Well...output stage plate to driver stage cathode feedback....
It certainly works, however using a pot to adjust the level of feedback is enlightening. I can achieve the same THD level and spectrum (very close to identical) as I do using only plate to grid local NFB at each stage of a 3 stage amp.
Downside....well now I encounter clipping much earlier, at only 1.5W output.
Scoping the driver cathode signal, and I have pretty much the same AC signal at the cathode as the quiescent Vgk.
I.e. 7V rms cathode feedback signal, and 6.2V Vgk. The waveform is a distorted sine with a 'nipple' where the grid is driven positive, a tiny bit.
I am not using a DC coupled driver stage, so driving the grid positive should be very limited in capability.
I would be as bold to say, the simple plate to grid FB over each stage, is more effective, AND gives me another 0.25 to 0.5W output at a similar level of THD.
How disappointing.
Perhaps now I might need to readjust the input stage to generate some 2nd, and cancel out the -15dB 2nd (not good at all)
However, given how well the global feedback worked, in terms of THD, more than halving the level of THD; I now suspect that the OPT is contributing a large amount of 2nd harmonic that just wont shift, without GNFB.
On this build I was hoping to avoid all SS devices, except diodes - just because, for the challenge of doing so.
But I may have to just try using a CCS on the 6P6S cathode, and see if that straightened things out a little. Ho Hum
It certainly works, however using a pot to adjust the level of feedback is enlightening. I can achieve the same THD level and spectrum (very close to identical) as I do using only plate to grid local NFB at each stage of a 3 stage amp.
Downside....well now I encounter clipping much earlier, at only 1.5W output.
Scoping the driver cathode signal, and I have pretty much the same AC signal at the cathode as the quiescent Vgk.
I.e. 7V rms cathode feedback signal, and 6.2V Vgk. The waveform is a distorted sine with a 'nipple' where the grid is driven positive, a tiny bit.
I am not using a DC coupled driver stage, so driving the grid positive should be very limited in capability.
I would be as bold to say, the simple plate to grid FB over each stage, is more effective, AND gives me another 0.25 to 0.5W output at a similar level of THD.
How disappointing.
Perhaps now I might need to readjust the input stage to generate some 2nd, and cancel out the -15dB 2nd (not good at all)
However, given how well the global feedback worked, in terms of THD, more than halving the level of THD; I now suspect that the OPT is contributing a large amount of 2nd harmonic that just wont shift, without GNFB.
On this build I was hoping to avoid all SS devices, except diodes - just because, for the challenge of doing so.
But I may have to just try using a CCS on the 6P6S cathode, and see if that straightened things out a little. Ho Hum
Last edited:
You should not get less power output unless you are just loading the output stage too much with the feedback resistor or something. There is a mistake somewhere.
Where did you have the pot? Was it the upper resistor or lower? Either one is going to throw off the bias of the input stage as you adjust. Or did you have a DC blocking capacitor in the feedback?
I once assumed the same about the OT generating lots of distortion, but when I measured the difference in distortion between primary and secondary, the difference was pretty small. Try measuring both.
Where did you have the pot? Was it the upper resistor or lower? Either one is going to throw off the bias of the input stage as you adjust. Or did you have a DC blocking capacitor in the feedback?
I once assumed the same about the OT generating lots of distortion, but when I measured the difference in distortion between primary and secondary, the difference was pretty small. Try measuring both.
Hi,
The pot was wired as VR and was Teed onto the voltage divider of 1k and 2x LEDs to give 7V at the cathode,
Feedback tapped via 0.47u connected to the anode.
I need to go back and look at square wave response with it set up like that. It was only a 5k pot, 1k to ground. 2x red LED to the cathode.
Loading is most definitely what I think is happening. I could replace the LEDs with a bypassed upper half of voltage divider, to get a similar operating point. It was somewhat higher at 7V in the driver stage, compared with before (5V). But even so, that resistance isn't likely to be more than...5k.
I elevated the driver stage voltage, more than I should really, to about 350V HT for the 6S6B triode. Va ends up about 240V roughly. It improved output a little.
So either way....that's a heavy load.
It would need a follower, and feedback to the grid instead, or...a more heavy duty driver than the 6S6B I'm using.
I.e. I think if I were to load the 6S6B, like the 6N16B, I'd want better than 20k. Probably 25k.
I will see what I can do to get the feedback to ground above 25k, as I am shunting 90% of the voltage to ground.
The pot was wired as VR and was Teed onto the voltage divider of 1k and 2x LEDs to give 7V at the cathode,
Feedback tapped via 0.47u connected to the anode.
I need to go back and look at square wave response with it set up like that. It was only a 5k pot, 1k to ground. 2x red LED to the cathode.
Loading is most definitely what I think is happening. I could replace the LEDs with a bypassed upper half of voltage divider, to get a similar operating point. It was somewhat higher at 7V in the driver stage, compared with before (5V). But even so, that resistance isn't likely to be more than...5k.
I elevated the driver stage voltage, more than I should really, to about 350V HT for the 6S6B triode. Va ends up about 240V roughly. It improved output a little.
So either way....that's a heavy load.
It would need a follower, and feedback to the grid instead, or...a more heavy duty driver than the 6S6B I'm using.
I.e. I think if I were to load the 6S6B, like the 6N16B, I'd want better than 20k. Probably 25k.
I will see what I can do to get the feedback to ground above 25k, as I am shunting 90% of the voltage to ground.
Last edited:
Regarding measuring THD on both primary and secondary, I once did exactly that, on this OPT, but with a different valve driving it (3 parallel 6P30B-R) and my recollection of that, was that THD was worse on the primary side, and better on the secondary.
It completely threw me, because I went into the test believing that the measure at higher voltages with appropriate scope probe, would resolve better low level information, but it didn't.
Since then I have measured on the secondary only, because its what the speaker sees anyway, and it gave me nicer (conveniently) THD results, when comparing, in that instance.
It completely threw me, because I went into the test believing that the measure at higher voltages with appropriate scope probe, would resolve better low level information, but it didn't.
Since then I have measured on the secondary only, because its what the speaker sees anyway, and it gave me nicer (conveniently) THD results, when comparing, in that instance.
At the high frequency end of the spectrum, an output transformer leakage inductance (some have more than others), the secondary will be rolled off above the audio spectrum.
In that case, for high frequencies, the harmonic distortion is higher at the primary than at the secondary.
Just one possible cause of what you got.
In that case, for high frequencies, the harmonic distortion is higher at the primary than at the secondary.
Just one possible cause of what you got.
If I'm understanding your circuit correctly, you are starting with a closed-loop gain of 6 and adjusting down from there. That's pretty low gain for two stages working together. Is the previous stage going to be able to drive this with low distortion?
Also, making the upper feedback resistor a 5k (or less) will present a very heavy load to the output tube plate. What's the OT primary impedance?
I used 220k so that I wouldn't load the output stage too much and I put the pot in the cathode of the driver tube to vary feedback (adjusting bias as necessary). I ended up with 500R in the cathode circuit for an overall closed-loop gain of about 440 across the two stages (measured at output tube plate). Then I made sure that the combined gain of the stages was well above that so that the feedback would be effective. Voila, a two stage amp with low distortion that can be driven to clipping by my phone.
Also, making the upper feedback resistor a 5k (or less) will present a very heavy load to the output tube plate. What's the OT primary impedance?
I used 220k so that I wouldn't load the output stage too much and I put the pot in the cathode of the driver tube to vary feedback (adjusting bias as necessary). I ended up with 500R in the cathode circuit for an overall closed-loop gain of about 440 across the two stages (measured at output tube plate). Then I made sure that the combined gain of the stages was well above that so that the feedback would be effective. Voila, a two stage amp with low distortion that can be driven to clipping by my phone.
I don't know about a closed loop gain of 6....
Driver stage gain is about 4 of 5 times, and output stage gain about 3 times, in the previous iteration, with only local plate to grid feedback.
I am trying to get back to the same output, so the same gain, using feedback over 2 stages, output plate to driver cathode. The input stage is unchanged. So between 12 and 15 V/V gain, is probably nearer the mark.
The feedback resistance is too low, if it is loading anything - I'll look into how much I can improve it.
Driver stage gain is about 4 of 5 times, and output stage gain about 3 times, in the previous iteration, with only local plate to grid feedback.
I am trying to get back to the same output, so the same gain, using feedback over 2 stages, output plate to driver cathode. The input stage is unchanged. So between 12 and 15 V/V gain, is probably nearer the mark.
The feedback resistance is too low, if it is loading anything - I'll look into how much I can improve it.
Perseverance...
Well... today I revisited the iteration that gave best THD vs power output...and took a frequency response.
I havent plugged all the numbers into the graph yet, but...the "resonances" above the audio band, don't appear as severe.
Since I tested this arrangement a few weeks back, I have changed a couple of things, mainly slightly different plate to grid feedback resistances, reduced values, increased FB, removing LEDs or bypass capacitors.
Now when I dial in the global feedback from the OPt secondary, I need less feedback signal, to arrive at a good result.
Output starts to decline at 25kHz, and it still in a slow shelf of decline, at 100kHz. But my fear of resonances, was a little "knee jerk" reaction.
Nice to say, I make 3.9V/8R2 with 1V rms input, and with 2nd and 3rd harmonics at -40dB!
That'll do...said the perfectionist (who's line is that anyway?)
Well... today I revisited the iteration that gave best THD vs power output...and took a frequency response.
I havent plugged all the numbers into the graph yet, but...the "resonances" above the audio band, don't appear as severe.
Since I tested this arrangement a few weeks back, I have changed a couple of things, mainly slightly different plate to grid feedback resistances, reduced values, increased FB, removing LEDs or bypass capacitors.
Now when I dial in the global feedback from the OPt secondary, I need less feedback signal, to arrive at a good result.
Output starts to decline at 25kHz, and it still in a slow shelf of decline, at 100kHz. But my fear of resonances, was a little "knee jerk" reaction.
Nice to say, I make 3.9V/8R2 with 1V rms input, and with 2nd and 3rd harmonics at -40dB!
That'll do...said the perfectionist (who's line is that anyway?)
Last edited:
Crucially, I have no bypass on the output stage cathode, so I have raised output impedance enough to damp the OPT resonance a little better than before. (I guess?)
I still need to try and decrease gain above 20kHz, and test to find the output impedance, but it's looking a lot easier to tame now.
So, thank you for your help and advice!
I still need to try and decrease gain above 20kHz, and test to find the output impedance, but it's looking a lot easier to tame now.
So, thank you for your help and advice!
Last edited:
Well, maybe I'm misunderstanding the circuit from the paragraphs. It wouldn't be the first time.
I'm picturing a feedback resistor of 5k from output tube plate to driver cathode, then a resistor of 1k from driver cathode to GND.
This would give closed-loop gain of A = 1+(5k/1k), assuming that you have enough open-loop gain. It's just series-applied voltage feedback so it is going to be similar to a non-inverting opamp configuration when calculating closed-loop gain.
Hi,
Yes as you imagine is accurate.
I replaced the LEDs with a single cathode resistor of about 3k3.
This meant that I could increase the Rf proportionally, to about 15k, connect to the cathode, rather than the junction of a divider.
It worked better but not great. So I went to Rf of 56k. Much much less FB, but less loading too. But it left me with a lot of gain to shed, and no recourse with plate to grid local feedback, as in those cases Rf is about as low as I want to go, 500k and 270k.
Yes as you imagine is accurate.
I replaced the LEDs with a single cathode resistor of about 3k3.
This meant that I could increase the Rf proportionally, to about 15k, connect to the cathode, rather than the junction of a divider.
It worked better but not great. So I went to Rf of 56k. Much much less FB, but less loading too. But it left me with a lot of gain to shed, and no recourse with plate to grid local feedback, as in those cases Rf is about as low as I want to go, 500k and 270k.
I am wondering if there's a rule of thumb regarding the unbypassed Rk value? If not we should come up with some.
I mean, even having global NFB, it is usually good to linearize each stage as much as possible, but we don't want to eat away too much gain so the global NFB cant work properly.
Those not using global NFB will have the compromise between higher output impedance, enough gain so typical input signal levels drives the amp to it's maximum, etc.
Higher output impedance damps transformer resonance, but also reduces bandwidth due to trasnformer parasitics.
This is very tube dependent. for instance low rp and low mu types like 2A3, 6C33, and 6AS7 etc can have relatively larger Rk values than higher rp and mu types such as 6V6, 6L6, EL34s, etc. I guess mu is the dominant parameter to use?
I'm thinking selecting unbypassed Rk which reduces gain by 1/2 and/or doubles output impedance by 2x is reasonable value. Compared to bypassed Rk.
Or is a factor of 1.41 and 0.71 (SQRT of 2) for a 3dB reduction better?
Any takes?
I mean, even having global NFB, it is usually good to linearize each stage as much as possible, but we don't want to eat away too much gain so the global NFB cant work properly.
Those not using global NFB will have the compromise between higher output impedance, enough gain so typical input signal levels drives the amp to it's maximum, etc.
Higher output impedance damps transformer resonance, but also reduces bandwidth due to trasnformer parasitics.
This is very tube dependent. for instance low rp and low mu types like 2A3, 6C33, and 6AS7 etc can have relatively larger Rk values than higher rp and mu types such as 6V6, 6L6, EL34s, etc. I guess mu is the dominant parameter to use?
I'm thinking selecting unbypassed Rk which reduces gain by 1/2 and/or doubles output impedance by 2x is reasonable value. Compared to bypassed Rk.
Or is a factor of 1.41 and 0.71 (SQRT of 2) for a 3dB reduction better?
Any takes?
This is a really good idea for HP filter?
Rk times mu, Rp goes up, with fixed Lpri, LF extension reduces.
Nice: 'another brick in the pole' - and benefit of degeneration upon linearity of the output tube at the expense of power output? ..
Could be a welcomed trade-off.
HK
Last edited:
Going back over my notes, I have similar result without bypass on the output stage, and no plate to grid FB, as I do with a bypass in place and also local plate to grid FB in the output stage.
So I reverted to using a bypass capacitor on the output stage cathode and reconnected the plate to grid, to bring gain back.
I've plotted the graph of frequency response, which is about -8dB at 150kHz.
Now working on reducing the gain bandwidth, so something more suitable for audio.
Ok, It's hard to follow all this in plain text. 56k is still a hefty load.
If you want to use a higher value of Rk (and allow a higher value feedback resistor), you can use a pot and tie the wiper to the grid bias resistor to adjust tube bias.