ecp audio Home  |  Products  |  Contact  |  F.A.Q.  |  D.I.Y.  |  Old Stuff

Sunday, August 22, 2010

The Caged Frog -- A Pentode Based Transconductance Amplifier for Headphones

Caged Frog -- Single Ended Pentode

The truth of the matter is that after years of playing around with all kinds of single ended triode (S.E.T.) amps, I'm not sure that they are really my cup of tea. Sure, they do a some things really well, they have that sweet midrange after all, but some other things just never seem quite there. Indeed, I generally feel like they lack coherence when things get complicated. The usual alternative is push-pull, which certainly has it's place, and indeed a good differential circuit is hard to beat. But, there are other solutions as well.

One solution that is often ignored is the lowly pentode.
Derided in tube circles as hardly better than solid state,* pentodes, if they are used at all, generally are either strapped into being triodes, or act as triodes' supporting cast members, as a CCS for instance. But, pentodes can be brought to the fore and in some instance they can really shine on their own.
Like a lot of my projects, this one started somewhere else. It started with trying to build a single ended, transformer coupled, pentode amp. Well, a single ended, transformer coupled pentode amp that was intendedly different from the one that actually came about.
The sort of triodes one would typically load with a transformer have plate impedances ranging from about 800 ohms for a 300B or 2A3 to perhaps 2500 for something like a 6DJ8. This allows loading them with a transformer 3 to 5 times their rp to provide a 2500 to 10K load or so, which is really the practical limit for transformers. Pentodes, on the other hand, have a high plate impedance -- 15000 ohms for an EL34, 60000 ohms for a 12HG7, much higher for small signal pentodes. There is no practical way to load them with a transformer that is 3 to 5X rp. If you load your EL34 with a 5K:8 transformer, you end up with an output impedance going to your speaker of about 25 ohms. For most modern speakers, you want the output impedance of your amp to be as low as possible (as near 0 ohms as possible), and 25 ohms is really pretty high. To overcome this high impedance, feedback is generally employed.

And that is where this project started -- I was trying to find a way to apply feedback to a single tube, transformer loaded, pentode. The transformer in question was 5K:32, which, when combined with the 12HG7 I wanted to use meant a Zout of around 400 ohms. I asked on the audio boards, and I came up with a few of options. The first was to run a feedback loop either from the output or from the plate (through a capacitor) back to the grid.

Feedback from the output to the Control Grid

Feedback from the Plate to the Control Grid

The problem with these options, however, is that without a driver of some sort to isolate it, the volume pot becomes part of the feedback circuit meaning that the amount of feedback applied changes with volume, which is probably not a good thing.

Another perhaps more interesting option is to put the output transformer's secondary in series with the cathode resistor.

Feedback in series with the Cathode
The interesting sort-of-upside to this is that regardless of the tube you use and the output transformer, this provides 100% negative feedback, leaving you with a unity gain stage -- a buffer. Since a buffer is all I really need for phones anyway, this would seem like a good option. However, because there is a DC voltage across the secondary (here, the DCR is 6 ohms, and I was running the tube at about 20ma meaning 120mV of offset), this ends up being across the attached headphones. For speakers that can take a little offset, this is not necessarily terrible, and indeed a transformer designed for 4 or 8 ohm speakers will have a lower secondary DCR than one for 32 ohm headphones, but for headphones it could be disastrous. 

One solution to this is to add a separate winding to the transformer just for feedback, and this is what the EAR headphone amp does to aparently very good effect. Indeed, it is a particularly good option as one can control the amount of feedback applied. That is, it need not be 100%. But, there is also an easier, and cheaper, solution which is to only run the AC though the secondary (George at Tubelab.com does this on one of his amps). This provides the feedback without the offset issues.
Feedback in series with the Cathode (AC Coupled)

So, this is what I built.

And it sounded awful. In some perverse sense, it was a "good" amplifier, but the sound was sterile and lifeless. Indeed, it sounded like the worst-critic-of-solid-state's nightmare amp.
This had all just been an evening's tinkering. The amp was built on breadboard, so it was not a big loss -- just another experiment that went south. But, as I was about to disassemble it and put the parts away, I wondered what the circuit would sound like without any feedback. That is, just a pentode with a transformer load. I figured it was going to be awful, so I was not prepared for what I heard, which was near sonic bliss. From note one, this was something special. 

Feedback Free
Turns out, I had built a transconductance amp more or less by accident. Most amplifiers that people use are voltage sources, with a low output impedance. A transconductance amp on the other hand is a current source, with a high output impedance. The only commercial amp of this sort that I am aware of is the First Watt F1, and you can read more about transconductance amps at Nelson's site. In a sense, it is a sort of solution looking for a problem. That is, most modern speakers do not work well with this sort of amplifier, and so the technology is one that is not often utilized. However, headphones, and particularly low impedance Grado headphones with their nearly flat impedance curve are a near perfect candidates. 

The only real difficulty here is that the tubes I used, 12HG7's, are really noisy. I needed a regulator for B+, another regulator for the screen grid, as well as one for the heater. Even with that, it is still not the quietest amp I've ever built. But, it is certainly one of my favorites to listen to.
Finally, a couple of construction notes. First, pentodes are not like triodes, which is maybe an obvious thing to say. But, why it's important is it influences power supply requirements. With a transformer loaded triode, you need to balance B+ with the amount of grid bias and the current through the tube. If B+ is too high, then current and bias increase. Pentodes, however, are a lot easier in this respect. Pentodes behave essentially like depletion mode mosfets in that they are constant current devices. The voltage between the control grid and the cathode determines the current across the tube, and any extra B+ will just be dropped by the tube. In practical terms, this means that so long as the tube has enough voltage, and so long as the tube is able to dissipate the heat, the actual B+ voltage is not that important. I have run my amp with power supplies ranging from 200V up to over 350V with no isssues, other than more heat.

Second, the overall tone of the amp can be greatly influenced by putting a resistor in parallel with the headphones. It changes the damping factor, and by adjusting the resistance, you can alter the amp from clean tight bass, to punchy full, to flabby. It's really pretty neat to play with.
Oh, and yes, the amp does look a lot like The Electric Avenue. I gutted that amp to build this one.
* I suspect that a lot of the bias against pentodes comes from the fact that they are more complicated to use, making them more work than many hobbyists may want to do.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.