I'm sure that had something to do with it. I did also have to tweak a couple settings even after I cleaned up all the detritus. Either way, it's still working, so I'll take it!
Physically prototyped the OT setup to get real-world figured and pulled those back into the sim. Again, I think I had something wrong with my math before, because with this setup (including proper series resistances) I got close to my original estimates. After warming up the bias a touch, I'm back at 22.5W at 1.2% for a 1VRMS input. 5% THD point is 1.2V in/25.5W out, 10% is 1.6V in/30W out.
Frequency response (prior to clipping) is basically unchanged.
And with that, I'm done with the LTSpice portion of our program. Next step is to actually build the thing and see what happens.
Frequency response (prior to clipping) is basically unchanged.
And with that, I'm done with the LTSpice portion of our program. Next step is to actually build the thing and see what happens.
Last bit on the power/distortion figures from the simulation:
0.3V in -> 1W @ 0.44%
1.0V in -> 22.5W @ 1.2% (no visible clipping)
1.2V in -> 25.5W @ 5% (soft power stage clipping)
1.6V in -> 30.0W @ 10% (hard power stage clipping)
The last of the parts arrive tomorrow. I'm always entertained by how poorly the simulation results match up with reality - it's new and different every time!
Haven't bought a PT yet. I've got decent bench supplies, and I usually end up adjusting voltages several times during prototyping. Not to mention that my current requirements are never what I expect, especially when I'm regulating. I guess that's one of the finer skills that takes more experience to nail down.
This will also be the shakedown cruise for my new prototyping board. Moveable octal relay sockets on DIN rails, euro-style terminal strips, heater rails with insulation cutouts for easy alligator clipping, etc. I'm almost more excited about that than the amp!
0.3V in -> 1W @ 0.44%
1.0V in -> 22.5W @ 1.2% (no visible clipping)
1.2V in -> 25.5W @ 5% (soft power stage clipping)
1.6V in -> 30.0W @ 10% (hard power stage clipping)
The last of the parts arrive tomorrow. I'm always entertained by how poorly the simulation results match up with reality - it's new and different every time!
Haven't bought a PT yet. I've got decent bench supplies, and I usually end up adjusting voltages several times during prototyping. Not to mention that my current requirements are never what I expect, especially when I'm regulating. I guess that's one of the finer skills that takes more experience to nail down.
This will also be the shakedown cruise for my new prototyping board. Moveable octal relay sockets on DIN rails, euro-style terminal strips, heater rails with insulation cutouts for easy alligator clipping, etc. I'm almost more excited about that than the amp!
Actually the simulation results can be quite accurate. However its only as good as the models - the OPT will be the biggest issue, you need to include primary resistance correct coupling factor and some primary capacitance if you want to get the GNFB correct.
There's something very wrong here. 7dB input difference is giving 14dB output difference.
All good fortune,
Chris
I honestly think you will learn more trying things out on an actual physical amplifier and measuring the results. Tubes are very resilient and I have yet to blow anything up trying different things and then seeing the results on my old techtronix scope.
A dim bulb and a variac are nice to have if you aren't sure what you put together will work lol. I would suggest using cheap parts as a starting point in case something melts, but so far I haven't had anything explode.
The fun part of working on a stereo amp is you can mod one channel, measure the results and if it looks reasonable on the scope, listen to one channel vs the other.
I feel like I've fairly quickly figured out the basics based on what sounds good. Using a sim, it appears that it can be fooled into showing a faulty design works great. Like others have said, it likely is useful tuning an already working design, I tune mine based on how they sound, which at the end of the day is what we use them for.
stephe,
I agree.
The late great Bob Pease in EDN magazine, in his column Pease Porridge, once said something to the effect that: analog designers needed to tinker and experiment in order to learn what no class can teach.
I agree.
The late great Bob Pease in EDN magazine, in his column Pease Porridge, once said something to the effect that: analog designers needed to tinker and experiment in order to learn what no class can teach.
Bob Pease and Jim Williams were probably the best analog guys ever. Jim Williams had a "Test your analog IQ" quiz every month with the answer in the next month's issue. Those were hard. And you couldn't find the answers in any textbook or on the internet. Anybody remember that?
At least part of that was an error on my part - that was 0.3V peak and 1.0V RMS. I also rounded by a touch.
Here's 1W output:
And here's 1Vrms input:
Here's the details, stepping from 0.1Vrms to 1.0Vrms on the input:
I'm not sure how to help you. What does "AVG" mean in the last screenshot? I suspect it's not what you think, but I can't tell.
All good fortune,
Chris
All good fortune,
Chris
Believe me, I do plenty of that too (to my wife's chagrin). I didn't even start using LTSpice until less than a year ago.
People learn in different ways. I learn best by integrating what I read, what I can simulate, and what I can build.
People also have different amounts and flavors of free time, and may be able to experiment on a simulator when they can't sit down at the bench. Or they may be waiting for parts to come in so they can start building ...
In any case, the parts are all in, and from here it's all taking place on the bench. Onward!
It's average power output to the speaker. Voltage at the speaker is V(GNFB). Current through the speaker is I(R16). So it's averaging V*I at the speaker.
Last edited:
Thanks. That part all looks fine, so I withdraw my objection, Your Honor.
All good fortune,
Chris
All good fortune,
Chris
Got 'er up and running. My OT isn't a perfect impedance match (11.5K vs 14K); I assume that this is why, when I'm swinging 730Vpp on a 400V supply, I'm not seeing the full voltage swing I want on the output. Since pushing any further would just make it clip harder, I backed off the screens to 250V and adjusted the front-end gain to get the same output with hopefully better tube life.
Some power and distortion numbers (inputs RMS):
1V -> 18W @ 2.75%
710mV -> 11.5W @ 2%
430mV -> 5W @ 1%
Distortion first rises out of the noise (about -50dB - this is all breadboarded with long leads, small resistors, etc.) at 220mV in / 1.8W out.
Other than the power output, it's working exactly like the datasheet says (allowing for individual tube differences) - biased at -22V, idling at 58mA plate / 1.85mA screen, and full-out at 78mA plate / 15mA screen.
GNFB is very small (2.1dB). I've got plenty more gain available on the front end; would it make sense to crank that up and throw more feedback at it?
Some power and distortion numbers (inputs RMS):
1V -> 18W @ 2.75%
710mV -> 11.5W @ 2%
430mV -> 5W @ 1%
Distortion first rises out of the noise (about -50dB - this is all breadboarded with long leads, small resistors, etc.) at 220mV in / 1.8W out.
Other than the power output, it's working exactly like the datasheet says (allowing for individual tube differences) - biased at -22V, idling at 58mA plate / 1.85mA screen, and full-out at 78mA plate / 15mA screen.
GNFB is very small (2.1dB). I've got plenty more gain available on the front end; would it make sense to crank that up and throw more feedback at it?
Last edited:
LOL! You want to put a 10 ohm resistor across 1000v Sure no problem! I always thought there should be an open source "commen sense" plugin for these simulator programs.
Like "do you realize this is 100 amps and 100kw!
Messed around with feedback today. Got it to about 1.5% at 18W by upping the feedback to 6.6dB (reducing the NFB resistor to 33k, and increasing the input stage load to 130k). Diminishing returns set in quickly, and I'm wary of instability, so that's where I'm calling it done.
Next up: build out a second channel so that I can dig into the power supply. Onward!
Next up: build out a second channel so that I can dig into the power supply. Onward!
Real-world frequency response isn't super great - it's down 4-5 dB at 20kHz. I'm not too worked up about that - I haven't heard anything above 14kHz in years. Also threw a square wave into it, with interesting results - I'll put that in the next reply.
Last edited:
So adding all that feedback made the waveforms inside the loop get pretty weird. Here's a 1kHz square wave input:
Without NFB, this is amplified pretty cleanly. With NFB, it goes wonky:
But by the time it makes it out the plates of the power tubes, all is well-ish (ignore the ghost trace - the 60Hz hum from the breadboard makes it bouncy):
There's clearly a little bit of HF and LF rolloff, but nothing terribly drastic. Here's something weird, though. If I feed it a 500Hz square wave, the waveform gets really unbalanced at the plate:
Any ideas on what might be causing that? Differences between the two output tetrodes? Or in the two toroids that make up the OT?
The OT is made up of two toroids, each with two primaries and two secondaries. I've got the primaries wired up so that each side of the CT has one coil from each of the toroids, in order to balance DC current. The secondaries are wired in series-parallel.
Without NFB, this is amplified pretty cleanly. With NFB, it goes wonky:
But by the time it makes it out the plates of the power tubes, all is well-ish (ignore the ghost trace - the 60Hz hum from the breadboard makes it bouncy):
There's clearly a little bit of HF and LF rolloff, but nothing terribly drastic. Here's something weird, though. If I feed it a 500Hz square wave, the waveform gets really unbalanced at the plate:
Any ideas on what might be causing that? Differences between the two output tetrodes? Or in the two toroids that make up the OT?
The OT is made up of two toroids, each with two primaries and two secondaries. I've got the primaries wired up so that each side of the CT has one coil from each of the toroids, in order to balance DC current. The secondaries are wired in series-parallel.
The transformers are Antek AS-0506. It's an experiment, for sure, but the possibility of getting a useable 50W OT for $38 was pretty enticing. I cribbed the wiring topology from kodabmx, in this thread I started a while back. Worst case, I end up using them for their intended purpose as filament transformers to go along with the industrial control transformers I also picked up cheap.
I don't know the absolute inductances. Instead, I figured out the turns ratio by wiring one primary to a (measured) 120V supply and measuring a secondary. I squared that to get the inductance ratio, then substituted in a guess of 20H as the primary to calculate a matching secondary inductance. I drew out all the coils in LTSpice and set up the proper couplings, then I double-checked by building out the physical transformer and testing against the simulated results.
It's not perfect, but it should be pretty close.