I use the back-to back zeners for protection out of reflex, but having been forced to think about it, I would want the inception for gate protection to be symmetrical and as far out of the way as possible, so that protection is only engaged where absolutely necessary. Engaging a protection zener is rather similar to a situation where a large capacitively coupled signal causes grid current and brings on blocking distortion due to asymmetric drive capability.
Zeners don't cost all that much, anyway. I bought a couple of reels of them from Electronic Goldmine for beer change (good beer, but still beer change). My actual motivation was gearing up the war chest for going into business, but the cost of one extra zener shouldn't inconvenience anyone especially - it absolutely doesn't hurt, and it might help a bit.
Zeners don't cost all that much, anyway. I bought a couple of reels of them from Electronic Goldmine for beer change (good beer, but still beer change). My actual motivation was gearing up the war chest for going into business, but the cost of one extra zener shouldn't inconvenience anyone especially - it absolutely doesn't hurt, and it might help a bit.
Cascode source follower
"you don't really need depletion mode fets if you're trying to make a current source loaded source follower"
Hi Wrenchone,
I havn't mentioned a current loaded source follower though I have talked about a cascoded source follower. I have read that cascoding fets solves several issues.To my limited knowledge depletion fets allow making a cascode source follower without additional biasing components. And the cascoded depletion fet constant current K&K boards were handy, I had the parts, they are right size, and they work.
Do a search for the EAR 859 circuit and you'll see the cathode follower I replaced. I can use a valve or my adapted K&K boards; easy swap though perhaps not identical biasing.
The deletions may not be perfect and may easily bettered but they do work well, very well.
tim
"you don't really need depletion mode fets if you're trying to make a current source loaded source follower"
Hi Wrenchone,
I havn't mentioned a current loaded source follower though I have talked about a cascoded source follower. I have read that cascoding fets solves several issues.To my limited knowledge depletion fets allow making a cascode source follower without additional biasing components. And the cascoded depletion fet constant current K&K boards were handy, I had the parts, they are right size, and they work.
Do a search for the EAR 859 circuit and you'll see the cathode follower I replaced. I can use a valve or my adapted K&K boards; easy swap though perhaps not identical biasing.
The deletions may not be perfect and may easily bettered but they do work well, very well.
tim
I would cascode a current source load in some circumstances, but I don't know that I would do so for a source follower. Adding the current source load reders immediate payback in decreasing distortion.Choosing the right topside mosfet (low reverse capacitance and gate charge) may make cascoding the top fet less important.
Just for grins, I ran a PSpice sim of a resistively loaded mosfet follower, one with a current source load as shown in my schematic, and one with an extra cascode fet perched on top. All three followers were biased at ~ 5mA drain current with +100V top supply and -40V bottom supply, and 1V drive signal. The CS loaded buffer was best at THD of 3.62 X 10^-4, the cascade was slightly worse at 3.77 X 10^-4, and the resistively loaded follower took third with 6.58 X 10^-4. The harmonic distribution looked similar in all three circuits.
Just for grins, I ran a PSpice sim of a resistively loaded mosfet follower, one with a current source load as shown in my schematic, and one with an extra cascode fet perched on top. All three followers were biased at ~ 5mA drain current with +100V top supply and -40V bottom supply, and 1V drive signal. The CS loaded buffer was best at THD of 3.62 X 10^-4, the cascade was slightly worse at 3.77 X 10^-4, and the resistively loaded follower took third with 6.58 X 10^-4. The harmonic distribution looked similar in all three circuits.
SLCF
Thanks for the update Wrenchone,
I think it is generally acceptable by everyone that a cathode (CF) or source follower (SF) is "better" if loaded by a constant current source, and the graph on page 438 of Valley and Wallman is often used to show this.
As you will have seen (if you checked) that in my venerable EAR 859 amp the output valve runs in A2 drawing grid current. A CCS becomes a little redundant as the CCS is in parallel to to the lower A2 grid impedance. 50 meg CCS (for example) in parallel with, say, 1 K of the grid is still pretty close to 1K. So little advantage in my case. A choke might work.
I also thought that it was accepted that the CF/SF also works best at constant current AND constant voltage--sometimes called the super linear cathode (source) follower.
I believe the cascoding (amongst other things) helps the CF/SF to work at closer to constant voltage. Its good that your simulation shows the significant advantage. You might like to simulate constant current and constant voltage.There are also likely to be capacitance advantages to cascoding. It seems that reducing non linear FET capacitances helps reduce the so called solid state sound of the SF.
best regards, tim
Thanks for the update Wrenchone,
I think it is generally acceptable by everyone that a cathode (CF) or source follower (SF) is "better" if loaded by a constant current source, and the graph on page 438 of Valley and Wallman is often used to show this.
As you will have seen (if you checked) that in my venerable EAR 859 amp the output valve runs in A2 drawing grid current. A CCS becomes a little redundant as the CCS is in parallel to to the lower A2 grid impedance. 50 meg CCS (for example) in parallel with, say, 1 K of the grid is still pretty close to 1K. So little advantage in my case. A choke might work.
I also thought that it was accepted that the CF/SF also works best at constant current AND constant voltage--sometimes called the super linear cathode (source) follower.
I believe the cascoding (amongst other things) helps the CF/SF to work at closer to constant voltage. Its good that your simulation shows the significant advantage. You might like to simulate constant current and constant voltage.There are also likely to be capacitance advantages to cascoding. It seems that reducing non linear FET capacitances helps reduce the so called solid state sound of the SF.
best regards, tim
The cascode (constant voltage) in my simulations shows no real significant advantage (in simulation) for a current source loaded mosfet source follower, possibly because the topside mosfet mimics (in simulation) ideal id vs. Vd "pentodish" behavior. I also didn't put significant impedance between the driving source and the follower stage.
If you check out Papa Pass's treatise on cascoding followers (readily available at Passdiy.com), it's interesting and instructive to note that the studies were done with a bipolar power transistor that had significant tilt in the Ic vs Vce department. Cascoding would obviously have an immediate benefit in that case.
If you are driving a source follower stage with something wimpy (12AX7s come to mind) cascoding will certainly help with removing the miller capacitance loading and ease matters for the previous stage. Otherwise, it can be incitement to oscillation. If you are cascoding a stage, it may be wise to bulk up on the input stopper resistors - I've been there/a word to the wise, and all that....
If you check out Papa Pass's treatise on cascoding followers (readily available at Passdiy.com), it's interesting and instructive to note that the studies were done with a bipolar power transistor that had significant tilt in the Ic vs Vce department. Cascoding would obviously have an immediate benefit in that case.
If you are driving a source follower stage with something wimpy (12AX7s come to mind) cascoding will certainly help with removing the miller capacitance loading and ease matters for the previous stage. Otherwise, it can be incitement to oscillation. If you are cascoding a stage, it may be wise to bulk up on the input stopper resistors - I've been there/a word to the wise, and all that....
Using the Cascode is much simpler with depletion mode. DN3545N3 on the bottom, and DN2540N5 on top. I want to use the IXYS 10Mxx on top as its gate-source voltage is higher( leaving the lower device with more voltage to work with). Higher voltage reduces its capacitance too...
Cathode-to-cathode 18V zeners have *COMPLETELY* stopped the device failure, both as plate loads and tail loads.
Using the cascode also leaves a consistent capacitive load which, IMO sounds better than the single device. I run the lower device with 3x 3.3V LiIon button cells so the R-set is bigger.
cheers,
Douglas
Cathode-to-cathode 18V zeners have *COMPLETELY* stopped the device failure, both as plate loads and tail loads.Using the cascode also leaves a consistent capacitive load which, IMO sounds better than the single device. I run the lower device with 3x 3.3V LiIon button cells so the R-set is bigger.
cheers,
Douglas
Using the Cascode is much simpler with depletion mode. DN3545N3 on the bottom, and DN2540N5 on top. I want to use the IXYS 10Mxx on top as its gate-source voltage is higher( leaving the lower device with more voltage to work with). Higher voltage reduces its capacitance too...
Using the cascode also leaves a consistent capacitive load which, IMO sounds better than the single device. I run the lower device with 3x 3.3V LiIon button cells so the R-set is bigger.
cheers,
Douglas
So I understand I can safely use a DN2540 as the follower (top) device?
I have a CCS based on DN2540/2SK170 @ 6 mA.
I drive a 300B from 5751 or ECC83. Currently with a Ra resistor but had gyrator in the past.
I had an IRF830 as follower (without zeners
) and it went to heaven 
on one channel; ain't got any left but do have DN2540 to use.
Yes you can use the DN2540 as a source follower device. You have to add protection zener(s) between gate and source as the maximum reverse voltage is 20V. I'd add a back to back 15V zeners if you are DC-coupling the gate of the source follower to a previous stage:
Here is an example: http://www.bartola.co.uk/valves/2016/01/09/slew-rate-part-iv-and-source-followers/
cheers
Ale
Here is an example: http://www.bartola.co.uk/valves/2016/01/09/slew-rate-part-iv-and-source-followers/
cheers
Ale
Thanks Ale. Very instructive.
I doubt I can hear the difference in the slightly higher H3 with the DN2540 compared to the a mosfet like the IRF830. Might even like it as I can appreciate the sound of a 6SN7 + Ra - harmonic degradation is important. There are these strange compensating effects with the 300B!
I will not forget the zeners (2x14V back to back).
I see from the sheets that the STP3NK60ZFP does have an integrated back-to-back zener while the IRF830 does not. So no extra external zener is needed with the STP3NK60ZFP I assume.
I cannibalized a IRF830 and will use it.
I doubt I can hear the difference in the slightly higher H3 with the DN2540 compared to the a mosfet like the IRF830. Might even like it as I can appreciate the sound of a 6SN7 + Ra - harmonic degradation is important. There are these strange compensating effects with the 300B!
I will not forget the zeners (2x14V back to back).
I see from the sheets that the STP3NK60ZFP does have an integrated back-to-back zener while the IRF830 does not. So no extra external zener is needed with the STP3NK60ZFP I assume.
I cannibalized a IRF830 and will use it.