Yes, if using a cascode the cascode transistor sees both Ic and Vce variation, the vas transistor sees mostly Ic variation. Andy's circuit appears to mitigate these effects on the cascode transistor.
Another thing is the error current due to the variation of Vbe of the VAS transistor, this variation appearing in series with the miller cap - the biggest capacitance in the region. To reduce variation in Vbe you want the vas transistor to have maximum beta.
Having said this, I must say I think the impact of the output stage, power supplies and so on are much, much bigger contributors to non-linear distortion than the VAS transistors themselves.
Another thing is the error current due to the variation of Vbe of the VAS transistor, this variation appearing in series with the miller cap - the biggest capacitance in the region. To reduce variation in Vbe you want the vas transistor to have maximum beta.
Having said this, I must say I think the impact of the output stage, power supplies and so on are much, much bigger contributors to non-linear distortion than the VAS transistors themselves.
I have built this VAS and trying to measure it by injecting a voltage at the output and measure the current drawn...
So far I don't get distorted curves like in the simulations, but I still get a lot less parasitic current with the cascode (so it kinda works).
My super pairs oscillate at around 50-100 MHz, will try the resistor as you said.
So far, disappointing bench results... I'm bored. I'll get back to it later.
So far I don't get distorted curves like in the simulations, but I still get a lot less parasitic current with the cascode (so it kinda works).
My super pairs oscillate at around 50-100 MHz, will try the resistor as you said.
So far, disappointing bench results... I'm bored. I'll get back to it later.
I imagine that the cascode circuit will oscillate without the miller cap. The single transistor is probably just ok due to Cbc. You might have to add a small miller cap, say 10pF, in both cases (to allow comparison) and then measure the current. How are you measuring the current, by the way?
(...)So far I don't get distorted curves like in the simulations, but I still get a lot less parasitic current with the cascode (so it kinda works).
Did you duplicate the +/-30 Volt swing on the output that you used in your simulations? If you're using lower voltages, I can see that your currents would be a lot less distorted than what the simulation indicates. If you did duplicate this swing, say with a power amp hooked up to the signal generator, maybe it's a case of pessimistic forward and/or reverse Early voltages in the model?
My super pairs oscillate at around 50-100 MHz, will try the resistor as you said.
On one hand, I'm sorry to hear that the circuit has this problem. On the other hand, this does speak well for the ability of SPICE to find potential problems such as oscillation or instability. Let's hope the fix that works in SPICE also works in the real world! Let us know what happens when you get back to working on this. I'm curious to find out.
Did you duplicate the +/-30 Volt swing on the output that you used in your simulations? If you're using lower voltages, I can see that your currents would be a lot less distorted than what the simulation indicates. If you did duplicate this swing, say with a power amp hooked up to the signal generator, maybe it's a case of pessimistic forward and/or reverse Early voltages in the model?
My super pairs oscillate at around 50-100 MHz, will try the resistor as you said.
On one hand, I'm sorry to hear that the circuit has this problem. On the other hand, this does speak well for the ability of SPICE to find potential problems such as oscillation or instability. Let's hope the fix that works in SPICE also works in the real world! Let us know what happens when you get back to working on this. I'm curious to find out.
OK, I now understand how the super pair works ; reducing the linearities by reinjecting the main transistor's base current into the main circuit.
However for me it oscillates ! I've opened a thread for that if you can help me :
http://www.diyaudio.com/forums/showthread.php?s=&threadid=25172
Thanks a lot.
However for me it oscillates ! I've opened a thread for that if you can help me :
http://www.diyaudio.com/forums/showthread.php?s=&threadid=25172
Thanks a lot.
LukasLouw said:
Lukas
I have to say that I use a sound card in preference to the Audio precision.
I have found several good sound cards. 1394 and PCI. I have not tested any USB cards.
Please be aware that you are talking about $1000
There are some things that a sound card won't do that the AP will... optical eye patterns for instance.
And somethings that the APs don't doe well..7.1 channel analysis is a completle B^%$y nightmare
Do we need to start a new thread, or even a new forum group for this one?
Brian
bscally said:
Agreed about the AP comments, I have a System One, but it's getting long in the tooth though.
The main problem with soundcards is of course limited input capability for starters.
A new thread is a good idea, I don't know if there is enough interest for a new forum group on this subject, but why not try?
Lukas
Quote:
"If only a single diff amp is used, the only way to obtain a complementary second stage is to take differential outputs from the diff amp and to use a current mirror to drive the other side of the second stage."
The only way ?
Leach seems to lack references. Just have a look to different circuits having the same name : Sansui diamond circuit and Hafler diamond circuit.
Quote
"Matching the gain and phase characteristics of these paths is difficult."
I think this is the real problem of the complementaray pair, with the additionnal fact that the same point is submitted to two different feedback paths. You can't use as much as feedback using a complementary pair input as with a single differential. I am not aware of any test intrument using a complementary pair for better performance.
Input complementary pairs are one of the greatest audio technical mysteries to me.
~~~~ Forr
§§§
"If only a single diff amp is used, the only way to obtain a complementary second stage is to take differential outputs from the diff amp and to use a current mirror to drive the other side of the second stage."
The only way ?
Leach seems to lack references. Just have a look to different circuits having the same name : Sansui diamond circuit and Hafler diamond circuit.
Quote
"Matching the gain and phase characteristics of these paths is difficult."
I think this is the real problem of the complementaray pair, with the additionnal fact that the same point is submitted to two different feedback paths. You can't use as much as feedback using a complementary pair input as with a single differential. I am not aware of any test intrument using a complementary pair for better performance.
Input complementary pairs are one of the greatest audio technical mysteries to me.
~~~~ Forr
§§§
although a really very interesting thread, as it start smartly, but alas,
erroneous conclusions are concluded about a so called superiority of the
simple differential over the symetrical differential, based on andy s simulations
which are badly flawed in matter of scientific protocol...
unfortunately, and it s pityful, he s no more among us...
so far, i remade the simulations by re setting some erroneous parameters,
and contrary to the common belief that did emerge in this thread, the symetrical
differential has lower distorsion than the simple differential...
i ll publish the datas soon..
erroneous conclusions are concluded about a so called superiority of the
simple differential over the symetrical differential, based on andy s simulations
which are badly flawed in matter of scientific protocol...
unfortunately, and it s pityful, he s no more among us...
so far, i remade the simulations by re setting some erroneous parameters,
and contrary to the common belief that did emerge in this thread, the symetrical
differential has lower distorsion than the simple differential...
i ll publish the datas soon..
These discussions overlook what should be the starting point of discussion , the human auditive sense. That has 2 almost independent ways of acoustical perceptions. The "acustical higher cortex" allows to identify and eventually determine pitch and to put objectively contingent sound field events in some sequential order thus that we can understand spoken human languages and somehow "understand" music.
The second mode is in terms of evolution much older it bypasses any understanding but it alerts. That old predator/prey mode analyses transients almost real time and as such
exceeds the frequency limit of "intellectual hearing" by far as it can distinguish between
rise times of 10msec down to 20usec. Thus it operates time mode not frequency mode.
This sense allows us without intellectual speculation to determine, alone in wilderness, whether a soft rattling in the shrubs is a rabbit, a bear, or a snake. It determines the size of the sound source the velocity of approaching and hence the distance.
This sense alerts when there is no time for pondering possibilities and their probabilities.
I.e. it works "subconsciously"
This explains how one amp sounds good and the other doesn't although the objective measurement does not reveal differences. Energetically very small low energy spikes such as switching distortions and higher harmonics trigger the alert mode but we don't know what is going on thus there is some subjective "stress" that cannot be related other than to "sounds not so good".
So what is the conclusion re amps, and electro-mechanical transducers?
Switching and other crossover distortion must be avoided by all means. This is item priority #1
The thd figure does not matter what matters is an approximately geometrically (or close to exponential base e ) falling energy of the harmonics.
If this "exponential law" is met I bet no one could detect any difference between 10% h2
and 1% not even a trained musician.
The second mode is in terms of evolution much older it bypasses any understanding but it alerts. That old predator/prey mode analyses transients almost real time and as such
exceeds the frequency limit of "intellectual hearing" by far as it can distinguish between
rise times of 10msec down to 20usec. Thus it operates time mode not frequency mode.
This sense allows us without intellectual speculation to determine, alone in wilderness, whether a soft rattling in the shrubs is a rabbit, a bear, or a snake. It determines the size of the sound source the velocity of approaching and hence the distance.
This sense alerts when there is no time for pondering possibilities and their probabilities.
I.e. it works "subconsciously"
This explains how one amp sounds good and the other doesn't although the objective measurement does not reveal differences. Energetically very small low energy spikes such as switching distortions and higher harmonics trigger the alert mode but we don't know what is going on thus there is some subjective "stress" that cannot be related other than to "sounds not so good".
So what is the conclusion re amps, and electro-mechanical transducers?
Switching and other crossover distortion must be avoided by all means. This is item priority #1
The thd figure does not matter what matters is an approximately geometrically (or close to exponential base e ) falling energy of the harmonics.
If this "exponential law" is met I bet no one could detect any difference between 10% h2
and 1% not even a trained musician.
quite wrong...
take a 500 HZ pure sine , listen to it a few seconds,
and then add 1 % of h2..
it will be clearly audible...let alone if that s 10 %..
anyway, my point was to mention that two amplifiers
having the same GBw product at 20 Khz, but with
different GBW product under this frequency can t
be compared fairly, for some simple reasons, thus
that the simulations that where drawn about the lower
distorsion of the simple differenatial was due to
inadequate implementation of one design compared to
the other...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.