"No Light District" P-P amplifier

[wrenchone]

This is the initial stab at a schematic for the "No Light District" push-pull amplifier inspired by SY's "Red Light District" amp and by my own cantankerous/contrarian nature. The inspiration was a conversation with SY at the first Burning Amp get-together regarding the RLD and active bias schemes to replace the LED array. As usual, it takes time for things to brew.

The circuit shown has a differential front end rather than the triode front end/concertina phase splitter used in the RLD in order to allow partial feedback experiments. A setup similar to Stu's would probably work ok as well, as the 6CW5 does not appear to be an exceptionally difficult beast to drive. The Pete Millett front end with differential pentodes would also work, so his "red board" could be used as a basis for experiments. I chose the 6CW5 outputs in order to properly use a batch of Baldwin organ transformers scored on E-pay. They were set up to work with 6L6 outputs, and have a 6K impedance working into a 16 ohm speaker. Used with an 8 ohm speaker, they will be just right for the 6CW5s and a lower B+ voltage.

The bias network I chose is a bit complex, but has all the bells and whistles to get rid of all the little peaks and flurbles I saw in my simulation experiments with simpler circuits. Seeing as the 6CW5 doesn't require an exceptionally high bias voltage, I may try an alternate TL431/Darlington active circuit that has a lower parts count. I will post that as time allows, as I want to do one more simulation regarding a possible enhancement.

At any rate, the circuit is a first hack at what I want to try, and will no doubt evolve a bit. All the ugly bits can easily go on on a hunk of perf board squirreled away under the chassis. I may use a roasting pan for the chassis. Ikea and Target have some possible candidates. The Ikea pans are especially cheap, and are made of stainless steel. The natural finish might look pretty cool, especially if brushed.

[Wavebourn]

Anyway I would shunt Q9 by a film capacitor.

[wrenchone]

An RC damping network with said film capacitor may be more appropriate. That worked well on the active bias circuit used with my "Mighty Mite" amp. I'll check my simulations when I get to work tomorrow.

[wrenchone]

Yes, it turns out that I did have an RC damper across the shunt regulator in my simulation at work. Somewhere around a few tents of a microfarad to a microfarad with 1 to 10 ohms in series does the trick. It's difficult to really predict what exactly will be needed to do the job, as the fuzz that it suppressing is very small in amplitude (much, much less that the gross voltage change due to the load step), and may not exist at all in a real circuit doe to the limited slew rate of the tube stage and the undocumented parasistic capacitances that exist in a real circuit. If you want to start with something, then 1uF and 2.2 ohms in series would be ok.

On another note, attached is a circuit using a TL431 that is a bit simpler. A cascode transistor protects the TL431, so it can be used in a situation where the required bias voltage is higher than the 35V maximum rated voltage of the TL431. I may start with this circuit to see if it works as well as simulation suggests.

[wrenchone]

Here are the simulation results, which indicate decent behavior and a static/dynamic resistance of less than one milliohm. The simulation looks a little grainy, but then, the step change in bias voltage is less than a millivolt....

[TheGimp]

I did some comparison distortion testing with a similar voltage sink, buffered LED Bias and standard (bypassed) cathode bias with my Firefly 6P1P amp and found that the voltage sink gave twice the distortion of cathode bias. I did not test a LED array for comparison, but wish I had. This results may be particular to the 6P1P and other similar small tetrodes.

Has anyone else actually built a circuit and tested it?

I think this was the version I tested with, except I ended up removing hthe 33uF cap.

[HollowState]

Should there be a resistor (or break) between the +250 on the left side and the +265 on the right? I don't like feeding the driver B+ directly from the output stage B+ without some isolation.

[wrenchone]

I don't understand why you felt compelled to sling a 33uF feedforward cap into your design. Also, since you are looking at a single-ended design, you'll need some keep-alive bias to keep the circuit from drying out and cutting off during the negative 1/2 cycle. An approprately sized resistor from B+ will work in a pinch. I would also feel more comfortable with some realistically sized PNPs as the pass elements. I don't think you can even get 2N3638s any more, except maybe someone's ancient parts stash.

[TheGimp]

I'm only running 35mA bias current so the 3638As are fine. I do have a stash, but they are virtually the same as a 2N3904 (IIRC) and very similar to 2N2907. You could also stick 2N2222As in there and it wouldn't make much difference.

The 33uF FF cap was an experiment that proved to be pretty much useless. I was trying to use FF to eliminate some voltage variation.

I never saw problems with cutoff, but I was more interested in small signal distortion and found it to be unacceptable. This may be due to the SE amp I'm working with, the particulars of my circuit or a general characteristic of Voltage Sink operation.

Just thought I'd point out my measured results and was wondering if anyone else had actually built a circuit and tested it yet.

[wrenchone]

I use a hefty pass transistor, as one also has to plan for the instance when the amp is driven hard. Silicon is cheap - toasted tubes and output XFMRs aren't.(Think like George...)

Anyway, attached is the next pass of the NLD with the promised TL431 bias regulator circuit. If I'm lucky, I'll get one of these done in time for Burning Amp 2010. To pile heresy on top of heresy, it'll use an SMPS, as it's the simplest way for me to get the voltages I want in the space I want.