EL84/6BQ5 Ultra-Linear Push-Pull Recomendation

[diegot]

I have some old parts in my stock and would like to make something usefull with then.

Here is what I have:
350-CT-350, 6,3V, 5V Power transformer
8K, 10W PP w/ UL taps Output transformer
RCA 5R4 Rectifier
RCA 8BQ5 (smoked glass) Power Pentodes
For de pre-amp/phase-splitter I have one Siemens E88CC, a lot of 12AX7, EF86, 12AU7 and 6SN7.

I think that I can do something very good with this parts, but I'm not convinced with design follow to build this amplifier.
My aim is 10W class A and 4 tubes, maximum 5. This is gonna be a mono amplifier.

Which design do you recommed?

Thanks in advance.

[alexmoose]

gorgeous EL84s, I have a pair of old Magnavox el84a, and there pretty cool.

Personally, look into the Red Light Disctrict, it pops up on the forum about once a week, I want to build it, but it'll have to wait a while, but it looks like a solid design.

you can do the simple el84 seen on http://www.diyparadise.com/simpleel84.html
although it is self splitting, and some people hate that. I have the line stage out of it, and it works as a great pre-amp

[d2134]

Hi,
I recomend you this paper:
http://www.artimusica.nl/wim/dc_coupled_amp.pdf

It was disscuted allready
el84 amp

[ray_moth]

That article is very interesting! I am a little puzzled as to how a triode (Figs 3 and 4) can be used as a CCS. With a pentode, you have the screen at a constant voltage, which guarantees that the plate current won't change significantly as the plate voltage fluctuates. A triode gives no such guarantee, since its current is strongly influenced by the plate voltage.

[Brian Beck]

Originally posted by ray_moth
That article is very interesting! I am a little puzzled as to how a triode (Figs 3 and 4) can be used as a CCS. With a pentode, you have the screen at a constant voltage, which guarantees that the plate current won't change significantly as the plate voltage fluctuates. A triode gives no such guarantee, since its current is strongly influenced by the plate voltage.

Ah yes, but it's all in the cathode resistor! That resistor provides feedback to the cathode that makes it all work. Viewed from the plate, the triode CCS’s dynamic resistance is (mu+1)*Rk + rp. In figure 4, Rk is 47K. The CCS resistance would be about 1.6 mega-ohms, which is not bad. A pentode can do even better, but there are diminishing returns. Also, you often see pentode CCSs that actually are wired as triodes, perhaps inadvertently. Since the cathode voltage of the pentode CCS is NOT grounded as it usually is in a common-cathode amplifier, the screen must float along with the cathode, and should NOT be tied to a supply if you want pentode behavior. This usually requires a capacitor from cathode to screen with a resistor supplying screen current that is connected to a more positive voltage. With the cap, the CCS won’t behave as such down to “DC”. A floating DC cathode-to-screen supply can be made, but it gets pretty complicated, pretty fast. If Rk can be made fairly large, and a moderately high-mu triode is selected, the “mu+1” multiplier of a triode CCS can drive the resistance up high enough for most applications, particularly for LTP "tail" duty.

[ray_moth]

Fascinating! Thank you very much!

[Eli Duttman]

Originally posted by Brian Beck


Ah yes, but it's all in the cathode resistor! That resistor provides feedback to the cathode that makes it all work. Viewed from the plate, the triode CCS’s dynamic resistance is (mu+1)*Rk + rp. In figure 4, Rk is 47K. The CCS resistance would be about 1.6 mega-ohms, which is not bad. A pentode can do even better, but there are diminishing returns. Also, you often see pentode CCSs that actually are wired as triodes, perhaps inadvertently. Since the cathode voltage of the pentode CCS is NOT grounded as it usually is in a common-cathode amplifier, the screen must float along with the cathode, and should NOT be tied to a supply if you want pentode behavior. This usually requires a capacitor from cathode to screen with a resistor supplying screen current that is connected to a more positive voltage. With the cap, the CCS won’t behave as such down to “DC”. A floating DC cathode-to-screen supply can be made, but it gets pretty complicated, pretty fast. If Rk can be made fairly large, and a moderately high-mu triode is selected, the “mu+1” multiplier of a triode CCS can drive the resistance up high enough for most applications, particularly for LTP "tail" duty.

Brian,

What is an appropriate cap. value to tie g2 and cathode of a pentode together? Sticking to the "usual" down to 5 Hz. seems reasonable. "Dead reckoning" indicates that a Xicon 1429-6474, 470 nF./630 WVDC MPP part could be OK, but a more rigorous discussion would be welcome.

FWIW, I recently suggested using the pentode section of a 6U8, here, in the current sink role to load a cathode follower.

[d2134]

Here is a basic design for an ultraliniar 10W using 6N2P and 2 x 6P14P ( equivalent 12AX7 and 2 x 6BQ5).
Caption reads: "Fig 8-16. Schematic for amplifier with ultraliniar PP final stage."

[Eli Duttman]

Originally posted by diegot
I have some old parts in my stock and would like to make something usefull with then.

Here is what I have:
350-CT-350, 6,3V, 5V Power transformer
8K, 10W PP w/ UL taps Output transformer
RCA 5R4 Rectifier
RCA 8BQ5 (smoked glass) Power Pentodes
For de pre-amp/phase-splitter I have one Siemens E88CC, a lot of 12AX7, EF86, 12AU7 and 6SN7.

I think that I can do something very good with this parts, but I'm not convinced with design follow to build this amplifier.
My aim is 10W class A and 4 tubes, maximum 5. This is gonna be a mono amplifier.

Which design do you recommed?

Thanks in advance.

You did say 8BQ5. Your biggest problem is going to be powering the 600 mA./8.0 V. heaters, whatever signal topology you settle on.
8BQ5 Data Sheet

10 W. from a PP pair of Class "A" 'BQ5s set up in pentode mode is a very realistic expectation. While I INTENSELY dislike the 12AX7 as a driver, it will be OK feeding pentode mode "finals". The basic signal topology of "El Cheapo" using a 'X7 as the splitter/driver will work. If you can scrounge up a 12AT7, you will be BETTER off.

You could get double duty from the negative supply. A LM337 adjustable negative regulator can be set up to O/P a steady 600 mA. Wire the 8BQ5 heaters in series and feed them 600 mA. of DC. 24 VAC/2.0 A. bridge rectified and 'lytic filtered, will feed the LM337 and an additional RC section for the splitter/driver.

Finally, don't forget to regulate 'BQ5 g2 B+ for lowest possible open loop distortion.

[Brian Beck]

Originally posted by Eli Duttman
Brian,
What is an appropriate cap. value to tie g2 and cathode of a pentode together? Sticking to the "usual" down to 5 Hz. seems reasonable. "Dead reckoning" indicates that a Xicon 1429-6474, 470 nF./630 WVDC MPP part could be OK, but a more rigorous discussion would be welcome.

Eli,

That’s a great question. I don’t have the analysis at hand. Nor do I recall seeing a rigorous treatment anywhere. I’m sure someone has grunted through it; I just haven’t stumbled across it. Hopefully someone else will have the exact equations to post here. Exact equations will involve some rather tedious math because of the screen’s involvement and because of the little feedback loop that’s formed by the cap. If I have time (ha!) maybe I can grind through the analysis. I’d use a technique using the “hybrid admittance matrix”. But it might get unwieldy on me. They used to teach this analysis technique in EE colleges; I don’t know if they still do. But don’t hold your breath waiting on me.

Anyway, this might be better solved by breadboarding and testing, or by a simulation. Be aware, though, that most simulation models of pentodes don’t treat the screen with sufficient complexity or fidelity. The screen has its own transconductance, and its own resistance, like a plate resistance. But a sim might get you in the ballpark, at least. My gut feeling is that you’ve picked a too-small capacitor. Most designs would uses capacitors in the several-microfarad range.

I can tell you from inspection that the output Z as a function of frequency would probably have a stepped shape when plotted log-log. See my little sketch. The goal is to push both f1 and f2 to below the audio range. However, the CCS doesn’t disintegrate below f1; it becomes a triode-connected-pentode CCS with a degenerated screen. The overall circuit’s response to this reduced output Z may only be slight.