Interesting experiment - Quickie as CF/buffer!

[Dr. Toobz]

I've been playing around with the Quickie, as it's an easy and low-voltage way to learn more about tubes. I recently had an idea - if no gain is needed, but one still wanted to "color" the sound or provide a better match between components, why not run the Quickie as a cathode follower? Obviously, it cannot swing much voltage, but we're only taking about CD-level signals (2V) going into a 100k load. I can report that this actually does work, and IMO, sounds better than the Quickie when used in plate mode. Distortion appears low, the bass is very solid, and microphonics are not as big of a deal. Some gain is lost, of course. This makes me wonder if my circuit is optimal, or if a tube with a higher transconductance would be a better choice (such as the 3V4, which is around 2000 microhmos).

Basically, I ran the B+ to the plates of the tubes (I have a red indicator LED in between; this drops the B+ by 1.5 or 2V), disconnected the output capacitors from the plates, and replaced the 1k resistor/1000uF cathode bypass with a 150H, 3700 ohm choke on each channel (i.e., as to load the cathodes). From there, I hooked up the output capacitor/parallel 470k resistor to the junction in between one leg of the choke and the cathode pin.

Any way to optimize this to get closer to unity gain? I estimate about a 3 or 4dB loss right now. I've played around with cathode resistor values and 3 or 4k seems optimal, but I'm not sure why - I'm guessing it might it have to do with the fact that the ideal load for this tube is about 4k on the plate, since the Rp in triode mode is 4k? So, the inverse of this would be a similar load on the cathode? Also, would the 3V4 work better? I'm also assuming that there is no way to make use of the pentode screens here, since what matters is the signal on the cathode (interestingly enough, the tubes in pentode mode are no different - even disconnecting the screens from the anodes makes no difference here).

[cpaul]

Sorry to resurrect such an old thread, but I've been pondering this approach for a while since, built as stock, I have too much gain.  (I've tried to buffer the gain using resistors in the inputs, but that hasn't done anything noticeable yet.)  But I have a few questions.  First,  how did you arrive at 150H, 3700 ohm for the choke, and why a choke in the cathode end at all?  Second, I currently have the PJCCS installed.  I assume that has to go if I do this change, no?  I don't suppose there's a way to use IT instead of the choke, stock or modified (at plate or cathode???).  I'm going to do some reading up on chokes and CCS in tube circuits, but I'm interested in any wisdom from all of you. 

Thanks,
Carl

[Paul Joppa]

In our experience, cathode followers sound better with a high impedance loading the cathode - presumably they are more linear. A choke is an easy way to provide a high impedance at audio frequencies. There is a Hammond choke of 150H/3700 ohms which is a popular choice to replace the plate resistor, because it is very close to the stock 4000 ohms DC resistance - so the operating point is unchanged. I estimate a good range of inductances for the 4K plate resistance to be 50 to 200 henries, so this choke works well. A lower inductance will audibly damage the deep bass, and by 200 henries you have passed the point of diminishing returns.

Certainly a current source would be even better; that would entail replacing the PNP transistors (MJE350) with NPN (MJE340), reversing the diode, and reversing the input and ground leads. In order to maintain the grid at ground (so as to not require a extra capacitor) you could reference the ground between the first and second of the 9-v batteries; this gives a bit over 9v of compliance for the current source.

The reason for the 3S4 is that is includes triode curves in the specification, and they are actually quite linear. Since there is no such spec for the 3V4, I don't know how linear it may be. They are otherwise pretty similar. It's an easy enough experiment, you just have to re-wire the socket.