MikeB said:
Mike,
I asked both questions in my earlier post, and you answered "I checked output from squarewavegenerator, rise/fall was barely visible on scope. I will check more details the next days.
"
How did it go?
I interpret your answer as the slewrate of generator square wave signal to have enough steep flanks.
You mentioned in post #63 the scope timedivision setting was 1uS, is it the smallest timedivision?
If you have 1 volt with such a steep flank from your generator you can't see any angle with 1 uS timedivision you have sufficient slewrate on your squarwave from generator, but the question is still how it looks like at the diff stage trany.
BTW I think you have very stable squarewave from your amplifier, there is not a single sign of ringing, which is not uncommon.
BTW, I checked again your schematic, you have a trim resistor at the input attenuating your generator output signal level!?
Is this e real life setup, and what is the value of it?
Regards Michael
Hi !
I will do more checks the next days, i had not too much time yet...
Michael, the smallest timestep on the scope seems to be 0.05uS per div,
the selector goes down to 0.01, but the last two steps did not show
any change...
I think the generator is far from perfect, but i had to build a generator
in short time, and for 3 bjts it's not bad i think. It's a modified
multivibrator with a gainstage followed.
The trim at input is the volumecontrol, i normally connect the amp
directly to the cd-player.
What i've meant with rise/fall barely visible is that the flanks are
very steep.
In diffamp i can only measure the inputs, the output does not
really contain voltageswing due to the cascoded nature.
But as the scope has 20pF inputcapacitance, these measurements
might not be very accurate.
But i could measure the currents through vas at the lower 150ohms.
Mike
I will do more checks the next days, i had not too much time yet...
Michael, the smallest timestep on the scope seems to be 0.05uS per div,
the selector goes down to 0.01, but the last two steps did not show
any change...
I think the generator is far from perfect, but i had to build a generator
in short time, and for 3 bjts it's not bad i think. It's a modified
multivibrator with a gainstage followed.
The trim at input is the volumecontrol, i normally connect the amp
directly to the cd-player.
What i've meant with rise/fall barely visible is that the flanks are
very steep.
In diffamp i can only measure the inputs, the output does not
really contain voltageswing due to the cascoded nature.
But as the scope has 20pF inputcapacitance, these measurements
might not be very accurate.
But i could measure the currents through vas at the lower 150ohms.
Mike
Mike,
Im just laughing when you say your generator is far from perfect BUT you can't see the flanks even with timedivision of 0,05 uS???
You have a radio transmitter!
btw I checked several scopes we have around where I am at the moment and theres several 20 MHz scopes of diffrent brands and the smallest timedivision is 0,2 uS on all, you must have a quite fast scope there with time division down to 0,01 uS?!
Well, the question about your attenuator at the input, the resistance will have an impact on the LF filter formed together with the 100 pF and input capacitance of the diff stage.
If the probe is 20 pF I think it won't have too much impact on your meassurements...
If you are going to measure voltage over the Re, it can be a problem if you are going to connect the earth of the probe to the high impedance CCS, be carefull.
Edit: One mor thing I made a fast calculation, you have about 2,2 mA through the VAS if I made the calculation right...
Anyhow my guess is that with doubble darlington you have bigger Miller load with MJE 15030/1 for the VAS than with tripple darlington 2n5551/401, which perhaps make the speed diffrence.
Cheers
Im just laughing when you say your generator is far from perfect BUT you can't see the flanks even with timedivision of 0,05 uS???
You have a radio transmitter!
btw I checked several scopes we have around where I am at the moment and theres several 20 MHz scopes of diffrent brands and the smallest timedivision is 0,2 uS on all, you must have a quite fast scope there with time division down to 0,01 uS?!
Well, the question about your attenuator at the input, the resistance will have an impact on the LF filter formed together with the 100 pF and input capacitance of the diff stage.
If the probe is 20 pF I think it won't have too much impact on your meassurements...
If you are going to measure voltage over the Re, it can be a problem if you are going to connect the earth of the probe to the high impedance CCS, be carefull.
Edit: One mor thing I made a fast calculation, you have about 2,2 mA through the VAS if I made the calculation right...
Anyhow my guess is that with doubble darlington you have bigger Miller load with MJE 15030/1 for the VAS than with tripple darlington 2n5551/401, which perhaps make the speed diffrence.
Cheers
Oops, you misunderstood me...
The scope can 0.05uS, this does not mean that i looked to the
signal at this resolution...
I looked to the signal with 10uS,setting like the first shot,
there i did not see the flanks and was satisfied...
I used this resolution when i hunted some RF, i thought it was
some oscillation, a ~6mhz signal with amplitude of 0.5mv.
But as the signal was still there when amp was switched off...
EMI-wasted world...
Bodo, who borrowed me the scope, said that it is a 200mhz-scope.
It's a tektronix (425 i believe)
Mike
The scope can 0.05uS, this does not mean that i looked to the
signal at this resolution...
I looked to the signal with 10uS,setting like the first shot,
there i did not see the flanks and was satisfied...
I used this resolution when i hunted some RF, i thought it was
some oscillation, a ~6mhz signal with amplitude of 0.5mv.
But as the signal was still there when amp was switched off...
EMI-wasted world...
Bodo, who borrowed me the scope, said that it is a 200mhz-scope.
It's a tektronix (425 i believe)
Mike
Hi Mike,
Your amplifier looks good and adaptable.
Presently it is fixedly biased as a low level class-A running into class-B; ie. class-AB.
I have gone through the posts but cannot see your choice for quiescent current.
Mike, it occurs to me that you could make this amplifier run without ever dropping out of class-A so that it will never actually 'crossover' into fully class-B operation.
Remove the cascode and mirror drives from the bias setting BD135, then add two resistors,
one between collector of BD135 and +ve rail,
one between emitter of BD135 and -ve rail.
select correct values to give say 50 to 100mA of steady quiescent flow with zero offset at output.
Reconnect cascode and mirror drives for normal operation.
The fixed resistors should maintain a 'never-off' output stage conduction for both halves, and extended the zero ac current 'conduction' base.
This would be an easy modificaion for checking whether there is indeed any genuine audible improvement or not.
You added 2x 100k resistors to make it sound 'sweeter'. I find this interesting, but wonder what is happening with such high value resistors. Are they introducing 'never-off' conduction at higher amplitudes ?
Cheers ........... Graham.
Your amplifier looks good and adaptable.
Presently it is fixedly biased as a low level class-A running into class-B; ie. class-AB.
I have gone through the posts but cannot see your choice for quiescent current.
Mike, it occurs to me that you could make this amplifier run without ever dropping out of class-A so that it will never actually 'crossover' into fully class-B operation.
Remove the cascode and mirror drives from the bias setting BD135, then add two resistors,
one between collector of BD135 and +ve rail,
one between emitter of BD135 and -ve rail.
select correct values to give say 50 to 100mA of steady quiescent flow with zero offset at output.
Reconnect cascode and mirror drives for normal operation.
The fixed resistors should maintain a 'never-off' output stage conduction for both halves, and extended the zero ac current 'conduction' base.
This would be an easy modificaion for checking whether there is indeed any genuine audible improvement or not.
You added 2x 100k resistors to make it sound 'sweeter'. I find this interesting, but wonder what is happening with such high value resistors. Are they introducing 'never-off' conduction at higher amplitudes ?
Cheers ........... Graham.
Hi graham !
Thanks for this suggestion, this would be great if working !
I choosed 100ma quiscent current. The 2 100k resistors at these
currentlevels (2.2ma) are enough to greatly increase OLBW,
and/or reduce Z-out of vas.
I am not sure which of these both effects is responsible, but
has an immediate positive audible effect.
Mike
Thanks for this suggestion, this would be great if working !
I choosed 100ma quiscent current. The 2 100k resistors at these
currentlevels (2.2ma) are enough to greatly increase OLBW,
and/or reduce Z-out of vas.
I am not sure which of these both effects is responsible, but
has an immediate positive audible effect.
Mike
Hi Mike,
Thinking more about my class-A suggestion.
It might well be necessary to split the central 33ohm resistor between the output bases to acheive this, ie. use two series connected 18 or 22ohm resistors with their junction connected to the central output/NFB line.
This would introduce a compromise situation between no crossover versus pull-down.
Cheers ........... Graham.
Thinking more about my class-A suggestion.
It might well be necessary to split the central 33ohm resistor between the output bases to acheive this, ie. use two series connected 18 or 22ohm resistors with their junction connected to the central output/NFB line.
This would introduce a compromise situation between no crossover versus pull-down.
Cheers ........... Graham.
Hi Mike,
I'm wondering how the 'NoCap' circuit is working out for you
Through attempting to optimise my class-AABB output stage with a folded cascode splitter/driving circuit like yours, I have noted that matching the output stage to the high impedance common base cascode stage is most important. I did not copy your 330 ohm cascode resistor value, but found that this is best for open loop phase linearity and control.
I also noted that when setting output stage bias to minimise the levels of simulated high odd harmonic distortion when driving the virtual 'Ariel' loudspeaker, the resistor measured thd could subsequently not better 0.1% at all AF frequency/levels.
Many designers (though not all) aim for circa 0.002% at say 20kHz and use complex nfb arrangements to acheive this, but then nfb loop induced internal current draws due to asymmetrical music waveforms and crossovers/dynamic loudspeakers can simultaneously cause fractionally asymmetrical high speed storage/discharge current effects within an amplifier due to the additional 'stabilising' components that become necessary. This is why I recently showed some of the asymmetrical current draws in my AABB thread, and of course a 'NoCap' amplifier cannot fail in being less afflicted.
In my circuit I keep a higher 1A standing current so that the (low delay) nfb loop assisted output stage will more phase linearly resist crossover/loudspeaker back emf induced deflection wrt on-going signal amplification. The ratio of cascode to fixed bias at the output stage can be optimised for extremely low virtual loudspeaker induced harmonic generation, though still without a good resistor loaded specification being measurable !
I also found it necessary to use a 10uF/72k/10pF compensation network (values still need to be checked in real life) on the static negative driving 2N5401 cascode, to compensate for internal feedback within the dynamically operating positive driving 2N5401 cascode, this to balance internal cascode transistor losses and minimise harmonic distortion.
Something like this might further balance your own circuit; a check may be made for virtually identical output stage half current waveforms and peak +/- voltages at say 10kHz with a resistor load.
Hi Jan,
Re your Post#24 question and series output choke inclusion ... I prefer to NOT use them. I also prefer to site power amplifiers as monoblocs directly behind the loudspeaker using short low capacitance leads.
Choke inclusion at the ouput of a global nfb loop controlled amplifier can be audibly detectable, and via the X-Y Monitor test I posted in Wimms' Distortion Microscope I illustrated the electrically measurable effect from a 6uH component.
Mike, are you driving capacitive loads such that you must use a choke ?
2uH is not much, but it might still be worth trying short circuit switching it in and out of circuit to see if you can hear any change to transient and sibilant clarity. However the presence of a series output choke will not necessarily show up on all loudspeakers/crossover/cable systems, and maybe even less so on this low nfb delay design.
Look forwards to hearing of any new developments.
Cheers ............. Graham.
I'm wondering how the 'NoCap' circuit is working out for you
Through attempting to optimise my class-AABB output stage with a folded cascode splitter/driving circuit like yours, I have noted that matching the output stage to the high impedance common base cascode stage is most important. I did not copy your 330 ohm cascode resistor value, but found that this is best for open loop phase linearity and control.
I also noted that when setting output stage bias to minimise the levels of simulated high odd harmonic distortion when driving the virtual 'Ariel' loudspeaker, the resistor measured thd could subsequently not better 0.1% at all AF frequency/levels.
Many designers (though not all) aim for circa 0.002% at say 20kHz and use complex nfb arrangements to acheive this, but then nfb loop induced internal current draws due to asymmetrical music waveforms and crossovers/dynamic loudspeakers can simultaneously cause fractionally asymmetrical high speed storage/discharge current effects within an amplifier due to the additional 'stabilising' components that become necessary. This is why I recently showed some of the asymmetrical current draws in my AABB thread, and of course a 'NoCap' amplifier cannot fail in being less afflicted.
In my circuit I keep a higher 1A standing current so that the (low delay) nfb loop assisted output stage will more phase linearly resist crossover/loudspeaker back emf induced deflection wrt on-going signal amplification. The ratio of cascode to fixed bias at the output stage can be optimised for extremely low virtual loudspeaker induced harmonic generation, though still without a good resistor loaded specification being measurable !
I also found it necessary to use a 10uF/72k/10pF compensation network (values still need to be checked in real life) on the static negative driving 2N5401 cascode, to compensate for internal feedback within the dynamically operating positive driving 2N5401 cascode, this to balance internal cascode transistor losses and minimise harmonic distortion.
Something like this might further balance your own circuit; a check may be made for virtually identical output stage half current waveforms and peak +/- voltages at say 10kHz with a resistor load.
Hi Jan,
Re your Post#24 question and series output choke inclusion ... I prefer to NOT use them. I also prefer to site power amplifiers as monoblocs directly behind the loudspeaker using short low capacitance leads.
Choke inclusion at the ouput of a global nfb loop controlled amplifier can be audibly detectable, and via the X-Y Monitor test I posted in Wimms' Distortion Microscope I illustrated the electrically measurable effect from a 6uH component.
Mike, are you driving capacitive loads such that you must use a choke ?
2uH is not much, but it might still be worth trying short circuit switching it in and out of circuit to see if you can hear any change to transient and sibilant clarity. However the presence of a series output choke will not necessarily show up on all loudspeakers/crossover/cable systems, and maybe even less so on this low nfb delay design.
Look forwards to hearing of any new developments.
Cheers ............. Graham.
Hi Graham !
I already "discarded" this design, it had too much 3rd harmonic distortion
for my taste, giving some unpleasant sound. It was a very interesting
circuit, and prooved that "fast" amps with small phaseshift deliver
better dynamics and details. In this discipline that circuit was very
impressive. But the overall sounding was not ok.
The last weeks i have been busy with "sniffing" into ClassD, so this
amp just collected some dust.
Cany you show your "10uF/72k/10pF compensation network" in schematic ?
I have another circuit in mind, having even less phaseshift and
more simple, but having low thd, mainly 2nd harmonics.
I will try this one today... (A quasi-complementary)
My problem with achieving low thd is, that it is mostly done by
increasing feedback / openloopgain, handicapping the amp with
large feedbackstabilizingcaps, creating extra large phaseshifts, or
even worse, large dynamic phaseshifts, ruining the whole sound,
just to be able to reproduce the perfect sinewave.
I think that this is the reason why most good sounding amps have
not very low thd. They sound good inspite of thd, not because of... (IMHO)
Mike
I already "discarded" this design, it had too much 3rd harmonic distortion
for my taste, giving some unpleasant sound. It was a very interesting
circuit, and prooved that "fast" amps with small phaseshift deliver
better dynamics and details. In this discipline that circuit was very
impressive. But the overall sounding was not ok.
The last weeks i have been busy with "sniffing" into ClassD, so this
amp just collected some dust.
Cany you show your "10uF/72k/10pF compensation network" in schematic ?
I have another circuit in mind, having even less phaseshift and
more simple, but having low thd, mainly 2nd harmonics.
I will try this one today... (A quasi-complementary)
My problem with achieving low thd is, that it is mostly done by
increasing feedback / openloopgain, handicapping the amp with
large feedbackstabilizingcaps, creating extra large phaseshifts, or
even worse, large dynamic phaseshifts, ruining the whole sound,
just to be able to reproduce the perfect sinewave.
I think that this is the reason why most good sounding amps have
not very low thd. They sound good inspite of thd, not because of... (IMHO)
Mike
Hi Mike,
That compensation circuit is on the circuit in my class-AABB thread. I have yet to investigate temperature compensation. PS. The top half choke should be 22uH and not 150uH.
In your second ( Darlington only ) circuit, the rightmost 2N5401 has degenerative internal feedback due to it having a swinging collector potential.
However the left 2N5401 has a fixed collector voltage and thus no internal feedback.
This may be compensated by a low 'C' of 4.7 to 10pF between the amplifier's output at the live end of the choke and the emitter of the left 2N5401. In parallel with this fit a resistor of probably about 120k in series with a 2.2 to 4.7uF dc blocking capacitor.
This will balance the dynamic characteristics of the inverting cascode transistors, and can further reduce low level harmonic distortion.
Maybe you could try simulating it first.
I've just tried simulating my circuit without the compensation, and the distortion goes from 0.02% at 80Vp-p 10kHz across the virtual loudspeaker, to 0.07%. (160kHz measuring bandwidth)
Simulation of my proposed circuit shows the distortion characteristic to be mostly third, and circa 0.01% when driving the virtual loudspeaker load at about 15Vrms. But third harmonic is what you get as soon as you go push-pull beyond class-A anyway, and this is why I am using a 1A quiescent for better quality at more normal listening levels.
I quite agree with your last paragraph, and like you I don't want to use too much nfb, but with class-D are you going to get the same imaging accuracy and transient clarity when the use of significantly valued series output chokes are so unavoidable, such that their effects vary independently on a per channel basis?
Another point regarding your 'NoCap' testbed relates to amplifier input resistance between the first base and the signal source.
I have for long now prefered low impedance amplifier drive via quality TV downlead coax and BNC connectors or filed out metal RCA phonos.. Your amplifier might well distort a little less if you do away with the 'input filter' which is also effectively in series with the first base as far as nfb loop control is concerned. The forward gain will not change, but the AF nfb response does improve slightly, and when compared to the 2k maybe by 2 to 3dB.
Another thing I prefer, is to use a resistor in the differential tail because its value tends to be so very high when compared to the generated differential emitter impedance. A current source can lead to unusual effects on transients and at high output.
I have just simulated my circuit with a current source in place of the passive biasing resistors; the simulated distortion went up from the 0.02% figure to 0.055% with some seriously undesirable cross-conduction on 10kHz sine, which in your circuit would attempt to raise the bias voltage on a per half cycle basis.
It should be very easy for you to work out the correct value for a fixed resistor and substitute it for listening tests.
Thus I am hoping that discarding your 'NoCap' design, has been only temporary.
Cheers ............. Graham.
That compensation circuit is on the circuit in my class-AABB thread. I have yet to investigate temperature compensation. PS. The top half choke should be 22uH and not 150uH.
In your second ( Darlington only ) circuit, the rightmost 2N5401 has degenerative internal feedback due to it having a swinging collector potential.
However the left 2N5401 has a fixed collector voltage and thus no internal feedback.
This may be compensated by a low 'C' of 4.7 to 10pF between the amplifier's output at the live end of the choke and the emitter of the left 2N5401. In parallel with this fit a resistor of probably about 120k in series with a 2.2 to 4.7uF dc blocking capacitor.
This will balance the dynamic characteristics of the inverting cascode transistors, and can further reduce low level harmonic distortion.
Maybe you could try simulating it first.
I've just tried simulating my circuit without the compensation, and the distortion goes from 0.02% at 80Vp-p 10kHz across the virtual loudspeaker, to 0.07%. (160kHz measuring bandwidth)
Simulation of my proposed circuit shows the distortion characteristic to be mostly third, and circa 0.01% when driving the virtual loudspeaker load at about 15Vrms. But third harmonic is what you get as soon as you go push-pull beyond class-A anyway, and this is why I am using a 1A quiescent for better quality at more normal listening levels.
I quite agree with your last paragraph, and like you I don't want to use too much nfb, but with class-D are you going to get the same imaging accuracy and transient clarity when the use of significantly valued series output chokes are so unavoidable, such that their effects vary independently on a per channel basis?
Another point regarding your 'NoCap' testbed relates to amplifier input resistance between the first base and the signal source.
I have for long now prefered low impedance amplifier drive via quality TV downlead coax and BNC connectors or filed out metal RCA phonos.. Your amplifier might well distort a little less if you do away with the 'input filter' which is also effectively in series with the first base as far as nfb loop control is concerned. The forward gain will not change, but the AF nfb response does improve slightly, and when compared to the 2k maybe by 2 to 3dB.
Another thing I prefer, is to use a resistor in the differential tail because its value tends to be so very high when compared to the generated differential emitter impedance. A current source can lead to unusual effects on transients and at high output.
I have just simulated my circuit with a current source in place of the passive biasing resistors; the simulated distortion went up from the 0.02% figure to 0.055% with some seriously undesirable cross-conduction on 10kHz sine, which in your circuit would attempt to raise the bias voltage on a per half cycle basis.
It should be very easy for you to work out the correct value for a fixed resistor and substitute it for listening tests.
Thus I am hoping that discarding your 'NoCap' design, has been only temporary.
Cheers ............. Graham.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.