No surprise that it is oscillating – the frequency compensation is waaaayyyyyy off. Firstly, that 330R/15pF (R15/C11) is a BAD idea – it is causing your loop gain roll-off to plateau above unity gain.
With R15 and C11 removed things are much improved, but the unity loop gain frequency is still way too high at almost 2MHz. The problem here is that your emitter degeneration on the LTP is too light.
Delete both R15 and C11 and increase R1 and R2 from 30 ohms to 150 ohms. That should give you a stable amp.
Also, 10 ohms is a little small for the degeneration on the current mirror – 100R would be better.
With R15 and C11 removed things are much improved, but the unity loop gain frequency is still way too high at almost 2MHz. The problem here is that your emitter degeneration on the LTP is too light.
Delete both R15 and C11 and increase R1 and R2 from 30 ohms to 150 ohms. That should give you a stable amp.
Also, 10 ohms is a little small for the degeneration on the current mirror – 100R would be better.
Last edited:
Thanks for the replies!
khodeir
Of course no oscillation will penetrate the output inductor, but this is not dealing with the problem, but with the consequence.
GK
The emitter degeneration on the LTP was prevously changed to 100R, but this did not help too. I also have tried removing R15/C11, did not work. I have modified the circuit according to your suggestions. Also grounded the input, just in case. Result: faster oscillation.
Schematics
Waveform
Regards,
Rait
khodeir
Of course no oscillation will penetrate the output inductor, but this is not dealing with the problem, but with the consequence.
GK
The emitter degeneration on the LTP was prevously changed to 100R, but this did not help too. I also have tried removing R15/C11, did not work. I have modified the circuit according to your suggestions. Also grounded the input, just in case. Result: faster oscillation.
Schematics
Waveform
Regards,
Rait
That's the Zobel, already in place (at least in the schematic).
Rait, you said when you touch the heatsink the oscillation changes - are the heatsinks grounded? They should be and if not, this can be a contributing factor.
Are you still using the 2 pole comp?
It's very unlikely that the board layout is causing the oscillation. I was with Glen, I thought that the lead comp in parallel with your feedback was causing the problem so I'm surprised that it didn't help.
Is it connected to a load when it oscillates?
Seriously, make the degeneration 150R and DONT put R15/C11 back. It looks like there are multiple issues here. If oscillations persist then you probably have an issue with parasitic oscillation; but the loop compensation has to be right before you can work on eliminating that.
Attached are the loop gain sims for an ideal front-end starting with your circuit values - the first plot. As you can see, the loop gain does not go through unity until after 10MHz! The plateu in the plot above ~1MHz is caused by that R15/C11 combination - this will never be stable.
The second plot is with R15/C11 removed. Now things look good, but the unity loop gain frequency is still way too high for a bipolar output stage, at 2MHz.
The 3rd plot is with both R15/C11 removed and the emitter degeneration increaded to 150 ohms -that would be comfortably stable.
Attachments
Is not that what he has already done? The schematic of his current setup in post 164 shows 150R LTP degeneration and the RC link removed. Feedback is taken directly from the VAS too without the OPS connected. Grounding the heatsinks properly could be a good idea, like MJL21193 says, but it may not be enough to solve the oscillation problem.
Well it would appear so. I just thought it might be constructive to actually show the effects of altering the frequency compensation components and why the previous configuration was unworkable.
Anyway, the 10MHz oscillation with the fixed loop compensation is definitely parasitic oscillation in the output stage.
If I read the scope horizontal scale correctly is 50ns per div = circa 20MHz oscillation. Is this the complete circuit? If it is, I get the feeling you may have layout problems because of the very high frequency of oscillation.
Any reason why you are running the diff amp tail current at 3.3mA? quite low in my view (but this is not the cause of your oscillation -just an observation.)
Of course, Glen's suggestions should be followed first before trying to sort these other issues out - you definetly have a few things going on here.
Any reason why you are running the diff amp tail current at 3.3mA? quite low in my view (but this is not the cause of your oscillation -just an observation.)
Of course, Glen's suggestions should be followed first before trying to sort these other issues out - you definetly have a few things going on here.
But it still oscillates:
(CH1, ~10 MHz, 4 Vpp)
When i touch the heat sink, then the oscillation changes.
Heat sink is not conncted in any way so this is some capacitive effect. Is it possible the problem lies only in pcb design or is the schematics causing the instability. Could the LONG feedback track cause the problem? Can anyone comment the pcb design?
Without a clue,
Rait
Tere Rait, kuidas käsi käib?
The high frequency of the oscillation make me think you might have some local oscillation.
Your PCB design is not very compact and have some questionable layout but let's try to solve the problem.
Also the use of wire wound emitter resistor for the output transistor can add up to the problem as they might have a not wanted inductance depending on the wounding technic used.
You could measure with your oscilloscope on the emitter of the positive half and compare with the amplitude of the emitter of negative half because some times only half of the output stage is the cause of oscillation.
Also check the Collector side and compare upper and lower half, you might think it might be DC only but oscillation can still be very clearly visible, check now which side has the highest oscillation amplitude.
What I would do then is try to experiment with a small capacitor around 10-220 nF and connect it with shortest path from the heatsink close to your output transistor to the GND plane on your PCB, (you might perhaps do it on both sides but try first with one side at a time and depending if you saw oscillation more dominantly on onside try to concentrate on that side first).
Also you could try connecting the small capacitor from from the heatsink between Q17 and Q18 to the nearest positive voltage rail on PCB (again try it on one side first and then the other side).
Cheers Michael
Last edited:
If I understood his post correctly he has disconnected the output stage and is taking the feedback directly from the VAS. If it still oscillates like that, the emitter resistors can't be the culprit. I'd rather suspect those floating heatsinks and/or that something is happening in the buffered VAS.
While looking at the board layout I discovered that the collector of Q7 is connected to input ground instead of decoupling ground. This detour to the power supply and back could be a problem; there is essentially an inductor (the PSU wiring) in series with the collector of Q7 before it finds it way back to something that is a proper HF ground for the VAS circuit. My suggestion, apart from grounding the heatsinks as has already been suggested, is to try either moving the collector connection to decoupling ground or strapping the two grounds together low-inductively on-board. A collector resistor on Q7 might also work if this is what is causing the oscillation.
While looking at the board layout I discovered that the collector of Q7 is connected to input ground instead of decoupling ground. This detour to the power supply and back could be a problem; there is essentially an inductor (the PSU wiring) in series with the collector of Q7 before it finds it way back to something that is a proper HF ground for the VAS circuit. My suggestion, apart from grounding the heatsinks as has already been suggested, is to try either moving the collector connection to decoupling ground or strapping the two grounds together low-inductively on-board. A collector resistor on Q7 might also work if this is what is causing the oscillation.
Last edited:
Hey!
It seems we have a winner, grounding the beta enchancement transistor to the decoupling ground stopped the oscillation!
Thanks for the help everybody!
First, I must excuse for the confusion, i have disconnected the output stage, so the fault was not there.
MJL21193
Heatsinks are not grounded, but I'll ground them! Compendation is 1 pole, as in schematics.
GK
The changes suggested were made, as was shown on the schematics. Thanks for explaining your points! Very nice graphs! Which software do you use for that?
Bonsai
Is the tail current really that low? Designs I have seen are using even lower setting, e.g. Fotios's amp shown in this thread few pages back.
Ultima Thule
Hyvää päivää! I know, that the PCB is not the best one out there, but it's my third amp (previous two were chip amps) and I hope next design will be better! If needed, this board will be redesigned. I'll try your suggestions when I connect the output stage.
megajocke
Thanks fof noticing Q7-s grounding. Used a wire to connect it to decoupling ground and seems that oscillation is gone. Is it a good idea to keep two grounds separate?
Thanks,
Rait
PS! I'll connect the output stage now and see, if the stability remains untouched.
It seems we have a winner, grounding the beta enchancement transistor to the decoupling ground stopped the oscillation!
Thanks for the help everybody!
First, I must excuse for the confusion, i have disconnected the output stage, so the fault was not there.
MJL21193
Heatsinks are not grounded, but I'll ground them! Compendation is 1 pole, as in schematics.
GK
The changes suggested were made, as was shown on the schematics. Thanks for explaining your points! Very nice graphs! Which software do you use for that?
Bonsai
Is the tail current really that low? Designs I have seen are using even lower setting, e.g. Fotios's amp shown in this thread few pages back.
Ultima Thule
Hyvää päivää!
I know, that the PCB is not the best one out there, but it's my third amp (previous two were chip amps) and I hope next design will be better! If needed, this board will be redesigned. I'll try your suggestions when I connect the output stage.megajocke
Thanks fof noticing Q7-s grounding. Used a wire to connect it to decoupling ground and seems that oscillation is gone. Is it a good idea to keep two grounds separate?
Thanks,
Rait
PS! I'll connect the output stage now and see, if the stability remains untouched.
To not caused again a misunderstanding I must point out that in mine design any increase of tail current above 3mA does not offers any decrease of rise time (or increase of slew rate). Instead, the LTP transistors are overheated and can be burned very easy in any input overloading. The rise time it is restricted from the dirty MJE15033 VAS transistors because their big Cob (i think 30pF). GK calculated a slew rate of 30V/μs, I insisted a slew rate of 40V/μs. The related measurement obtained with all feedbacks closed (included the VI limiter, the protection limiter etc.) Never mind.
The issue is what use have in your mind for this amplifier. For example, me I had a clear intention to use my amplifier for driving P.A. speakers.
P.A. speakers are extremely sensitive from 300Hz and above. Instead, from 40 up to 250Hz to produce the same SPL with good control, those need big amounts of power and a very low output impedance of amplifier. For this reason i use 7 pairs of output transistors, to drop down the Zo, and not so much for sharing-out the total power. The lower the Zo, the higher the damping factor which ensures a very tight bass for 12"-18" PA speakers of hard suspension, loaded either with large horns or band-pass chambers.
To be more precise, i did the circuit implementation based on the design of Peavey CS1200X and by knowing its performance in practice. Peavey CS1200X uses a more sophisticated input circuitry from mine. Its LTP tail current it is 2,4mA and uses separate VAS for the positive and negative signal swing, which are the obsolete MPSU-10 and MPSU-60 with Cob=3pF. Although its circuit - from a first view - shows that it can obtain a very high slew rate, finally it is restricted just in 40V/μs. Rise time figures of mine amplifier and CS1200X are similar. The extremely sensitive Mid - High horns of PA speakers, can produce enough SPL driven only from the harmonics of signal!!! So, it is reasonable the restriction of slew rate.
Summarizing, the main possibility of PA amplifiers it is focused on the good reproduction in low frequencies.
For home use, with insensitive domestic speakers, there is a big difference as for the amplifier implementation.
Regs
Fotios
I'm a bit surprised that my suggestion actually worked! I'm glad to be able to help. Keeping those grounds separate can be useful in avoiding ground loops as it makes the input pseudo-balanced where common mode signals are only amplified by 1 instead of by the gain of the amplifier. Many designs use a ~10 ohm resistor between the two grounds, some also a capacitor in parallell. Doing that is probably a good idea as it makes sure there is a relatively low-impedance path at high frequencies while preventing hum from ground loops.
MJL21193: Thanks for the kind words!
AAAAAAAAAAAAAAAAAAAAARGH!
Connected the output stage, flipped the power switch and ...
... some weird lines on the oscilloscope, little flash as the fuses burned. Quick measurement - both legs shorted. And I had all 5 pairs connected...
Desoldered the trannies today, to great releaf only one pair has gone to Valhalla, others seem to be OK. Will be testing tomorrow, what caused this. Then I'll try stage by stage, first predrivers, then drivers and then outputs.
Thanks for the explanations fotios and megajocke, again something to consider in the future!
Until then,
More Power!
Rait
Connected the output stage, flipped the power switch and ...
... some weird lines on the oscilloscope, little flash as the fuses burned. Quick measurement - both legs shorted. And I had all 5 pairs connected...
Desoldered the trannies today, to great releaf only one pair has gone to Valhalla, others seem to be OK. Will be testing tomorrow, what caused this. Then I'll try stage by stage, first predrivers, then drivers and then outputs.
Thanks for the explanations fotios and megajocke, again something to consider in the future!
Until then,
More Power!
Rait
AndrewT's suggestion of using a light bulb tester is a very good one. Resistors instead of rail fuses is another possibility for minimizing damage in case of problems. (but watch out for shrapnel if they explode )
Some suggestions if you have local oscillation in the output stage:
* Try connecting input stage decopling and output stage decoupling ground.
* Try base resistors for the driver transistors. Many commercial designs that use large driver transitors have around 10 ohms in this position.
)Some suggestions if you have local oscillation in the output stage:
* Try connecting input stage decopling and output stage decoupling ground.
* Try base resistors for the driver transistors. Many commercial designs that use large driver transitors have around 10 ohms in this position.
AAAAAAAAAAAAAAAAAAAAARGH!
Connected the output stage, flipped the power switch and ...
... some weird lines on the oscilloscope, little flash as the fuses burned. Quick measurement - both legs shorted. And I had all 5 pairs connected...
Desoldered the trannies today, to great releaf only one pair has gone to Valhalla, others seem to be OK. Will be testing tomorrow, what caused this. Then I'll try stage by stage, first predrivers, then drivers and then outputs.
Thanks for the explanations fotios and megajocke, again something to consider in the future!
Until then,
More Power!
Rait
Now I'm confused. When you said that the oscillation problem was fixed, was that with or without an output stage connected? From where are you picking off the nfb now (speaker output or the VAS collector?).
A quickly blown power output device usually means local oscillation in the output stage itself. Also, a power output stage can quite happily oscillate all by itself (collpits oscillator), within the global feedback loop or not.
I did those loop gain sims in LTspice.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.