Hello guys. I am not versed in diagnosing to component level with amps, but I hooked up the tranny wrong, and cooked a diode or two until the fuse blew. After rebuilding the PS, there is a small but noticeable hum that is steady and independent of volume level.
This is a scene on the scope at the speaker terminals with no input:
I am wondering if anyone could point to the possible culprits. During diagnostics, I had unmounted the amp from the sink, and powered up with quiescent current only. There are 4 power transistors, which became slightly warm after a minute. After 2 minutes or so, only one of the devices became very hot--close to the danger point, so I shut it down.
Any pointers are welcomed. I could simply rebuild the amp with all new components, but that would be undesireable.
Thanks.
This is a scene on the scope at the speaker terminals with no input:
I am wondering if anyone could point to the possible culprits. During diagnostics, I had unmounted the amp from the sink, and powered up with quiescent current only. There are 4 power transistors, which became slightly warm after a minute. After 2 minutes or so, only one of the devices became very hot--close to the danger point, so I shut it down.
Any pointers are welcomed. I could simply rebuild the amp with all new components, but that would be undesireable.
Thanks.
Assuming this is a class AB amplifier like most discussed here, its safe and stable operation will depend on thermal coupling between the output devices and the bias controller. Thermal coupling means leaving the output transistors etc. properly mounted on the heatsink. The heatsink isn't just a means of removing waste heat - it's an integral link for the stable operation of the amp, so don't operate the amplifier without it unless you either really know what you are doing or you're prepared to replace the fried output transistors and any other damaged parts as a consequence. This isn't the way to troubleshoot the amplifier.
We need to know what type of amplifier and at least reference to the schematic you have there in order to offer specific advice. Generally, you could start with building a bulb tester to use in series with the mains supply or say, 100R 1/2 watt resistors in the power rails to preventing damaging high currents and then measure any DC at the output terminals and measure the currents through the various stages of the amp. You could also do worse than read articles like this one
We need to know what type of amplifier and at least reference to the schematic you have there in order to offer specific advice. Generally, you could start with building a bulb tester to use in series with the mains supply or say, 100R 1/2 watt resistors in the power rails to preventing damaging high currents and then measure any DC at the output terminals and measure the currents through the various stages of the amp. You could also do worse than read articles like this one
Last edited:
True, and true. The transistors were not roasted, and the exact same hum is there now that there was before. Here is the schematic:
And the power supply, except each diode is clamped with 100n, and the storage caps are bypassed with the same:
Thank you for your response, Ian.
And the power supply, except each diode is clamped with 100n, and the storage caps are bypassed with the same:
Thank you for your response, Ian.
Last edited:
It's a Quasi circit being biased by Q3, Q13 and Q14, I would check these. Yu say 1 transistor gets "HOT" on which side? This will tell you whether its neg. or pos. side
Err..no, this is Rod Elliott's P27 design and is actually a CFP design. (complementary feedback pair) Note the output devices are also complementary (TIP35/36). The bias controller is Q12 but has a slightly different quirk, since it needs to track drivers Q6 and/or Q7 temperature, not that of the output devices.
This doesn't affect the hum problem though, if you are using the original PCB and layout. As it's mains related hum, it would seem there's a problem with your power supply or grounding arrangements. What was the DC measurement at the output terminals?
Check the supply voltages, diode/bridge and possibly the decoupling caps, C6,7 if they are incorrectly returned to star ground or likewise the signal input ground. Perhaps a pic would help? Otherwise, I guess the available wiring guidelines have been followed?
It is Q9 on the negative. Thanks, that was what I was hoping--to narrow it down to a few components.
Knowing this one power transistor is running extremely hot compared to the others, may narrow it down further?
Thank you for your interest and guidance.
Q9, Q11 are in parallel so you have one answer in a very narrow area. One of them is not operating correctly. Compare the small voltage drops across all 4 emitter resistors to confirm.
I am using the official board, and yes, two transistors are physically back to back with conductive paste.
0.00 volts DC.
C6 and C7 may have seen a voltage level beyond their rating. I have experimented with the input wire. It must be grounded at both ends.
As far as grounding scheme, this is a roll-your-own hybrid tube pre, so as you can imagine, I have spent many hours hashing that out. It is certainly possible there is still a loop. But my "seat of the pants" says no.
What is bugging me is this is more like a buzz rather than mains hum. I have a 120v resistive load circuit with one wire separated that I pass around all parts of the chassis. I also have a toroid with nothing but the primary winding feed directly by a mains plug. Moving that around does nothing. The P27A picks up nothing audible to the ear, unlike V1, or now U1, an opamp which is the pre-preamp. Those on high gain will pick up an exhale if you are too close!
There is no way a picture would help in this regard. I should make a dedicating grounding schematic, or rather, a physical wiring diagram so the paths may be clearly seen, since it is quite complicated due to the hybrid design. Trial and error has largely determined what is now there.
There are two intentional loops: One in the P27A design, and another in the FX send of the preamp. At the present, everything is dead, except for the power amp. I disconnected one by one to try and find a possible source.
I suppose more experiments with the grounding scheme could be tried. Surely, I have not exhausted every possibility since there are quite a number of possible arrangements.
Incidentally, the power amp works perfectly, even with this buzz. It is the quietest noise so far, having eliminated all the unacceptable ones one by one.
Thank you sir.
Yes, there are parallel pairs on each leg. It is an intentional overdesign.
Ah, I see, R16 through R19....
Also, as far as the heatsink goes, I now see why it has to be mounted and grounded (as both you and Mr. Elliot said). While touching the back plates to monitor temps, one picked up some kind of interference..
The reason I unmounted was to check the input coax. The one there had been spliced so I replaced it. I also slid a stainless steel braid over it, and grounded that just to make sure that line was not picking up mains field. Steel is better for 50-60Hz than copper, but copper is better for RFI.
All dead even at 0.6mV. The quiescent current is also dead on.
I suppose it should be remounted to the heatsink, and the ground scheme tinkered with again?
No one commented on the the waveform. I don't know but it looks really wierd. No particular frequency has increased magnitude. I assumed 60Hz or 120Hz, or some facsimle thereof, would stand out if mains-related.
It looks like PSU capacitor charging current shifting a ground somewhere. Do you have a second scope channel? If so, overlay the transformer secondary voltage to compare.
The PSU capacitors are only charged when the transformer secondary voltage exceeds the voltage on the capacitors. This charging current will be very high and for a short time, and if you have any stray ground impedance, a voltage will appear on the ground somewhere often appearing at the input terminals of the amplifier.
If that's the case, good star grounding would be the best bet. Can you post a photo of your grounding layout?
HTH,
Brian
A second scope channel would require an expensive sound card, or the Daqarta USB sound card, which would also permit DC readings. I cannot do that until next month. I have a decent DVMM which I would place at the ends of each wire or wherever this drop may occur (everywhere possible in fact) scanning several fixed ranges to inusure whatever is there may be picked up, since "Auto" would probably miss everything. All leads should be close to zero. I do have some undersized wires I noticed, but everything is soldered. I have read the 0V leg should be oversized because of the charging duty-cycle.
This stray ground impedence would have to be very high since the amp is only idling and is fed by a 200VA transformer, I am thinking...
Photos even with the best photography would be very hard to follow. There are boards all over the place for various functions (these will be consolidated later) so the chassis real estate is tight. One photo would not cover all of it.
I have seen excellent physical diagrams showing the points, and ideally the wire lengths and sizes, IMO. This will be time-consuming in my case. I really need at least the ability to show DC on the scope. The Daqarta modified DC sound card does that, and gives the second channel, too.
The headphone-philes use low value dropping resistors between the transformer secondary and the rectifier. These flatten the charging peaks and reduce the effective load on the transformer, too. For unregulated HT, I have seen 10R for each diode. Yet for regulated heaters, nothing, despite the higher current. I do not have these on the power amp PSU at the moment.
I did unearth Mr. Elliot's chassis grounding scheme with no effect either way.
I did experiment with 10R on the RG174u input coax shield. This produced considerably more audible noise. The preamp pcb is grounded at both ends of the ground bus come to think of it. So even though there are no other loads since it is dead at the moment, the ground connections are still intact. This bolsters your "stray ground impedence" theory, since added impedence has a deleterious effect. If that shield is ungrounded, the amp is so noisy it is unusable.
For the moment, I will try to find something with the DVMM, and perhaps experiment with different routes and wire size.
The preamp pcb is the thickest board with the heaviest traces I have ever seen. It is the bootleg, Soldano SLO-100 board from C3 Amps. There is no impedence there. But the board is grounded on one end only anyway.
The power amp signal input (or output from the preamp if you will) is at the end where the PI and power tube circuitry would have been. The unecessary board length has been cut of so as to eliminate the possible RFI antenna effect the empty traces may have introduced, and is adjacent to the HT supply input.
The HT supply does not have a path to STAR as it is an isolated supply from another transformer. Same goes for the opamp pre-preamp feeding the V1 input on the opposite side, also fed by a dedicated transformer. This scheme actually eliminated some loop problems. This opamp is so sensitive at high gain that a cheap resistor would afford much hiss, and the input decoupling cap could be said to be an antenna! But all that is dead for the purpose of the instant diagnostics. Their respective transformers are also disconnected.
At least there is a smaller realm to consider now. I really need to see DC ripple on the scope (if any). I have been putting it off, but now I may be at the end of the line with that.
Thank you brian.
It seems you are familiar with most problem areas and have put in a lot of work on the project already. The separate problem of one output device overheating but not the others suggests something strange, if the idling current is equal at 6mA through them all. That's very low bias current, even for a CFP design and the transistors should not even feel warm but who am I to argue, and the application isn't hi-fi anyway. This problem should not have anything to do with the hum, so it would be better to disconnect the preamp stages completely (you could then ground the input via a 10k resistor, for certainty)
There are a couple of things to try there: Swap the 2 NPN output devices to see if the heating goes with the particular transistor. Test for oscillation, though it will be at a much higher frequency like 300KHz to maybe even 10 MHz - certainly not 60Hz or low harmonics, so you won't see this on an PC audio card. You may have to beg, borrow etc. a 'scope - even a cheap 20MHz 'scope should display BJT output stage oscillation.
Your FFT graph shows a fundamental of 60 Hz with harmonics (odd dominating) running out, well beyond the audio range. That means it is mains related, not from the rectifier, as long as that is operating correctly. Brig's comment about ground shifting with the rectifier charging pulses sounds interesting, though I don't claim much experience testing it.
There are a couple of things to try there: Swap the 2 NPN output devices to see if the heating goes with the particular transistor. Test for oscillation, though it will be at a much higher frequency like 300KHz to maybe even 10 MHz - certainly not 60Hz or low harmonics, so you won't see this on an PC audio card. You may have to beg, borrow etc. a 'scope - even a cheap 20MHz 'scope should display BJT output stage oscillation.
Your FFT graph shows a fundamental of 60 Hz with harmonics (odd dominating) running out, well beyond the audio range. That means it is mains related, not from the rectifier, as long as that is operating correctly. Brig's comment about ground shifting with the rectifier charging pulses sounds interesting, though I don't claim much experience testing it.
Last edited:
I started on this one year ago. Funds are limited so I went in stages; researching and testing each subsystem. The project is patently too complicated for a beginner, but it's too late now. Good news is I have the basic sound that I like: An MXR Distortion Plus up front feeding a Fender Super Reverb. It's sounds even better and has the versatility of two channels. The sustain is endless on the highest note on the guitar (the most difficult to sustain), at max gain and low volume. I rang a note and stopped it because I got tired of waiting for it to die after 30 seconds.
The pre-preamp has a gain of 3 and has quite an effect. Some attenuation is required in the stages that follow, although the large grid stoppers limit any deleterious effects. The reason for this pre-boost was to "beat" the noise floor of the first valve stage, which could not be taken out, less a noise gate, which is undesireable, IMO, except for recording session, but if it's quiet at say 30w, it should not be necessary.
It works quite nicely as long as the opamp circuit is noise free, of course. And it will be.
Rod claims HiFi quality may be achieved by doubling the bias current and using more linear (and more expensive) output transistors. I believe this since when testing the unit early on, an audio input was sent in and it sounded about medium-end quality--my subjective opinion only: A $200-300 stereo on one channel.
When I said "warm" I meant just noticeably to the touch. As I said, just the one started to come close to breakdown. In other words, it was shut down just before the point a heatsink would be necessary. Quite hot, though. And yes, I find that strange, too.
The preamp is disconnected all this time. I also suspect there is HF oscillation because I noticed the cathode resistors on the cathode followers run a little hot even though they are heavily overrated. The current is normal on the DVMM. Yet the amp works perfectly, and all controls work as expected.
The chassis has a hinged lid because I knew I would be going in there frequently, yet the ability to close it for testing is also necessary. Most of the time it is open. There is a line filter with with two small caps tied to ground after the common mode choke, in addition to the one up front. There is also TVS on the mains but that just protects against line spikes.
Are you implying a swap test will reveal a damaged transistor, or the fact of HF oscillation?
I don't see any fundamental in that mess. I did check for AC at the PSU output. I get zero but if I switch the leads, 0.5 VDC shows steady on each leg. I am assuming it's the DVMM since all the parts of the supply are new. There are TO-220 schottky rectifiers in there, if that makes any difference. At least they are overrated in terms of PIV and current.
Thank you Ian. These ideas and thoughts are good. This is how I went along: armed with a basic knowledge of what each circuit does, some common sense logic, and extensive trial and error.
Actually, some of the mistakes yielded good object lessons. For instance, there was an arresting oscillation at low gain and volume settings (ground loop caused) in the beginning. I hooked up a speaker and made changes hot, listening for the results. Aside from a few 400 VDC blasts and 4 burned out speakers, I found out why builders stress the importance of good solder connections and clean valve pins! Long wires invite trouble, too, I found.
The cheap, chinese ceramic sockets are no good for hi gain designs, especially in first and second position. Belton is the absolute, bare minimum here. The CNC machined, gold plated types are more appropriate.
OK, I should have realized that you referred to the temperature getting hotter through one device, only when removed from the heatsink. That would be normal without a heatsink to add its thermal inertia and equalize temperatures. Otherwise, one device just slowly builds up a differential via its slight physical differences and a limited thermal runaway commences. Without a heatsink, there is little to ensure current sharing.
'An unusual problem but one you wouldn't have in normal operation. A swap test would find device differences alright but probably not reveal anything subtle, unless on the heatsink.
That is precisely what I was thinking. All that is needed is a slightly higher resistance on one device for that one to commence thermal runaway. The low current prevents that but it should not be left on too long without a sink. And yes, a common heatsing with device in close proximity with each other should average out. But we would not be able to tell short a very accurate digital thermometer.
This reminds me of series-connected valve heaters, or incandescent light bulbs for that matter. Upon energization at full voltage, the filament with the most cold resistance will begin to warm first, and during the process, inrease its resistance due to the thermal expansion--hot resistance. It then commences its own thermal runaway, since it is robbing all the voltage, and the filament burns open as the voltage is twice its design rating.
I am not using series heaters at the moment, but my headphone-phile tube pre design called for the LM317 with a massive load storage cap (for a resistive load) that takes several seconds to build up. The actual preamp design, I think, considered the possibility of series heaters. With AC heaters direct from the transformer, a small value series resistor is used, to limit slow "inrush". This does not sound too promising to me, and I certainly would not use it on expensive NOS valves.
I have detected that you realize my limited knowledge and equipment situation, and the resultant methodology I am compelled to use. In my defense, it should be noted that the established amp companies, although they have a great head start with their established designs, still have to do the chopstick thing for days or weeks in order to remove noises and oscillations. So in the end, as it has been said, amplifier design is more of an art than a science. However, the rules of physics cannot not be broken, and we all must come to terms with these, eventually.
My convenient "black box" view of each subsystem must be jettisoned from time to time....
Here is the grounding scheme:
Bear in mind everything is dead except for the power amp.
Mr. Elliot, in his solid state amplifier diagnostics page, states faint hum/buzz that requires placing the ear next to the speaker is essentially impossible to remove by established methods.
Mine is a little louder than that and must be addressed.
Bear in mind everything is dead except for the power amp.
Mr. Elliot, in his solid state amplifier diagnostics page, states faint hum/buzz that requires placing the ear next to the speaker is essentially impossible to remove by established methods.
Mine is a little louder than that and must be addressed.
Did you replace all the rectifier diodes after connecting the transformer wrong? It is possible that one or two are not turning off correctly giving you a momentary short on each half cycle. This high current could be causing your problem.
Brian
Brian
Yes, I replaced them with some old U820 from a 2005 projection TV. At least one of the original diodes was damaged because when I replaced them, the hum was attenuated considerably.
I set up a 1k pot across the legs, with the wiper set so my sound card would not blow. AC voltmeter under this load (35mA) says:
4.2v AC on the Neg
2mV AC on the Pos
Confirming with the scope, although noise increased when the leads were connected, a clear sine wave was seen. The voltmeter leads were reversed just to make sure, and the values were exactly the same either way.
Before all this, I fiddled with ground wire changes, eliminated an unecessary loop, doubled wire sizes, to no avail.
So, there is AC going into the power amp. I see ultra-fast diodes used in switch-mode power supplies are used by some for this application, clamped with 100n ceramic or film caps. MUR8100EG I saw on one.
Any recommendations for diodes that don't leak too much, or some other way to block all the AC?
The original design for this PS does not call for anything special. Just a regular bridge with no clamping, and is said to be the quietest amp the designer knows of. I am baffled.
Thanks for your response!
Before doing anything else, I tried adding a 10R resistor in series with the AC0 input (the transformer center tap) and the PSU, in hopes to reduce the charging current peaks spoken of earlier in the thread.
The buzz is cut in half. So as it has been mentioned, this points to a ground loop. Or did it just reduce the effects of AC leakage?
The design will not withstand a total resistance of not much more than 1R here, since there would be too much voltage drop at full load. It draws 3A per leg on the DC side.
I will fiddle more with the grounding scheme in case something was overlooked (quite possible), by connecting, disconnecting, and rerouting again. 😡
The buzz is cut in half. So as it has been mentioned, this points to a ground loop. Or did it just reduce the effects of AC leakage?
The design will not withstand a total resistance of not much more than 1R here, since there would be too much voltage drop at full load. It draws 3A per leg on the DC side.
I will fiddle more with the grounding scheme in case something was overlooked (quite possible), by connecting, disconnecting, and rerouting again. 😡
- Status
- Not open for further replies.