@Retrofocus
Same concept, implemented in a different way. French include output transformer in the loop, japanese don’t. We discussed about it and how to improve at post #200 and following.
Same concept, implemented in a different way. French include output transformer in the loop, japanese don’t. We discussed about it and how to improve at post #200 and following.
We should keep gnfb strong, the local feed back is just an option to lower the global open loop gain at high frequency to obtain stability. Not the other way around. Unless you want to build a tone(distortion) generator.
Here is a good example. Dynaco ST-70. The highlighted portion is the local negative feedback that doesn't include the output transformer. The component count is so small that the local loop is almost unconditional stable. The local nfb is through the 390pf cap. It engages only at high frequency to lower the gain before the global negative feedback.the local feed back is just an option to lower the global open loop gain at high frequency to obtain stability
I have worked through most of the accuracy issues with my uTracer6.You're not alone, it seems the FUtracer was a frustrated reponse to that issue.
https://futracer.com/about/
It turns out a main source of the uTracer6 accuracy issues were related to available GUI software from Ronald and to a lesser degree uTmax.
GUI software from Ronald has a bug in the Va=Vs code that can result in gross errors with the uTracer6 below 100 volts at high currents.
GUI software from Ronald and uTmax both fail to provide correct calibration values for the MOSFET switch losses in a uTracer6 resulting is accelerating error levels below 100V at high current. These errors can exceed 50% under some conditions.
After working with Ihor and the uTracerJS GUI both the above issues have been overcome resulting is very good accuracy in uTracer6 measurements under most conditions when using utracerJS.
Anyone using a uTracer6 should investigate the latest versions of uTracerJS. It can take a utracer6 to the next level up in accuracy of you measure high current low voltage devices like sweep tubes, high current triodes or rectifers.
The one renaming accuracy issue with the uTracer6 is the reduced resolution on the Va and Vs AD subsystem compared to a utracer3.
This is mainly a issue below 100 volts. The workaround is the take averaged measurements below 100V.
The uTracer3+ will also benefit from uTracerJS, just not to the degree the uTracer6 does.
In my experience the main improvement in the FU design comes from the improved provided software in both the PIC and the uTmax GUI.
This improved software's importance in providing performance improvement should not be underestimated.
Ronald's PIC and GUI software are weak points in the uTracer family in my experience. Ronald's hardware is generally quite good.
The uTracerJS GUI has allowed one of these weak points to be overcome.
Perhaps some day uTmax will also be updated to incorporate the uTracer6 accuracy improvements that are now available in uTracerJS.
I would love to see a open souce PIC code option for the uTracer family.
I have found frustration in getting low distortion at high frequencies, clean 10khz square waves and good stability in designs where all or almost all feedback is through the output transformer. Maybe I am just too picky, but is low distortion and good square waves asking too much?@jxdking If you read through the thread you can find alot of examples where local feedback is not only what you wrote here, but can be used alone without being a tone generator.
This is not news to the experienced.
That output transformer is SO important when global feedback is added and all I have tried have leakage inductance and/or stray capacitance levels that really hurt at 20Khz if a lot of global feedback is added for pentode output stages. It does not help the transformer vendors generally provide inadequate information on leakage inductance and stray capacitance levels in their products.
Then there is the little discussed issue of variation in leakage inductance between each half of a push pull transformer primary and the secondary.
To make it ever more complex there is the leakage induction between each 1/2 section of the primary.
All this causes some very complex feedback phase issues as soon as you leave class A operation and head into AB1 operation when one tube stops conducting for part of the output cycle.
As late my thinking is heading toward linearaszing the amplifier outside the transformer as much as possible by careful design and a strong local feedback loop (not cathode degeneration) and then providing only a light touch of global feedback through the transformer.
The jury is still out with me however if this can deliver low distortion, snappy square waves and good stability all at the same time.
Last edited:
I have found frustration in getting low distortion at high frequencies, clean 10khz square waves and good stability in designs where all or almost all feedback is through the output transformer. Maybe I am just too picky, but is low distortion and good square waves asking too much?
This is not news to the experienced.
My fomula is to get >20dB NFB, and let it roll off (1st order) so that the open loop gain drops below 1 at 100KHz.
Low distortion with a good square wave is hard, because output transformers over 100KHz bandwidth are rare and expensive. There will be overshoot or ringing. To get a clean square wave, you have to either drop the NFB ratio or let it roll off from a lower frequency. Both of them will harm your distortion figure. If let me pick, I will pick low distortion over the sqare wave.
Thanks for the tip. I will see how my project measures compared to your suggestion. Seems a good baseline to start with.My formula is to get >20dB NFB, and let it roll off (1st order) so that the open loop gain drops below 1 at 100KHz.
My concern of acceptance of square wave testing that shows over shoot or ringing is does this not suggest marginal stability at high frequency's?
Marginal stability can then lead in my experience to strange behavior on some program material transients that may not show up in steady state tone testing.
Marginal stability can also lead to extra sensitivity to the load impedance creating problems driving "difficult" speakers.
I my experience many (most) contemporary commercial speaker designs assume it is the power amplifier's job to deal with speaker impedance variation or swings in the input current phase angle. That can be troublesome for many tube amplifier designs.
Getting to a tube amplifier that is speaker agnostic, with low distortion, good stability and transient response is the holey grail not yet even approached by myself.
I does seem logical that if the output transformer is the limiting factor in the feedback loop then putting as much of that feedback as possible outside the transformer could be helpful.
Last edited:
When I verify the stability with tube amps, I test with square wave without any speaker connected. The speaker at the output would reduce your open loop gain, and it becomes more stable.Marginal stability can also lead to extra sensitivity to the load impedance creating problems driving "difficult" speakers.
For most speakers (let's rule out electrostatic speakers), they become inductive at high frequency. The resonance(/troublesome) frequencies won't go beyond the audio frequency. As long as it stays inductive at high frequency, it won't affect your high frequency stability.
- Home
- Amplifiers
- Tubes / Valves
- Strong local feedback + light global or the opposite?