I would say pnp+cap, to avoid the biasing of the input.
One suggestion would be to use a CFP in the input ?
It seems that you use heavy currents through all parts (~40ma).
This excludes the use of small signaltransistors, if i'm right.
Why not reduce currents below 2ma and use a predriver ? (Tripledarlington output)
This enables the use of signaltransistors, which have much better
linearity, speed and gain.
Which transistortypes are you using ?
Mike
One suggestion would be to use a CFP in the input ?
It seems that you use heavy currents through all parts (~40ma).
This excludes the use of small signaltransistors, if i'm right.
Why not reduce currents below 2ma and use a predriver ? (Tripledarlington output)
This enables the use of signaltransistors, which have much better
linearity, speed and gain.
Which transistortypes are you using ?
Mike
Hi, Mike,
The transistor I use for front end is D669 and B649 attached in small heatsink. The final uses C5200.
I use manual adjusting, using 100K VR attached to +/-15V, because my servo type didn't work as expected.
About triple darlington, it still confuses me. Some designer said to advoid triple bipolar darlington, and use double bipolar darlington, or use double darlington with mosfet as driver. Triple darlington gives too much shift, and prone to oscilation, they say (I dont know in SIMS)
CFP in input? Wouldn't that be the original design at post #1 ? (I put 25K as CFP resistor)?
The transistor I use for front end is D669 and B649 attached in small heatsink. The final uses C5200.
I use manual adjusting, using 100K VR attached to +/-15V, because my servo type didn't work as expected.
About triple darlington, it still confuses me. Some designer said to advoid triple bipolar darlington, and use double bipolar darlington, or use double darlington with mosfet as driver. Triple darlington gives too much shift, and prone to oscilation, they say (I dont know in SIMS)
CFP in input? Wouldn't that be the original design at post #1 ? (I put 25K as CFP resistor)?
Hmm, the leachamp uses tripledarlingtonoutput...
In my sims the tripledarlington helped increasing bandwidth
and reduce oscillation. Only with tripledarlington i was able to
use 10pF-cdoms. And from the audiophile suspect, i think you can hear
that the amp has "more" power. It drives my 4ohm speakers with ease,
like it is "bored" with that high impedance. 😉
Your ckt in the 1st post is quite similar to a cfp, but different.
You should use a 2k at the collector of the input, connect emitter/base
of the complementary to this 2k, and connect collector of the CFP-bjt
to the emitter of the input-bjt.
You should look to a CFP as a single transistor.
Mike
In my sims the tripledarlington helped increasing bandwidth
and reduce oscillation. Only with tripledarlington i was able to
use 10pF-cdoms. And from the audiophile suspect, i think you can hear
that the amp has "more" power. It drives my 4ohm speakers with ease,
like it is "bored" with that high impedance. 😉
Your ckt in the 1st post is quite similar to a cfp, but different.
You should use a 2k at the collector of the input, connect emitter/base
of the complementary to this 2k, and connect collector of the CFP-bjt
to the emitter of the input-bjt.
You should look to a CFP as a single transistor.
Mike
Hi, Mike,
What happens if we put double darlington, but use mosfet (like IRF540-9540) as the driver for the final transistor (bipolar?)
Mr.John Curl uses this configuration, he said this configuration can "block" what ever happens in the final stage (like heavy/strange speaker load) from entering the front stages of amplifier.
Triple bipolar transistor can make VAS need only low current, but still can be affected by the strange speaker load.
Yes, you are right. At first I design it like Graham Maynards design, the upper 25K is attached to the emitor of second transistor.
But then I see the design of Cyrus amp. It has the 25k attached to rail instead to the emitor of the second transistor.
What happens if we put double darlington, but use mosfet (like IRF540-9540) as the driver for the final transistor (bipolar?)
Mr.John Curl uses this configuration, he said this configuration can "block" what ever happens in the final stage (like heavy/strange speaker load) from entering the front stages of amplifier.
Triple bipolar transistor can make VAS need only low current, but still can be affected by the strange speaker load.
Yes, you are right. At first I design it like Graham Maynards design, the upper 25K is attached to the emitor of second transistor.
But then I see the design of Cyrus amp. It has the 25k attached to rail instead to the emitor of the second transistor.
Hi !
Hmm, i can do some simus with this config. But you will loose
some power, due to the 3.6v Vgs-treshold.
I do not fully agree that a mosfet blocks what ever happens in the
final stage. This is true for low frequencies, but with high frequencies
mosfet's no longer block. With higher frequencies the inputimpedance
of mosfet can get very low.
I need to check, because in the driver stage you don't have large
currents. But with these low currents the mosfets operate in a very
unlinear range, but on the other hand, if driverstage operates ClassA
with high enough currents, this looks different.
I need to simulate...
In theory, the predriver stage reduces loads from output by ~300.
(300 = hfe of bc546/556)
For CFP, you might have a look at "my" design. It's easy to implement
and gives very good results. (You can simply replace the jFet by bjt)
I was able to increase bandwidth AND ol-gain in my ckt, this reduces
thd by factor 5 to 2 (1khz-20khz), and imd by factor 3. I will apply
these changes the next days for listeningtest. 3rd harmonics are
completely gone for lower freqs. This means that i was somehow
completely wrong about currentmirror... (damn it !)
THD at 1khz nearly full power into 4ohm is now 0.00012%.
(2nd harmonic only) Whatever this means...
Mike
Hmm, i can do some simus with this config. But you will loose
some power, due to the 3.6v Vgs-treshold.
I do not fully agree that a mosfet blocks what ever happens in the
final stage. This is true for low frequencies, but with high frequencies
mosfet's no longer block. With higher frequencies the inputimpedance
of mosfet can get very low.
I need to check, because in the driver stage you don't have large
currents. But with these low currents the mosfets operate in a very
unlinear range, but on the other hand, if driverstage operates ClassA
with high enough currents, this looks different.
I need to simulate...
In theory, the predriver stage reduces loads from output by ~300.
(300 = hfe of bc546/556)
For CFP, you might have a look at "my" design. It's easy to implement
and gives very good results. (You can simply replace the jFet by bjt)
I was able to increase bandwidth AND ol-gain in my ckt, this reduces
thd by factor 5 to 2 (1khz-20khz), and imd by factor 3. I will apply
these changes the next days for listeningtest. 3rd harmonics are
completely gone for lower freqs. This means that i was somehow
completely wrong about currentmirror... (damn it !)
THD at 1khz nearly full power into 4ohm is now 0.00012%.
(2nd harmonic only) Whatever this means...
Mike
Thanh!,
Yes, it works with good sound😀
Mike,
What happens if you put all CFP transistor in your design (diff already CFP, then put CFP for VAS and output stage, instead of darlingtons)? Will the harmonics (especially the 3rd cancelation) be better
Yes, it works with good sound😀
Mike,
What happens if you put all CFP transistor in your design (diff already CFP, then put CFP for VAS and output stage, instead of darlingtons)? Will the harmonics (especially the 3rd cancelation) be better
Hi lumanauw !
I tried CFP in the VAS, but had no big change. (In simus)
I played with CFP-outputstage, but it seems to increase 3rd harmonic,
and also seems to be bad choice for ClassAB, as turnon/off times
increase and heavy crossoverdistortions occured. (In simus)
I wanted to try a ClassA with CFP-output, CFP-VAS needs listeningtest.
Mike
I tried CFP in the VAS, but had no big change. (In simus)
I played with CFP-outputstage, but it seems to increase 3rd harmonic,
and also seems to be bad choice for ClassAB, as turnon/off times
increase and heavy crossoverdistortions occured. (In simus)
I wanted to try a ClassA with CFP-output, CFP-VAS needs listeningtest.
Mike
OK, so there is no need for CFP in VAS
Really? So CFP works in opposite ways if put in final stage than in differential?
Yes, I read this in DougSelf book. He didnt mentioned about the fit for class A, class B, class AB. But is it really bad for Class B or AB? Randy Slone usually using CFP output stage for class AB amp. He's using K1058 as the final pairs.
lumanauw said:
I haven't seen any design with CFP in VAS, seems not to be good.
The CFP in outputstage helps with increasing dampingfactor,
reduce thermal drift and so on. It does not sound bad.
I have built one with mosfets, and it had some very interesting
sound. It sounded like having 10times the power, crying for
pumping up the volume. But i had audible crossover. Sadly i
detected later (when reusing the parts) that the mosfets were
completely mismatched (vgs: 3.2v to 4.8v,irf540n irf9540n)
I would say, if mosfet in output, then with CFP.
I will try again, as the sound was really fascinating.
thanh said:
I did, i like the symetrical ones for their stability and reliability.
They have perfect switchon/off behaviour, eliminating the need
for relays. My latest simus resulted in a symetrical again, it showed
up the best results. (lowest 3rd harmonic at 20khz, best stability)
I will build this one these days. (I must be crazy, 12 bjts for input
only, will be much work matching these)
Mike
😀
What? 12 device for input only? You built full complementary differential with CFP or you parrarel those? What's the design look like?
Thanh!
Yes, I've built full complementary audio power amp. Infact, quite many of them.
Perfect design and preferable actual listening design maybe quite different. If you design with SIM or electronic theory, you will end with very complicated design, full complementary design. Many buffer stages, many cascodes, etc.
If you design with your speaker as your benchmark, you will find "the minimum component/minimum stages" stream is prefereable. You will end up with designs like Aleph.
After building several kind of design, I began to search for "perfect" design. I tried "Optimum" power amp by Randy Slone. If you read this chapter on his book, you will feel that this is the one. But after I finished it (with modif in the current mirrors), it feels no great difference with other design. (Maybe not so, a master said Detail is everything, I never explore/experimenting with this design more deeply)
So, I continue to search for this "perfect" amp. At first I was inspired by numbers, like minimum THD, 1000 damping factor, etc (you can read this in pro-audio amps, like MC2, stunning numbers) Trying to get close to these numbers.
(you should read DestroyerX thread on "Blameless" amp. This amp gives perfect measurement, but he complained about the sound)
BUT,
then, I found out that simple design or classic designs (that people overlooked) tends to sound better. Then I post the thread "single or dual differential" for knowing why is that. This thread never finish with exact conclusion. Different masters gives different opinion, each with strong argument. Like Mr.John Curl insisted on full complementary design, because the VAS will be perfect push-pull. He's right about this.
For now, I concluded (for me) that single differential tends to sound better. It will not as clean as full complementary design (yes, you can hear it gives not so clean reproduction compared to full complementary designs), but this non-symmetrical design have distortions that makes the amp sound better (to me).
I dig up this fact and look at some master's design, and try to find out why they make cct like that. Like Mr.Pass Aleph, AKSA (never find out the real thing, just the approximation by DestroyerX thread). Then I make experimental amps, and somehow I feel that the more "imbalance" design, it will give preferable sound (again, to me). Maybe this is why JLH or NAD3020 is so strange if you look at the schematic.
This is the conclusion for small power (like 100W or less). I still feel that if you built big rating amp, (250W upwards) you will need full complementary design. You cannot power up the speakes in the stadium with 20.000 watt with not so clean sound. Almost all of pro-audio amp cct uses full complementary, while smaller ratings you called "hi-end" at home audio took different path.
Read about the man who changes career from audio designer to aircraft intrumentation designer after he hears the Single Ended Triode sound. Whats so symmetrical in SET? What's so perfect about SET measurement results? But why it sounds good?
lumanauw said:
Hi !
It sounds more than it is. Yes, it is full complementary diffamp with CFPs.
That makes already 8. The other 4 are the 2 ccs to the diffamp.
Then 2 for VAS, and 6 for output.
I think it's still "simple". It shows great measurements in the sims,
now i want to know how important it is, to have "ultimate" low THD
at high frequencies and extreme bandwidth. Somehow i was not
really satisfied with my previous amp, i had the feeling that something
is not as it should be. The version with the currentmirror was some flop,
it was very hard to stop it from oscillating. I would never trust this one.
If i am not fully satisfied with this one, i will try completely different
topologys, as simple as possible. (except the CFP-mosfet one)
Mike
Hi, Mike,
What kind of CFP mosfet output stage are you using? Bipolar as driver and mosfet as final stage?
Actually I'm thinking about making discrete CFP-IGBT, that is mosfet as driver (but L mosfet like K1058) and rugged bipolar (C2922) as final stage.
Which is better?
What kind of CFP mosfet output stage are you using? Bipolar as driver and mosfet as final stage?
Actually I'm thinking about making discrete CFP-IGBT, that is mosfet as driver (but L mosfet like K1058) and rugged bipolar (C2922) as final stage.
Which is better?
FAB,
I just read article of Lavardin power amp with its "Heat Memory" effect. You have built one. What does it sounds like?
Where can I see Lavardin schematic?
I just read article of Lavardin power amp with its "Heat Memory" effect. You have built one. What does it sounds like?
Where can I see Lavardin schematic?
lumanauw said:
lumanauw
I have not built a Lavardin amp but used the concept described in the patent (for my personal use of course) on reduding the memory effect caused by thermal variation in transistor.
I Have incorporated the concept in an old DH-200 amp and the result was quite noticeable (before and after comparison). There was more details which I could not notice before the mods. Almost all the peoples I who did not know about the mod told me that my sound system sounded "better" than before after noticing that the sound seemed more "real".
I do not have a Lavardin schematics but if you have a look at the US patent 5,635,874 it says everything you need to know. However, on my side I did not use the DC servo connected at the VAS. The low memory concept is also used in Halcro amp patents.
Fab
Lavardin
It is really interesting topic. I learn about "Heat Transfer" when I was a student in university. It is true, that in small dissipating area (like TO-92case), small heat will stack. But this is a slow process (compared to electric speed). Heat are moving very slowly, like having big dampers. There is "Thermal Inertia" that you cannot add or substract instantly (like heatsink cooling down, you cannot force it cool or to heat it up naturally in 1 second).
From the solutions offered, I saw it more as solutions to minimize transistor distortions.
From what I learn, transistor distortions coming from 2 point, the fluctuation of current and fluctuation of voltage.
Output stage suffer from both alot. VAS have big bias current, so the voltage fluctuation is dominant here.
The remedy of voltage induced distortions is to put cascode. The remedy of current induced distortions is to make the bias stands bigger than the fluctuating signal current.
I see all the "Thermal Inertia" theory remedy is the same as "Transistor distortion-elimination" theory.
If you can make both Voltage and Current steady in a transistor, you will get steady dissipation (P=VxI), would't you?
Lavardin theory call it steady dissipation, but I also can see it as steady voltage+steady current=minimum transistor distortion. No thermal consideration needed here.
It is really interesting topic. I learn about "Heat Transfer" when I was a student in university. It is true, that in small dissipating area (like TO-92case), small heat will stack. But this is a slow process (compared to electric speed). Heat are moving very slowly, like having big dampers. There is "Thermal Inertia" that you cannot add or substract instantly (like heatsink cooling down, you cannot force it cool or to heat it up naturally in 1 second).
From the solutions offered, I saw it more as solutions to minimize transistor distortions.
From what I learn, transistor distortions coming from 2 point, the fluctuation of current and fluctuation of voltage.
Output stage suffer from both alot. VAS have big bias current, so the voltage fluctuation is dominant here.
The remedy of voltage induced distortions is to put cascode. The remedy of current induced distortions is to make the bias stands bigger than the fluctuating signal current.
I see all the "Thermal Inertia" theory remedy is the same as "Transistor distortion-elimination" theory.
If you can make both Voltage and Current steady in a transistor, you will get steady dissipation (P=VxI), would't you?
Lavardin theory call it steady dissipation, but I also can see it as steady voltage+steady current=minimum transistor distortion. No thermal consideration needed here.
Oh! My input stage includes 20 bjt!😀
I am building it but I have not succeeded in it yet
Larvadin? do google with the word "peufeu"
I am building it but I have not succeeded in it yet
Larvadin? do google with the word "peufeu"
lumanauw said:
Yes, i use bipolar as driver and mosfet as final stage. This way
you should have best of both "worlds".
Hmm, it seems a strange idea to me to exchange this, but often
strange ideas are good...
Mike
Re: Lavardin
lumanauw
This is only my understanding and interpretation of the Lavardin patent:
All in all you are saying about the same thing as Lavardin finally. Lavardin says that in order to avoid thermal variation you need to keep the power constant in a device (thus both voltage and current as constant as possible). The thermal variation makes the operating point of the transistor to change a bit. This bad effect is amplified by the subsequent stage(s) with their gain. After all, you have also a higher DC offset variation at amp output that the overall feedback send to the diff input stage that will try to correct it. Of course, the higher the DC offset variation the "harder" it will be to correct it.
The temp. inertia of a small device is not comparable with a big dissipator. The thermal variation is named "memory" by Lavardin because of the thermal inertia since when the large signal has ended the device does not come back quickly enough to its "previous" temperature.
Fab
lumanauw said:
lumanauw
This is only my understanding and interpretation of the Lavardin patent:
All in all you are saying about the same thing as Lavardin finally. Lavardin says that in order to avoid thermal variation you need to keep the power constant in a device (thus both voltage and current as constant as possible). The thermal variation makes the operating point of the transistor to change a bit. This bad effect is amplified by the subsequent stage(s) with their gain. After all, you have also a higher DC offset variation at amp output that the overall feedback send to the diff input stage that will try to correct it. Of course, the higher the DC offset variation the "harder" it will be to correct it.
The temp. inertia of a small device is not comparable with a big dissipator. The thermal variation is named "memory" by Lavardin because of the thermal inertia since when the large signal has ended the device does not come back quickly enough to its "previous" temperature.
Fab
- Status
- Not open for further replies.