Re: Jfet Cascode + Cascode Current Source
I don't think anyone can answer that question for you. I have experimented a little with those active devices and I cannot hear the differrence from a resistor and a current mirror (or a current source on the tail).
But then I cannot hear golden feet from rubber feet so who knows.
Maybe we should do a survey to see who can / cannot hear current mirrors?
LineSource said:
I don't think anyone can answer that question for you. I have experimented a little with those active devices and I cannot hear the differrence from a resistor and a current mirror (or a current source on the tail).
But then I cannot hear golden feet from rubber feet so who knows.
Maybe we should do a survey to see who can / cannot hear current mirrors?
LineSource,
presumably you intend to run the (current) output of the gm stage into a low impedance so the appropriate measure of the output performance is to measure the current in a low impedance, typically the output conductance and linearity vs output voltage are measured separately
cascoding fets with identical devices doesn't give the linearity and go improvement you might be expecting, the "inner" device is guaranteed to be biased in triode mode when cascoded with a matching device, the "outer" devices should have larger Idss and operate with vg > vg_off of the inner fet
bjts are even better cascoding devices with respect to stabilizing the inner device Vds, if you can figure out a good bias arrangement they will have much lower Z at the emitter than even properly selected fets at similar currents
Fred,
You are quite correct that the substrate diodes will have negligible effects in typical current mirror apps, I just am annoyed that they are seldom shown in part schematics and often the substrate node is not brought out or is unlabeled for the (few) occasions you might want to terminate or bootstrap it – and this hobby horse of mine had little relevance to the larger issues, can I email my future posts to you for editing?
presumably you intend to run the (current) output of the gm stage into a low impedance so the appropriate measure of the output performance is to measure the current in a low impedance, typically the output conductance and linearity vs output voltage are measured separately
cascoding fets with identical devices doesn't give the linearity and go improvement you might be expecting, the "inner" device is guaranteed to be biased in triode mode when cascoded with a matching device, the "outer" devices should have larger Idss and operate with vg > vg_off of the inner fet
bjts are even better cascoding devices with respect to stabilizing the inner device Vds, if you can figure out a good bias arrangement they will have much lower Z at the emitter than even properly selected fets at similar currents
Fred,
You are quite correct that the substrate diodes will have negligible effects in typical current mirror apps, I just am annoyed that they are seldom shown in part schematics and often the substrate node is not brought out or is unlabeled for the (few) occasions you might want to terminate or bootstrap it – and this hobby horse of mine had little relevance to the larger issues, can I email my future posts to you for editing?
To current mirror or not
Hi Line source,
You should take into account if you have a "normal" PNP stage after the input or a folded cascode. The latter has a low input impedance so I believe the current mirrors loading the first stage don't make much sense. I tried allmost all conceivable configurations but abandoned the current mirrors after the first stage.
Contrary to your approach I did not simulate a thing but listened to all implementations......
I found the folded cascode better sounding than a second PNP differential stage like in the OPA627.😎
Hi Line source,
You should take into account if you have a "normal" PNP stage after the input or a folded cascode. The latter has a low input impedance so I believe the current mirrors loading the first stage don't make much sense. I tried allmost all conceivable configurations but abandoned the current mirrors after the first stage.
Contrary to your approach I did not simulate a thing but listened to all implementations......
I found the folded cascode better sounding than a second PNP differential stage like in the OPA627.😎
Hi Charles and Nelson.
I was refering to rf mirroring, not just DC; thereby the potential for reducing amplifier distortion due to more accurate NFB activity.
I have not used 'A' in my own circuits because it is better with a slightly higher minimum voltage than 'B'.
Increasing the collector voltage of the first transistor in 'B', with a diode or LED paralleled with a small C in series with the bases, is also better for NFB controlled accuracy.
This is the version I tend to use in amplifier circuits. You could try it.
Cheers ............. Graham.
I was refering to rf mirroring, not just DC; thereby the potential for reducing amplifier distortion due to more accurate NFB activity.
I have not used 'A' in my own circuits because it is better with a slightly higher minimum voltage than 'B'.
Increasing the collector voltage of the first transistor in 'B', with a diode or LED paralleled with a small C in series with the bases, is also better for NFB controlled accuracy.
This is the version I tend to use in amplifier circuits. You could try it.
Cheers ............. Graham.
Charles Hansen said:
first cycle distortion?
🙂
Graham Maynard said:
what's RF mirroring? and how would it better at "RF mirroring"?
Re: Jfet Cascode + Cascode Current Source
Many people apparently can. Myself, I prefer a resistor loading
the diff pair on an amplifier like the Aleph, but I've had people
who preferred current sources or current mirrors, presumably
because they boost the open loop gain (particularly at low
frequencies), increasing the amount of feedback and creating
a different spectral balance to the sound.
LineSource said:
Many people apparently can. Myself, I prefer a resistor loading
the diff pair on an amplifier like the Aleph, but I've had people
who preferred current sources or current mirrors, presumably
because they boost the open loop gain (particularly at low
frequencies), increasing the amount of feedback and creating
a different spectral balance to the sound.
Charles and Nelson,
'Why would a single mirror be better ?' has been asked.
Let me ask another question in return;-
Would I be so foolish as to make such a statement without having actually tested each (and other similar) sub-circuit in isolation for hf phase accuracy, to thus find the best for minimising the inaccuracy that such a stage would introduce.
Also then make doubly sure by running all types within amplifier topologies to see which best supports phase coherent NFB.
I have.
Has anyone else ?
Like you Millwood who delights in trying to belittle me ?
Because I cannot get away from Millwoods distracting bullying in diyAudio, I am signing off.
'Why would a single mirror be better ?' has been asked.
Let me ask another question in return;-
Would I be so foolish as to make such a statement without having actually tested each (and other similar) sub-circuit in isolation for hf phase accuracy, to thus find the best for minimising the inaccuracy that such a stage would introduce.
Also then make doubly sure by running all types within amplifier topologies to see which best supports phase coherent NFB.
I have.
Has anyone else ?
Like you Millwood who delights in trying to belittle me ?
Because I cannot get away from Millwoods distracting bullying in diyAudio, I am signing off.
1 ohm and class A, sounds like Apogee and Krell
Hi Line Source,
do you remember the Apogee Ribbon speakers famous for their
'lo-lo impedance' ?
They were often combined with Krell 50, 100W amps.
BTW I built a 50W Class A amp recently based on a french design from L'audiophile magazine(80ies) . The transistors were out of production and the compare type was the same motorola type Krell used.
Thomas
LineSource said:
Hi Line Source,
do you remember the Apogee Ribbon speakers famous for their
'lo-lo impedance' ?
They were often combined with Krell 50, 100W amps.
BTW I built a 50W Class A amp recently based on a french design from L'audiophile magazine(80ies) . The transistors were out of production and the compare type was the same motorola type Krell used.
Thomas
Graham (assuming you are reading this),
Perhaps my post gave the wrong impression. I was not trying to harass you, but instead trying to learn something from you. That is because it is clear (to me, at least) that you have a lot to offer.
I was confused by your original post because you said that a Wilson mirror was "better" than a standard mirror, which in turn was "better" than a cascode mirror. I haven't tried all three of these extensively, which is why I was hoping to learn something from you.
There are (at least) two sources of error found in current mirrors:
a) DC inaccuracy due to current "lost" to drive the bases. This can be improved with a Wilson mirror or by using a 4th type of mirror where an emitter follower drives the bases.
b) DC inaccuracy due to the Early effect, whereby Vce is unequal across the two devices. This can be improved with either a Wilson mirror or a cascoded mirror.
I was confused by your ranking order because (at DC, at least) both the Wilson and cascode mirrors should be better than the standard mirror.
In your original reply you stated that you use a (slightly modified!) standard mirror due to its HF improvement, presumably over both the Wilson and cascode mirrors. Again this is somewhat confusing to me because I would expect an improvement at HF from the cascode mirror (similar to cascoding any circuit).
Again, if I thought you were wrong about this, I would either say so or ignore your post. In this case I believe that you have a lot of practical experience, and I'm hoping to learn from you. Anyway, I hope you are still reading this and willing to provide an answer. If you don't want to post here, perhaps you could send a private message.
Thanks,
Charles Hansen
Perhaps my post gave the wrong impression. I was not trying to harass you, but instead trying to learn something from you. That is because it is clear (to me, at least) that you have a lot to offer.
I was confused by your original post because you said that a Wilson mirror was "better" than a standard mirror, which in turn was "better" than a cascode mirror. I haven't tried all three of these extensively, which is why I was hoping to learn something from you.
There are (at least) two sources of error found in current mirrors:
a) DC inaccuracy due to current "lost" to drive the bases. This can be improved with a Wilson mirror or by using a 4th type of mirror where an emitter follower drives the bases.
b) DC inaccuracy due to the Early effect, whereby Vce is unequal across the two devices. This can be improved with either a Wilson mirror or a cascoded mirror.
I was confused by your ranking order because (at DC, at least) both the Wilson and cascode mirrors should be better than the standard mirror.
In your original reply you stated that you use a (slightly modified!) standard mirror due to its HF improvement, presumably over both the Wilson and cascode mirrors. Again this is somewhat confusing to me because I would expect an improvement at HF from the cascode mirror (similar to cascoding any circuit).
Again, if I thought you were wrong about this, I would either say so or ignore your post. In this case I believe that you have a lot of practical experience, and I'm hoping to learn from you. Anyway, I hope you are still reading this and willing to provide an answer. If you don't want to post here, perhaps you could send a private message.
Thanks,
Charles Hansen
Graham Maynard said:
You're over-invested in what XXXXXXXX may think. Cut you're losses...NOW.
JF
Graham Maynard said:
By that do you mean that you have been so foolish, or have
actually performed the tests you describe? 😉
I have built up a variety of amplifiers with current mirrors for the
past 30+ years, and while I admit that they work as advertised,
that doesn't mean that any one of them is necessarily the best
in any given circuit.
Let's face it, the first thing and primary thing current mirrors do
is increase the open loop gain. Sometimes you want that,
sometimes you don't.
I agree with you, Nelson. I prefer resistors to current mirrors, especially on my input stages. The high impedance of the current mirrors lowers the open loop bandwidth and actually puts stress at high frequencies on the second stage as far as nonlinear distortion generation is concerned.
WHEN, I HAVE to worry about significant distortion reduction, I have found that complex current mirrors work better than simple current mirrors to reduce distortion over a broad range of frequencies. Sometimes, I will use a current mirror in a second stage for practical reasons.
WHEN, I HAVE to worry about significant distortion reduction, I have found that complex current mirrors work better than simple current mirrors to reduce distortion over a broad range of frequencies. Sometimes, I will use a current mirror in a second stage for practical reasons.
Current events
"You're over-invested in what XXXXXXXX may think. Cut you're losses...NOW."
Yes, so he can go back to torturing me full time.........
There are several steps in between current mirrors and resistor loading. Just loading the diff pair output with a single current source instead of a current mirror might be one interesting approach. Also a current source in parallel with a load resistor would give the flexibility of setting the diff pair bias current fairly high while retaining the desired voltage drop across the resistor to bias the second stage as desired. Sufficient power supply voltage for the current source would lower device capacitance and might even allow a series resistor to help isolate the current source capacitance.
'Cause there's a million ways to go. You know that there are ..........
"You're over-invested in what XXXXXXXX may think. Cut you're losses...NOW."
Yes, so he can go back to torturing me full time.........
There are several steps in between current mirrors and resistor loading. Just loading the diff pair output with a single current source instead of a current mirror might be one interesting approach. Also a current source in parallel with a load resistor would give the flexibility of setting the diff pair bias current fairly high while retaining the desired voltage drop across the resistor to bias the second stage as desired. Sufficient power supply voltage for the current source would lower device capacitance and might even allow a series resistor to help isolate the current source capacitance.
'Cause there's a million ways to go. You know that there are ..........
john curl said:
that has been my preferred approach as well. I don't preface to be able to tell a current source from a resistor (unless the CS goes into oscillation) but I have found resistor loading to be simple and stable, much like bootstraps.
Has anyone tried to use a differential pair in a JLH1969-like amp? I have been simulating one for quite some time and will attempt to build one shortly.
Folks, I forgot to add that there is one other SERIOUS PROBLEM with current mirrors. They amplify their own self noise! This is very important and why I always use emitter degeneration resistors, unless working at extremely low power supply voltages. This greatly reduces the active gain of the mirrors and the extra noise can be minimized. Resistors have noise too, but no active gain to make it even worse. It took years for me to learn this, so don't be too surprised if you haven't thought it through, yourself. 😉
JFET and BIPOLAR PAIRS FOR CASCODES
JCX NOTED: "cascoding fets with identical devices doesn't give the linearity and go improvement you might be expecting, the "inner" device is guaranteed to be biased in triode mode when cascoded with a matching device, the "outer" devices should have larger Idss and operate with vg > vg_off of the inner fet
bjts are even better cascoding devices with respect to stabilizing the inner device Vds, if you can figure out a good bias arrangement they will have much lower Z at the emitter than even properly selected fets at similar currents"
TO JCX or experienced cascode designers...
What N and P JFETs work best together for the input FET and cascode FET for cascode paring, especially for a complementary differential cascoded front end?
What PNP input and cascoded PNP work best for an input differential cascoded front end?
The 2N5911 N-JFET looks like a good N matched pair, but I can't find a similar 5 mS gfs matched P-JFET.
JCX NOTED: "cascoding fets with identical devices doesn't give the linearity and go improvement you might be expecting, the "inner" device is guaranteed to be biased in triode mode when cascoded with a matching device, the "outer" devices should have larger Idss and operate with vg > vg_off of the inner fet
bjts are even better cascoding devices with respect to stabilizing the inner device Vds, if you can figure out a good bias arrangement they will have much lower Z at the emitter than even properly selected fets at similar currents"
TO JCX or experienced cascode designers...
What N and P JFETs work best together for the input FET and cascode FET for cascode paring, especially for a complementary differential cascoded front end?
What PNP input and cascoded PNP work best for an input differential cascoded front end?
The 2N5911 N-JFET looks like a good N matched pair, but I can't find a similar 5 mS gfs matched P-JFET.
- Status
- Not open for further replies.