update
Update :
I shorted Z1 and Z2 in my previous schematic and it changed something. I have no more RF oscillations all the way up to 10MHz. The trace looks always clean (I mean without RF oscillations).
I replaced the Cascode Vrefs Zeners Zcc1 and Zcc2 by LM4040 4,096V shunt regulators. This didn't change anything except the DC setting is now more balanced and stable (with zeners one was 3,7V and the other 3,5V and not really stable).
I removed the TMC compensation again to a single miller cap but leaving the 10 ohm resistor in series with it. This reduces the ringing on top of the squarewave signals quite a lot but it is not completely removed.
I don't like this ringing , it also overshoots a little bit ... why? I would have expected this with TPC compensation but not witrh TMC and certainly not with simple miller compensation.
My signal gets still visually worse at 140-150KHZ (this means more than 1% thd I guess maybe even 5% ... don't have thd analyser here).
According to the simulation my HF rolloff point should be around 800KHZ but in reality it is only 380KHZ this is a significant difference that i dislike ... were it 700KHZ i wouldn't consider it an issue ... but this ?
I still believe this amp has an issue and I cannot lay my hands on the problem... all i do is get it worse or better by changing this or that but never really adressing the undrlying problem ...
Help would be heavily appreciated ...
Olivier
Update :
I shorted Z1 and Z2 in my previous schematic and it changed something. I have no more RF oscillations all the way up to 10MHz. The trace looks always clean (I mean without RF oscillations).
I replaced the Cascode Vrefs Zeners Zcc1 and Zcc2 by LM4040 4,096V shunt regulators. This didn't change anything except the DC setting is now more balanced and stable (with zeners one was 3,7V and the other 3,5V and not really stable).
I removed the TMC compensation again to a single miller cap but leaving the 10 ohm resistor in series with it. This reduces the ringing on top of the squarewave signals quite a lot but it is not completely removed.
I don't like this ringing , it also overshoots a little bit ... why? I would have expected this with TPC compensation but not witrh TMC and certainly not with simple miller compensation.
My signal gets still visually worse at 140-150KHZ (this means more than 1% thd I guess maybe even 5% ... don't have thd analyser here).
According to the simulation my HF rolloff point should be around 800KHZ but in reality it is only 380KHZ this is a significant difference that i dislike ... were it 700KHZ i wouldn't consider it an issue ... but this ?
I still believe this amp has an issue and I cannot lay my hands on the problem... all i do is get it worse or better by changing this or that but never really adressing the undrlying problem ...
Help would be heavily appreciated ...
Olivier
update2
I just soldered off my millercaps of 100pF since I guessed my charge current for them is too high and the distortion came from current stravatioon in the IS... (?)
result? -> I just powered it up w/o any miller caps. This should be unstable like hell right? Well it was stable !? Signal distortion (visual) starts from 180KHZ instead of 140KHZ ... only thing is that it now shows signs of RF oscillations when input reaches 400KHZ over 400KHZ the ouput shows a large gain which is sign of nyquist instability or phasemargin down to 0 or positive feedback. However i would have expected a total uncontrolable amp instead it was more stable than believed at first.
What should i think of this? Is this just normal or is there something in the circuit that declares this behavior and maybe explains my problem?
I will just check what it does with square too .... let me see ... hm this is less nice ..; at 100KHZ squar it goes all crazy. At 10KHZ it is still bad ... at 1KHZ square the upper side is stable but with quite an overshoot spike (very short and no ringing) the lower side shows RF oscillations...
I continue ...
I just soldered off my millercaps of 100pF since I guessed my charge current for them is too high and the distortion came from current stravatioon in the IS... (?)
result? -> I just powered it up w/o any miller caps. This should be unstable like hell right? Well it was stable !? Signal distortion (visual) starts from 180KHZ instead of 140KHZ ... only thing is that it now shows signs of RF oscillations when input reaches 400KHZ over 400KHZ the ouput shows a large gain which is sign of nyquist instability or phasemargin down to 0 or positive feedback. However i would have expected a total uncontrolable amp instead it was more stable than believed at first.
What should i think of this? Is this just normal or is there something in the circuit that declares this behavior and maybe explains my problem?
I will just check what it does with square too .... let me see ... hm this is less nice ..; at 100KHZ squar it goes all crazy. At 10KHZ it is still bad ... at 1KHZ square the upper side is stable but with quite an overshoot spike (very short and no ringing) the lower side shows RF oscillations...
I continue ...
Hi Edmond, good to hear from you !
No I did not implement a tiny OPS yet ... Could I use one of my VAS Q's for that since I have no output devices yet (only heavy gauge Mosfets)? It will be difficult however because my board did not expect for it ... but I will do some flyby wiring... do you think this is essential to my issue?
I will just leave a signle miller cap for now ... I could still do the TPC (not done yet) because the pcb is accomodated for TMC and TPC (depending on which resistor you install).
When you see the traces is there something that strikes you?
I will send the MC9 file in the next reply. It is exactly how it is in front of me right now (except I placed 47pF miller caps on the schematic and not yet onto the pcb)
Ps : I also test without any load. The output of the VAS goes straight into the feedback circuit and back to the bases of the IS)
Olivier
No I did not implement a tiny OPS yet ... Could I use one of my VAS Q's for that since I have no output devices yet (only heavy gauge Mosfets)? It will be difficult however because my board did not expect for it ... but I will do some flyby wiring... do you think this is essential to my issue?
I will just leave a signle miller cap for now ... I could still do the TPC (not done yet) because the pcb is accomodated for TMC and TPC (depending on which resistor you install).
When you see the traces is there something that strikes you?
I will send the MC9 file in the next reply. It is exactly how it is in front of me right now (except I placed 47pF miller caps on the schematic and not yet onto the pcb)
Ps : I also test without any load. The output of the VAS goes straight into the feedback circuit and back to the bases of the IS)
Olivier
Hi Olivier,
VAS trannies are okay, but please pay some attention to the quiescent current (about 10mA or so). Just sim it first to find the right component values.
This tiny dummy OPS is really important, as otherwise the VAS is loaded by the FB resistor, which disturbs the whole setup.
Regarding the MC9, I can't guarantee that I've enough time to look at seriously in the coming days. You know X-mas and N-Y, busy days!
Cheers,
E.
VAS trannies are okay, but please pay some attention to the quiescent current (about 10mA or so). Just sim it first to find the right component values.
This tiny dummy OPS is really important, as otherwise the VAS is loaded by the FB resistor, which disturbs the whole setup.
Regarding the MC9, I can't guarantee that I've enough time to look at seriously in the coming days. You know X-mas and N-Y, busy days!
Cheers,
E.
The basic flaw in this design has been unrelentlessly pointed,
as it lies on the insufficent voltage drop in the current mirrors
to reach the threshold of the darlington vas, that is, 1.2V plus
the voltage drop at the vas voltage gain stage emitter resistor,
thus, the standing current is undefinied although negative feedback
will allow to rapidly get out of this non linear zone, but at the expense
of stability and performances.
Since you didnt adress this fundamental issue, you embarked
on complexe means to corner the problem by the outside ,
while it can be only resolved by getting rid of the obstacle..
Here a schematic wich is based on this exact topology,
hope it will help..
I havent built it, but the sims show that standing
current is defined and easyly adjustable by setting
the relevant components , while overall behaviour
show no components dependance for the global functionning.
47PF
Edmond,
I will start to construct a tiny OPS right away.
Please enjoy X-mass and N-Y as it should be !
I just would like to know where C101/R105 and C102/R505 stand for in the CMCL circuit? As I removed them it showed better HF stability with quite some impact. What are their importance? Why is it better w/o? Or not? It is possible I just nulled a nasty ground loop by removing them and thus losing their benefit too ... I don't know that. You can find them on the initial schematic a few posts earlier.
Another point is the zeners I had to bypass for the same reason (Z1 & Z2). They are there to prevent some Q's going to saturation in case of a failure. They are the ref voltage as the one setting up the VAS Cascode Q's. You rather opted to not install cascode Q's but it should work with them too right? I cannot return to a situation without them since they are heatsinked and the others not because they are isolated from hv swings by the QCC's. It could work for some minutes to see what happens but I would rather like to do this as last resort. Is it dramatic to bypass those zeners? I tried also replacing them with LM4040 precision Vrefs. Same thing. However the pos side did not show difference the lower side showed RF when applied ... it left as soon the device was bridge by wire.
From my previous tests it seems my lower side is more tricky than the upper side but this is mere hazard... lots of parameters PNP, NPN, ...
I now tested with 47pF miller caps and i must say it looks nice. See pics in annexe.
What I still don't like is the seeming spike on the squarewave. I want my signal to go UP and then bend nicely to horizontal w/o spike, dips, ringing, blubber and other artifacts ...
It never shows RF all the way to 10MHZ
LF roll-off (-3dB) : 1 HZ
HF roll-off (-3dB) : 550 KHZ (there seems to be some some -3V dc offset here ? This is not the case on lower freq, it starts when amp starts rolls off at HF)
SLEWRATE : 65V/us
Wahab, I know this circuit in its basics is doomed but I think it is a promising one if we can get rid of her main flaw. I will check your circuit right away and let you know ...
Greetz
Olivier
Edmond,
I will start to construct a tiny OPS right away.
Please enjoy X-mass and N-Y as it should be !
I just would like to know where C101/R105 and C102/R505 stand for in the CMCL circuit? As I removed them it showed better HF stability with quite some impact. What are their importance? Why is it better w/o? Or not? It is possible I just nulled a nasty ground loop by removing them and thus losing their benefit too ... I don't know that. You can find them on the initial schematic a few posts earlier.
Another point is the zeners I had to bypass for the same reason (Z1 & Z2). They are there to prevent some Q's going to saturation in case of a failure. They are the ref voltage as the one setting up the VAS Cascode Q's. You rather opted to not install cascode Q's but it should work with them too right? I cannot return to a situation without them since they are heatsinked and the others not because they are isolated from hv swings by the QCC's. It could work for some minutes to see what happens but I would rather like to do this as last resort. Is it dramatic to bypass those zeners? I tried also replacing them with LM4040 precision Vrefs. Same thing. However the pos side did not show difference the lower side showed RF when applied ... it left as soon the device was bridge by wire.
From my previous tests it seems my lower side is more tricky than the upper side but this is mere hazard... lots of parameters PNP, NPN, ...
I now tested with 47pF miller caps and i must say it looks nice. See pics in annexe.
What I still don't like is the seeming spike on the squarewave. I want my signal to go UP and then bend nicely to horizontal w/o spike, dips, ringing, blubber and other artifacts ...
It never shows RF all the way to 10MHZ
LF roll-off (-3dB) : 1 HZ
HF roll-off (-3dB) : 550 KHZ (there seems to be some some -3V dc offset here ? This is not the case on lower freq, it starts when amp starts rolls off at HF)
SLEWRATE : 65V/us
Wahab, I know this circuit in its basics is doomed but I think it is a promising one if we can get rid of her main flaw. I will check your circuit right away and let you know ...
Greetz
Olivier
Attachments
Wahab,
I checked your circuit. If i am correct the diodes in the IS serve to bring down the output voltage of the IS far enough in order to set the current in the vas depending on this oitput voltage. This is perfectly predictable... as iit should be ... but does the circuit still get the benefits from the current mirrors? I once read someone degenrating the mirror Q's so much that the emitter voltage is significantly lower then the rail voltage (usualy just a few hundred mV but they dropped say 2V) ... it worked stable but the benefit of the current mirror was gone... don't know if this is the case here?
But something else :
I notice that the darlington VAS Q's have their collector not direct to ground as I have ... but they pass a resistor (cannot read value) which is in turn shunted by a 100pico cap... HF stability adressing? RF adressing or something else?
Thnx
Olivier
I checked your circuit. If i am correct the diodes in the IS serve to bring down the output voltage of the IS far enough in order to set the current in the vas depending on this oitput voltage. This is perfectly predictable... as iit should be ... but does the circuit still get the benefits from the current mirrors? I once read someone degenrating the mirror Q's so much that the emitter voltage is significantly lower then the rail voltage (usualy just a few hundred mV but they dropped say 2V) ... it worked stable but the benefit of the current mirror was gone... don't know if this is the case here?
But something else :
I notice that the darlington VAS Q's have their collector not direct to ground as I have ... but they pass a resistor (cannot read value) which is in turn shunted by a 100pico cap... HF stability adressing? RF adressing or something else?
Thnx
Olivier
Thank you. The same to you!
I can't locate them, but I think you mean the 100pF caps and 100R resistors at the top respectively bottom on the schematic, between the supply rail and the VAS input, right?
Well, these were optional components to tame the internal Miller loops, in case the were instable. If the amp is stable, no need to install them.
Wait, correction!. Now I find them. Indeed they set the roll-off of the CMCL circuit. I'm amazed that they have such an impact, because they only have a common mode effect (thus not differential, i.e. don't affect the audio signal). Just like R102 and R103, you can connect them to a dirty ground. Anyhow, I wouldn't recommend to delete them.
Indeed, as you already remarked, these Zeners set the correct input voltage of the VAS, in order to prevent saturation of the VAS trannies (during clipping). BTW, the higher the voltage of the cascode Zeners, the higher the voltage of Z1 & Z2.
But first, please install the dummy OPS and see what happens. Without this dummy the front-end is derailed anyhow.
Regards,
Edmond.
Hi, Olivier
You got it right , the diodes allow a voltage shift of the IPS output
dc level, while letting the current mirrors work as they are supposed to.
This doesn t change at all the dynamic behaviour of the CMs contrary
to a degeneration in the said CMs emmitters.
A byproduct of this is that with rising temperature, the diodes gap
voltage decrease lightly , but this has no consequence as since there s
two junctions drived by this voltage , two diodes will compensate
for the VAS and its buffer , letting only one remaining diode that is not
compensated, though this will allow a very slight deacrease of the vas
current with temp. increase , a perhaps beneficial purpose for the global
temp stability, but i didn t check yet this aspect.
Whatever, as the vas are now both conducting with stabilised currents,
the feedback loop can act properly without transiting in non linear domain.
To reduce the vas buffer thermal dissipation, i did add a resistor
in serie with its collector , which has the undesirable byproduct of
making the ac voltage appear in the said collector , which is annoying
at high frequencies thanks to this device Cob.
So the caps short the collector to ground at high frequencies.
cheers,
w
i agree, with shorts on q5 and q8 it'll never work. you can't tie 2 pins on a transistor and expect it to work, there's no way.
Hi, Kbeist
The vas is already a darlington, i.e, a buffered vas.
Or did i miss something ?.....
Wahab,
I wonder how your circuit with the diodes in the IS leg compares to the one I am building for the moment? Yours seems simpler but what about the rest? I guess it's hard to tell. Anyways I will write out all my findings on this one ...
Anyway, on the DC side it is very very steady. Clipping behavior is very nice too. While powering down the amp you see the signal smoothly clipping down to zero w/o crazyness. Same thing when it's powered up, signal grows until it is full swing and out of clipping.
I just can't tell why I am stuck with overshoot in my squarewave ? I need to find out...
Is it normal for instance that a sine wave starts to look quite like a sawtooth wave (more sawtooth than sine) when the freq is around 250KHZ? I mean ... is this :
a) Off course it is ... no amp could do better (at 100Vpp )
b) It is quite normal , don't worry , the amp is more than very good
c) mmm it is normal but it could be better
d) mmm it is normal if the amp is rather low end
e) Grrr only poor amps do
f) Did you drink while soldering the pcb?
Good night
Olivier
I wonder how your circuit with the diodes in the IS leg compares to the one I am building for the moment? Yours seems simpler but what about the rest? I guess it's hard to tell. Anyways I will write out all my findings on this one ...
Anyway, on the DC side it is very very steady. Clipping behavior is very nice too. While powering down the amp you see the signal smoothly clipping down to zero w/o crazyness. Same thing when it's powered up, signal grows until it is full swing and out of clipping.
I just can't tell why I am stuck with overshoot in my squarewave ? I need to find out...
Is it normal for instance that a sine wave starts to look quite like a sawtooth wave (more sawtooth than sine) when the freq is around 250KHZ? I mean ... is this :
a) Off course it is ... no amp could do better (at 100Vpp )
b) It is quite normal , don't worry , the amp is more than very good
c) mmm it is normal but it could be better
d) mmm it is normal if the amp is rather low end
e) Grrr only poor amps do
f) Did you drink while soldering the pcb?
Good night
Olivier
Those in the know? Let me tell you buddy, I have my degree in electronics, computer repair, and now i'm going for my telecommunications degree, not to mention I build my own things too, and I have 17 years in it. For those in the know. Ha
OPSat
Edmond,
I inserted the dummy OPS.
Very simple one.
VAS + output to base of upper OPS Q
VAS - output to base of lower OPS Q
Collectors of OPS Q's to corresponding rails
Emitters tied together with 2 x 47 ohms in between.
Feedback pickup from between those resistors.
IQ ops 6mA
(found some Q's for it 2SA1930 and 2SC5171)
At first glance it does not change too much. Since it is flyby wire it is touchy when some wires get close to one another ... however it works and that what counts.
I reinserted the input cap of 2200pF because I wonder if my amp is not picking up just rubbish in the air or maybe my function generator is not good enough at square waves...
But did you check the NO SIGNAL trace on the scope (see pictures previous posts) I always have some 20-30mV rubbish when no input is connected or the generator is OFF. Is this acceptable? Outrageous? What's more of a problem is I think the freq. If it were just very very HF noise unable to trigger on my 40mhz scope i would say the amp is going to ignore it because it can't amplify those freq. But I have quite some period wave (not realy sine) ... If i measure between what seems to be a period it comes down to 500-600 KHZ ... well in the bandwidth of the amp ... Is that the ringing I see?
What I see on the 100KHZ square wave now is not really an overshoot but when the signal reaches nearly to the top there seems a higher freq component added to it...it bends to horizontal still with the higher freq on top of it which goes away when reaching 1/3 of the square (top) idem on the lower side...
Without signal I also probed on the pcb at many points with the scope set to AC and scaled to 50mV... well the noise i told about is not only at the output but it is rather on every node. GND, SIG GND, RAILS, between components, just everywhere. There is no point on the PCB that does not reflect this signal except for non connected items like island ...
On the input side that signal is about 60% of the value of that on the output. Don't know if this is relevant?
Another strange thing is that I have a shield on midlayer 1 ... it is not connected on its output terminal. Because grounding it did not seem OK. If I am right it does not touch anything on the pcb not a single coponent ... yet that noise trace is visible also there ...
I will post it now but let me check on the pcb too if it touches anyway somewhere ....
greetz
Olivier
Edmond,
I inserted the dummy OPS.
Very simple one.
VAS + output to base of upper OPS Q
VAS - output to base of lower OPS Q
Collectors of OPS Q's to corresponding rails
Emitters tied together with 2 x 47 ohms in between.
Feedback pickup from between those resistors.
IQ ops 6mA
(found some Q's for it 2SA1930 and 2SC5171)
At first glance it does not change too much. Since it is flyby wire it is touchy when some wires get close to one another ... however it works and that what counts.
I reinserted the input cap of 2200pF because I wonder if my amp is not picking up just rubbish in the air or maybe my function generator is not good enough at square waves...
But did you check the NO SIGNAL trace on the scope (see pictures previous posts) I always have some 20-30mV rubbish when no input is connected or the generator is OFF. Is this acceptable? Outrageous? What's more of a problem is I think the freq. If it were just very very HF noise unable to trigger on my 40mhz scope i would say the amp is going to ignore it because it can't amplify those freq. But I have quite some period wave (not realy sine) ... If i measure between what seems to be a period it comes down to 500-600 KHZ ... well in the bandwidth of the amp ... Is that the ringing I see?
What I see on the 100KHZ square wave now is not really an overshoot but when the signal reaches nearly to the top there seems a higher freq component added to it...it bends to horizontal still with the higher freq on top of it which goes away when reaching 1/3 of the square (top) idem on the lower side...
Without signal I also probed on the pcb at many points with the scope set to AC and scaled to 50mV... well the noise i told about is not only at the output but it is rather on every node. GND, SIG GND, RAILS, between components, just everywhere. There is no point on the PCB that does not reflect this signal except for non connected items like island ...
On the input side that signal is about 60% of the value of that on the output. Don't know if this is relevant?
Another strange thing is that I have a shield on midlayer 1 ... it is not connected on its output terminal. Because grounding it did not seem OK. If I am right it does not touch anything on the pcb not a single coponent ... yet that noise trace is visible also there ...
I will post it now but let me check on the pcb too if it touches anyway somewhere ....
greetz
Olivier
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.