Q5/10 and R5/8 aren't doing anything at the moment. They are there just for flexibility, so I can change it to a differential output (which I probably would do if I built it).
And yes, the zeners and current sources are a daft way around, as was pointed out before. The updated schematic later in the thread has them the right way.
And yes, the zeners and current sources are a daft way around, as was pointed out before. The updated schematic later in the thread has them the right way.
Hi !
I had a visit from Madonna yesterday !
At least it felt like it, listening with my FC-prototype, i really had
this feeling like having madonna standing in front of me, feeling
her voice, her breath...
So i got rid of the audible distortions, they had gone after adding
voltagestabilizing to the frontend. So FC seems to be very sensible
to bad supplyvoltage ?
So far, the sound from FC is impressive, not perfect yet, but very
promising. I would prefer a bit smoother sound, but finetuning
should enable this.
I am surprised what the amp does to my speakers, they are really
not highend, but the amp makes them highend...
Keep on Konrad, you might be surprised by the sound of FC !
BTW, i think it's not a good idea to have extra gain in the outputstage,
this way you add an extra stage to your amplifier.
Mike
I had a visit from Madonna yesterday !
At least it felt like it, listening with my FC-prototype, i really had
this feeling like having madonna standing in front of me, feeling
her voice, her breath...
So i got rid of the audible distortions, they had gone after adding
voltagestabilizing to the frontend. So FC seems to be very sensible
to bad supplyvoltage ?
So far, the sound from FC is impressive, not perfect yet, but very
promising. I would prefer a bit smoother sound, but finetuning
should enable this.
I am surprised what the amp does to my speakers, they are really
not highend, but the amp makes them highend...
Keep on Konrad, you might be surprised by the sound of FC !
BTW, i think it's not a good idea to have extra gain in the outputstage,
this way you add an extra stage to your amplifier.
Mike
Hi,
error correction will not work with folded cascode as expected. Using this way error correction will not decrease distortion, it only shifts low order to high order distortion. There are some statements who do not prefer eror correction sound. For me I'd also increase the number of output transistors to minimise distortion. When you look into the datasheet of the MJL3281A the current gain drops above 4A @ 100°C. This means that the MJL3281 is able to produce 60W @ 8 Ohms and only 30W @ 4 Ohms without increasing of distortion.
error correction will not work with folded cascode as expected. Using this way error correction will not decrease distortion, it only shifts low order to high order distortion. There are some statements who do not prefer eror correction sound. For me I'd also increase the number of output transistors to minimise distortion. When you look into the datasheet of the MJL3281A the current gain drops above 4A @ 100°C. This means that the MJL3281 is able to produce 60W @ 8 Ohms and only 30W @ 4 Ohms without increasing of distortion.
lumanauw said:Hi, Mike,
She's there?
Mike, something interesting about "Error correction". Someone submitted "new EC sound" based on NP-PMA cct. The comment is very good.
Why don't you complete your amp with one of these? Maybe will result in "perfect" amp.
No, she left me when i pulled the plug...
Yes, i already thought about that as thd of my circuit is not really low.
But i was not able yet to sim a functional EC, maybe mjl3281a
doesn't like it ? I also tend to try the mosfets sj201 and co.
bocka said:
Hmm, in the datasheet, hfe is still above 100 for 7A,Vce=5v,100° ?
But anyway, hfe is never flat as vce changes much in an amp.
Ok, doubling devices is always good for increasing safety...
But 30W into 4 ohm is much ! My amp is fed with +/-21v, giving
~32Wrms into 4 ohm. The power already delivered is painful !
But i wanted to increase to +/-36volt, so doubling the devices
is necessary to keep junctiontemp low. (My boxes are 4ohm,
the tweeter driven undamped, means no resistor in series)
Mike
Probably Bocka is right. I did that scematic just for funn y know. But shifting the distortion up above hearing range or into the range above where most tweeters roll off anyway may not bee so bad sounding after all.
I can see that output stage distortion will bee dominant as the open loop gain is "low". Making a amp with dual differensial inputs and not taking advantage of the 4 signal taps would bee a waste. Here the goal is to make sym folded cascode for audio. It may still not bee the best performing amp. The best preformer of audio amps is probably the 20-50w klass A range with a "unknown" configuration. Many of wich use litle global or lokal only feedback. Output resistance in the speaker impedance range before a often low amount of global feedback is aplyed. As i see it the lokal feedback is doing the same ie redusing lower order harmonics and not nessity increase higher order, but they will show up more as others are redused.
There is also several ...audio.... aspects that is commonly overlooked in debating a spesific amp configuration, as the configuration is in fokus.
The output of a folded cascode is comparable to a voltage in current ouput, OTA . Making the input stage to a OTA there is many ways to linearise and som may as well bee better than "simple" folded cascode (lokal feedback linearising).
How abt cascoded folded cascode ota output and preferred output stage (?) im shore there will bee some nice performers, sounding good among em.
I can see that output stage distortion will bee dominant as the open loop gain is "low". Making a amp with dual differensial inputs and not taking advantage of the 4 signal taps would bee a waste. Here the goal is to make sym folded cascode for audio. It may still not bee the best performing amp. The best preformer of audio amps is probably the 20-50w klass A range with a "unknown" configuration. Many of wich use litle global or lokal only feedback. Output resistance in the speaker impedance range before a often low amount of global feedback is aplyed. As i see it the lokal feedback is doing the same ie redusing lower order harmonics and not nessity increase higher order, but they will show up more as others are redused.
There is also several ...audio.... aspects that is commonly overlooked in debating a spesific amp configuration, as the configuration is in fokus.
The output of a folded cascode is comparable to a voltage in current ouput, OTA . Making the input stage to a OTA there is many ways to linearise and som may as well bee better than "simple" folded cascode (lokal feedback linearising).
How abt cascoded folded cascode ota output and preferred output stage (?) im shore there will bee some nice performers, sounding good among em.
Hi,
If we want to make perfect amp, we must perfectly design every stages. Folded cascode has make a very good front end, so what is left is only the final/output stage.
People usually "dont look much" for class AB final stage, because there is not much topology option here. EC is one option.
It is sad, because in class AB power amp, most of the distortion actually occurs in the final stage itself, because the current and the voltages at the output stage fluctuates very much.
Bocka is right. EC itself will generate high order distortion, because it is an "active cct" itself. It acts like a complete amp, comparing input and output of the final stage. So it must generate high order distortion.
But if EC gives better sound (if it does so), why not use them? Some commercial amp uses these, and customers likes it.
EC gives an interesting property, that it can make the damping factor of the whole power amp infinity (zero or negative output impedance). I've heard a power amp which has 1000 damping factor, it sounds good, especially in bass control.
If we want to make perfect amp, we must perfectly design every stages. Folded cascode has make a very good front end, so what is left is only the final/output stage.
People usually "dont look much" for class AB final stage, because there is not much topology option here. EC is one option.
It is sad, because in class AB power amp, most of the distortion actually occurs in the final stage itself, because the current and the voltages at the output stage fluctuates very much.
Bocka is right. EC itself will generate high order distortion, because it is an "active cct" itself. It acts like a complete amp, comparing input and output of the final stage. So it must generate high order distortion.
But if EC gives better sound (if it does so), why not use them? Some commercial amp uses these, and customers likes it.
EC gives an interesting property, that it can make the damping factor of the whole power amp infinity (zero or negative output impedance). I've heard a power amp which has 1000 damping factor, it sounds good, especially in bass control.
"most of the distortion actually occurs in the final stage itself"
AMEN !
Yes, it's annoying, most of the feedback is used to compensate
the distortions created by the outputstage, linearizing the frontend
seems easy in compare, as these stages operate fully ClassA...
But ClassAB operates transistors always in the on/off region, the
most distortive region of transistors.
If i understood correctly, EC tries to reduce exactly this problem ?
I tried several EC's, but all of them increased distortion, am i too
dumb for EC ? I haven't checked PMA's version with mosfets yet,
i wanted to keep BJTs... (I tried converting it to bjts)
I tried several outputstagetypes, but simple tripledarlington still
seems unbeatable ?
Dampingfactor is not a big issue with tripledarlington, my previous
symamp has DF of > 4000 into 4ohms... (with 62db feedback)
BTW, how much sense makes a cascoded cascode ?
Mike
AMEN !
Yes, it's annoying, most of the feedback is used to compensate
the distortions created by the outputstage, linearizing the frontend
seems easy in compare, as these stages operate fully ClassA...
But ClassAB operates transistors always in the on/off region, the
most distortive region of transistors.
If i understood correctly, EC tries to reduce exactly this problem ?
I tried several EC's, but all of them increased distortion, am i too
dumb for EC ? I haven't checked PMA's version with mosfets yet,
i wanted to keep BJTs... (I tried converting it to bjts)
I tried several outputstagetypes, but simple tripledarlington still
seems unbeatable ?
Dampingfactor is not a big issue with tripledarlington, my previous
symamp has DF of > 4000 into 4ohms... (with 62db feedback)
BTW, how much sense makes a cascoded cascode ?
Mike
Hi, Mike,
You said an important thing there. If the distortions mainly comes from the output stage, why we always have to feedback to the front end (which have been linear)? Shouldn't we think something that have "local" distortion elimination, not global one? So each stage doesn't have to correct errors that is made by other stage.
People dont talk so much about class AB output stage, because there is not much can be done there. Class AB output stage will always distort. Inspite of it, maybe EC will help in "sound", even if the measurement (of higher order distortion) is not liked.
In concept, EC should give lower distortion, not higher. It will act to level what is in and what is out of the output stage. Hawksford type is the common type. We can apply it to any voltage. I dont understand much of the NP-PMA style. For any EC cct, maybe it is the higher-order one that people don't like.
As for cascoding a folded cascode, it is doable. I've asked Steven about this. You can cascode the right transistor, to give it minimum voltage fluctuation, which will reduce the distortion caused by voltage fluctuation.
You said an important thing there. If the distortions mainly comes from the output stage, why we always have to feedback to the front end (which have been linear)? Shouldn't we think something that have "local" distortion elimination, not global one? So each stage doesn't have to correct errors that is made by other stage.
People dont talk so much about class AB output stage, because there is not much can be done there. Class AB output stage will always distort. Inspite of it, maybe EC will help in "sound", even if the measurement (of higher order distortion) is not liked.
In concept, EC should give lower distortion, not higher. It will act to level what is in and what is out of the output stage. Hawksford type is the common type. We can apply it to any voltage. I dont understand much of the NP-PMA style. For any EC cct, maybe it is the higher-order one that people don't like.
As for cascoding a folded cascode, it is doable. I've asked Steven about this. You can cascode the right transistor, to give it minimum voltage fluctuation, which will reduce the distortion caused by voltage fluctuation.
Hi lumanauw !
"which will reduce the distortion caused by voltage fluctuation"
Yes, that was the idea ! Maybe simply try ? Devices creating v-gain
shouldnt have big voltage fluctuation.
The concept of multistagefeedbackloops exists, and i heard that it can
greatly improve sound.
Of course it's easy to use global feedback to reduce distortions in
outputstage. But it's not easy to apply local feedback in the outputstage,
as there's no gain or diffamp etc. I mean a feedback that really does
reduce distortions...
Now i will try to understand some about this "PIM"...
Maybe a feedbacknetwork can be adjusted to sound good, even with
very high openloopgain ? Its "very" easy to make amps with OLgain
above 1:1.500.000 (stable), but it's hard to get good sound from
these, even with their ultralow thd... (< 0.0002%)
The more i think about phaseshifted feedback, the more i wonder
that these amps work at all ?
But now i am changing topic...
The next days i will try to linearize my FC-circuit.
Mike
"which will reduce the distortion caused by voltage fluctuation"
Yes, that was the idea ! Maybe simply try ? Devices creating v-gain
shouldnt have big voltage fluctuation.
The concept of multistagefeedbackloops exists, and i heard that it can
greatly improve sound.
Of course it's easy to use global feedback to reduce distortions in
outputstage. But it's not easy to apply local feedback in the outputstage,
as there's no gain or diffamp etc. I mean a feedback that really does
reduce distortions...
Now i will try to understand some about this "PIM"...
Maybe a feedbacknetwork can be adjusted to sound good, even with
very high openloopgain ? Its "very" easy to make amps with OLgain
above 1:1.500.000 (stable), but it's hard to get good sound from
these, even with their ultralow thd... (< 0.0002%)
The more i think about phaseshifted feedback, the more i wonder
that these amps work at all ?
But now i am changing topic...
The next days i will try to linearize my FC-circuit.
Mike
I think, that easy to make open loop gain about 1:1500000.
You can make it with -say- two stages, but in this case the bandwidht will be small. There will be easy to make stable this amplifier with closed loop, and it will be low THD amp, but only up to few kHz, where the open loop gain starts to fall.
If You make this large open loop gain with more stages, You can get wide bandwidht, but there will be hard to apply the feedback.
So I think that the best solution to apply local feedbacks for every stage, and finally use lower amount of global feedback. I always plan my amplifiers to get the open loop bandwidht 100kHz (-3dB), or higher.
sajti
You can make it with -say- two stages, but in this case the bandwidht will be small. There will be easy to make stable this amplifier with closed loop, and it will be low THD amp, but only up to few kHz, where the open loop gain starts to fall.
If You make this large open loop gain with more stages, You can get wide bandwidht, but there will be hard to apply the feedback.
So I think that the best solution to apply local feedbacks for every stage, and finally use lower amount of global feedback. I always plan my amplifiers to get the open loop bandwidht 100kHz (-3dB), or higher.
sajti
Hi sajti !
The "< 0.0002%" was for 20khz in this case. This one is multistage,
using nested diffamps. I think about actually trying this one, just to
know how this sounds. If it doesn't sound, it's still suitable for a
subwoofer...
For 1khz, i had to increase accuracy of sims, just to get harmonics
visible... (2nd harmonic = 1.5uV at 12v output into 4ohm)
It was not too easy to stop this one from oscillating.
This is a 4stage, 2 diffamps+vas+output. With cascoding the vas,
using cfp in diffamps etc, it's easily possible to reduce thd even more.
openloopbandwidth -3db@100khz ? Phew ! Is this with or without
cdoms and other stabilizingcaps ?
What does this tell us ? It's more complicated to make a lowgain amp ?
Mike
The "< 0.0002%" was for 20khz in this case. This one is multistage,
using nested diffamps. I think about actually trying this one, just to
know how this sounds. If it doesn't sound, it's still suitable for a
subwoofer...
For 1khz, i had to increase accuracy of sims, just to get harmonics
visible...
(2nd harmonic = 1.5uV at 12v output into 4ohm)It was not too easy to stop this one from oscillating.
This is a 4stage, 2 diffamps+vas+output. With cascoding the vas,
using cfp in diffamps etc, it's easily possible to reduce thd even more.
openloopbandwidth -3db@100khz ? Phew ! Is this with or without
cdoms and other stabilizingcaps ?
What does this tell us ? It's more complicated to make a lowgain amp ?
Mike
Hi mikeB,
my experiences, that there is no audible difference between the 0.00002%, and 0,2% harmonic distortion. But if You increase the nfb You increase the danger of the TIM, especially if Your amplifier is slow.
The 100kHz open loop bandwidht is not so difficult if You apply local feedback. If You need no high power, You can get 1MHz...
I made it in 30W amplifier. So to make this amplifier stable I used 20dB step type compensation fot the 20dB nfb. With this compensation I get 110 degree phase margin. When I tried the amp with pure capacitive load, there was no any overshot or ringing on the squarewave...
sajti
my experiences, that there is no audible difference between the 0.00002%, and 0,2% harmonic distortion. But if You increase the nfb You increase the danger of the TIM, especially if Your amplifier is slow.
The 100kHz open loop bandwidht is not so difficult if You apply local feedback. If You need no high power, You can get 1MHz...
I made it in 30W amplifier. So to make this amplifier stable I used 20dB step type compensation fot the 20dB nfb. With this compensation I get 110 degree phase margin. When I tried the amp with pure capacitive load, there was no any overshot or ringing on the squarewave...
sajti
Hi sajti,
this is not my experience. With my previous amp, any change intended
to lower thd, gave a cleaner and smoother sound. But i am lacking
the ability to make real mesurements, maybe distortions were that
high, that these modifications were necessary. So i can't say which
thd-level my amp has in realworld. In sims thd got far below 0.01%.
This clean and smooth sound is now missing in my foldedcascodeamp.
The problem is, what kind of distortions are bugging me ?
Maybe the recordings are not clean and smooth, so my amp simply
does his job ? (I hate this question)
So in my other amp i simply added smearing ?
Or i have unmusical harmonics ?
What i definitely know, is that 3% 2nd harmonics is nearly inaudible,
3% 3rd harmonic make you instantly puke... (Verified by applying
these harmonics to a .WAV)
Maybe i should try down to which level i can detect 3rd harmonics ?
Mike
this is not my experience. With my previous amp, any change intended
to lower thd, gave a cleaner and smoother sound. But i am lacking
the ability to make real mesurements, maybe distortions were that
high, that these modifications were necessary. So i can't say which
thd-level my amp has in realworld. In sims thd got far below 0.01%.
This clean and smooth sound is now missing in my foldedcascodeamp.
The problem is, what kind of distortions are bugging me ?
Maybe the recordings are not clean and smooth, so my amp simply
does his job ? (I hate this question)
So in my other amp i simply added smearing ?
Or i have unmusical harmonics ?
What i definitely know, is that 3% 2nd harmonics is nearly inaudible,
3% 3rd harmonic make you instantly puke... (Verified by applying
these harmonics to a .WAV)
Maybe i should try down to which level i can detect 3rd harmonics ?
Mike
Ush
I still vander why using singel ended drive to the folded cascode?
Here we have low input impedance to the 'cascode'. And current mirrors easily lets you apply more available current swing for the same innput (and the current is still from the input stage. And that's done without altering its bias current. Also the little noise (at af) contribution from the current sorces is balanced away.
as an exsample the Leach amp is 'singel ended cascode' drive: http://users.ece.gatech.edu/~mleach/lowtim/graphics/ckt.pdf
Lets just modify it so that its current mirror drive ( ya know there is bouth current sink and sorce available WHY NOT use it ?????). Like Slone "failed" to use double mirrors just because he used high input impedance to the next stage. Mirror input to the common base stage not make the biaspiont of 'vas stage' unstable as there is a low value resistor making the bias point.
I still vander why using singel ended drive to the folded cascode?
Here we have low input impedance to the 'cascode'. And current mirrors easily lets you apply more available current swing for the same innput (and the current is still from the input stage. And that's done without altering its bias current. Also the little noise (at af) contribution from the current sorces is balanced away.
as an exsample the Leach amp is 'singel ended cascode' drive: http://users.ece.gatech.edu/~mleach/lowtim/graphics/ckt.pdf
Lets just modify it so that its current mirror drive ( ya know there is bouth current sink and sorce available WHY NOT use it ?????). Like Slone "failed" to use double mirrors just because he used high input impedance to the next stage. Mirror input to the common base stage not make the biaspiont of 'vas stage' unstable as there is a low value resistor making the bias point.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.