Help me troubleshoot this design, please

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello everybody. I'm in the middle of assembling this amplifier but it seems to have some stability problems. When I first tested it, it oscillated on a low frequency for few seconds (not very loudly though) and then short circuited. I had the amp hooked up on a light bulp current limiter so the output devices did not die. No component seems overheated or blown and the amp ran very cool... Of course, I shut it down immediately when I noted the short. This is where the things go beyond my comprehension...

If I remove the wire link marked in the schematic the output devices indicate no short circuit: When powering on, the light bulp burns on a dim glow for a second and then fades out - during short it is very bright. Now, this would indicate that the short is on the bias/class-A driver side. However the DMM shows no shorted tracks or components there. As far as I understand, the only thing to cause such drastic current draw would be a falsely operating VBE multiplier - more specificly a short between it's base and emitter or something similar. However, it's resistor values test ok, so does the transistor - I even deliberately ran the amp underbiased when I powered on the first time. The vbe mult transistor is mounted on the same heatsink as the power transistors and drivers.

The circuit works fine on spice simulation but sometimes this has little meaning in reality - perhaps I have something wrong in the design in the first place? This has gone beyond my comprehension: My test results indicate that everything should work but still something is wrong. Perhaps someone can offer a fresh idea for what to troubleshoot and help me out on this...

I should add that during the first test the zobel was omitted and the output plus vbe mult transistors have a quite long wiring. The driver and VBE multiplier transistors (except the class-A driver) are BD179 and BD180 instead of ones shown in the schematic and the power transistors are TIP141.
 

Attachments

  • problem.gif
    problem.gif
    7.1 KB · Views: 434
How about a pic of the setup.

Lf oscillations could be due to ( ground) wiring. What appears to be LF oscillations could also be really RF oscillations! Are you looking at a waveform on a scope ? A picture of the set up might give us a better perspective to your problem.
In addition the BD140 driver generally had a resistor and diode on its emitter in more recent circuits. Supposedly to duplicate the emitter load on the upper driver. I might be wrong ...check D Self.
Cheers.
 
mrshow4u said:
pardon my ignorance, but what is that 200k, 47uF stuff yanking on the output line??


....and how is the input transistor biased??

That trimmer is used to set the DC level to half of Vcc at the output (for symmetrical drive). Together with the input transistor's base resistor it also handles input transistor biasing since the 47uF capacitor bypasses AC signals. The following resistor/cap sets the negative feedback.

As requested here's a picture of the set - I haven't taken a newer picture, which would show the circuit left from wire link in the schem. It's built on a small veroboard mounted on the top left bolt holding together the heatsinks. The two blue + yellow wires come from the vbe transistor and they were shortened to about 1/3 of the length shown in the picture. The grey wire (on left)goes to speaker terminal (where the output cap - and later the zobel - are mounted) and the red and black wires go to power supply. The "lower current" circuit board gets the supply and ground connections by other wires not shown in the picture. All black ground wires go to star ground point. Granted, the layout is messy and the wiring too long but currently I have no possibility to make a proper PCB. Unfortunately, I do not own a scope but it will probably be a future investment.
 

Attachments

  • setti.gif
    setti.gif
    59 KB · Views: 308
Re: How about a pic of the setup.

ashok said:
In addition the BD140 driver generally had a resistor and diode on its emitter in more recent circuits. Supposedly to duplicate the emitter load on the upper driver. I might be wrong ...check D Self.
Cheers.

Ah, you are meaning a circuit that I know as "Baxandall-diode". (Took a while to comprehend that - at first I thought you were speaking about the internal structure of the transistor itself). I am well aware of this circuit but I decided to omit it for simplicity's sake, nice that you mention it though. Leaving it out will increase distortion a bit but as far as I know it should not cause issues with stability.
 
davidallancole said:
Have you measured your voltage across the Vbe multiplier. From what you have there, would the bias voltage not be 4V (1900/470=4)?

Hmm... The spice simulation shows 1,7 V across it rather than four when the trimmer is set at 1900 ohms. I haven't tested this since I'm not too keen on keeping the amplifier on during a short circuit for a too long time. The light bulp limiter is a crowbar safety measure allowing me to flip the power on and notice a problem without blowing things up but I fear a long term short circuit (such as making the voltage measurement over vbe mult) would be pushing it too much.

I know little about vbe mult circuits; only that the bias voltage requirement is about 0,7 times the number of BE-junctions. If the transistor has a high gain the ratio between R2 (collector to base) and R1 (base to emitter) should be around (Vrequired / 0,7) - 1. In this circuit the result would be four resulting, for example, into 4 x 470 ohms = 1880 ohms. In spice simulation this value was proven insufficient for BD139. A higher beta transistor model (2N222A) indeed shows a figure around 4 V @ 1900 ohms. Perhaps the spice model I have has an error (it's from ON Semi)and the bias current is indeed set too high. But I gather that the Hfe figure counts pretty much and BD139 has a pretty low one. However, I do not suspect an error in the amount of bias voltage because....

I mounted the Zobel network plus did the very simple test mrshow4u adviced and shorted the vbe mult base to collector (before that I changed the BD179 to brand new BD139) - still a short circuit.

To me all this indicates that the problem can not be in the bias circuit but I just can't believe it. I can not understand any other way the circuit would make the power transistors draw such as high current as in a short. The Vcc supply for the "low current" circuit goes through very high resistance - if there is a short instantly from wire link to ground then at least my DMM does not show it. Also, I do not have a load attached, which should be a reason good enough to rule out the theory of oscillation in the input that would drive the amp at maximum level. Still, even without the load attached an error in the vbe mult circuit would push the output hard.
 
only that the bias voltage requirement is about 0,7 times the number of BE-junctions.
The word "about" causes a little concern. If you plot the I/V curve of the B/E junction, or any diode, you'll see that after the forward voltage is reached, the current increases rapidly. Differences in transistor doping, temperature, etc will have a large impact on Vf. There's another thread just on getting bias to track well. Sim first, but then real world it. Under bias first, then adjust the bias until you get the current you desire through your output devices. After that, check to see how close the real measurement is to your sim.
 
mrshow4u said:
The word "about" causes a little concern. If you plot the I/V curve of the B/E junction, or any diode, you'll see that after the forward voltage is reached, the current increases rapidly. Differences in transistor doping, temperature, etc will have a large impact on Vf. There's another thread just on getting bias to track well. Sim first, but then real world it. Under bias first, then adjust the bias until you get the current you desire through your output devices. After that, check to see how close the real measurement is to your sim.

Very true indeed. However, even when the vbe mult transistor's base is shorted to collector (should cancel the effect of the circuit) the amp shows a short. I'm believing the problem has to lie elsewhere. There was a mention about supply polarity and internal "protection" diodes of TIP142. They were good points but surely not the problem here. I am 100% sure the circuit is operating on correct polarity.
 
teemuk said:


If I remove the wire link marked in the schematic the output devices indicate no short circuit: When powering on, the light bulp burns on a dim glow for a second and then fades out - during short it is very bright.



Hi teemuk

It seems that the glow at power on is the charging current of the output cap 1.000 uF. Try to power on ,without a load connected to the output...
 
I have tried this and it has no effect - ever since I noticed the short I have been making the tests without a load. The light is definitely not a "charging glow" - usually that one is just a bright flash after which the lamp either settles to dim orange glow or no light at all. Now it constantly burns very bright unless I disconnect the wire link shown in the schematic.

Is it safe to disconnect the "power" section (before the wire link) and test the "low current section" individually to see if there's indeed a short or whether it drives the power section to constantly open state? If this is safe, which is the best measure; to disconnect both supply and ground wire or just one of them?
 
Update: When the supply of the "power section" is disconnected and the supply of the "low current section" connected the light bulp current limiter shows no short - as I was afraid it would do. It is now obvious that something saturates the output transistors but what could do it? The effect of vbe multiplier should be canceled since I connected it's base to collector. At this point I'm completely out of ideas...
 
teemuk said:


Is it safe to disconnect the "power" section (before the wire link) and test the "low current section" individually to see if there's indeed a short or whether it drives the power section to constantly open state? If this is safe, which is the best measure; to disconnect both supply and ground wire or just one of them?

The best way to test the low current section is disconnecting the two bases of the BD139/140 drivers and connecting the 220k pot to the emitter of the BD139 VBE transistor.

I suspect that the output stage is oscillating , because with the high gain of the output darlingtons, stability is not easy to achieve , specially in the lower part of the output stage...
 
Tube_Dude said:
The best way to test the low current section is disconnecting the two bases of the BD139/140 drivers and connecting the 220k pot to the emitter of the BD139 VBE transistor.

I don't quite follow.... What do you mean by connecting 220k potentiometer to the emitter of BD139? I did the previous test simply just by disconnecting the supply of the power section.

I guess I will test whether the "short" goes away by adding an insanely large miller cap to the circuit. This should tell me whether the problem is really oscillation, right? If this will not help then the only possibility I can think of has to be that the power transistors have become so stressed that they simply fail under any loading.
 
teemuk said:


I don't quite follow.... What do you mean by connecting 220k potentiometer to the emitter of BD139?

You need to connect the upper pin (in the schematic) of the 220k pot to the the emitter of the BD139 VBE (that regulate the bias ), because only that way the input transistor is properly biased and the input stage working in a normal condition...
 
No. Not as I see it. That trimmer is used to bias the input together with the base resistor. At the same time it sets the output voltage at half the supply voltage. As you see the trimmer bypasses most of the AC signals. The 33k resistor with HF bypass then forms the rest of the feedback loop. The VBE mult circuit's purpose is to make a sufficient resistance between the bases of the drivers in order to lower crossover distortion. These two circuits basically operate very separately. If I would connect the 200k trimmer to VBE mult's emitter I would break the feedback loop and the input transistor would be left unbiased. The result: little or no gain at all. Perhaps I have misunderstood you though, can you provide a schematic example?

I have also become almost certain that the "short" can not be caused by oscillation. Without a load attached even an oscillating amplifier has hard times to draw enough current to light the current limiter bulp. The output load is basically infinite so the power required by the output stage is very minimal. I shall make a test with a large miller cap to be certain about this.

At this point it seems pretty evident that I have to try with new power transistors next week. In that place I might as well improve the design to get rid of the LF oscillation that the amp experienced for few secs before it blew. Do "base stopper" resistors help or are they just reserved for lowering HF oscillation. How should they be chosen? Perhaps adding some capacitance closer to power transistors or the low current board... Suggestions, anyone?
 
I finally found out the problem: One of the lower power transistors had shorted collector to base. As an effect the circuit will oscillate and draw a very high current. Resistance at base (such as from a cold solder joint) will add a very low frequency component to oscillation, which will be heard as low volume bursting sound - this was what I heard on the first power up. I guess when the joint warmed up the amp "blew". If the low current part is disconnected the current draw will drop a massive amount and the light bulp limiter will act like I described.

I replaced all the power transistors and drivers, added some base resistors to prevent further oscillations and powered up, which again showed a short circuit! This time in one of the upper power transistors. I'm beginning to think that perhaps the TIP141 transistors I have are counterfeit; 125W transistors shouldn't die just because of soldering leads to them - plus they were on a heatsink when I did that. The drivers are once again ok although they should be less heat-tolerant. Has anyone bumped into counterfeit TIP141/142? How common would it be to counterfeit these - at least TIP142/147 seem to have become quite rare - out of production? Any ideas how to verify whether I have real devices or fake ones?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.