voltage regulator using ECL84 + OB2

[ErikdeBest]

It has been a while since I build Frank’s pre amp with regulated PS, a nice and good sounding project. I build it with the regulator as from Frank’s schematics, but I had some trouble sourcing the tubes used – I admit, they aren’t that hard to find, but can cost quiet a bit when you need more regulators. So I started thinking about an alternative and stumbled upon the regulators based on the ECL85 – I found at least three, but all aimed at a 250VDC, while I needed (at least) 300VDC. Surely it’s possible to make a regulator for 300VDC with a ECL85, but I am also charmed of the idea of cascoding triodes for the error amp (just like in Frank’s regulator) to get more amplification, resulting in a (even) lower output impedance.

In my search for cheap tubes I found, beside the ECL85, lots of ECL84. The transconductance from this one seems to be at least as good as from the EL86, widely applied in regulators. Problem of the ECL84 is it’s relative small plate dissipation, in triode mode about 5W, I believe. Hum…what to do? Use two!! Parallel the pass devices (the pentodes of this tube) and put the triodes in a cascode. Even before consulting any theory on the matter, I realised that paralleling the output of the pentodes could give problems because of the very low output impedance. But I went further: I read the articles from Broskie, from steve bench, some books from Elektor, compared schematics of regulators, but beside a small mention from steve bench and frank’s regulator, I didn’t find anything more using a cascode error amp in a regulator. And my search for cascode circuits wasn’t that productive too. So I decided to go for the trial and error, and the support of diyaudio members.

I designed the regulator using two ECL84 and a OB2. I adjusted the resistor which sets the current for the OB…there is now 7mA running through the OB2 (plus any current coming from the triodes of the ECL84!!??). The OB2 regulates at 105V, that way I settled the voltage at the grid of the lowest triode at 100V. I read somewhere that the grid of the upper triode must be at approximately a fixed potential of 80VDC. I added the value of the OB2 to this, and to make things easier set the upper grid voltage on 200VDC, allowing the use of three resistors of the same value. The plate resistor for the lowest triode I copied from Frank (the use of this plate resistor allows some more current flow through the lowest triode, giving better results…so I read). For the plate resistor of the upper triode I took a 470k, half of what Frank uses (a lower value here gives less amplification, but some more current through the circuit).

The value of the capacitor from the regulated output to the grid of the lowest tube I defined according through this formula, which I took from www.tubecad.com. Don’t know if it is right (Frank’s regulator uses almost a decade less)

C >> (R1 + R2)/(2*pi*f*(R1*R2)
C>> (66.000 + 33.000) / (2*pi*2*(66.000*33.000)
C>> 3,61*10-6
C>> 3,61muF
(f is the lowest frequencies for which the regulator will be used, maybe 2Hz is to low).

Well, that was it. I greatly appreciate anyone’s input on this.

[ErikdeBest]

Just a note: in the above schematic, the ECL84 pentodes are wired as triodes, that's why I've drawn triodes.

[ErikdeBest]

Of course that what I have done with the resistor setting is a stupid thing - it will not just settle itself on 300VDC at the output.

I have got the following formula to calculate Vout of a cascode circuit.

Vout = -Vin * mu 1 * Ra (Ri 1 + ((Ra + Ri 2)/(mu 2 + 1)))

In my schematic Ra is 470.000. Values of mu 1, mu 2 and Ri (internal resistance) are dependent on the conditions in which the valve is operating. If I take them from data sheet, they are as follow.

Typical characteristics triode section as AF amplifier
µ = 65 at Va = 200V, Vg = -1.7V, Ia = 3mA, S = 4mA/V and Ri = 16.200 ohms. Filling this in gives

Vout = 1296 * -Vin, what seems a bit too much.

So...ahh... I'm stuck. Maybe prototyping, using potentiometers... but how would this be done in the 'theoretical' way?

[Sch3mat1c]

Negative feedback loop. Don't consider the gain (BTW I think your equation is for AC signal ), consider the voltage divider chain.

Tim

[ErikdeBest]

Hi Tim

Now you talked about negative feedback I see that this circuit really uses it. And about that equation, indeed, is meant for AC

So, you think it should work using the pot to adjust the desired voltage?

And something else...I hope to join you in the MP amplifier idea (and implementation)! I've already been thinking about some issues about it (which tube to use, bias arrangements for correct balancing), but haven't gone further than that.

Erik

[Sch3mat1c]

If you want to vary voltage on that circuit, change the middle and bottom 33k resistors to an, uh, 66kohm pot. Good luck with that; 47k for the top resistor and 100k for the pot will be easier to find. Keep the capacitor from output to pot wiper -- couples ripple to the error amplifier, improving rejection.

Tim

[ErikdeBest]

Hi Tim

Thanks for your help! I greatly appreciate it. It's not that I want to vary te output that much...just trying some fine adjustments here.

Erik

[Sch3mat1c]

Yep, either is possible. If you want to restrict range of the potentiometer, just subtract some from the ends and add it back in as fixed resistors, for example taking my previous values, you could use a 20k pot with two 40k resistors.

Tim

[ErikdeBest]

Hi

When I checked duncanamps for the exact pinout of the ECL84 i found this note: g3 of pentode connected to triode cathode. Checking some of the datasheets I found that g3 is connected to a (kind of) screen between triode and pentode. I've also just measured impedance between g3 and cathode, but there is now connection between them. So I will go further with this one.

A fast question, actually two questions. I want to use perforated board for the prototype. I'm leaving one track between the high voltage (~350VDC) and other voltages. Any idea about possible sparks?

And about the needed reference voltage. I will be using a OB2 now because they seem to be "better" than stacked zener diodes, which produce too much noise. I wonder if reduced noise, but still using SS, could be obtained from the combination of a voltage reference as the REF02, amplified to about 60VDC through a power opamp supplied from a single rail (like a opa541 that can take 70VDC). In the original circuit the OB2 (or zener) will have some current from the cascoded triodes flowing through it - can a (power) opamp handle this current?

Thanks, many thanks

Erik

[Sch3mat1c]

I don't know where people are getting these ideas from. Zeners are way quieter BECAUSE you can put massive capacitance across them. Try that with a 0B2, hope you have a lot of spares!

Tim