If you are cascoding an amplifier, what should the voltage relationship be for the best linearity? Should there be equal voltages across the input MOSFET and the cascode MOSFET, or should one have more voltage across it than the other? I assumed that the cascode MOSFET should have more voltage across it, because it is the one that is swinging all the voltage (isn't that the whole point of the cascode- to eliminate the voltage swing on the lower transistor, shielding it from the output voltage swing?).
My other question is about biasing the upper MOSFET... what is the best way to do this? Any help or clarification would be very much appreciated. Thanks.
Steve
My other question is about biasing the upper MOSFET... what is the best way to do this? Any help or clarification would be very much appreciated. Thanks.
Steve
Thanks for the specifics...
So, if you're building a differential input stage, you use a current-sourced setup (under the input transistors, I would assume) and you just put a zener bypassed with a cap on the gate on the cascodes(remembering the all important gate resistor
). That's what I was looking at doing also. Would it beneficial to use an actual regulated supply in this position, or is that overkill?
I'm still not clear as to how much voltage to leave across the upper transistor, as compared to the lower one (I'm looking for a general guideline here- maybe a suggested ratio) It seems like when I ran a simulation using a modified Pass BLS circuit with 80 volt supplies, current sourced at 80mA per side, with 500 ohm load resistors, I had pretty good results when I had a ratio of:
(measured from ground)
B+ 80v
source of cascode, bottom of load resistor- 40V
source of input Mosfet, drain of cascode-20V
I am wondering exactly how much voltage should be across the bottom Mosfet in particular... In my simulations it dosen't swing a lot of voltage, something like .5V max p-p (I assume that this is the cascode in action). Does it still need to have a lot of voltage headroom, even though it's not really doing a lot, as far as voltage is concerned? I have noticed in some situations that the signal present in between the lower and cascode mosfets starts looking really wierd, but the output is supposedly correct... is this something to be concerned with?
So, if you're building a differential input stage, you use a current-sourced setup (under the input transistors, I would assume) and you just put a zener bypassed with a cap on the gate on the cascodes(remembering the all important gate resistor
). That's what I was looking at doing also. Would it beneficial to use an actual regulated supply in this position, or is that overkill?I'm still not clear as to how much voltage to leave across the upper transistor, as compared to the lower one (I'm looking for a general guideline here- maybe a suggested ratio) It seems like when I ran a simulation using a modified Pass BLS circuit with 80 volt supplies, current sourced at 80mA per side, with 500 ohm load resistors, I had pretty good results when I had a ratio of:
(measured from ground)
B+ 80v
source of cascode, bottom of load resistor- 40V
source of input Mosfet, drain of cascode-20V
I am wondering exactly how much voltage should be across the bottom Mosfet in particular... In my simulations it dosen't swing a lot of voltage, something like .5V max p-p (I assume that this is the cascode in action). Does it still need to have a lot of voltage headroom, even though it's not really doing a lot, as far as voltage is concerned? I have noticed in some situations that the signal present in between the lower and cascode mosfets starts looking really wierd, but the output is supposedly correct... is this something to be concerned with?
Steve,
I would bias the cascode mosfet one third to one half the supply rail unless your differential pair were jfets, in which case I would set it at 10-15v, for lowest noise and distortion.
Current diodes or two transistor current sources work well to bias the zener. Regulation might help but I think it would be overkill, like you suggest.(Remember the zener is a regulator and when supplied with a current source and bypassed with a capacitor is very quiet)
That being said regulating the tail of the differential has important sonic benefits. This can be accomplished be inserting a regulator between the supply rail and the current source of the differential.
Jam
P.S. I hope all this talk about current sources won't get Grey upset.
I would bias the cascode mosfet one third to one half the supply rail unless your differential pair were jfets, in which case I would set it at 10-15v, for lowest noise and distortion.
Current diodes or two transistor current sources work well to bias the zener. Regulation might help but I think it would be overkill, like you suggest.(Remember the zener is a regulator and when supplied with a current source and bypassed with a capacitor is very quiet)
That being said regulating the tail of the differential has important sonic benefits. This can be accomplished be inserting a regulator between the supply rail and the current source of the differential.
Jam
P.S. I hope all this talk about current sources won't get Grey upset.
Jam,
THanks A LOT for all the info!
So, you actually are adding a current source for the zeners that are biasing the cascode transistors? That's something I wouldn't have thought of. I read in one of the Pass labs articles that about 4mA is good for zener current. Would you agree with that? If I'm running the current sink on the tail from a regulated negative supply, would that be the same thing as what you were refering to?
On another note, if someone were to use bjt's for the cascode transistors, would the character of the sound lean more toward the BJT or toward the input mosfet's sound?
BTW, Grey told me once when I asked him a question on current sourcing to do a search on this forum for your name... looks like he was right.
Thanks again.
Steve
THanks A LOT for all the info!
So, you actually are adding a current source for the zeners that are biasing the cascode transistors? That's something I wouldn't have thought of. I read in one of the Pass labs articles that about 4mA is good for zener current. Would you agree with that? If I'm running the current sink on the tail from a regulated negative supply, would that be the same thing as what you were refering to?
On another note, if someone were to use bjt's for the cascode transistors, would the character of the sound lean more toward the BJT or toward the input mosfet's sound?
BTW, Grey told me once when I asked him a question on current sourcing to do a search on this forum for your name... looks like he was right
.Thanks again.
Steve
Steve,
Yes, current sourceing the zeners help, I prefer the sound of a mosfet cascode to a bipolar but the differences are fairly small.
Nelson Pass is correct in stating that you need a current of at least 4ma for lowest zener noise.
Yes, regulate the supply to the tail (current sink)of the differential,
you will be surprised how much this helps. (Screams of horror from Grey!)
When using mosfet cascodes look for stablity issues but correct layout and gate resistors should correct any problems you encounter.
Regards,
Jam
Yes, current sourceing the zeners help, I prefer the sound of a mosfet cascode to a bipolar but the differences are fairly small.
Nelson Pass is correct in stating that you need a current of at least 4ma for lowest zener noise.
Yes, regulate the supply to the tail (current sink)of the differential,
you will be surprised how much this helps. (Screams of horror from Grey!)
When using mosfet cascodes look for stablity issues but correct layout and gate resistors should correct any problems you encounter.
Regards,
Jam
I have been messing around with the simulator some more, and I have found something that is a little strange-
when I have the cascode mosfet biased so that there is very little voltage across the lower Mosfet, the circuit seems to respond by compressing the signal instead of clipping it- it looks a lot like what a tube will do when pushed. it will eventually clip, of course, but the signal starts to have the peaks squashed before it gets there. I wonder if this is just in the simulation, or if this would really happen... seems like it could be beneficial- give a little extra headroom, and maybe a tube-like characteristic? Does anyone know anything about what I'm seeing?
Steve
when I have the cascode mosfet biased so that there is very little voltage across the lower Mosfet, the circuit seems to respond by compressing the signal instead of clipping it- it looks a lot like what a tube will do when pushed. it will eventually clip, of course, but the signal starts to have the peaks squashed before it gets there. I wonder if this is just in the simulation, or if this would really happen... seems like it could be beneficial- give a little extra headroom, and maybe a tube-like characteristic? Does anyone know anything about what I'm seeing?
Steve
For those of you who might be interested, I do have
a nice article on Cascoding which appeared in Audio
Magazine many years ago and which is posted on the
passlabs.com web site, I believe also on passdiy.com.
It does not deal with Mosfets, but I would like to offer up
the observation that the Mosfets I have worked with
generally require at least 2 volts from Drain to Source
when cascoded. When dissipation of the gain device is
at a premium, say as with a Zen amp, 2 volts can be seen
to be enough to get the performance you want. If dissipation
is not an issue, as with most input stages, a few volts more
is appropriate. This is an excellent approach to getting a
lot of wattage out of a Zen and also improving performance,
and will show up as a future project in the Zen Variations.
Also, it is not typically that critical to regulate the reference
to the cascoding transistor. It doesn't hurt, but the benefits
of cascoding a device do not depend heavily on the
stability or noise of the reference.
a nice article on Cascoding which appeared in Audio
Magazine many years ago and which is posted on the
passlabs.com web site, I believe also on passdiy.com.
It does not deal with Mosfets, but I would like to offer up
the observation that the Mosfets I have worked with
generally require at least 2 volts from Drain to Source
when cascoded. When dissipation of the gain device is
at a premium, say as with a Zen amp, 2 volts can be seen
to be enough to get the performance you want. If dissipation
is not an issue, as with most input stages, a few volts more
is appropriate. This is an excellent approach to getting a
lot of wattage out of a Zen and also improving performance,
and will show up as a future project in the Zen Variations.
Also, it is not typically that critical to regulate the reference
to the cascoding transistor. It doesn't hurt, but the benefits
of cascoding a device do not depend heavily on the
stability or noise of the reference.
Zen Variations part 2 will be posted at www.passdiy.com next
week, and is also going to Audio Express. Parts 3 and 4 will
follow in another week or so. All three make up the
"penultimate zen" and with part 4 there will be a PC board.
As to tracking source voltages on a cascode, we weigh the
additional complexity against the theoretical advantage of
truly constant voltage across the diff pair transistors. I
could go either way, but usually I just give the cascodes a
DC to ground reference.
week, and is also going to Audio Express. Parts 3 and 4 will
follow in another week or so. All three make up the
"penultimate zen" and with part 4 there will be a PC board.
As to tracking source voltages on a cascode, we weigh the
additional complexity against the theoretical advantage of
truly constant voltage across the diff pair transistors. I
could go either way, but usually I just give the cascodes a
DC to ground reference.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.