I guess I don't know how to figure Cm in this case. Do I use the plain pentode gain or the gain that I will get with feedback to multiply by the grid-to-plate capacitance to figure Cm? If I take a pessimistic approach, Cm could be up to 30pF in a KT88?
If I want a corner frequency at 200kHz, that would give me a feedback resistor of 26.5k. This seems low and likely to load the driver down. I don't know what this will do to the source follower. I have 5mA idle current through it right now with a 50k load resistor. It would be easy enough to bump that up.
On the other hand, a 100k feedback resistor would give me a corner frequency of 53kHz which means I may not be as flat as I would like at 20kHz.
Then I will have a capacitor in series with the feedback resistor to block DC. I'm thinking I should shoot for a 2Hz cutoff frequency there, does that sound right?
If I want a corner frequency at 200kHz, that would give me a feedback resistor of 26.5k. This seems low and likely to load the driver down. I don't know what this will do to the source follower. I have 5mA idle current through it right now with a 50k load resistor. It would be easy enough to bump that up.
On the other hand, a 100k feedback resistor would give me a corner frequency of 53kHz which means I may not be as flat as I would like at 20kHz.
Then I will have a capacitor in series with the feedback resistor to block DC. I'm thinking I should shoot for a 2Hz cutoff frequency there, does that sound right?
Of course, you should calculate for the real gain, with feedback applied!
Speaking of 2 Hz corner, it depends on other LF poles in the global feedback loop; also power filter time constants must be considered, otherwise you may easily get LF instability through B+ chain, since OPT on such a frequency is almost short.
Speaking of 2 Hz corner, it depends on other LF poles in the global feedback loop; also power filter time constants must be considered, otherwise you may easily get LF instability through B+ chain, since OPT on such a frequency is almost short.
That's an easy one: two words: cost cutting. UL is a lot cheaper than an active screen regulator. Since this is not a concern of mine, the two designs I did using pentode finals include active screen regulation. When using 807s, I found (as suggested by Schade) that parallel feedback off the plates worked as advertised. The other project, using 6BQ6GTBs, didn't need any local NFB at all, since the 6BQ6GTBs produced way less of the nasty high order harmonics. All these needed was a bit of gNFB to take the edge off.
UL basically does the same thing as parallel NFB, but it avoids the necessity for screen regulation, and allows the use of types whose screen and plate operate at the same voltage.
There is just one commercial design that used active screen regulation, and that was a theatre amp, not consumer electronics.
We don't have to since we don't answer to the pencil-pushers in the Accounting Dept.
Yuppers, you got it.
No, that anode-to-grid resistor is in parallel with the reverse transfer capacitance. Cmiller occurs at the input. Simple fix for that: don't return the far end of the resistor to the grid. Take it back to the driver stage, and skip over Crt. That's how I dunnit.
Right. I knew there was something funny about that. We don't care about the
Miller effect, which is a multiplication of the feedback capacitance by the plate
swing relative to the control grid, rather we are worried about where the
feedback reactance starts to become significant in parallel with the feedback
resistor. It's independent of the plate voltage swing.
PS
The MOSFET between the pentode driver plate and the control grid isolates the
feedback capacitance when the feedback goes to the driver plate
hey-Hey!!!,
The circuit is a means of delivering plate to grid FB around the output stage at minimum complexity. No resistor network needed. Single-node PS is also a plus I think; everything gets fed to the OPT's ct( in the PP implementation ).
The output stage's g2 isn't really part of it; it can be riding along in U-L mode, or done pentode, or rigged as triode. It is just plate to grid, at U-L tap ratio of plate signal. One can tap the OPT where you want, or stick with end-of-layer points if y'all want someting other than a specific U-L ratio( anywhere from 20-50% deending on who made the OPT ). It could even be done with a cathode FB topology; fixed g2, cathodes to the tertiary winding and the E-Linear front end attached to taps on the plate winding. While winding the coil it could even be tapped to maintain pentode operation( g2 to tap on the opposite side of the anode CT, but you'd loose the ability to run g2 at other-than-B+ ).
cheers,
Douglas
Yes, it is.
My E-linear is derived from Pete Millette's as shown in the
below URL.
http://ja1cty.servehttp.com/E-LINEAR/E-Linear-amp.png
Due to deep NFB of G2 > G1 in the original Pete's, I reduced it
by means of two pcs of 39Kohm which are connected to the
plate of C3g.
E-Linear is set-fraction output plate to grid FB. That the output g2 is able to get attached to the same tap( creating U-L ) is not relevant. E-Linear can be done pentode, U-L, or triode output stages.
cheers,
Douglas
OK, I think I get it. E-linear is plate-grid feedback using an OPT tap instead of a
divider or shunt network.
What you do with the screen grids is a separate issue.
That was my question, and whether there is any advantage, other than
feedback is feedback, to using g2 modulation in addition to the plate-grid
local feedback.
Thanks, Douglas!
Michael
hey Michael,
I think there is. U-L does not linearize a pentode; its g1 spacing is pretty much like a pentode's IMO. What I do find useful is it reduces its output Z, evidenced by the slant the grid lines now take on. Given the requirements to drive the inductive load( at LF, more or less ), the steeper( v. pentode ) plate curves are useful.
I think that the optimum will be done with a CFB output TX, equipped with taps on its anode winding for the E-Linear connection. Stuff like the Chicago BO-14 comes to mind, but I think I'll have Heyboer wind versions of it with taps at less than original 37% to compliment the 10% cfb tertiary coil.
The CFB method of implementing U-L also offers application to power tubes like TV sweeps with low g2 voltage limits.
cheers,
Douglas
The attached schematic is the original McIntosh Amplifier's
published in 1949. It uses Bi-Filer winding for
P and K windings. Therefore 100% feed back is applied to the
Cathode. Since the Cathode should be the referenced potential,
resulting 100% feed back is applied to both G1 and G2.
The attached schematic uses 100% positive feed back to G2
to cancel G2 negative feed back. Therefore only 100% feed back
to G1 is remained as the beam connected 6L6.
E-Linear uses 50% tapped feed back to both G1 and G2 if
UL tap is 50%.
published in 1949. It uses Bi-Filer winding for
P and K windings. Therefore 100% feed back is applied to the
Cathode. Since the Cathode should be the referenced potential,
resulting 100% feed back is applied to both G1 and G2.
The attached schematic uses 100% positive feed back to G2
to cancel G2 negative feed back. Therefore only 100% feed back
to G1 is remained as the beam connected 6L6.
E-Linear uses 50% tapped feed back to both G1 and G2 if
UL tap is 50%.
Attachments
Last edited:
The McIntosh design sets out to deliver large cathode FB while maintaining pentode operation of the output tubes. They also ran the driver tubes B+ supply from the opposite end of the plate winding( positive FB ) to deliver the required swing to the output tube grids.
E-Linear tapping is up to the output TX winder. I am using multi-tap designs so the amout of FB can be adjusted. 20-30% is adequate, but with OPT designs running 40% U-L taps it is not a large handicap.
The larger tap percentage requires more voltage headroom from the amplifier stage, and would work best with AB designs instead of the B+ limited Class A designs...unless we go to large power finals that can take Class A dissipation at higher voltage...
cheers,
Douglas
A nice advantage of the McIntosh circuit is that it allows great bandwidth cheaply, by using a low-ratio output transformer. It scores you a huge amount of feedback, but it's just as easy to give some of that back with positive plate-to-grid feedback, via a bootstrapped driver stage. Or, run it without positive feedback and you have what behaves like a ~mu=2 triode output stage with vanishingly low distortion and an acceptable open-loop damping factor. Just need a ballsy high voltage driver stage.
There's a Krohn Hite unity-coupled amp out there with a bootstrapped pentode driver good for ~0.0015% THD at 35 watts, using 80 dB total feedback.
I'd go one further, they do it with simple bobbin geometry too; no complex interleave needed.
cheers,
Douglas
As folks explained, E-Linear has similar concept of G1, G2
feed back to McIntosh without special winding transformer,
but needs UL tap. As Bandersnatch says, value of feed back
can be adjusted by the percentage of UL tap. In my case,
I added two pieces of 39Kohm to the plate of the driver,
it reduces G1 feed back to the half, resulting, the driver's
operating condition could be relaxed. Please refer to my schematic
as below.
http://ja1cty.servehttp.com/E-LINEAR/E-Linear-amp.png
'73 de JA1CTY & JA2DHC/1 Kazuo Ohashi
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.