an other incarnation
The drain resistors for the jfet front end should be tied directly( or AC coupled) to the negative rail junction of the output stage at the source resistors. You have done this pretty well with your coupling cap between the gate of the J74s and the negative 16V rail. I would suggest anyone using a seperate negative supply for the front end take note of this requirement. I would use larger drain resistors the J109s to get the same voltage gain for the front end and a ZVN3310 as a follower to drive the IRFP 244s. No extra supply needed for front end and lower high frequency distortion for IRFP244s. You wouldn't need doubled up J109s. I would use 1K gate resistors for the VZN330s and 392 ohm source resistors. I think you will get better performance and greatly reduced parts count. If you are going to parallel J109s use seperate gate resistors for each one. I would not tie the input gates to ground with the 220K resistors if the amp is ever to be driven single ended. A very well thought out implementation though.
H.H.
The drain resistors for the jfet front end should be tied directly( or AC coupled) to the negative rail junction of the output stage at the source resistors. You have done this pretty well with your coupling cap between the gate of the J74s and the negative 16V rail. I would suggest anyone using a seperate negative supply for the front end take note of this requirement. I would use larger drain resistors the J109s to get the same voltage gain for the front end and a ZVN3310 as a follower to drive the IRFP 244s. No extra supply needed for front end and lower high frequency distortion for IRFP244s. You wouldn't need doubled up J109s. I would use 1K gate resistors for the VZN330s and 392 ohm source resistors. I think you will get better performance and greatly reduced parts count. If you are going to parallel J109s use seperate gate resistors for each one. I would not tie the input gates to ground with the 220K resistors if the amp is ever to be driven single ended. A very well thought out implementation though.
H.H.
Ergo, an interesting variation on the X theme and one that looks pretty thoroughly engineered at that. Given that you asked for comments, I have to say that it is a little complicated for my taste. More specifically I am unclear as to the benefits of the input cascades (Q1 and Q6) and the current sources (J3 and J4) in this application. Nor am I sure why the additional negative power supply is required. Let me try to explain my reasoning.
The Aleph output stage has voltage as well as current gain and therefore the signal required to drive it is pretty small. It can also be driven to the power rails without requiring an additional Vgs worth of input drive. This would seem to suggest that the additional negative supply is superfluous.
I appreciate that cascading is somewhat in vogue at the moment and many would argue that in any case it does not constitute an additional gain stage. Nevertheless, I subscribe to the principal that if is not required then don’t include it – a principle often referred to as KISS (keep it simple stupid – no offence intended). Cascodes are useful in improving linearity when signal excursion is significant and also in improving frequency response by virtual of their low input impedance. However, in this application the signal is small and I believe the frequency response would be adequate without it. Please feel free to correct me if I am wrong or if I have misunderstood the intention.
I admit to liking the idea of using the J109 (I intend to do this myself) for the input and I have no problem with the notion of doubling these up. In fact, I would prefer to do this rather than include the follower suggested by Harry.
I also note your novel approach to common mode stabilisation by applying a servo (at least I believe that is what it is) to control the current sources (J3 and J4). I would be interested to know what advantage this has over applying a servo to the input pair tail current source (M1).
Ian.
The Aleph output stage has voltage as well as current gain and therefore the signal required to drive it is pretty small. It can also be driven to the power rails without requiring an additional Vgs worth of input drive. This would seem to suggest that the additional negative supply is superfluous.
I appreciate that cascading is somewhat in vogue at the moment and many would argue that in any case it does not constitute an additional gain stage. Nevertheless, I subscribe to the principal that if is not required then don’t include it – a principle often referred to as KISS (keep it simple stupid – no offence intended). Cascodes are useful in improving linearity when signal excursion is significant and also in improving frequency response by virtual of their low input impedance. However, in this application the signal is small and I believe the frequency response would be adequate without it. Please feel free to correct me if I am wrong or if I have misunderstood the intention.
I admit to liking the idea of using the J109 (I intend to do this myself) for the input and I have no problem with the notion of doubling these up. In fact, I would prefer to do this rather than include the follower suggested by Harry.
I also note your novel approach to common mode stabilisation by applying a servo (at least I believe that is what it is) to control the current sources (J3 and J4). I would be interested to know what advantage this has over applying a servo to the input pair tail current source (M1).
Ian.
Banned
Joined 2002
Uhh, finally some time to reply to the thread. We had a big national holiday
here in Estonia and making fires and eating meat took all the time
Anyway, first a bit about how this circuit came about. It started with
experimenting to replace transistors in JLH'96 schematic with fets and
simulating the behaviour. Then there was a all fet schematic with aleph
current source as that gives better performance and modulates the current
more.
Then the circuit took a fully balanced mode. It was after the cross feedback
idea from Grey that it took the Su-Sy form.
***
Now the answer to Harry about using the follower, well as in the Hood
schematic there was one, but as it seemed also to perform well without it, it was dropped.
The drain resitors certainly seem a good idea...
The transistors J3 and J4 were included as without those the 4th harmonic raised from
the noise floor. With these transitors in place it doesn't. With Su-Sy distortion decrased
additional 10x.
As for the 220K resistors, I'm not sure either if those would be needed at
all. They are planned there to keep the circuit from overcompensating the DC
conditions.
***
Ian, yes the schematic is a bit complicated. But there are two things to
consider, first it's a work in progress so it might happen that with some
good ideas it will become much simpler. The other is that most of the
complicating stuff is in relation with DC offset compensation. This schematic
performs well in this respect (but remember only simulations this far)
and mostly thanks to the additional efforts and complications.
The cascode here gives two advantages. The more important one being that it
helps to increase the phase margin, which is not that great without it. The
second one being that it very slightly decreases distortion.
***
JasonL, short answer is "don't know jet".
As it is seen from the previous comments there are ideas to implement and
rethink it some more and then perhaps build it. One of the problems with
actual building will be the IRFP244 availability here. If anyone knows good
sources in Europe or Scandinavia please let me know.
here in Estonia and making fires and eating meat took all the time
Anyway, first a bit about how this circuit came about. It started with
experimenting to replace transistors in JLH'96 schematic with fets and
simulating the behaviour. Then there was a all fet schematic with aleph
current source as that gives better performance and modulates the current
more.
Then the circuit took a fully balanced mode. It was after the cross feedback
idea from Grey that it took the Su-Sy form.
***
Now the answer to Harry about using the follower, well as in the Hood
schematic there was one, but as it seemed also to perform well without it, it was dropped.
The drain resitors certainly seem a good idea...
The transistors J3 and J4 were included as without those the 4th harmonic raised from
the noise floor. With these transitors in place it doesn't. With Su-Sy distortion decrased
additional 10x.
As for the 220K resistors, I'm not sure either if those would be needed at
all. They are planned there to keep the circuit from overcompensating the DC
conditions.
***
Ian, yes the schematic is a bit complicated. But there are two things to
consider, first it's a work in progress so it might happen that with some
good ideas it will become much simpler. The other is that most of the
complicating stuff is in relation with DC offset compensation. This schematic
performs well in this respect (but remember only simulations this far)
and mostly thanks to the additional efforts and complications.
The cascode here gives two advantages. The more important one being that it
helps to increase the phase margin, which is not that great without it. The
second one being that it very slightly decreases distortion.
***
JasonL, short answer is "don't know jet".
As it is seen from the previous comments there are ideas to implement and
rethink it some more and then perhaps build it. One of the problems with
actual building will be the IRFP244 availability here. If anyone knows good
sources in Europe or Scandinavia please let me know.
Ian, I'm using MicroCap 7 evaluation.
http://www.spectrum-soft.com/index.shtm
The software itself is expensive, but the evaluation version is very sensible. Allows 50 components (not nodes!!!). So it's possible to simulate quite a few things with it. It is also great for schematic input and it accepts standard SPICE models.
Also great are the help scripts that get you going in no time
Ergo
http://www.spectrum-soft.com/index.shtm
The software itself is expensive, but the evaluation version is very sensible. Allows 50 components (not nodes!!!). So it's possible to simulate quite a few things with it. It is also great for schematic input and it accepts standard SPICE models.
Also great are the help scripts that get you going in no time
Ergo
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.