I wonder if this is not just noise caused by the signal input level that is low. When I inject 1kHz the oscillation observed is still there but only if I stretch the time base to the extreme...
Then, the 100MHz seems to be common mode pick up. Try reducing the scope bandwidth to 20MHz.
However, I'm not in a hurry to say the amp is stable or without other issues, as you still do pick up a burn smell, and VAS transistors unexpectedly heat up.
Any devices that smell burnt or suspected of prior stressed conditions should be replaced, as they don't worth your time.
For the time being leave C3, R301/401 out, until the unstable DC operation condition has been dealt with.
To safeguard the VAS transistors, insert a 1K ohm resistor between the collector of Q301 and GND. The same for Q401 as well.
Temporarily dial down the VAS current by a few mA, by changing resistor value from 220-ohm to 100-ohm for R102/104/202/204.
I don't see any major bypass capacitors to the power rails in your schematic, except them two 0.1uF. Do you have any on the PCB at all? I'm afraid they are needed.
However, I'm not in a hurry to say the amp is stable or without other issues, as you still do pick up a burn smell, and VAS transistors unexpectedly heat up.
Any devices that smell burnt or suspected of prior stressed conditions should be replaced, as they don't worth your time.
For the time being leave C3, R301/401 out, until the unstable DC operation condition has been dealt with.
To safeguard the VAS transistors, insert a 1K ohm resistor between the collector of Q301 and GND. The same for Q401 as well.
Temporarily dial down the VAS current by a few mA, by changing resistor value from 220-ohm to 100-ohm for R102/104/202/204.
I don't see any major bypass capacitors to the power rails in your schematic, except them two 0.1uF. Do you have any on the PCB at all? I'm afraid they are needed.
Ok, I can replaced all VAS 2N5401 and 2N5551 again, and do the suggested modifications above. No there are no more bypass capacitors. I know now that I should have put more after reading my new book. But seeing that I use two Rigol DP712 Power Supply for now I though that with these regulated supply it would be sufficient. Although these two capacitors come from the original mistake to blindly copy the schematic from the bad book.
As for the amp stability, don't worry I don't think it is ready yet. And I really appreciate all the time you invest analyzing and helping me in this issue. I must say that seeing all the modifications that the new PCB will require, I already have though about starting from the begining with a much more easy design.
Do you thing that this actual design; Paralleled Complementary Feedback, worth the effort to try to make it work as a first experience? In the contrary, which best quality design would you suggest for a beginner?
As for the amp stability, don't worry I don't think it is ready yet. And I really appreciate all the time you invest analyzing and helping me in this issue. I must say that seeing all the modifications that the new PCB will require, I already have though about starting from the begining with a much more easy design.
Do you thing that this actual design; Paralleled Complementary Feedback, worth the effort to try to make it work as a first experience? In the contrary, which best quality design would you suggest for a beginner?
I did yesterday without effect.
Meaculpa
I can hardly believe it myself! When I tried to desolder the transistors, I realized I'd forgotten to connect the base of Q301 to the collector of Q302, and the same for Q401 and Q402. No wonder I smelled something!
This time I used the good old-fashioned method of making a copy of the schematic in Gimp, then hatching each of the connections one by one after checking their continuity.
All VAS transistors 2N5401 and 2N5551 replaced with new.
Checked!
Done...
Also done!
I decided to wait another time for this one. I got pretty better results with the modification above. I am now at +/-35 VDC rails with 1.490 VDC across R503 and absolutely no oscillations at the output.
To open a parenthesis, the sale manager of Electro-Meters just call me and it was my mistake to think that the generator of my new scope wasn't giving the right voltage. The generator setting is peak to peak while I though it was RMS. 10 Vp-p on the generator is measured as 10 Vp-p on the scope channel.
Here is a picture of the actual measures with the modification above. The DVM of the scope measure -526uVDC at the output. My Multimeter measure -5.979mV. The noise at the input; 1.8291mVrmsrms give 1.0487mV at the BIAS TP5 and 207.96uVrms at the output. Is it time to test with a signal?
I tried to inject a 1 kHz signal at the input. It look's good. I was able to raise the input to 593.7mVrms with a clean output of 11.170Vrms. Raising it more started to distort. The voltage across R503 is stable around 1.4 VDC until I raise the frequency. I don't remember at witch frequency the voltage start to raise across R503, but at 20 kHz I measured 4.131 VDC across R503. It started to be hot at the finger there but no transistors become enough hot to be apparent to my finger.
The DC voltage at the output was at -35mVdc but I am not totally confident with this DVM precision on the scope yet.
The DC voltage at the output was at -35mVdc but I am not totally confident with this DVM precision on the scope yet.
You probably did not realize the Zobel network (10-ohm and 0.15uF) at the output can become a significant load to the driver transistors, now de-facto output transistors, at higher frequencies and at sufficient signal amplitude. At 20KHz 10Vrms the load current would be about 160mA rms, far greater than enough to load the driver transistors out of class A, into class B operation. That was why the rise in measured DC voltage over R503.
At 1KHz I would expect a distortion-free swing at least -25V - +25V at the output. I don't know why it distorted at a much lower level.
By probing the new 1Kohm resistors protecting the VAS stage you can tell if clipping is actually taking place.
This is an expected waveform at +/-25V p-p output, without clipping. The blue trace would be the voltage at collector of Q301. It is a straight line at 2V, meaning working at a constant 2mA
Ic.
This is an expected waveform when clipping starts to kick in. Notice during clipping the Q301 would either shut off, or try to overdrive the VAS but gets clamped by Q302 (Q302 action not shown).
Likewise, you can probe Q401 and see what's going on there when the distortion takes place.
I probed Q401 instead of Q301 because the collector is more accessible. Resistors 1K are soldered under the PCB and aren't accessibles. The first screenshot is with 250mV at the input and the voltage across R503 is stable at 1.062 VDC. The voltage across Q401 collector is at -28.3 VDC (VDM measurement at C2 Cyan).
EDIT: I just noticed that my probe was 1X while set to 10X. That would explain the 28.3 VDC instead of 2.83 VDC. To confirm...
If I measure only AC and raise the sensitivity of the channel, I can already see a little sine wave on the Q401 Collector but no distortion.
At 500mV at the input, I measure 68mV at Q401 Collector. The output sin wave isn't distorted yet but I can see a beginning of distortion at Q401 Collector.
At 550mV at the input. similar , but if I raise it 10 mV more, the Q401 Collector is raising high and the Power Supply current raise also. So I raised it temporally and stop the scope acquisition, then lowered it down to 130mV. Here is the capture.
EDIT: I just noticed that my probe was 1X while set to 10X. That would explain the 28.3 VDC instead of 2.83 VDC. To confirm...
If I measure only AC and raise the sensitivity of the channel, I can already see a little sine wave on the Q401 Collector but no distortion.
At 500mV at the input, I measure 68mV at Q401 Collector. The output sin wave isn't distorted yet but I can see a beginning of distortion at Q401 Collector.
At 550mV at the input. similar , but if I raise it 10 mV more, the Q401 Collector is raising high and the Power Supply current raise also. So I raised it temporally and stop the scope acquisition, then lowered it down to 130mV. Here is the capture.
Last edited:
If I change the Frequency for 400 Hz, I am able to raise the voltage input to a point where the output reached 21.598Vrms.
Back at 1 kHz but turning the RV1 BIAS fully clockwise, the voltage across R503 is around 0.3 VDC and I can raise the input to reached 16.814 Vrms at the output before distortion.
Above 1 Vrms at the input, the distortion came in...
Back at 1 kHz but turning the RV1 BIAS fully clockwise, the voltage across R503 is around 0.3 VDC and I can raise the input to reached 16.814 Vrms at the output before distortion.
Above 1 Vrms at the input, the distortion came in...
@Sevy
What we want is the DC voltage ref GND at Q401-C, so that the static DC operating current could be easily made out by Icq401 = Vcq401/1K
However, assuming power rail is at 35V, assuming Q402 did not kick in (it appears so from the scope shot), at 28.3V Vce Q401-C would have been too much apart from GND, or too high a standing current. By design you should see -2V at Q401-C, or 2mA standing current. Something is not right.
Also, the clipping response is much more violent at that small overdrive level than would anticipated, and it lasted way too long before recovery, about 1/4 cycle of 1KHz! Something is really not right.
Not saying it's part of the problem, but I would put in a couple rail bypassing capacitors (100uF/50V) at this point and see what they would do to the long lasting clipping response. (not removing C101/201)
I was thinking about next troubleshooting steps, but since you have no problems firing that parts canon please go ahead. More often than not that canon works.
Some basics: always make sure DC operation condition at all transistors are at anticipated value before attempting an AC test, by measuring DC voltage, and make out current with the sounding resistances.
Another thing you could help yourself with is getting rid of D1. That diode was there to protect the polar capacitor from being backwards biased, all because a polar capacitor was used to begin with. At development/troubleshooting phase, D1 can make things more complicated than necessary. So get rid of it, and replace C2 with a non-polar electrolytic variety, or two 470uf polar e-cap in series with negative leads tied together.
As I mention above, the 28 VDC across Q401 Collector is wrongly measured with the probe set to 1X while the scope was set to 10X on that channel. I just measured voltage accross Q301 and Q401 Collectors and I get:
Q301 2.74 VDC
Q401 -2.81 VDC
Zut... I cannot find 220uF non-polarized in my stock, neither something close to 470uF polarized. I do have 4 x 470uF but 50 VDC, Tying them together is to space consuming I am affraid...
I added two 100uF/100 VDC decoupling capacitors across D301-K and ground and D401-A and ground respectively.
Then while doing new tests, I realized that all preceding tests was done by measuring the rms voltage instead of the peak-to-peak voltage, that is more representative to the rails supply voltage limits.
Starting at low +/-10 VDC for the rails voltages, the distortion was flat and symmetrical.
At 10 Vpp at the output, third harmonic is at -50dB:
At 10.864 Vpp at the output, third harmonic is at -45dB:
At 16.874 Vpp at the output, third harmonic is at -30dB and we start to see the flat at each top:
Pushing the rails at +/-25 VDC, the distortion was still flat and symmetrical.
At 25.326 Vpp at the output, third harmonic is at -42dB:
Pushing the rails at +/-35 VDC, the distortion was still flat and symmetrical.
But raising it above, the weird non-flatten distortion came in... I don't have this picture...
Then while doing new tests, I realized that all preceding tests was done by measuring the rms voltage instead of the peak-to-peak voltage, that is more representative to the rails supply voltage limits.
Starting at low +/-10 VDC for the rails voltages, the distortion was flat and symmetrical.
At 10 Vpp at the output, third harmonic is at -50dB:
At 10.864 Vpp at the output, third harmonic is at -45dB:
At 16.874 Vpp at the output, third harmonic is at -30dB and we start to see the flat at each top:
Pushing the rails at +/-25 VDC, the distortion was still flat and symmetrical.
At 25.326 Vpp at the output, third harmonic is at -42dB:
Pushing the rails at +/-35 VDC, the distortion was still flat and symmetrical.
But raising it above, the weird non-flatten distortion came in... I don't have this picture...
It is better...
I reach 45,757 Vpp at the output before distortion. This is 2,4852 Vpp at the input. Voltage DC across R503 is stable around 0.3 VDC. Nothing is hot. The output of my generator is 900 mVrms. The current sinked from the two Power Supply are +0.05 A and -0.04 A respectively at +/-35 VDC. If I raised the input signal to 910 mVrms at the generator, the Power Supply current goes up around 0.1 A and -0.1 A, the voltage across R503 raise above 3 VDC and the distortion came in. See second picture.
I didn't removed D1 yet seeing that I cannot find a substitute for C2
I noticed that the distortion above came later if I turn the RV1 BIAS trim-pot to it's maximum value, So I though maybe the total resistor value of R6+RV1 wasn't high enough. I replaced R6 from 620 Ohms to 1.1 kOhms. Then I lowered the rails voltage from +-35 VDC to +/-25 VDC for security. With these I am able to raise much more the input voltage before getting a signifiant distortion. Also the distortion is now flat and symetrical. Even if I raise the input voltage above, the current from the Power Supply is stable at 40 mA and the voltage across R503 is also stable around 600mV. No heat on the finger from all parts, including the heat-sink of the two Q304 and Q404 back to back.
It may appear better, it malfunctions nonetheless. It should be able to put out a distortion-free swing between -30V to +30V before it clips at +/35V rail voltage.
At 1KHz D1 is unlikely an issue.
It appears the output stage over-current protection did not kick in, and therefore has nothing to do with the early clipping.
Since I ran out of suspects, I have to ask if you got your 2SB649/D669 from a reputable source, and suggest temporarily replace them with 2N5401/5551 and see if things change.
You earlier tests at a few incremental rail voltage showed consistent peak output increase with raised rail voltage at lower rail voltages. That increase stopped when the peak hit 20V. Something has put a limit to the output peak. Today, you also noticed the rail current shoots up when the clipping took place, and that would be consistent with what happens when a transistor breaks down under voltage greater than its rating.
At 1KHz D1 is unlikely an issue.
It appears the output stage over-current protection did not kick in, and therefore has nothing to do with the early clipping.
Since I ran out of suspects, I have to ask if you got your 2SB649/D669 from a reputable source, and suggest temporarily replace them with 2N5401/5551 and see if things change.
You earlier tests at a few incremental rail voltage showed consistent peak output increase with raised rail voltage at lower rail voltages. That increase stopped when the peak hit 20V. Something has put a limit to the output peak. Today, you also noticed the rail current shoots up when the clipping took place, and that would be consistent with what happens when a transistor breaks down under voltage greater than its rating.