Hi Andrewbee,
An interesting collection of patents, thanks. I was just thinking about how to get linear gain from a cascode (instead of 3/2 power law) the other day and I came up with the same thing as Johnson's patent. (and Reiffin's re- patent, virtually the same thing)
The Ketchledge 1956 patent, 2,751,442, looks a little reminiscent of the present day error correction scheme used by Hawksford. Hawksford cites Llewellyn, 1941, patent # 2,245,598 as the origin of the error correction idea (it was done originally in tube technology, but is popular in high end SS today). I think Llewellyn's patent went over everyone's head back then.
Don
An interesting collection of patents, thanks. I was just thinking about how to get linear gain from a cascode (instead of 3/2 power law) the other day and I came up with the same thing as Johnson's patent. (and Reiffin's re- patent, virtually the same thing)
The Ketchledge 1956 patent, 2,751,442, looks a little reminiscent of the present day error correction scheme used by Hawksford. Hawksford cites Llewellyn, 1941, patent # 2,245,598 as the origin of the error correction idea (it was done originally in tube technology, but is popular in high end SS today). I think Llewellyn's patent went over everyone's head back then.
Don
The NFB loop used in the Johnson and Reiffin patents to linearize a cascode brings up an interesting question. Can one get linearity without sacrificing gain, ie, something for nothing?
(the scheme uses a resistive divider to ground, from the final output, typically after a CF, back to the grid of the top cascode tube, which effectively controls the AC voltage on the plate of the bottom tube)
As the voltage variation on the top grid is increased (by tweaking the resistive divider), a point will be reached where the bottom tube's plate is varying by its Mu ratio (with respect to the input drive signal) and so would just put out a constant current.
So if we adjust this grid feedback to just below the Mu ratio, the current in the cascode will be changing just slightly. By using an active load for the cascode's load (as Reiffin does in his 2nd patent), we can still get voltage gain arbitrarily larger than the Mu (well, up until capacitances finally kills the gain). So overall high gain operation is still possible and the gain should be linear instead of the usual 3/2s power law.
Going even farther, suppose we feed back GREATER (NFB) than the bottom tube Mu ratio, to the top grid. Then tube current DECREASES with increasing input signal and we have a non-inverting stage. (use a high Z or CCS active load to get high gain still) To derive the top NFB grid voltages we will need two of these weird cascodes in a P-P type setup and cross-couple the feedbacks to get the inverted NFB phases. I have NO idea what happens to the linearity with THIS setup. (well, I'm hoping it might generate INVERSE tube distortion, like 2/3 power dist. at 2xMu NFB, so as to cancel the distortion of a following normal pentode or cascode stage) (maybe could use an LTP configuration on the bottom P-P tubes then too if all is still near linear.)
(I dreamed this one up so you all can lay awake tonight puzzling over this!
)
Don
(the scheme uses a resistive divider to ground, from the final output, typically after a CF, back to the grid of the top cascode tube, which effectively controls the AC voltage on the plate of the bottom tube)
As the voltage variation on the top grid is increased (by tweaking the resistive divider), a point will be reached where the bottom tube's plate is varying by its Mu ratio (with respect to the input drive signal) and so would just put out a constant current.
So if we adjust this grid feedback to just below the Mu ratio, the current in the cascode will be changing just slightly. By using an active load for the cascode's load (as Reiffin does in his 2nd patent), we can still get voltage gain arbitrarily larger than the Mu (well, up until capacitances finally kills the gain). So overall high gain operation is still possible and the gain should be linear instead of the usual 3/2s power law.
Going even farther, suppose we feed back GREATER (NFB) than the bottom tube Mu ratio, to the top grid. Then tube current DECREASES with increasing input signal and we have a non-inverting stage. (use a high Z or CCS active load to get high gain still) To derive the top NFB grid voltages we will need two of these weird cascodes in a P-P type setup and cross-couple the feedbacks to get the inverted NFB phases. I have NO idea what happens to the linearity with THIS setup. (well, I'm hoping it might generate INVERSE tube distortion, like 2/3 power dist. at 2xMu NFB, so as to cancel the distortion of a following normal pentode or cascode stage) (maybe could use an LTP configuration on the bottom P-P tubes then too if all is still near linear.)
(I dreamed this one up so you all can lay awake tonight puzzling over this!
)Don
This stuff is mostly over my head but I can see Don loves a challenge!
Sometimes when I get really lucky some of it actually makes some sense to me.
There are a lot of bright sparks out there I think moreso than those that seek patents. There are quite a few people on this board who are obviously quite knoweldgable, present company included
Andrew
Sometimes when I get really lucky some of it actually makes some sense to me.
There are a lot of bright sparks out there I think moreso than those that seek patents. There are quite a few people on this board who are obviously quite knoweldgable, present company included
Andrew
> Can one get linearity without sacrificing gain, ie, something for nothing?
TANSTAAFL.
Nonlinearity is "gain variation over the signal swing". So you want to reduce gain when it is high, increase it when gain is low. If you could increase gain, you would have. so the best we can do is decrease the excess gain. But how does a dumb collection of parts know when gain is too high? The only way (other than first-order cancelleation of second-order nonlinearlity, AKA push-pull) is to amplify the error... i.e. more gain.
Anyway: if you could improve linearity without sacrificing gain any simple way, it probably would have been done 70 years ago. The only fer-sure way to get non-trivial linearization is (avoiding misbiasing, using push-pull, and) Black's paper on Negative Feedback (wash my mouth with soap) which demands we sacrifice some gain. Back in the 1930s, strong men cried over lost gain, but they did it a lot.
TANSTAAFL.
Nonlinearity is "gain variation over the signal swing". So you want to reduce gain when it is high, increase it when gain is low. If you could increase gain, you would have. so the best we can do is decrease the excess gain. But how does a dumb collection of parts know when gain is too high? The only way (other than first-order cancelleation of second-order nonlinearlity, AKA push-pull) is to amplify the error... i.e. more gain.
Anyway: if you could improve linearity without sacrificing gain any simple way, it probably would have been done 70 years ago. The only fer-sure way to get non-trivial linearization is (avoiding misbiasing, using push-pull, and) Black's paper on Negative Feedback (wash my mouth with soap) which demands we sacrifice some gain. Back in the 1930s, strong men cried over lost gain, but they did it a lot.
There ain't no such thing as a free lunch: tanstaafl
Seems to be the case.
The cascode circuit, with the top tube grid just connected to a voltage level, almost seems to cry out to be put to some use. Thats why the Johnson and Reiffin patents seem intriguing, since they put the top grid to work. But they obviously sacrifice current gain for linearity by NFB.
I was thinking maybe something like the VT current mirror could be implemented somehow, using the top grid to linearize things, maybe using a thermionic diode as a corrector. But, even the normal current mirror just reduces gain down everywhere to the lowest tube gain point to get uniformity.
Maybe by putting a thermionic diode in a positive feedback path around a gain stage, one could boost the gain where needed instead of dropping it. I haven't figured out any workable configuration yet though. But seems like it might be do-able.
(succeeding at this might get Black to jump out of his grave maybe!)
Don
Seems to be the case.
The cascode circuit, with the top tube grid just connected to a voltage level, almost seems to cry out to be put to some use. Thats why the Johnson and Reiffin patents seem intriguing, since they put the top grid to work. But they obviously sacrifice current gain for linearity by NFB.
I was thinking maybe something like the VT current mirror could be implemented somehow, using the top grid to linearize things, maybe using a thermionic diode as a corrector. But, even the normal current mirror just reduces gain down everywhere to the lowest tube gain point to get uniformity.
Maybe by putting a thermionic diode in a positive feedback path around a gain stage, one could boost the gain where needed instead of dropping it. I haven't figured out any workable configuration yet though. But seems like it might be do-able.
(succeeding at this might get Black to jump out of his grave maybe!)
Don
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.