Discrete audio amplifier project PT14 - BIG PROBLEM!

[@zoomernw]

Hey John, great video as always. Sorry to see the amplifier isn't going as you pleased. I've also recently built an amplifier that exhibited a very similar result to what yours is producing. I tried some things to alleviate the oscillation which helped but didn't quite get rid of it. Make sure your bypass caps are connected. You could've missed a solder joint or something, can't hurt to check. Since the oscillation peak-to-peak voltage is around 3v, it probably means that its being amplified. Change your feedback resistance and see if the oscillation voltage changes proportionally to the resistance. If it does, then it means that its before the output stage. If the problem lies in the input stage or vas stage then I would suggest you begin probing with only the input stage connected and gradually connect portions of the circuit until you find the issue.You can also add an rc lowpass filter on the input and vas stage rails. If the problem is on the output stage id suggest doing the same with the input and vas stage. Disconnect the input and vas stage, probe and gradually connect portions of the circuitry back together. Finally, if all that does not work, you might want to consider that your amplifier has a fundamental design fault. I do not want to believe this is the case because it makes sense to me that the amplifier should work. Also remove any unnecessary circuitry such as the zobel network, output choke and protection diodes. Thats all the "help" I can give unfortunately, I'm sure you tried most of this already or probably thought about it. Just letting you know that I'm not an engineer, just a guy who puts up with these problems after every build. Lol

[@fredfabris7187]

zoomernw, it sucks but I find that it always turns out to be a learning experience!

[@TheTrueVoiceOfReason]

Well, since it was "working" (that you know of) at the end of the last video in this series, I suggest undoing what you did and verify operation at that level again. Then do 1 change at a time and test at each step. Slow and methodical helps keep the vision clear.

[@envisionelectronics]

You will need an RC filter at the input known as an RF stopper: Place a 0.001uF to GND from the base of Q2 to GND on the left side of the DIFF AMP. This sets a corner frequency of about 160kHz which will eliminate the amplification of RF ingress.

[@johnyang799]

I think he mentioned he shorted input to ground. So it can't be rf being picking up. Also 3V p-p is really a lot, instead of 30mV.

[@envisionelectronics]

John Yang Sure it can. He grounded the input - where? At the end of his input cable? It’s still a big antenna.

[@johnyang799]

@@envisionelectronics Well. Apart from what I said, he already has a capacitor in the front end. It's shown on both the schematic and the board(yellow ceramic at top).

[@envisionelectronics]

John Yang But it’s not fitted for RF blocking. There is yet a 1k resistor between it and the base of Q2. It is also 1/5 the value I recommend and paralleled with a 22k ohm resistor. Trying to figure out what that’s all about.

[@fabiotrevisan8922]

Hey John, greetings for the great job. At such high frequency, I would put my bets in emitter follower oscillation, which means either the drivers or the output devices. I'd say more the drivers, because they "see" the quite high base collector capacitance of the output transistors (while the output transistors themselves are not seeing any capacitance). And among all, this one is relatively easy to rule out... Just lift the bases of the driver transistors and insert a small value resistor there. As for the value, consider the base collector capacitance of the drivers and set a time constant that sets the cutoff to - say - two or three octaves below the oscillation frequency you're getting. Well, I may be wrong about how to calc the resistance all along... As what it needs to do is to counter and swamp the eventual negative resistance created at the base due to the capacitive load at the emitter. Ah, in time, it just crossed my mind as well... Did you ground the heat sink? I just thought of a possible capacitive coupling between the collectors of the output transistors to the Vbe doubler.

[@MichaelBeeny]

John, we have all built amplifiers over the years that are better transmitters than audio amplifiers. Assuming you don't have a faulty part, I think I would look at the feedback loop layout. As it's such a high frequency it must be very lightly coupled capacitively. Try re routing that lead. Good luck, in my experience you get get stability problems on ready made amplifiers that are fine until you connect it up and you need to spend hours messing with earth connection.

[@NotMarkKnopfler]

Before you go down rabbit holes, check that what you have built actually matches your schematic! In your last video you stated that you pretty much threw it together. Be sure that what you are debugging is what you INTENDED to build! :-)

[@KissAnalog]

NotMarkKnopfler I think this is the first step. Maybe take a fresh schematic and highlight each connection and part as you trace your board. Nice video!

[@cassvirgillo3395]

Hey John, Not counting the amp, I hope all is good with you. Where a Runoffgroove, 1/2 watt, LM 386, Ruby amp is no where close to the amp your building, I have found 'stray' voltage on the actual phenolic material the board is made of. Judging by the nice work you do, I'm sure you clean your board after parts install, I do too, and, I use a utility knife to score around the pins of FET. I would bread board a Ruby and would work great, on perf, poor performance. After cleaning and scoring around pins, the amp would work perfectly. Solder paste residue that cleaning did not get. I use HEET, ethanol alcohol, gas drier. Stray as it were voltage, just a thought. Take care, Cass.

[@MrAlFuture]

Thanks, John! Following your trouble shooting is nearly as good as seeing the working amp. Did the hf oscillation occur when you ran the board with no heat spreader/sink? What's the twin core speaker cable doing running under the board? Is it connected to anything?

[@johnyang799]

Did you try the output stage outside with the 1n capacitor across the bias spreader ? Could cause instability. Touch up each solder joint again to make sure everything is connected, especially the bypassing/decoupling capcitors. Check dc values of each point, a feedback amplifier could actually "work" with such issue.

[@johnyang799]

One another thing. Can you still play music through it? If not, some component may be damaged. If yes, then try tests with signals. Can give a bit more of information.

[@johnyang799]

Another thought, set gain to around 1000 times then probe around and see where the issue is.

[@JohnAudioTech]

Yes the amp sounds good with music. With the +/- 15v rails, it produces 18 watts into a 4 ohm load and low distortion. No oddities noted in the waveform at clipping. It just has the 40MHz oscillation riding on the test signal

[@johnyang799]

@@JohnAudioTech It's certainly weird situation. It would make more sense if it doesn't work. To me, i have to start with changing gain settings and see if the amplitude changes, and try open loop probing with input shorted to ground, then disconnect the 1n cap across the vas. Try capacitive load stability and square wave(clipped sine should be fine) just like when you did the EF CFP output stage, but with this "working" amp. There can be a lot to do, but since it's "working" it's really hard to diagnose.

[@johnyang799]

And my last hope is to build another one...

[@Cesar_III]

So now you have an "unconditionally stable oscillator". I had the same problem, but with CFP output. I fixed it redesigning the PCB and modifying the Zobel network. A 20pF capacitor in the FB sometimes work, but is not the solution, the amplifier *must not* oscillate without any cap in the FB. By the way, a radio filter in the input may help, it may be a strong radio signal near the house.

[@fredfabris7187]

I don’t remember it doing that prior to mounting it, (not an engineer)I would be curious if removing it from the aluminum makes it go away. Is the aluminum grounded? Maybe it should be. Seems like the collectors of all those transistors with the Mica spacers would be a capacitor.

[@killcar5nbike2]

Current mirrors on the input stage have given me many a headache. I generally run LTP inputs with the inverting side straight to the rail and 470-820 ohm as the collector load in the non-inv side (base to emitter VAS) with DC offset determine the final value. Not sure what TO220 transistors your using but I've found the TIP range to be gits for oscillation. TIP42C etc seem to be buggers for it. Did you connect the heatsink to 0v, I've had floating heatsinks cause me oscillation before.

[@n.shiina8798]

so, a floating heatsink could cause oscillation too?. i have my own designed amp oscillate on high bias (higher than 400mA per rail) and the heatsink was floating. never thought of grounding them

[@killcar5nbike2]

It becomes a plate capacitor with the heat pads of the transistors bolted to it, it some circumstances high frequencies can couple between devices through the heatsink.

[@n.shiina8798]

@@killcar5nbike2 makes sense. i'll try to ground the sink later. could my Al2O3 thermal pad induce the problem even worse?. it's 0.6mm thick

[@killcar5nbike2]

It shouldn't do, there's so many reasons an amp can start oscillating. It may help, it may not.

[@humpadinger]

Hi John, Im sure you will get this working. So the amp was working, no oscillation, before fitting the heatsink? The only change was the Vbe pot. I cant see this being an issue. Try grounding the heatsink to stop it trying to impersonate an antenna. If this doesn't work, remove the heatsink and check again. All the best.

[@TheTrueVoiceOfReason]

Also, is it possible there is some sort of capacitive coupling of the transistor leads with the "heat spreader" due to how close they are?

[@envisionelectronics]

FYI I sent you an email under envisionelec about a PCB layout. I’ve designed a few amps over the years. :)

[@mrjohhhnnnyyy5797]

To me, that Vbe multiplier bypass cap looks suspicious. It might create a delay long enough to turn negative feedback into positive. Or it can be the CCS indeed, try putting Miller cap on Q7. Your transistors have nice gain bandwidth, it very well can be the entire loop oscillating. With slow hogs like 2n3055 and such - yeah, 39Mhz wave will never pass through them, they will attenuate the oscillation like a snubber :D

[@bernardmarx6705]

Did you connected heatsink to gnd?

[@fredfabris7187]

Borko Markovic I’m wondering that too!

[@EngineeringEssentials]

I think the problem is right from the input differential stage. And the oscillation is not sustaining means this oscillation is generated directly from the signal side. It is not the problem of stability, for an unstable system the oscillation should not damp itself out. I think the 40mhz noice source is outside the circuit.

[@envisionelectronics]

MY TECH WORLD AND KSRTC Correct.

[@johnyang799]

@@envisionelectronics Only thing that I can see is the output inductor.

[@envisionelectronics]

John Yang I don’t understand what you are saying about the zobel compensation inductor. That has no play in this oscillation issue. I can’t see the whole schematic or the actual layout but I suspect the amplifier is barely stable and some external signal is causing it to break into oscillation. The unshielded, twisted-only pair of input leads isn’t helping. There are lot of things at play in causing an amplifier to oscillate and I suspect it will be difficult to determine from a video.

[@Dazzwidd]

I think it's likely to be related to inductances in the ground circuit. If you made it on a pcb with better ground integrity then it probably wouldn't do it. However there may be a solution to get it working fine on the perf board... I'm sorry I don't have a clue on how to achieve that other than doing your best to lower the ground impedance. It would be interesting to see how any changes change the amplitude and frequency of the oscillation before curing the problem

[@mrjohhhnnnyyy5797]

Yeah, but I have built quite a few amps on perfboard, they didn't oscillate. It might be something sneaky :/

[@envisionelectronics]

“Inductance in the ground circuit” Is not likely at all. How much inductance in the ground circuit would it take to make it oscillate at RF? And what is causing the LC tank to resonate? This is a linear amplifier with no switching mechanisms.

[@JohnAudioTech]

The ground traces amount to only a couple cm in length excluding the trace that goes over to the input (high impedance) and the shield that doesn't really carry current. High current grounds "star" into the ground plane around the bypass caps on their own trace. It is hard for me to believe the grounds are an issue.

[@Dazzwidd]

you may be surprised. try to improve the decoupling at radio frequencies. Maybe even try ferrite beads at certain locations

[@eugenepohjola258]

Howdy. One cause of parasitics is the grid - cathode capacitance or base - emitter capacitance. Together with a grid stopper resistor or base stopper resistor they form a RC filter. The RC filter shifts the phase a little. Hopefully out of the oscillation criterion region. If the transistors are fast one may try to "worsen" them with separate capacitors between grid and cathode or base and emitter. This should shift the RC kink frequency toward lower values taking the oscillation frequency out of unity gain. One may check the inherent capacitance from data sheets and try adding an outer capacitor of about the same value. Regards.

[@rochtourigny8140]

Hi John, if you have Bob Cordell book, get to read chapter 9. I am not an engeneer, so i will not try to pretend i understand the advanced solutions and can help you. What i understand is that you need a low pass filter in the feedback loop. Low enough to stop the oscillation, high enough not to hurt the slew rate. Then monitor phase relationship of input and output across the whole bandwith of the amp.

[@kirknelson156]

i'd disconnect the feedback to see if it goes away, remember transistors also have capacitance, its small but you since all it takes to make an oscillator is a RC network and a feedback loop, you might have built an amplifier/oscillator. if you disconnect the feedback and it goes away then perhaps you can put an inductor on the feedback loop to filter out the high frequency signal. good luck :)

[@n.shiina8798]

We got the same problem lol. mine works fine only if i don't bias the final stage higher than 400mA per rail. idk what went wrong but i've blown a couple sets of NJW0302/0281 and TTA/TTC004B transistors. It runs fine at high bias current if i increase the miller cap value, though

[@johnsenchakinternetsecurit8935]

Interference from your cell phone or the microchip in Snickers brain is sending out high frequency oscillations

[@johnyang799]

To check this it's pretty easy, just turn up supply voltage and see if the amplitude changes.

[@fredfabris7187]

John Senchak Internet Security Guru it is probably the cats fault!!😜

[@linorocha6319]

Greetings, can you make a discrete amplifier using MOSFET instead of bipolar transistors? Thank You!!

[@bjtaudio]

It's more a pcb layout issue, you have amp just love to oscillate, and go pop!

[@1959Berre]

I agree. Perf board is not the most stable basis in the first place. The ground connections should be real wide wherever possible, so capacitance knocks down HF oscillations.

[@fredfabris7187]

I suspect that it’s not in the output section because it does it with or without a load.

[@Dazzwidd]

I think you're right, especially if the frequency remains relatively unchanged with the load removed

[@Pico_Farad]

Try if it persists without negative feedback

[@JohnJackson66]

You can split the amplifier sections for testing. Disconnect the driver transistors from the VAS , reconnect the feedback to the VAS emitter. Obviously this won't drive a load but it should be DC and AC stable and be a half decent amp stage in it's own right. I've seen the current source stabilization with a 1K and 100p. If the input stage is OK it might be your CFP output, maybe your 220R between the op bases could be split into two 100Rs that base to emitter? I was a little concerned about the common ground for input and output section but you have plenty of decoupling so maybe it's OK.

[@JohnJackson66]

I just double checked and noticed your op is Darlington not CFP my bad.

[@1959Berre]

John perf board is not the way to go. You need broader & wider ground surfaces for shielding and capacitance. Can you put the whole thing in a metal box and rule out external sources?

[@JohnAudioTech]

Perf board is certainly not ideal but allows me to change the circuit without redoing a board. In the video after this, I located the oscillation problem.

[@1959Berre]

@@JohnAudioTech I have seen the follow up video and the solution. Seems I was wrong about an external influence. Good to know you nailed it.

[@swaaapy1]

Driving output or VAS transistors close to ratings, had my basic amp oscilating.

[@Markoul11]

input stage differential amplifier stage and feedback loop. Seems to me that you don't have negative feedback but a positive one turning your amplifier into an oscillator. Also some lead can not be connected or bad soldering or you made a connection it shouldn't be there.

[@johnyang799]

The amplifier works with signal and has low distortion near clipping, and no wierd behavior at clipping. Only has 3vp-p 39mhz signal riding on the signal.

[@jamespalmer7629]

I hate putting complex circuits on perfboard. Things I have designed and tested on a breadboard seem like they never work correctly. I would etch a board if I were you. Good luck!

[@boonedockjourneyman7979]

Probably your best video. No one has the guts to keep it real on KZbin. What did you change from the last test? Heat sink is all I could see. To me, a digital guy, it seems like an issue between collector and base. People call it parasitic until the TO220 bodies are connected by a sink. Don't know what to call that effect. Some sort of coupling through the mica? Just a guess.

[@JohnAudioTech]

It may have been oscillating before. It doesn't really affect the sound. This was the first time it was on the scope. I think I'm getting somewhere with it. I hope to get a video up with changes as I go along.

[@boonedockjourneyman7979]

JohnAudioTech - Take your time. The thoughtful content of your work is what distinguishes you from the hoard of "mail bag" EE videos.

[@HillsWorkbench]

Smith Kerona is doing a very similar amp design. link for part 1: kzbin.info/www/bejne/hoTCoWaFbtZnj6s In part 2 he gets deep into the math, part 3 he will get into the thermals. Will be interesting to see if it oscillates when he gets to building it.

[@stephanc7192]

Interesting!

[@natecontarino1748]

Could you have unintentionally made an inductor with the two jumpers?

[@joohop]

We Know Its Gonna Be Good Lad Just Take Your Time And Put It Out Your Mind For The Time Being Take Care Brother

[@markweikle5645]

Definitely measuring the cosmic background radiation!

[@iceberg789]

that does not oscillate in exactly a single sine wave i think.

[@markweikle5645]

@@iceberg789 correct. (I might have just been playing a little silliness there, adding to the pile of speculation... )

[@DAVIDGREGORYKERR]

What about a 5W Wide Band class A amp using 3X BC182L transistors in a long tailed pair driving 2X BD131 power transistors could be used as a preamp driving a wire-wound Variable resistor which feeds the main output stage just Google for it, what about a 470pf between the inverting side of the amp and ground.

[@snaprollinpitts]

Hi John, the high frequency oscillations you are seeing are called ringing, your board/ circuit is resonating. I saw this a lot when I was working on a hydrogen generator project. the 1 micro Henry coil is doing what? whenever you have capacitors and coils together you just might see this ringing. what would happen if you removed that coil? I know this is not what you want to hear, but what if you added a low pass filter? to attenuate the high freq?! okay maybe that's a bad idea. mike

[@aljivenalejo3904]

try to replace a shifter transistor is a preoutput stage of the amplifier somehow i make ones i somehow i put larger amperage to the circuit and start to get noise then somehow the noise it louder on the sound itself til it the music gone and the oscillation got more compromised and wont work so i decided test if theres a leak and yes!! the positive supply shifter damage because of pulls lots of power

[@aljivenalejo3904]

somehow did you try to filter much on the power supply?? ithink thats the one too because some how i make a mistake that i dont filter much on the supply side and the noise comes from

[@BigIron_06]

Just put a 1K resistor parallel to the one that is in there.

[@Margarinetaylorgrease]

So much drama : )