I tried that with music, also. With single tone it is easy to detect harmonics, especially at lower end. Music is a different story. In my listening tests I applied 1% and 5% to a wave file. When comparing the original and 5% distorted wave file, all that I could conclude is that it sounds different but, I couldn't distinguish what is the original recording and what is the distorted recording. When comparing 1% distorted music with the original recording I could barely detect any differences. Maybe I'm deaf 🙂
fairly true...
i only wanted to point that in certain circumstances,
distorsion is clearly audible..
Now try it with a square wave and a pure sine... and tell me what at what frequency you are unable to differentiate the "timbre" of the sine and square.
Yes yes my reference is to reproduction of sound fields i.e. in this case musical performance. The mental image of the visual as well as of acoustical is a computation. Certainly not a formal computation ( a von Neuman digital computer is a formally deciding linear system) otherwise undecidability, unidentifiability ( Tarski) and halting ( Turing) would apply. All of these may apply to thinking. But at least I do not think about music.
This said computation is dependent on "mood" and "mood" is dependent on computation. So if there is some alert overshadowing the no-firstness mood computation interaction one could indeed start to think about what you hear
and that is not what music is intended for.
back to the thread, here thd figures for the two contendings
topologies....
bandwith and ol gain are adjusted so none of the topology
is blatantly favoured, with both having same differential and vas
bias currents , and same degeneration resistors....
the symetrical amp had to be restrained and brought to the level
of the simple differential for fair comparisons...
the symetrical seems clearly better in cancelling distorsions...
topologies....
bandwith and ol gain are adjusted so none of the topology
is blatantly favoured, with both having same differential and vas
bias currents , and same degeneration resistors....
the symetrical amp had to be restrained and brought to the level
of the simple differential for fair comparisons...
the symetrical seems clearly better in cancelling distorsions...
May we see what schematic have you used for symmetrical differential input with current mirrors? Did you used common mode control like Edmond in his designs or something like GK have used in his A class amp posted here on forum?
neither of the two, i think..
i think that the edmond s amp you re talking about is the
one that syn08 did (co) design..
elegant amp, for sure, with many good implementations,
although i wouldn t venture building it...
i m recent by there, so i still don t know what andy proposed,
i will check the site....
for sims, i simply replaced the collectors load by current mirrors
to enhance the ac loading to the level of the simple differential
that was loaded in such way, providing vastly higher open loop
gain..
provided the amp don t clip, it work, but if there s a rail sticking,
the vas would blow out if ever the time recovery was too long..
anyway, it s enough to have some clues about the dynamic behaviour
of the topology..
of course, in my practical tests, i didn t use this simplified design,
though it corroborate that the symetrical topology has a better
behaviour than the single ended version..
The problem with symmetrical differential input with simple current mirrors is the ill defined standing current in VAS. It sometimes work wonderfully in simulations but it utterly fails in practice. I think it's better, and if you have time, to choose a method for controlling the VAS standing current and re-simulate it and post the conclusions here. GK had to give up off some OLG in his symmetrical design to control the VAS current. I haven't done deep analyses of Edmond's implementation, but you can look up for an inspiration at his site (not Syn08's site).
yes, there s some issues, but imho, this is the best topology to date,
it can be made to work in real world with no extreme difficulties...
Hi wahab,
In practice, not in a simulator, the simple differential pair is something that can be manufactured. The symmetrical differential pair requires very careful matching before it will deliver good performance and does not lend itself to a production line. Each transistor must be matched to the same beta and have the same tail current, if not the sound quality does suffer. With the simple differential pair, you're only tied to matching one pair and they are the same polarity. Some mismatch does not seem to create the audio penalties that mismatches cause with the symmetrical design. You can't even rely on THAT Semiconductor matched pairs in the 340. They guarantee a match between PNP pairs and NPN pairs, but state that there is no match attempted between the two polarities of pairs.
Once everything is working well, comparing a simple differential pair to the symmetrical differential pair, I have to say that I prefer the simple differential pair when listening. That represents a best case scenario for the symmetrical design, and not what you see in real life. Add to that the problems in matching J-Fets and you're really up the creek! N and P channel J-Fets are not complimentary, period.
The measurements I have made over many years back up what I hear, so I'm pretty confident about my opinion.
-Chris
In practice, not in a simulator, the simple differential pair is something that can be manufactured. The symmetrical differential pair requires very careful matching before it will deliver good performance and does not lend itself to a production line. Each transistor must be matched to the same beta and have the same tail current, if not the sound quality does suffer. With the simple differential pair, you're only tied to matching one pair and they are the same polarity. Some mismatch does not seem to create the audio penalties that mismatches cause with the symmetrical design. You can't even rely on THAT Semiconductor matched pairs in the 340. They guarantee a match between PNP pairs and NPN pairs, but state that there is no match attempted between the two polarities of pairs.
Once everything is working well, comparing a simple differential pair to the symmetrical differential pair, I have to say that I prefer the simple differential pair when listening. That represents a best case scenario for the symmetrical design, and not what you see in real life. Add to that the problems in matching J-Fets and you're really up the creek! N and P channel J-Fets are not complimentary, period.
The measurements I have made over many years back up what I hear, so I'm pretty confident about my opinion.
-Chris
in my experience, i rarely encountered problems to make
a simetrical diff. working the good way no matter the output
stage being with bipolar or mosfets..
as far as i can remember, i sticked with it once i did built one
in the late 70s...
there s no need for exterme matching, contrary to common beliefs,
if the implementation is cleverly thought..
but i concede that a simple differential is more tolerant in the final
result if not well designed than a symetrical variant..
a simetrical diff. working the good way no matter the output
stage being with bipolar or mosfets..
as far as i can remember, i sticked with it once i did built one
in the late 70s...
there s no need for exterme matching, contrary to common beliefs,
if the implementation is cleverly thought..
but i concede that a simple differential is more tolerant in the final
result if not well designed than a symetrical variant..
Hi wahab,
I have never had any problems servicing or building these designs either, but then it helps to know what is important to begin with. The outputs have nothing to do with the voltage amplifier stage though. The J-Fets I mentioned where for the diff pair(s).
If it counts, I have found BJT output stages to be far more linear, and easier to drive and control than the run of the mill mosfet stage.
Personally, I find the extra complexity and aggravation in matching the extra parts a no starter. A ton of extra work would have to sound much better to be worth the effort. The best symmetrical sounds about as good as a single diff pair in my book. I will admit that I don't have as good a feel for this type of design as someone like John Curl does, maybe his sound far better than mine.
-Chris 🙂
I have never had any problems servicing or building these designs either, but then it helps to know what is important to begin with. The outputs have nothing to do with the voltage amplifier stage though. The J-Fets I mentioned where for the diff pair(s).
If it counts, I have found BJT output stages to be far more linear, and easier to drive and control than the run of the mill mosfet stage.
I'll disagree with you on this. You don't have to carefully match anything, but if you do, the performance is noticeably improved. If you are building an amplifier, why wouldn't you match transistors when it makes so much of a difference? That and the common belief is actually one where matching is considered not important by most people.
Personally, I find the extra complexity and aggravation in matching the extra parts a no starter. A ton of extra work would have to sound much better to be worth the effort. The best symmetrical sounds about as good as a single diff pair in my book. I will admit that I don't have as good a feel for this type of design as someone like John Curl does, maybe his sound far better than mine.
-Chris 🙂
fact is that a simple differential using a current mirror
and a currrent source as vas load has enormous gain..
the current mirror does increase the gain
well over the commonly believed doubling, as it
act in fact as negative resistors, thus, there is
lot of negative feedback available...
of course, the collectors of the differentials can now
swing the full relevant rail voltage, but the following
vas is protected against such swings by its current source
collector loading..
this technic is not as easily implementable in a
symetrical differential, as the vas are connected symetricaly,
each acting as a negative resistance vis a vis the side that is
conducting, but as in the simple diff., there s no vas current limitation
in this topology, and implementing one can reduce to zero the benefits
if not well implemented..
anyway, even without current mirrors, wich is a comparatively
huge disadvantage, a symetrical diff. using resistors as collector
loads still display as good perfs as a well designed simple differential..
what make me reluctanty with the latter is stability and among aother
things, it s lack of symetry in matter of slew rate...
and a currrent source as vas load has enormous gain..
the current mirror does increase the gain
well over the commonly believed doubling, as it
act in fact as negative resistors, thus, there is
lot of negative feedback available...
of course, the collectors of the differentials can now
swing the full relevant rail voltage, but the following
vas is protected against such swings by its current source
collector loading..
this technic is not as easily implementable in a
symetrical differential, as the vas are connected symetricaly,
each acting as a negative resistance vis a vis the side that is
conducting, but as in the simple diff., there s no vas current limitation
in this topology, and implementing one can reduce to zero the benefits
if not well implemented..
anyway, even without current mirrors, wich is a comparatively
huge disadvantage, a symetrical diff. using resistors as collector
loads still display as good perfs as a well designed simple differential..
what make me reluctanty with the latter is stability and among aother
things, it s lack of symetry in matter of slew rate...
Hi wahab,
Did you built amplifier like that in the first post: http://www.diyaudio.com/forums/solid-state/16796-unstable-vas-current-amp-slone-book.html#post195526
not this particular one, but i experimented this kind of implmentation,
although i soon abandonned the cascoded vas scheme...
it surprise me that they had trouble making the thing work
decently...
i will read the thread, but i suppose that too much open
loop gain along with vast slew rate were the target...
It is worth mentioning that using one simple differential input doesn't automatically mean unsymmetrical slewing. Furthermore, I think, but didn't investigated into it a lot, that you can achieve symmetrical slewing even with standard Self topology amplifier, only if your current source in the upper part of the VAS is adequate.
no, in such designs, the impulse response is always faster on
the rail where the vas emitter is connected, as this side has no
current limiting contrary to the current source rail side..
increasing the cuurent source current will change nothing,
apart from increasing distorsion and heatings..
some have tried to remedy this by implementing a differential
vas whose added collector is used to boost the current source
current during the relevant cycle, but its action is marginal,
as it s device and thus temperature dependant...
there s still the possibility of using a current mirror as
the differential vas load, and that s a better solution..
it has been extensevely used particularly to drive mosfets..
OK, I should said almost symmetrical. In my opinion, the difference between positive and negative slewing can be so small that the significance is academic only.
yes, generally thanks to a (big) cdom that will decrease
enough the slew rate of the fast side to bring it to parity..
enough the slew rate of the fast side to bring it to parity..
Hi wahab,
I don't think that having a symmetrical slew rate is at all necessary. I can't figure out why this would even seem to be important to begin with. Let's reason this out and you can let me know where I've strayed.
The slew rate limit will be extremely high compared to the bandwidth normal operation runs at. In other words, you should never even approach slew rate limiting.
Normally, one would restrict the upper frequency response at the input. This serves more that one purpose, as it protects against some RFI. The zobel and inductance in the output helps protect the other input (feedback network). The attenuation of 24 ~ 35 dB doesn't hurt either.
Now what you have is a bandwidth limited, wide band amplifier. That seems to be a good idea as signal induced slew rate limiting is far less likely to occur. The only possible drawback I could possibly imagine is a possible link between audible "jitter" as the high frequency poles move around, or interaction between the variable slew rates and time domain errors. Until then, I'll not worry about it. Just make the slew rates semi-equal and don't sweat it past that.
Let's imagine that it is discovered that the time domain errors do turn out to be very important. Even a symmetrical diff pair setup may easily have differing slew rates. PNP and NPN transistors can be close compliments, but they are not mirror images in performance. Asymmetrical slew rates are still very likely unless care is taken to match the slew rates. We're back to similar issues with a single diff pair.
I still don't think this is really important right now. It's an effect that occurs well outside the normal operating areas. Another example where some things we obsess or worry about have little to no impact on what is actually going on.
As far as Cdom is concerned, many well designed amplifiers either buffer the Vas, or they run it in a cascode arrangement. A variable Cdom just dropped out of view. Yamaha tends to run the Vas in cascode for example.
-Chris
I don't think that having a symmetrical slew rate is at all necessary. I can't figure out why this would even seem to be important to begin with. Let's reason this out and you can let me know where I've strayed.
The slew rate limit will be extremely high compared to the bandwidth normal operation runs at. In other words, you should never even approach slew rate limiting.
Normally, one would restrict the upper frequency response at the input. This serves more that one purpose, as it protects against some RFI. The zobel and inductance in the output helps protect the other input (feedback network). The attenuation of 24 ~ 35 dB doesn't hurt either.
Now what you have is a bandwidth limited, wide band amplifier. That seems to be a good idea as signal induced slew rate limiting is far less likely to occur. The only possible drawback I could possibly imagine is a possible link between audible "jitter" as the high frequency poles move around, or interaction between the variable slew rates and time domain errors. Until then, I'll not worry about it. Just make the slew rates semi-equal and don't sweat it past that.
Let's imagine that it is discovered that the time domain errors do turn out to be very important. Even a symmetrical diff pair setup may easily have differing slew rates. PNP and NPN transistors can be close compliments, but they are not mirror images in performance. Asymmetrical slew rates are still very likely unless care is taken to match the slew rates. We're back to similar issues with a single diff pair.
I still don't think this is really important right now. It's an effect that occurs well outside the normal operating areas. Another example where some things we obsess or worry about have little to no impact on what is actually going on.
As far as Cdom is concerned, many well designed amplifiers either buffer the Vas, or they run it in a cascode arrangement. A variable Cdom just dropped out of view. Yamaha tends to run the Vas in cascode for example.
-Chris
- Status
- Not open for further replies.