (ie. the light bulb on my bulb test/current limit box comes on), and the upper driver transistor ref 623 (2N5320) fails short. The PCB in my amplifier differs slightly from the circuit diagram in that there is a 22R resistor in the collector of the 2N5320, which starts smoking.

I have replaced all the semis on the board, as well as the output devices, and I have tested every device and satisfied myself that they are all okay. I've not tested the resistors, other than to measure them in circuit to check that I don't get any suspicious readings. I have replaced all of the electrolytics.

I am completely stumped! Has anyone got any ideas?

Because of device obsolescence, I have replaced the BC297 with a BC327, the BC267 with a BC337, and the ITT44s with 1N4148s. I have also replaced the BF258 with a KSC3503 and the output devices with 2N3773. The other channel is working fine with these substitutions.

This is driving me nuts...!

 

[JonSnell Electronic]

If you link the base of 623 to the base of 624, does it still fail? If not, take a close look at the biasing circuit, check the preset.

 

[Ian Finch]

There is also the possibility of a short to the heatsink so check and compare the DC conditions around the output transistors, such as 40V at the midpoint of the 0.47R resistors.

 

[Sassenach]

Thanks for your suggestions and apologies for the delay in replying. I have checked the output devices for shorts to the heatsink and all appears to be well. I have replaced the mica insulators with new sil pads.

The preset appears to be fine and gives sensible readings if I measure its resistance whilst adjusting it. I've given it a good going over with contact cleaner just to be on the safe side. If I link the bases of 623 and 624, the circuit still draws excessive current. The DC level on the output does not rise when I wind up the variac carefully, and it looks as though the lower output device turned on and holding it low. I tested this theory by shorting the base of 624 to its emitter and checking that the output voltage rose when I increased the supply voltage. I've again used the diode checking setting of my DVM to confirm that all the transistors are still good (which they indeed appear to be).

I'm now checking the board for any visible PCB damage, as well as visually inspecting all the passives.

Any other theories?

 

[Old'n'Cranky]

Be super wary of the bias/dc trimpots.
Any slight wear/cracks in the track will potentially blow the outputs and drivers.
I speak from experience here.

I replaced the pots with 10 turn bournes.

How are the two 4,000uf caps ?
Unit I worked on needed those replacing too.
With the age of the unit, I'd do a full recap for future reliability

 

[Sassenach]

I've already replaced all the electrolytics, including the 4000uF output caps. I agree that it would be a good idea to replace the trim pots just to be safe, although it doesn't explain why the circuit still draws high current when the bases of the driver transistors are shorted together. My present line of enquiry is to understand why the upper output Darlington pair isn't behaving like a Darlington, ie. Why the emitter of the output device is not following the base of the driver. I've been looking for a PCB trace fracture or dry solder joint but everything looks okay.

And I thought this would be a quick repair Job...!

 

[indianajo]

Other quirks of these old designs:
Be sure that you noticed any TO220 semi's are BCE and not EBC, and twisted the leads around.
Some To92 products are the revised lead order, too, compared to TO5.
Modern output transistors do not have the base resistance of an RCA homotaxial part, even though the part # is similar. You need to install 10 ohm >1/2 watt resistors between the drivers and the bases of the output transistors.
I find drivers hold up better with a heat sink.
I've had the upper driver transistor get very hot on the similar Apex AX6 board and found the O.T. base driver solder joint was open, making the upper driver work very hard trying to pull the output capacitor center line up. This has the light bulb limiting.
Also without a base resistor new output transistors with high Ft (>200khz) can oscillate ultrasonically. Check for AC voltage with an analog VOM with a 390 or 220 pf capacitor in series with the ground lead. Music won't go through a capacitor that small, ultrasound will. Or use a scope and look for fuzzy waveform that won't hold still.
If oscillating, dynaco in their copy ST120 circuit (I don't know who came first) added 50 pf disk capacitors between base and collector of the two driver transistors. See the "TIP mod" on greg dunn's dynaco website. they also put a 1k series .047 uf snubber between output and ground to keep RF from sneaking in that way. There is a lot more RF floating around the house these days than in 1966, ie cell phones. (The dynaco ST120 bias spreader circuit was stupider than armstrong, I hope armstrong was later design)
What I like about these capacitor output amps, no matter what happens they won't blow up your speaker.
If your boards are getting torn up ( my ST120 has been repaired too many times) see the Apex AX6 thread for modern boards that fit modern transistors that will do the job. Nobody is selling the boards, you have to make them. I've built one on perf board, I'm trying today to see if it oscillates or not.
The AX6 also has a 27 ohm resistor in the lower driver emitter line, perhaps to reduce thermal runaway. You see armstrong found a 22 necessary in 623 the upper driver. And the AX6 has the .51 resistor in the emitter lead of the lower output transistor instead of the collector, probably to also limit thermal runaway. Newer output transistors aren't as slow and don't have the base resistance of the RCA ones.
I clamped my bias trim pot with a parallel diode, to keep idle current from running away if the wiper fails. If the voltage across it is low enough after prototyping I may clamp with a schottky diode instead of a 1n4148.

 

[Sassenach]

Thanks for your detailed comments!

The only replacement device with a different pin-out is the KSC3502. I needed to joggle their legs to get them in the board but they've gone in fine. The output drivers (2N5320 and 2N5322) have heatsinks. My frustration is that the left channel works absolutely fine with all the same device substitutions as the dead right channel. I've checked it thoroughly for instabilities (using the scope and a spectrum analyser) and it seems to be absolutely fine. You make a good point about adding base resistors to the output drivers - presumably these should be fitted as close to the output device as possible (the output transistors are mounted off the PCB with wired connections). I've re-made all of the wired connections because the original was solid cored and prone to fracturing. And after 40 years has become very tarnished and difficult to solder to.

I've just worked out that the lower driver (ie the PNP in the CFP) has a 22R resistor in its collector, presumable to reduce the risk of thermal runaway. A bit of adhesive aluminium tape was loose in the chassis when I first opened it, and traces of dried adhesive on the board suggests that it once provided a thermal connection between the upper driver heatsink and the two diodes in the bias chain. I've looked at a few pictures of Armstrong 621s on line, some of which show this piece of tape and others which don't. I suspect is was a manufacturer's bodge to keep the bias current more constant with temperature, but I haven't decided whether to replace it (I'm worried about relying on the tape adhesive to provide electrical insulation).

So I'm still not much further forward - these sort of jobs should be easy!

 

[indianajo]

I've re-made all of the wired connections because the original was solid cored and prone to fracturing. And after 40 years has become very tarnished and difficult to solder to.

A bit of adhesive aluminium tape was loose in the chassis when I first opened it, and traces of dried adhesive on the board suggests that it once provided a thermal connection between the upper driver heatsink and the two diodes in the bias chain. I've looked at a few pictures of Armstrong 621s on line, some of which show this piece of tape and others which don't. I suspect is was a manufacturer's bodge to keep the bias current more constant with temperature, but I haven't decided whether to replace it (I'm worried about relying on the tape adhesive to provide electrical insulation).

I mounted the bias diodes and the pot over on a cinch solder terminal strip over on the output transistor heat sink, trying to make the bias more stable. don't know if those 4" long wires are an oscillation problem or not yet. I did put an 8 turn inductor in series with the drive end, salvage from a dead PCAT power supply. Once the amp works, I'll put a blob of heat sink grease over there connecting the diodes thermally to the aluminum flange.
The output transistor bias resistors, I replaced the flying wires over to the heat sink with actual long lead 10 ohm 1 W resistors for the base connection. I did this in 1985, just trying to provide something to blow up other than $10 each (then ) output transistors. Turns out it was the right thing to do! For a different reason. My output transistors were NTE60, probably white box MJ15003 clones.
If this is a customer job, you may want to put in an LM3886 instead of the armstrong board, using the single supply option in the datasheet. Or djoffe/akita sells a board for the LM3886. I don't want to convert because the NTE60 can pump out 15 to 20 amp peaks during drum hits in classical music (70 db s/n disks), whereas the LM3886 is limited to about 7 amps. RMS wattage LM3886 is about the same due to heat sink limitations on these old amps.

 

[Sassenach]

I'm fixing the Armstrong for a friend ("It'll be a simple job", I foolishly said). It sounds really lovely when it's working - almost valve-like with predominantly even-harmonic distortion - so it would be a shame to replace the whole board. Getting it working is the part of the "fun"...

 

[indianajo]

Yeah, I like the sound of the copy ST120 design, after the djoffe bias mod to get rid of the ST120 cold crossover distortion problem. With those high gain MJ15003 clone output transitors, and an 80 v power supply, it never clips. Sounds very similar to the $1000 Peavey CS800s I bought for ~$80, only it won't put out the high wattage all day. I listened to the ST120 for 18 hours a day (or left it on that long) until I kicked the board putting away a record and knocked a driver loose. Then the circuit board traces started lifting, and I decided to try AX6 which is about the same but probably more thermally stable. I had fans on the St120 for the couple of years with the djoffe bias mod, hiding it under the couch to kill the fan noise. Those ST120 ehat sinks are NOT suitable for PA use, I discovered in a Christmas cantata, where the solder melted and some resistors went up in flames. I wonder how good the Armstrong 621 output transistor heat sinks are- the ST120 ones are only a 1/8" 3 1/2" x 5" aluminum sheet. Good for about 2 W average 18 hours a day.
Trace your output and bias wiring with an ohmmeter if you've already checked for oscillation. Could be you have a cold solder joint or cracked board trace on the board with the overheating 623. check on the component side to the leads. As I say, my overheating upper driver was a cold joint on the wire out to the output transistor.

 

[Sassenach]

The heatsink is actually pretty adequate. Even with the volume set at a high level, it only gets to 55degC or so. Here's a picture the output board and a bit of the heatsink prior to me starting work on it.

 

[indianajo]

Thanks for the picture. That's a nice pair of heatsinks.
I couldn't get the cover of my ST120 on with auxillary TO220 heat sinks atta ched to the top of the transistors, that interfered with the cover. Since the new output caps are so much smaller than the dynaco's (1.2" dia instead of 2.5) I'm thinking of buying a 2.5 " x 7" heat sink with horizontal fins from heatsinkusa down in vendor forums. I can't find anything suitable at newark or digikey . Then I could put all 4 output transistors across the amp instead of on the outside edges, and blow one PCAT fan on it instead of two from the outside like mickey mouse ears.
The AX6 thread is this one:http://www.diyaudio.com/forums/soli...amp-50w-single-supply.html?highlight=apex+ax6
He has board layout drawings in there, several revisions, the last with a bias trim pot.
I find twisting the leads around on new transistors and having them support heat sinks, the leads tend to break. Besides with those 50 pf caps on the bottom and the djoffe bias daughter board correction, I had patches all over the board.

 

[Ian Finch]

You have one good channel, right? There must be DC differences between the two to indicate that not everything is correct, in addition to the high idling current. Perhaps marking up the schematic posted in your OP with any voltage discrepancies would be useful to all of us.

As you are telling us that everything you looked at checks out correct, it is difficult to get a handle on your problem, other than as some have already suggested, that the bias control circuit must simply be faulty. So let's look instead for the differences between the channels with say the bulb limiter or better, a fixed ~ 1k limiting resistor in series with the supply to each channel - and there will be some differences even if you have to compare every circuit node in each channel to find them.

PS: I'd simply use a pair of current limiting bench supplies for this but not everyone has a pair of supplies that are good for 80V or so. Most dual supplies seem to only go up to 50V but that should still be enough for the purpose.

 

[Sassenach]

I finally found the fault - the 2N5322 had a failed even though its diode drops measured okay on the DVM. Its junction had gone leaky, which I spotted when I measured the resistance across the output of the channel and found it didn't match the good channel. I traced it back through the circuit until I got to the 2N5322. I've now put a new one in and it's working fine!

I'm going to replace the trim pots, even though all is now well. These old carbon types worry me - I've known the tracks to crack and the wiper contact to become corroded.

Thanks to everyone for your help - you've kept me motivated to find the problem!

 

[indianajo]

Congratulations. Many thousands of hours of pleasant listening ahead for your friend, I hope.
I found the upper driver transistor shorted across today, after it was cooked by the bad solder joint on the wire to the output driver yesterday. I spent all evening fooling around with the light bulb box which wuuldn't let the VCC-gnd be bigger than 4.5 v. You know you can't buy 100 W incandescent bulbs anymore? The halogen 1200 lumen "equivalents" now at family dollar have a cold resistance at 2 v of infinity.
I've bought 500 or so LP's for a quarter apiece since I kicked this circuit board 2 years ago and broke it. I'm Looking forward to the amp operating again and listening to them.

 

[Old'n'Cranky]

Good news.

I love seeing pretty units like this live on.
Way more style than a lot of the ugly plastic monsters of today.

 

[AndrewT]

The 2n5322 driver may have failed because they have selected a 75Vceo device to run on supply rails of 82Vdc +-mains voltage variations.

You really need a higher SOA device with Vceo >= 100V
The 2A rating is OK. But gain can be a bit low (30 minimum) unless they have selected higher gain versions for the build.

 

[Sassenach]

Yes, I had wondered about the choice of driver devices. I suppose I could have upgraded them. Having said that, it worked for over 30 years with the existing device types. Armstrong did change these parts in later version of the product, though.

Any suggestions for good alternatives?

Re availability of 100W incandescent bulbs, this is a pet gripe of mine. The inefficiency of filament bulbs is in the form of heat, which adds to the warmth in the room and offsets the energy required to heat the house. So in that respect, they are 100% efficient. Compact fluorescents contain mercury and take ages to get to full brightness, and LED lamps have significant embedded energy in their manufacture. Fortunately I was able to stock up on incandescent bulbs before the ban, so I'm good for a few more years. My local hardware store still has them for sale "under the counter" for "special customers"...

 

[AndrewT]

Lot's of To220 devices and a few To126 that make very good drivers.
But since their specs will be different, they may require changed compensation.
That can be difficult to get right. It's one area where I am not competent.