I give you ... drum roll ... another dramatic slashing of THD!
Proto 054
THD vs. power 1kHz - 054 (dark) vs. 052 (light)
Proto 054
THD vs. power 1kHz - 054 (dark) vs. 052 (light)
Okay, hold my beer, this even works with just a 100K resistor from the grid to GND on the 2nd LTP triode? What? How? I'm not exactly sure why? MrPeakyMcPeak was no match for a simple RC (100pF/47K) on the EF86' anode resistor though, so that was sorted quickly. Can anyone explain why a 100K on the 2nd triode grid shows this behavior?
Proto 055
Frequency response - Proto 055
Proto 055
Frequency response - Proto 055
Last edited:
Alright, I'm chasing a red herring my tubular friends and I let my enthusiasm get the better of me. In my defence, I'm still a young gun, I'm entitled to that right? Obviously what is happening is that the added network caused an additional unbalance on the anode resistors of the LTP which just further reduced the THD. After removing the Rs and C and putting the 1K to GND back on the 2nd triode I tinkered with the AC balance pot a bit more, I used the infamous 'magic touch' the missus likes so much and low and behold I can get the same results. I'll go stand in the corner now, facing the wall, for the next hour or so ... thank you!
One interesting thing is that since you have a lot of NFB. The signal driving the grids of the KT88 is actually the signal + pre-distortion needed to linearize the output. When I looked at this before you find its a sine wave but with the tops and bottoms flattened slightly. You will see this in ltspice too as the tube models are quite accurate. Your cold LTP will start to produce the kind of distortion needed.
So I was looking at the KT88 output on its own and comparing it with the ECC88 followed by the LTP. In all cases the distortion of the two stage together is much lower than that of just the output stage. So yes distortion cancellation occurs. Its best with no cathode resistors on the LTP and with the LTP cold which is what you have found. The improvement in the 3rd harmonic is nearly 15dB. Now that's a simulation. Interesting I left of the grid stopper off the output and got 2MHz oscillation - which show LTspice modeling of tubes can be quite accurate. However the amount you get is quite unpredictable. I was trying to find a way of adjusting but its tricky.
I did find if you increase the KT88 cathode resistors (22R) and make the output more linear the cancellation improves which means the LPT is not quite giving enough cancellation. That will drop your output power a bit but that's not an issue. If anybody got ideas on how this could be controlled please say.
I did find if you increase the KT88 cathode resistors (22R) and make the output more linear the cancellation improves which means the LPT is not quite giving enough cancellation. That will drop your output power a bit but that's not an issue. If anybody got ideas on how this could be controlled please say.
Last edited:
The other thing I tried was an ECC85 which may be even better when biased cold. Its all quite interesting but how reproducible I don't know.
I've actually tried a few different tubes for the LTP on the prototype, ECC82, ECC99, 12BH7, 6N23P and found the 6922 fared best. Its distortion profile must be the inverse of the KT88, so they partially cancel out?
As requested and per the schematic noted below, but obviously with the feedback disconnected from the OPT:
Proto 057
Frequency response 057 - Open loop
1kHz square wave 057 - Open loop
THD vs. frequency 057 - Open loop
THD vs. output power 057 - Open loop
I always want to try and get the most out of it, the 'get 'er done' attitude I reserve for stuff I don't really enjoy doing, like garden work etc.
