I went to check for crossover distortion and I don’t have any at 15 mV (22.4 mA). I do however have some odd oscillation on the bottom side of the waveform.
A real bummer. Why only on the bottom side? Would this have anything to do with my trimmers being on long leads?
Dan
A real bummer. Why only on the bottom side? Would this have anything to do with my trimmers being on long leads?
Dan
If your pot is measuring 940 ohms, then likely the cold voltage is .55 volts or higher. Perhaps even .65 volts. In that case, clamping the pot with a 0.3 v schottky or germanium, will cut idle bias current down to zero. Ie maybe a bad idea.
You could make a 940 fixed resistor by putting a 910 in series with a 33 ohm resistor. Or some parallel combination. There are calculators for the parallel combinations.
You can remove the long wires to the pots and see if the oscillation disappears. If it is still there, you will have to probe forwards in the circuit to find where the oscillation originates.
My best friend in 1976 had a Heathkit AR-15 receiver, Bose 901 speakers, a Gerrard turntable. I was not impressed. Perhaps the problem was the cartridge, perhaps the speakers, perhaps the problem was the 5 sided bedroom he lived in. Perhaps I could not tell anything definitive listening to Genesis, Black Sabbath, etc. My dynaco ST70 tube amp, LWEIII speakers, Grado FTE cartridge were more pleasing to me.
You could make a 940 fixed resistor by putting a 910 in series with a 33 ohm resistor. Or some parallel combination. There are calculators for the parallel combinations.
You can remove the long wires to the pots and see if the oscillation disappears. If it is still there, you will have to probe forwards in the circuit to find where the oscillation originates.
My best friend in 1976 had a Heathkit AR-15 receiver, Bose 901 speakers, a Gerrard turntable. I was not impressed. Perhaps the problem was the cartridge, perhaps the speakers, perhaps the problem was the 5 sided bedroom he lived in. Perhaps I could not tell anything definitive listening to Genesis, Black Sabbath, etc. My dynaco ST70 tube amp, LWEIII speakers, Grado FTE cartridge were more pleasing to me.
Last edited:
You might just need a small cap collector to base on the PNP driver. It’s a local oscillation and usually won’t respond to treatment in the global loop.
BTW, you cannot properly see crossover distortion directly on a scope. If you don’t have an audio analyzer or sound card you can capture FFT with (same thing really) then you need to look at the RESIDUAL on the scope. Subtract out the fundamental on a low distortion differential amp, and you minimize the result. Do it at a mid band frequency where there is little phase shift, or phase shift the reference signal to match the amp at the test frequency (variable 1st order LPF). It can take some tweaking of levels/phase to find the minimum.
BTW, you cannot properly see crossover distortion directly on a scope. If you don’t have an audio analyzer or sound card you can capture FFT with (same thing really) then you need to look at the RESIDUAL on the scope. Subtract out the fundamental on a low distortion differential amp, and you minimize the result. Do it at a mid band frequency where there is little phase shift, or phase shift the reference signal to match the amp at the test frequency (variable 1st order LPF). It can take some tweaking of levels/phase to find the minimum.
Very good thinking, looks like I’ll get an exact measurement on the resistance and make up what I can with multiple resistors if the diode will not work. Lol, I would bet the ST70 would sound better, did you build it, a Dynakit? My father’s best friend’s older brother had just got his Heathkit in 1970 and had built it. From what I understand he was just 13 at the time, the older brother that had built it. My father was 11. He said he was in awe at the time, but I guess for an 11 year old that probably wasn’t hard to do. So he’s always wanted to live that again so he bought a couple cheap units obviously broken in hopes that I would fix them. We’ll see, I’ll let you know what he thinks once they’re done. Considering he has a huge collection on Kenwood, L-07M, L-07C, L-07C II, L-09M, Supreme 600, 700C, 700M, Nakamichi Stasis, etc, etc, etc, all fully restored by me and exceeding their distortion spec I’m guessing this will be “meh”.
Excellent, I will try that, do you have a recommendation as to value? I can play around, maybe start at 10 pF and then go up in increments until the oscillation is gone.
I had no idea, I had always seen crossover distortion seen on a scope, but clearly they weren’t just using the scope. Most of what you said went over my head unfortunately. The only audio analyzer I have is a Quantasylum Q403, it has differential inputs, will that work? I also habe differential probes for my scope, not sure if those would be handy in this situation.
Thank you for the assistance all!
Dan
The QA is what you need. If you have one, that’s great. Much better hardware than a sound card. You can fake it with one (I have), but I’ve managed to ruin every one that I’ve ever used for that except an old AWE64 that I no longer have an ISA slot machine to use it in. I’ve been considering one since my E-mu finally quit. Probably be another release or two of the QA by the time the lab is operational again anyway.
With a QA you can just look at the spectrum and beat it down. You can continue to beat it down long after you stop seeing distortion on the scope. Looking at the residual on a scope was/is just a practical way of increasing the dynamic range of the distortion you can “see” on a scope - when other analysis methods are unavailable. When the large fundamental is nulled out, you see only the distortion products. In the 1970’s and 80’s that’s all there was, unless you had 30 grand to shell out. Modern PCs changed all that.
With a QA you can just look at the spectrum and beat it down. You can continue to beat it down long after you stop seeing distortion on the scope. Looking at the residual on a scope was/is just a practical way of increasing the dynamic range of the distortion you can “see” on a scope - when other analysis methods are unavailable. When the large fundamental is nulled out, you see only the distortion products. In the 1970’s and 80’s that’s all there was, unless you had 30 grand to shell out. Modern PCs changed all that.
I had been waiting well over two years for the QA404 to be released and finally gave up and just grabbed the QA403. I’ve owned the QA403, I don’t know, seems like at least two years and the 404 is still not out. It’s very likely that by the time your lab is ready the Q403 will still be the one to get lol. Appreciate it!
Dan
I would start with 47 or 51 pf c to b on Q205. Ceramic disc or mica. That is what dynaco recommended in the "TIP mod" on the ST120 in that location. With scope, you could see what it is doing. I just did it because the factory said so.
The dynakit ST70 was built in 1961 when I was mowing lawns for $5 a day. Way out of my budget then. I was still testing TV tubes at the corner store. No soldering iron. I had no stereo until 1965 when Mother got a RCA record ripper from Top Value stamps. I could hear the highs gone on the second play of an LP. 2" speakers, a 12AX7 powered each side. New stylus did not help. My 3rd year in college I had enough spare cash to buy the ST70, PAS2, and AR turntable from a minister going to Africa to be a missionary. I was Inspired by the dynaco ST35s on headphones in the college library. MUCH better. However, the McIntosh clinic salesmen told me the ST70 was producing 1.5% HD at 7 watts per channel. Took new filter caps, cathode bias cap, rectifier tube and output tubes to get it to 35 w/ch. Still 1% HD, which was good for 1971 but not these days. Any repaired Peavey PA amp sounds better.
Last edited:
Wow, great story and that was quite a nice pick up from the missionary. I just picked up a ST-70 from a local electronics flea market/ham fair. It was 300, in fantastic shape and I tested all of the tubes yesterday and all extremely strong on my Hickok 650. The outputs are the original Dynaco. Tube is just going to have more distortion, but the ST-70 still has quite the following, many out there offering upgrades and such. I had one I rebuilt about 5-6 years ago and thought it sounded pretty decent. Was my first tube amp. Once I rebuild this one I’ll measure the distortion on it.
So here are the waveforms with no alterations. I do have the the bias at 15 mV with resistors that have been installed.
I wanted to see what differences would be made so I left one channel as is and added the 47 pF cap to one channel, to B-C of Q205. Just soldered it legs for testing purposes.
I powered it up on the dim bulb tester to start off with and it looked to have slowed the oscillation a lot, but oddly it was affecting both channels.
Once I put it to full power the oscillation went to something a bit more familiar.
The oscillation definitely looks like it’s sped up with the cap in place, but why is it changing both channels? I pulled the cap off and it slowed back down.
Dan
Because it is actually a supply decoupling/ grounding issue. Causing a local oscillation in the negative follower. If the amplifier was following all of Self’s rules, to the letter, without changing the schematic or the BOM, it wouldn’t be doing it. There is an issue with the way it’s wired, that doesn’t show up with the original parts.
When I make the assertion that “you can change output and driver transistors out with faster ones, and all it does is make it more stable” it is with the assumption that everything was done correctly in the first place. If there were design layout issues, they may have to be corrected for “better” parts to work at all. If such issues exist - they are worth correcting because all it will do is make the amp perform better.
When I make the assertion that “you can change output and driver transistors out with faster ones, and all it does is make it more stable” it is with the assumption that everything was done correctly in the first place. If there were design layout issues, they may have to be corrected for “better” parts to work at all. If such issues exist - they are worth correcting because all it will do is make the amp perform better.
Okay, try 10 ohm 1 watt resistor instead of wires, driver collectors to output transistor bases. All 4 places. Mooly said this helped his problems with fast transistors installed on a Masic? Mavic? Maplin? Brit amp with 40411's. I did it on my ST120 before I had internet access, because I wanted some part to blow out when the output transistors shorted from overheating, instead of the driver transistors & zener diodes. Since the output transistors are on an external heatsink, it is a natch to change the wires to long leg resistors. I had carbon comp, but metal film or carbon film resistors should be fine. A little inductance (they have a spiral track) certainly can't hurt ultrasonic oscillation. Add 10 ohms emitter of Q204 to base Q206, also collector of Q205 to base of Q207. Etc.
Make sure the wires to the external transistors are not close, and non-parallel if you can arrange them that way. Close parallel runs are capacitively coupled.
If that does not help, Surplussales of Nebraska had homotaxial 2n3055's last time I ordered from them. Allegedly. Definitely East European.
Make sure the wires to the external transistors are not close, and non-parallel if you can arrange them that way. Close parallel runs are capacitively coupled.
If that does not help, Surplussales of Nebraska had homotaxial 2n3055's last time I ordered from them. Allegedly. Definitely East European.
It should be possible to track down the parasitics that are causing this. There is no universal fix, other than making a new single board solution that follows the grounding and decoupling rules to the letter. Collector-base “slowdown” caps, base stoppers, rerouting the grounding - whatever it takes.
Hometaxial 3055’s won’t help unless they are the voltage-graded versions. Electronicsurplus.com had RCA1B01’s last time I was browsing. 95 volt. Hometaxial “3055’s” go in and out of stock all the time.
Hometaxial 3055’s won’t help unless they are the voltage-graded versions. Electronicsurplus.com had RCA1B01’s last time I was browsing. 95 volt. Hometaxial “3055’s” go in and out of stock all the time.
So are you implying that it was my choices of replacement drivers? I tried to go as close as I possibly could as far as Ft and such. Unfortunately, there isn’t a whole ton of info on the original devices that were in this thing.
I looked through my stash of 10 ohm resistors, and I’m not sure what the best option would be. I’m not sure how hot these resistors are going to be getting. I have these Vishay one watt resistors, they are rated one watch, but they are pretty small. Here they are next to a standard one watt resistor.
My other option are these 3 W rated resistors. Better to go bigger?
The the tiny ones are metal film, and the bigger ones are oxide and the other wire wound.
Obviously, the wire is going from the driver collectors to the outputs are fairly long, so even if I put the resistors in they’re still gonna be a good length of wire.
You say try not to run the wires in parallel, that’s exactly how they have them. All wire wrapped.
A little confusion, apologies. So you say to connect a 10 ohm resistor to the emitter of Q204 and base of Q206 as well as the collector of Q205 and base of Q207. So that would be four resistors in total. Two resistors in the series between the driver and the output, equaling a total of 20 ohms resistance between the emitter of Q204 and base of Q206, is that correct? So did you both channels I would need eight resistors in total? Or is there supposed to be only one 10 ohm resistor in series between each driver and its output?
Would you recommend that I eliminate one channel altogether? Remove its input, outputs and power? I would absolutely love to pinpoint this parasitic oscillation, I just don’t know where to start.
Dan
Disconnecting one channel may be one necessary step in tracking down the problem. You can even have push-pull oscillations involving both channels beating against one another. Not saying it IS that here, but I’ve seen it.
Of course choices of output and driver affect things. What happens is that with crummy ones the original designer can sweep problems under the rug. CFPs (the lower half of a quasi-comp) are sensitive to implementation. Better faster outputs will generally improve overall global NFB stability because the poles they create get shoved even further out of your way. But can reveal problems that won’t show up in a simulation unless layout parasitics, magnetic coupling, and realistic power supply impedances are included in the model. Fix any problems that they’ve hidden with slowpoke transistors, and the resulting performance will always be better than it was. Even if you go back to original types.
Of course choices of output and driver affect things. What happens is that with crummy ones the original designer can sweep problems under the rug. CFPs (the lower half of a quasi-comp) are sensitive to implementation. Better faster outputs will generally improve overall global NFB stability because the poles they create get shoved even further out of your way. But can reveal problems that won’t show up in a simulation unless layout parasitics, magnetic coupling, and realistic power supply impedances are included in the model. Fix any problems that they’ve hidden with slowpoke transistors, and the resulting performance will always be better than it was. Even if you go back to original types.
Just 10 ohms between driver and base of output transistor. Some people use 4.7 ohm base stoppers. I wanted the resistor to burn out if the output transistor shorted, so I went with 1 watt. At gain 5 and output transistor putting out 4 a, current should be 0.8 amp or 6.4 watts. So maybe 3 watt resistor would be better. My 1 watt base stoppers only burned when the output transistor shorted after a 3.5 hour choir rehearsal.
Nobody is making 400 khz Ft transistors anymore. The communist block factories held on for a while, but the RCA factory that made homotaxials is now a county park in New Jersey. These products were designed for those transistors, and failure to perform properly with newer parts due to layout is not their fault. If the wires to the temp sense diodes are run parallel to the base wires to the output transistors, definitely unbundle them and run them some other way.
Nobody is making 400 khz Ft transistors anymore. The communist block factories held on for a while, but the RCA factory that made homotaxials is now a county park in New Jersey. These products were designed for those transistors, and failure to perform properly with newer parts due to layout is not their fault. If the wires to the temp sense diodes are run parallel to the base wires to the output transistors, definitely unbundle them and run them some other way.
Any hometaxials anywhere, even from the communist bloc, would be old stock. There still are RCAs floating around, too. Few and far between, and generally nobody wants them.
Unlike 2SD424’s, where every single one that was ever made has already been bought up or used up, and there is enough demand that fakers are everywhere.
I haven’t seen hometaxials anywhere that look fresh off the production line. You wouldn’t want that anyway - you would want them to LOOK like they’ve been sitting in a warehouse for 40 or 50 years.
Unlike 2SD424’s, where every single one that was ever made has already been bought up or used up, and there is enough demand that fakers are everywhere.
I haven’t seen hometaxials anywhere that look fresh off the production line. You wouldn’t want that anyway - you would want them to LOOK like they’ve been sitting in a warehouse for 40 or 50 years.
The 2n3055h I bought from surplussales.com were dull, magnetic, RCA logo, date code of 8528. #6 holes instead of 3 mm. They have a little stamping on the top of the long end, "+H" Probably in some USAF depot for 40 years. That is where he gets his stuff, IMHO.
Don't know if 2n3055H would be good enough for +-46 v rails.
Last edited:
If tested to 100V. It’s not guaranteed. SOA will be good enough if it doesn’t just zener on you or have too much ICBo.
But far more straightforward to just use MJ15003’s. Oscillation problems should just be fixed.
But far more straightforward to just use MJ15003’s. Oscillation problems should just be fixed.
Last edited:
Well I had to step away from this for a few days and then my daughter brought another “something” home from preschool so we all got sick for several days. Now that I’m not feeling horrible I jumped back onto it and also picked up a second AA-15. Between the two I was able to get enough working original transistors to complete a channel.
I started by replacing the drivers and the VAS transistor with the originals. Absolutely zero change, exact same oscillation on the output.
I wanted to see if I saw any oscillating at the input of the amplifier board and I did, but on the top side of the waveform.
So then I got the bright idea to remove the input wire from the amplifier board and measure at that wire to see if it was being introduced in the preamp. With the input disconnected from the amplifier board the signal was as clean as could be. This told me the issue was associated with the amplifier. The only transistors that weren’t original were the outputs being MJ15003 and then the input pair Q201 and Q202. These were originally 2N3393 and I had swapped in a pair of 2N5551. I put the 2N3393 back in and wouldn’t you know it all of the oscillating was completely gone on the output.
I’m guessing that maybe the 300 MHz Ft of the 2N5551 is a little too high? I went with them since the 2N3393 had a higher than average current rating. I’m going to try swapping in some other options, any recommendations?
Would a KSC1845 survive here? The gain is quite a bit higher, the originals having a gain of around 100-120. Also thought about the KSC1815, lower gain and higher current rating. These both have much lower Ft (50 MHz and 80 MHz). The 1815 will be closer to the originals, but the 1845 lower noise. Didn’t think it would have been these that caused the issue.
Dan
I started by replacing the drivers and the VAS transistor with the originals. Absolutely zero change, exact same oscillation on the output.
I wanted to see if I saw any oscillating at the input of the amplifier board and I did, but on the top side of the waveform.
So then I got the bright idea to remove the input wire from the amplifier board and measure at that wire to see if it was being introduced in the preamp. With the input disconnected from the amplifier board the signal was as clean as could be. This told me the issue was associated with the amplifier. The only transistors that weren’t original were the outputs being MJ15003 and then the input pair Q201 and Q202. These were originally 2N3393 and I had swapped in a pair of 2N5551. I put the 2N3393 back in and wouldn’t you know it all of the oscillating was completely gone on the output.
I’m guessing that maybe the 300 MHz Ft of the 2N5551 is a little too high? I went with them since the 2N3393 had a higher than average current rating. I’m going to try swapping in some other options, any recommendations?
Would a KSC1845 survive here? The gain is quite a bit higher, the originals having a gain of around 100-120. Also thought about the KSC1815, lower gain and higher current rating. These both have much lower Ft (50 MHz and 80 MHz). The 1815 will be closer to the originals, but the 1845 lower noise. Didn’t think it would have been these that caused the issue.
Dan
fT may not matter with that 220 pf hanging off the collector. Changing transistor types may need adjustment of C210, or at worst case, C204. It does appear to be a global loop issue, which explains why it doesn’t blow up when oscillation starts. More modern topologies tend to be more predictable. This has multiple nested feedback loops….R228 and 222, within whatever that 330k is (can’t read the designator).