While researching variations, and being inspired by Nelson's paper on Cascode and use of same, I ran into some interesting idea's. I believe the reason cascode would be less than spectacular to apply to the basic Zen philosophy, Zen4 being my choice, is this topology would have high output impedance. Is this correct? Their are also issues like the extra power being burned etc. etc.
Somewhat before I arrived on this planet, David Hafler and Herbert Keroes published a paper titled "An Ultra-Linear Amplifier" that dealt with the issues of pentodes/tetrodes and the like as power amps. They noted that the triode amp is superior for sound quality but lacks the power to be an output stage. Their paper appears to be worded carefully to reflect the same findings, without actually claiming they found them, as a british dude from years earlier who applied for a patent regarding this "Ultra-Linear" topology.
Before someone puts me into another forum catagory, in a nutshell, they added a percentage of plate signal to the screen grid to create a tetrode output stage, with the qualities of a triode amp stage. They also effectively lowered the output Z. MOSFETS being the modern equivalent of a tube, if there is one, I believe it is possible to run a cascode topology, with a small percentage of output feedback to reduce the output Z of such an arrangement and create the Ultra-Linear Zen...
Does this raise any eyebrows???
Somewhat before I arrived on this planet, David Hafler and Herbert Keroes published a paper titled "An Ultra-Linear Amplifier" that dealt with the issues of pentodes/tetrodes and the like as power amps. They noted that the triode amp is superior for sound quality but lacks the power to be an output stage. Their paper appears to be worded carefully to reflect the same findings, without actually claiming they found them, as a british dude from years earlier who applied for a patent regarding this "Ultra-Linear" topology.
Before someone puts me into another forum catagory, in a nutshell, they added a percentage of plate signal to the screen grid to create a tetrode output stage, with the qualities of a triode amp stage. They also effectively lowered the output Z. MOSFETS being the modern equivalent of a tube, if there is one, I believe it is possible to run a cascode topology, with a small percentage of output feedback to reduce the output Z of such an arrangement and create the Ultra-Linear Zen...
Does this raise any eyebrows???
Yes, Exactly. I have seen that dwg and that represents what I was thinking. I beleive the more feedback, the less the cascode properties. I would prefer to use a smaller ratio than the 50% of that dwg, maybe 25% for a starting point. But maintian something like 12Vds avg. (maybe + or - 2.5V) on the gain transistor with a zener under the voltage divider. The biggest gains I'm thinking will be the increased linearity due to the reduction of gate to drain capacitance variation. The biggest drawback has to be the avg. Pd of the gain transistor @ 12Vds or is that the cascode transistor Pd. There seems to be this anti Zen, if you will, about the contribution of an extra transistor in terms of complexity, temperature, efficeincy or something. That really is my question, is it worth it???
Idealy I would be like to simulate this stuff but I cant' seem to get my Pspice going...
I've built some test circuits, it does work, but I need a real prototype pcb to really critically test the idea. I'm working out a board as we post...
Are there any MOSFET designs out there utilizing this idea??? I can not be the only one to try this???
Idealy I would be like to simulate this stuff but I cant' seem to get my Pspice going...
I've built some test circuits, it does work, but I need a real prototype pcb to really critically test the idea. I'm working out a board as we post...
Are there any MOSFET designs out there utilizing this idea??? I can not be the only one to try this???
You might wanna browse trough this thread: http://www.diyaudio.com/forums/showthread.php?s=&threadid=50267
/Kari
/Kari
Circlomanen wrote:
'Ultralinear feedback is possible in a Zen Cascode typ amp.
I have been fooling with it a bit but did not see any real benefits.'
Indeed.
There won't be any benefits (that I can see), because unlike triodes, MOSFETs (and pentodes) have a very high drain (plate) resistance. With triodes, the plate resistance of the lower valve is moderate, and hence it's possible to get some gain (ca 10 dB or so) with low-mu triodes for feedback from the upper grid. This will work much like a pentode in UL connection (as Broskie has discussed).
With MOSFETs, the upper gate won't have much voltage gain (if any), so the only effect you get is the one you don't want: increased Miller effect, due to the lower device seeing voltage swing across it. (With small triodes, the increased Miller effect is not as serious.) I suppose paralleling the lower MOSFET with a resistor would give the upper device something to bite on, but if this is practical or has any benefits on distortion - no idea.
Morgan
'Ultralinear feedback is possible in a Zen Cascode typ amp.
I have been fooling with it a bit but did not see any real benefits.'
Indeed.
There won't be any benefits (that I can see), because unlike triodes, MOSFETs (and pentodes) have a very high drain (plate) resistance. With triodes, the plate resistance of the lower valve is moderate, and hence it's possible to get some gain (ca 10 dB or so) with low-mu triodes for feedback from the upper grid. This will work much like a pentode in UL connection (as Broskie has discussed).
With MOSFETs, the upper gate won't have much voltage gain (if any), so the only effect you get is the one you don't want: increased Miller effect, due to the lower device seeing voltage swing across it. (With small triodes, the increased Miller effect is not as serious.) I suppose paralleling the lower MOSFET with a resistor would give the upper device something to bite on, but if this is practical or has any benefits on distortion - no idea.
Morgan
There's an interesting patent about taking the bias for the cascode tube from the resistor under the next stage's cathode, which happens to be a follower. This puts a little audio into the cascode grid in addition to setting the DC for the cascode's grid. It's a clever idea.
I don't remember the patent number, but it was granted to William Johnson and/or Audio Research. You should be able to find it fairly easily.
Take a look and see if that sparks your creativity.
Grey
I don't remember the patent number, but it was granted to William Johnson and/or Audio Research. You should be able to find it fairly easily.
Take a look and see if that sparks your creativity.
Grey
> MOSFETS being the modern equivalent of a tube
Not equivalent enough.
> cascode topology, with a small percentage of output feedback to reduce the output Z of such an arrangement and create the Ultra-Linear Zen...
Yeah but. The "but" is: a vacuum tetrode has two control grids. G1 and G2 both have large influence on cathode current, just as in a triode G1 and Plate both affect current. G1 has 10 times more effect, but the effect of G2 or P is not negligible.
In the common power MOSFET, grounded-source, G1 has large effect, Drain has almost no effect at practical voltages, and there are no other electrodes to fool with.
They used to bring the Substrate out. That was in effect a low-gain grid/gate with significant control. It was pretty wonky, and anyway modern processes don't seem to want to bring out the Substrate separate from Source.
As Morgan says, in the cascode, that top gate has almost no effect until it hammers the bottom Drain down against its Source. Consider the top FET as a voltage-amp: the load resistance is the load, the cathode/source resistance is the bottom FET's Source, nearly infinity. Say 10Ω load, 10K source resistance: gain is 10/10K= 0.001. Whanging that top gate around does almost nothing until side-effects pin some other electrode to a rail or ground, and then it just saturates.
The vacuum triode (G1+P in triode, G1+G2 in tetrode) has an internal miracle. The actual control voltage is the sum of G1 and G2/10. It is all done with electric fields. We can do the same thing externally with resistors.... but now we have a non-infinite input impedance. The vacuum triode does this sum without (much) lowering of input impedance. Such a thing could perhaps be done with MOS, but I suspect it isn't practical for about the same reason sand FETs "work better" than vacuum devices: their junction is much-much thinner and more conductive than the cathode-grid space of a vacuum device. To mix the electric fields you would have to open-up the gate electrode and interlace a G2 electrode, probably at dimensions smaller than masks can manage.
The other, and potentially more profitable, VT analogy is cathode feedback. If you transformer-load the FET, add a 10% or 20% winding and stuff it up the Source. Voltage feedback, output is damped, higher input voltage, no reduction in input impedance. But of course the high current of FETs negates the original purpose of transformers in VT amps: current multiplication. Tubes MUST have transformers to feed practical low-Z loads efficiently; FETs don't "need" iron so it is annoying to add iron "just for feedback". But it does work, has for decades. (I had a TEAC tape deck with headphone output that was one BJT with collector load and a feedback winding into the emitter.)
Not equivalent enough.
> cascode topology, with a small percentage of output feedback to reduce the output Z of such an arrangement and create the Ultra-Linear Zen...
Yeah but. The "but" is: a vacuum tetrode has two control grids. G1 and G2 both have large influence on cathode current, just as in a triode G1 and Plate both affect current. G1 has 10 times more effect, but the effect of G2 or P is not negligible.
In the common power MOSFET, grounded-source, G1 has large effect, Drain has almost no effect at practical voltages, and there are no other electrodes to fool with.
They used to bring the Substrate out. That was in effect a low-gain grid/gate with significant control. It was pretty wonky, and anyway modern processes don't seem to want to bring out the Substrate separate from Source.
As Morgan says, in the cascode, that top gate has almost no effect until it hammers the bottom Drain down against its Source. Consider the top FET as a voltage-amp: the load resistance is the load, the cathode/source resistance is the bottom FET's Source, nearly infinity. Say 10Ω load, 10K source resistance: gain is 10/10K= 0.001. Whanging that top gate around does almost nothing until side-effects pin some other electrode to a rail or ground, and then it just saturates.
The vacuum triode (G1+P in triode, G1+G2 in tetrode) has an internal miracle. The actual control voltage is the sum of G1 and G2/10. It is all done with electric fields. We can do the same thing externally with resistors.... but now we have a non-infinite input impedance. The vacuum triode does this sum without (much) lowering of input impedance. Such a thing could perhaps be done with MOS, but I suspect it isn't practical for about the same reason sand FETs "work better" than vacuum devices: their junction is much-much thinner and more conductive than the cathode-grid space of a vacuum device. To mix the electric fields you would have to open-up the gate electrode and interlace a G2 electrode, probably at dimensions smaller than masks can manage.
The other, and potentially more profitable, VT analogy is cathode feedback. If you transformer-load the FET, add a 10% or 20% winding and stuff it up the Source. Voltage feedback, output is damped, higher input voltage, no reduction in input impedance. But of course the high current of FETs negates the original purpose of transformers in VT amps: current multiplication. Tubes MUST have transformers to feed practical low-Z loads efficiently; FETs don't "need" iron so it is annoying to add iron "just for feedback". But it does work, has for decades. (I had a TEAC tape deck with headphone output that was one BJT with collector load and a feedback winding into the emitter.)
Well, thanks poeple. 
I appreciate the responses. I guess the gains are still a little questionable, as applied to FETs... Someone would have been there and found sucsses, or not, by this time. I'm going to build a design for a Pre-Amp using the cascode output topology but allowing for the insertion of F.B. to the cascode transistor. I cant' really play with the diff/balanced circuits without a reasonable single ended to balanced converter anyway. Something like a balanced Mini ZEN 4 with cascoded output... I don't know about using devices like the IRF610s though. I gotta go with a bunch of little SO8 type dual fets for the matching. Zetex makes a few good looking possiblities, ON does too. A good low C, P channel input is difficult to find. Something like Nelson's favorite ZVP3310 in a single die dual would be ideal???
I appreciate the responses. I guess the gains are still a little questionable, as applied to FETs... Someone would have been there and found sucsses, or not, by this time. I'm going to build a design for a Pre-Amp using the cascode output topology but allowing for the insertion of F.B. to the cascode transistor. I cant' really play with the diff/balanced circuits without a reasonable single ended to balanced converter anyway. Something like a balanced Mini ZEN 4 with cascoded output... I don't know about using devices like the IRF610s though. I gotta go with a bunch of little SO8 type dual fets for the matching. Zetex makes a few good looking possiblities, ON does too. A good low C, P channel input is difficult to find. Something like Nelson's favorite ZVP3310 in a single die dual would be ideal???
dual FETs for matching
I think that most dual power mosfets do not actually have a common substrate. They most likely come from the same silicon wafer, but are not necessarily from adjacent die sites. The device fingers are not interleaved. So while there is probably considerably better process tracking (they came from the same die) you won't get the kind of matching you could expect from common substrate devices (on the same die) like the dual JFETs 2SK389 ...
However, maybe there are some single substrate matched pair power mosfet devices out there that I am not aware of. If so, tell us what they are!
I think that most dual power mosfets do not actually have a common substrate. They most likely come from the same silicon wafer, but are not necessarily from adjacent die sites. The device fingers are not interleaved. So while there is probably considerably better process tracking (they came from the same die) you won't get the kind of matching you could expect from common substrate devices (on the same die) like the dual JFETs 2SK389 ...
However, maybe there are some single substrate matched pair power mosfet devices out there that I am not aware of. If so, tell us what they are!
Petent "Dude"
The British "Dude" was Blumlein, of EMI, Ltd. He patented the ultralinear circuit in 1937, shortly after filing patents on stereophonic sound - complete with drawings of sterophonic cutters, cartridges etc. An audio Hero at least twenty years ahead of his time - quite a dude indeed.
The British "Dude" was Blumlein, of EMI, Ltd. He patented the ultralinear circuit in 1937, shortly after filing patents on stereophonic sound - complete with drawings of sterophonic cutters, cartridges etc. An audio Hero at least twenty years ahead of his time - quite a dude indeed.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.