what the assymetric slew rate means is that the
transfer function of the amp is not linear in respect
of the dv/dt..
said more simply, in large signal mode, the amp
open loop gain and/or bandwith are variable with
the dv/dt, the amp will have less open loop gain
and bandwith in one side of the cycle...
of course, this gain/bandwith variation as a function
of the dv/dt will move the high frequency poles in
respect of the dv/dt...
transfer function of the amp is not linear in respect
of the dv/dt..
said more simply, in large signal mode, the amp
open loop gain and/or bandwith are variable with
the dv/dt, the amp will have less open loop gain
and bandwith in one side of the cycle...
of course, this gain/bandwith variation as a function
of the dv/dt will move the high frequency poles in
respect of the dv/dt...
It is.
You are correct.
What i'm looking for is a sort of symmetrical design which does NOT use a differential or complementar pair, but identical (=well matched or monolithic) transistors/FETs.
I still haven't seen one - maybe it's not possible...
Hi wahab,
Yes, I am aware what the mechanism is, and what effects it may have. I'm just trying to point out that if the HF response is very high, and we are not exercising these limits, the effects will be much less than other distortion mechanisms. I don't mean that as long as we aren't nearing slew rate limiting, I mean more that our maximum signal slew rate is << than the actual slew rate.
Hi Telstar,
There are also very successful designs that use a complimentary input pair (not a diff amp at all), matched of course! The average single diff pair performs extremely well. Search for the "symasym", there are a few variants available. Not only that, but I have built some and they do sound pretty good, much better than I assumed they would. So there is a proved DIY design with PCB files available as well.
-Chris
Yes, I am aware what the mechanism is, and what effects it may have. I'm just trying to point out that if the HF response is very high, and we are not exercising these limits, the effects will be much less than other distortion mechanisms. I don't mean that as long as we aren't nearing slew rate limiting, I mean more that our maximum signal slew rate is << than the actual slew rate.
Hi Telstar,
There are also very successful designs that use a complimentary input pair (not a diff amp at all), matched of course! The average single diff pair performs extremely well. Search for the "symasym", there are a few variants available. Not only that, but I have built some and they do sound pretty good, much better than I assumed they would. So there is a proved DIY design with PCB files available as well.
-Chris
Last edited:
300V/uS ?...That s quite generous, that would be less than
0.3 uS to swing from a rail to the other, assuming
a 100 w rms amp...
no wonder that the layout is critical...
a symetrical differential can yield those kind of speed,
but at the expense of stability, that i wonder how it is
stabilized..
i don t know this reference, but i ll check, for sure..
The 50 watt MOSFET amp with error correction that I did a long time ago, described on my website, did +/- 300 V/us. That's quite a bit for a 50W amp. It used a unipolar N-channel JFET input differential pair.
Often the key to symmetrical slew rate is to use a complementary VAS; the ones with only a current source load will usually not have symmetrical slew rate.
Complementary input differential pairs are only one way to drive a complementary VAS. Note that the APT1 also drives a complementary VAS with a unipolar BJT input LTP.
Keeping transistors out of saturation when slewing is also helpful to preserving symmetrical slewing behavior. The use of Baker clamps, as in the APT 1 and mine, is an example of this.
Cheers,
Bob
i have no doubt that such speed are do able, but this yield the
question : are these designs stable enough , what do the
impulse response look like?....and if a capacitive load is
added?..
eventually, a hard LR network at the output usually help,
but personnaly, i prefer to design for unconditionnal stability
without the use of these kinds of artefacts...
regards,
wahab
I think capacitive stability and RF rejection is more a function of the overall amp compensation and the way it is implemented not so much the output stage. It is important that each mosfet be compensated individually so that they don't disrupt the output and fb loop with sudden burst of RF oscillation.
Hi Wahab,
The MOSFET amp is very stable with excellent transient response. Much of the credit for the high slew rate goes to the unconventional compendation used. It is not straight Miller compensation where slew rate is limited by CLG, input stage gm and fc. I use Miller Input Compensation (MIC), which breaks these relationships. It is explained in my amplifier paper available on my web site.
As always, I do use an output L-R series circuit to isolate the load at high frequencies to preserve stability, but the coil is only 1 uH.
Cheers,
Bob
thanks for the infos, bob....
i ll check your design..
in my tries, no cdom compenstation was needed with
(lateral) mosfet output devices , provided the gates where
drived directly from the vas collector, as this constitute
a shunt compensation...
alas, distorsion is higher than the same amp with the vas
output buffered with a pair of emitter followers..
in this latter case, miller compensation was mandatory to
compensate the added phase shift,although no more than
5 to 7 pF was needed, as the input stage transconductance
was vonlontarly limited..
this only allow for slew rates in the range of 30 to 50 V/uS ,
as unconditionnal stability is one of my key requirement..
i did test lead compensations, but it happens that
it didn t yield the expected results..
slew rate is better, though, but that all..
the value of 1uH you mention is correct..
this is the exact value i use, with a 1R resistor..
regards,
wahab
Hi wahab,
It is for this reason that I never recommend driving even lateral MOSFETs directly from the VAS. Their input capacitance is nonlinear. Don't let the lateral MOSFET capacitance be a part of the compensation scheme. If you are willing to heavily degenerate the input stage to get its gm down, even Miller compensation can yield decent slew rates. Just watch the input-referred noise of the amplifier. JFETs are often a better choice overall, parly because they do not have input current noise.
Anything less than about 20 nV/rt Hz is decent noise performance for a power amplifier. 5 nV/rt Hz is considered stellar.
Cheers,
Bob
Hi Bob,
Just think, Counterpoint (it's just too easy) attempted to drive output mosfets using a 6DJ8 in the SA-100 / SA-220 amplifiers. *Unclear on the concept ...
Hi wahab,
The only thing I can really say to you is to not get hung up on a couple features in an amplifier design. A well executed "okay" design will outperform a poorly executed "great" design every single time. There are so many factors that come together to determine the way an amplifier sounds that focusing on one thing (or two) may completely mislead you.
Jan Didden also has designed a very good sounding amplifier that uses mosfet outputs. Personally, I'm not a fan of mosfets, but I have heard his amplifier and it sounds very good. Building a low cost design like the Symasym or one of it's variants may give you valuable insight as well. You can always reuse the case, heat sinks and power supply in another design if you don't like it. I think this is a good, low cost way to learn about amplifiers and may even be what you want at the end of the day. One thing it has going for it is that many have been built, and many variants as well. It was well received and is perhaps a good "known quantity". I would say the same for Bob's and Jan's amplifiers as well.
-Chris
Here you have hit the nail on the head as one of the major design errors made in mosfet amplifiers. Even using a buffer can be a joke here, I'm referring to those designs that drive the output stage with a pair of TO-92 transistors at light bias current. The message that you need to charge and discharge the gates seems lost even to this day.
Just think, Counterpoint (it's just too easy) attempted to drive output mosfets using a 6DJ8 in the SA-100 / SA-220 amplifiers. *Unclear on the concept ...
Hi wahab,
The only thing I can really say to you is to not get hung up on a couple features in an amplifier design. A well executed "okay" design will outperform a poorly executed "great" design every single time. There are so many factors that come together to determine the way an amplifier sounds that focusing on one thing (or two) may completely mislead you.
Jan Didden also has designed a very good sounding amplifier that uses mosfet outputs. Personally, I'm not a fan of mosfets, but I have heard his amplifier and it sounds very good. Building a low cost design like the Symasym or one of it's variants may give you valuable insight as well. You can always reuse the case, heat sinks and power supply in another design if you don't like it. I think this is a good, low cost way to learn about amplifiers and may even be what you want at the end of the day. One thing it has going for it is that many have been built, and many variants as well. It was well received and is perhaps a good "known quantity". I would say the same for Bob's and Jan's amplifiers as well.
-Chris
Member
Joined 2009
Paid Member
well, although driving lateral fets directly from the vas yield higher
distorsion, i measured it at 0.05 % at 10KHZ 58 V PP, wich is quite low
considering that there s not much NFB...
adding a pair of emitter follower bring it to about 0.015% with
the same settings...
noise is not an issue, as the contribution of the amp alone
is about 4.5 nV/sqrt hz , but of course, this will be ruined by
the source generator impedance...
reducing distorsion further imply using currents mirrors
for the differentials , and in this case, distorsion is reduced
to anectdotical levels, but at the expense of some gain and
phase margin...
regards,
wahab
hi , anatech...., i did stick for lateral mosfets
for a while , as well as symetrical differential topology; and it
just happen that you seems to dislike the two. !!....
Hi wahab,
You're decided, so there really isn't any point in talking about much else.
What I know comes from both theory and observations of real equipment over many years. However, in case you are lucky and the design is really operating this well, try some On-Semi or IR mosfets. That oughtta cure you of this. Remember that there was a well regarded brand (Musical Fidelity) in the 80's that designed and amplifier very similar to yours. The output bias was very high and it still didn't really sound that good. Look them up just for giggles and laughs. Some people swear by these still even today.
Come to think of it, this sounds like a Greg Ball idea. I mean - exactly like his SKA, so you may have a clone here.
-Chris
I wouldn't say that. I just wouldn't do this myself.
You're decided, so there really isn't any point in talking about much else.
Several early designs have done the same thing with similar claims. Did you test under load? 4 or 8 ohm with any reactance? If this is simulated, you may be in for some disappointment in real life.
What I know comes from both theory and observations of real equipment over many years. However, in case you are lucky and the design is really operating this well, try some On-Semi or IR mosfets. That oughtta cure you of this.
Remember that there was a well regarded brand (Musical Fidelity) in the 80's that designed and amplifier very similar to yours. The output bias was very high and it still didn't really sound that good. Look them up just for giggles and laughs. Some people swear by these still even today. Come to think of it, this sounds like a Greg Ball idea. I mean - exactly like his SKA, so you may have a clone here.
-Chris
hi chris,Several early designs have done the same thing with similar claims. Did you test under load? 4 or 8 ohm with any reactance? If this is simulated, you may be in for some disappointment in real life.
What I know comes from both theory and observations of real equipment over many years. However, in case you are lucky and the design is really operating this well, try some On-Semi or IR mosfets. That oughtta cure you of this.
-Chris
0.05% at10khz is not exactly what i was looking for,
but it was possible with a symetrical differential plus the two vas
that did drive the output stage that consisted of three pairs
of 2SJ48/2SK133...this are the same devices as 2SJ160/2SK1056
if we except the obsolete TO3 casing..
a buffer between the vas and output stage did reduce it by
three fold...
i don t use IRF devices, they doesn t suit the purpose, unless
it is for class A operation, and even in that case, i would stick
for the hitachi s...
nelson pass manage to use them properly at the expense of
vast amount of power losses; anyway, that s not the kind of designs
i m looking for, would it be only because they are anti ecological...
Hi wahab,
Did you measure these results, or are they the result of a simulation?
If you measured this figure, it's important to know what instrument you used. Many newer THD analyzers made by Kenwood and Leader (or similar) will give you overly optimistic readings at higher frequencies. I discovered this myself with a Leader that I paid $2,500 (or was it $2K, I bought a wow & flutter meter a the same time) for new. Why? The high frequency response was down 3dB at 30 KHz or so. Look at a 10 KHz test frequency, you will read the 2nd harmonic (already measuring low) and the 3rd will be 3 db down. Higher harmonics will not show up. Changing to an older HP THD meter revealed the truth. They will read the 10th harmonic easily enough, and the higher readings properly reflect what is going on. Measuring the THD at 10 KHz or 20 KHz is a great idea, but only if your equipment doesn't lie to you. Otherwise you should restrict your readings to 1KHz or 2 KHz. Any higher and you could be missing many harmonics.
I am absolutely not trying to talk your design down. It's just that I am more than a bit familiar with it. Nelson tends to take a simple circuit and get very good performance from it. I have a lot of respect for him, but you have to look at everything he does. There are very good reasons why he does anything. The higher bias current is one example. There may also be a much higher standing current in the Vas. These are things I haven't looked into with this design - my goals and ways of doing things are different than his. Also, Nelson is commercially successful whereas I am not. I just repair what other people have designed. So the way that Nelson looks at a design will not be the same as mine. I do intend to try some of his current designs out for myself, and I have already tried some of Nelson's earlier designs.
Anyway, I am interested to see how you came about the figures you quoted for distortion. Can you answer my questions in post #377? I do not intend to debate you, I'm only curious about the performance of your circuit. I'm sure other members might have some interest in this in case they would like to build your circuit.
-Chris
Did you measure these results, or are they the result of a simulation?
If you measured this figure, it's important to know what instrument you used. Many newer THD analyzers made by Kenwood and Leader (or similar) will give you overly optimistic readings at higher frequencies. I discovered this myself with a Leader that I paid $2,500 (or was it $2K, I bought a wow & flutter meter a the same time) for new. Why? The high frequency response was down 3dB at 30 KHz or so. Look at a 10 KHz test frequency, you will read the 2nd harmonic (already measuring low) and the 3rd will be 3 db down. Higher harmonics will not show up. Changing to an older HP THD meter revealed the truth. They will read the 10th harmonic easily enough, and the higher readings properly reflect what is going on. Measuring the THD at 10 KHz or 20 KHz is a great idea, but only if your equipment doesn't lie to you. Otherwise you should restrict your readings to 1KHz or 2 KHz. Any higher and you could be missing many harmonics.
I am absolutely not trying to talk your design down. It's just that I am more than a bit familiar with it. Nelson tends to take a simple circuit and get very good performance from it. I have a lot of respect for him, but you have to look at everything he does. There are very good reasons why he does anything. The higher bias current is one example. There may also be a much higher standing current in the Vas. These are things I haven't looked into with this design - my goals and ways of doing things are different than his. Also, Nelson is commercially successful whereas I am not. I just repair what other people have designed. So the way that Nelson looks at a design will not be the same as mine. I do intend to try some of his current designs out for myself, and I have already tried some of Nelson's earlier designs.
It's also possible with single differential, and probably also with a singleton and complimentary front end. There is not point made there. The excessive matching required to achieve good quality is beyond what you need to do with the other types. As I've said before, the differences between matched and unmatched devices is audible. That is a large difference.
If the front end is matched, you may find the difference the buffer makes is even more dramatic. Honestly, anything that allows for a 3 to 1 reduction in distortion would be implemented unless the quality of the amp was not important.
That may be, but there are several people who do use them. The point was a demonstration to show you how important a Vas buffer really is. The other comment I could make is that there have been several commercial designs sold using the IR components. They are commercially successful, so by that definition, they are suitable for use in an audio amplifier. In the real world, I tend to partially agree with you on this.
Anyway, I am interested to see how you came about the figures you quoted for distortion. Can you answer my questions in post #377? I do not intend to debate you, I'm only curious about the performance of your circuit. I'm sure other members might have some interest in this in case they would like to build your circuit.
-Chris
hi, chris,
these results were measured with distorsion meter and an
spectrum analyzer...
both of these instruments were costly at the time (1987,
and still are) so i had to rely on a friend who own an audiophile
store by there...i don t remember exactly the models, but it was
hewlett packard gears..
distorsion was measured under a 8R load at a voltage of
60V PP, the amp maximum ouptut being about 90V PP..
bandwith used for 10KHZ distorsion measurement was 200khz..
at 1KHZ in the same conditions, distorsion is about 0.01%,
not an exceptionnal value, as the design has not much more
open loop gain at this frequency..
in fact, it rely on a wide open loop bandwith to provide
enough, though modest globally, NFB at the higher frequencies,
making the compensation capacitor useless without harming
stability in any way...
adding a pair of EF buffer to drive the mosfets reduced the thd
substiancially, but as amazing as it appear, i retained the simpler
version, without these buffers..
the design proved a stunning reliability, as i used it for the next
ten years on stage for my set of keyboards; and i can tell
you that being a musician amp is not an easy life...
as i pointed, i don t share a lot with nelson pass, but there s an area
where i agree with him, power mosfets and simplicity of the designs..
i was about to publish these designs in this site, but it happens that
a after recent event about a member that did great technical contributions
and was banned from here, i realized that DIY is not exactly what i thought
it was supposed to be..this is to say the least...
so i ll refrain publishing more designs (i published only one, very simple by the way)
as long as the situation will appear to me as troubling...
that said, i m willing to share them privately with whoever respect intellectual
property, though it s not a revolutionnary thing, it s just that it s designed
with some ideas (which i didn t invented either) that are often forgotten...
regards,
wahab
these results were measured with distorsion meter and an
spectrum analyzer...
both of these instruments were costly at the time (1987,
and still are) so i had to rely on a friend who own an audiophile
store by there...i don t remember exactly the models, but it was
hewlett packard gears..
distorsion was measured under a 8R load at a voltage of
60V PP, the amp maximum ouptut being about 90V PP..
bandwith used for 10KHZ distorsion measurement was 200khz..
at 1KHZ in the same conditions, distorsion is about 0.01%,
not an exceptionnal value, as the design has not much more
open loop gain at this frequency..
in fact, it rely on a wide open loop bandwith to provide
enough, though modest globally, NFB at the higher frequencies,
making the compensation capacitor useless without harming
stability in any way...
adding a pair of EF buffer to drive the mosfets reduced the thd
substiancially, but as amazing as it appear, i retained the simpler
version, without these buffers..
the design proved a stunning reliability, as i used it for the next
ten years on stage for my set of keyboards; and i can tell
you that being a musician amp is not an easy life...
as i pointed, i don t share a lot with nelson pass, but there s an area
where i agree with him, power mosfets and simplicity of the designs..
i was about to publish these designs in this site, but it happens that
a after recent event about a member that did great technical contributions
and was banned from here, i realized that DIY is not exactly what i thought
it was supposed to be..this is to say the least...
so i ll refrain publishing more designs (i published only one, very simple by the way)
as long as the situation will appear to me as troubling...
that said, i m willing to share them privately with whoever respect intellectual
property, though it s not a revolutionnary thing, it s just that it s designed
with some ideas (which i didn t invented either) that are often forgotten...
regards,
wahab
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.