How much ringing is OK?

[justinis]

Hello again, folks. I performed a bunch of tests on my new tube amp this afternoon. Frequency response, power, and noise tests went great. Then I did some square wave testing. Below is a screenshot of a 1W output with a 10KHz input. I did 100 Hz, 1 KHz, and 10 KHz for both channels but they all looked pretty similar. This is the worst looking one:



Is this an acceptable amount of ringing? If not, there is a parallel R and C in the negative feedback connection. Should I play around with the values of these components to try to improve things?

Thanks!

[tomp]

David Berning could probably shed the most light on tube amp performance.

Tom

[SoundMods]

The C in the feedback network provides 100% feedback at a certain frequency setting the gain at that frequency at unity. Tell me what the values are and I can tell at what frequency.

You might want to try it out without the cap.

[justinis]

Tell me what the values are and I can tell at what frequency.

Values are 2.7K and 220pF. The impedance of the C is so big at audio frequencies that I doubt it's doing too much. Here is the driver schematic. The components are R9, R10, C3, C4. I have J12 and J13 run to the 16 ohm taps on the OPTs.

[SoundMods]

You are looking at 100% feedback at 270-kHz. This could have impact as far down as 27-kHz.

You might try your amp. without the C. Many tube-amp designs don't use a C in the feedback preferring to have the same amount of feedback through-out the entire available amplifier bandwidth.

The there are there are the arguments for no feedback at all.

Anyhow -- the feedback in the design may be too aggressive to begin with -- then there is OCD quotient -- did the amp sound good before you revealed its hidden secret of some ringing?

[randy warren]

the amount of ringing shown in your square wave response is pretty negligible, you are fully dampening the ringing after 1.5 cycles, and its level is also pretty small. For more information about testing amps in this fashion, I would recommend reading those articles by Dave Gillespie on audiokarma, he gives a full treatise on measurements and what they mean.

[justinis]

I would recommend reading those articles by Dave Gillespie on audiokarma, he gives a full treatise on measurements and what they mean.

This sounds great! Do you have a link?

I tried increasing/decreasing the feedback cap and wasn't able to make the ringing any better. I reduced R13/R14 and got the following result:



I'm gonna call this art and see how it sounds. Thanks everyone!

[randy warren]

look in the Fisher forums as well as the tube audio, he is a Fisher fan and has lots of restoration articles with in-depth information including how to measure and what that means. Be interesting to hear the 2 measured setup, that second one the waveform is much less damped, it goes out for several more cycles, that could be an issue with certain speakers...

[dberning]

The amount of ringing you show in the first waveform is quite acceptable into a resistive load. But problems can show their ugly side when the load is reactive. I always test my amps with a variety of purely capacitive loads to check stability. It appears that your design has quite a lot of feedback to successfully wrap around a transformer and the other stages and have unconditional stability. The network consisting of 330 pF and 5.6k is used to force stability by reducing gain at high frequency and modifying the phase. I prefer not to use this type of network because this type of diddling usually makes for a less than optimum sounding amplifier in my experience. But without it or something like it the amount of feedback that can be applied will be most likely reduced. As the design is, I suspect that it will behave badly with purely capacitive loads, and may oscillate. If it does, the fix would likely be to remove the diddle and lower the amount of feedback until you get unconditional stability with any kind of reactive load you can find. Ringing with a capacitive load is acceptable, but the ringing should damp at all power levels.

David

[dberning]

By the way, justinus, with the exception of the diddle (pair) and the feedback network, I thing your design is very clean and I like it. My first choice is to have no frequency-compensation capacitors, with all stages having as much bandwidth as practical with power consumption constraints. Of course the output transformer will limit your ability to apply resistive-only feedback and you will likely need a small compensation capacitor unless you make the feedback very minimal, and then your speaker damping goes to hell.

David