I thought it would be sporting to report back to everyone who helped out. I was able to get everything up and running last night with my measured results almost perfectly coinciding with my calculated ones. I tried a couple of different configurations and my real-world operating points were always within a few percent of the ones on paper. Ultimately, I don't think this arrangement is going to be suitable for my project — but I am really happy that I was able to figure out how everything interacted.
The key to understanding the circuit — for me, anyway — was not assigning a resistor value for the input stage cathode. I simply made a note of the plate current and grid voltage of the desired operating point and waited until I had established how much current would be flowing into the cathode resistor through the feedback resistor. I then divided my expected grid voltage by the sum of the two currents through the cathode resistor to get my component value. This may be blindingly obvious to some, but it had me stymied for a few days.
As it stands, the circuit offers far too much gain (and thus — far too little feedback and far too much noise) for my purposes. I can only lower the value of the feedback resistor so much without things getting really inefficient. Yes, I could simply AC-couple the feedback resistor to the output stage in order to get a lower gain ratio — but I think this might load the output unpredictably during AC conditions.
Thanks to everyone who helped me work through my thoughts — even if they suspected it wasn't the best approach in the end. I really learned a lot.
-Nick
The key to understanding the circuit — for me, anyway — was not assigning a resistor value for the input stage cathode. I simply made a note of the plate current and grid voltage of the desired operating point and waited until I had established how much current would be flowing into the cathode resistor through the feedback resistor. I then divided my expected grid voltage by the sum of the two currents through the cathode resistor to get my component value. This may be blindingly obvious to some, but it had me stymied for a few days.
As it stands, the circuit offers far too much gain (and thus — far too little feedback and far too much noise) for my purposes. I can only lower the value of the feedback resistor so much without things getting really inefficient. Yes, I could simply AC-couple the feedback resistor to the output stage in order to get a lower gain ratio — but I think this might load the output unpredictably during AC conditions.
Thanks to everyone who helped me work through my thoughts — even if they suspected it wasn't the best approach in the end. I really learned a lot.
-Nick
I'm glad you got it figured out in the end.
I think if I were to attempt this I would CCS load the 12AX7 and buffer the output with a follower. It would be extremely linear, have a low Zout, and you would get more gain (hopefully even enough gain).
The CCS would have to be high quality since you are dealing with small currents and high impedances. I'd use a cascode of depletion mode fets.
I think if I were to attempt this I would CCS load the 12AX7 and buffer the output with a follower. It would be extremely linear, have a low Zout, and you would get more gain (hopefully even enough gain).
The CCS would have to be high quality since you are dealing with small currents and high impedances. I'd use a cascode of depletion mode fets.
Great exercise in learning. I like your approach.
If you do try a fet-based CCS, you might also consider the IXYS 10M45 which is really trivial to setup, although ultimately may not give you quite the performance of two cascoded fets. Might not matter though within the audio bandwidth range.
Doesn't hurt to try it. I assume the output to the speaker would still be direct coupled.
This is why I suggested perhaps a dynamic plate load or paralleling the triode sections. Paralleling alone would not offer more gain, but paralleling sections (to use the otherwise unused triode in the glass envelope), in addition to a fet CCS plate load might be just the ticket.
If you do try a fet-based CCS, you might also consider the IXYS 10M45 which is really trivial to setup, although ultimately may not give you quite the performance of two cascoded fets. Might not matter though within the audio bandwidth range.
Doesn't hurt to try it. I assume the output to the speaker would still be direct coupled.
This is why I suggested perhaps a dynamic plate load or paralleling the triode sections. Paralleling alone would not offer more gain, but paralleling sections (to use the otherwise unused triode in the glass envelope), in addition to a fet CCS plate load might be just the ticket.
Last edited:
I made a couple of weak attempts to actively load the plate when I needed a break from figuring out the negative feedback scheme, but the results were underwhelming. I'm certain my efforts were sub-optimal — but I'm willing to try again!
The limitation I am facing is that the only device I have on hand that will readily drop 100v without evaporating is the other half of the 12ax7. If I am reading the load lines correctly, the amount of current the tube could source (while allowing for its voltage drop to change +/-50v in sympathy with the driver triode) is un-usably low for this application.
All of the standard-issue Radio Shack transistors I have on hand seem to be limited to a maximum collector-emitter voltage of 60v. I did have the bright idea of putting two LM317 voltage regulators (in CCS configuration) in series set to source 2.2mA and biased the output around 260v. In theory, these could swing +/-35.75v and withstand 40v without releasing the magic smoke. Everything biased correctly, but the output hard-clipped at much lower voltages than I would have expected. I don't remember if I used a bypass capacitor on the driver's cathode, however — I'm not sure if this makes a difference with a CCS. If there is nothing wrong with the idea on paper, adding a third LM317 in series might increase the headroom enough to be workable. I realize it's unconventional, but I'm trying to work within the capabilities of my parts bin before I order a specialized component. Is there any reason these regulators can't be run in series?
Sidebar: I experimented last night with grounding the grid and driving the cathode of the known-good circuit. I know auditory memory is pretty unreliable, but this has subjectively sounded the best of all of my efforts. A buffered-input, grounded-grid arrangement with an active plate load might be just the thing.
Last edited:
That's a differential amplifier, which is widely used. The two tubes would be directly coupled, but the voltage gain would still be about the same as a single 12AX7 stage.
You could avoid the need for an input capacitor by using either a resistor with a negative supply, or else an active current source, at the common cathode connection.
Ugh. The LM317 requires a bare minimum of 3.5mA (much more if it's slightly out of spec) to maintain regulation. :\
You might be giving me too much credit here. I was just thinking of configuring a JFET buffer to share the cathode resistor as a load. This would depend on getting some sort of CCS to work as an active load for the triode because the original lack of gain is still a problem.
It might make more sense to simply use a JFET to provide a small amount of gain and drive the low impedance of the grounded grid triode. The coupling capacitor isn't a problem since I need to filter the lows anyway.
Sigh . . . there's always input transformers.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.