This can be a tough challenge to one´s skills; the 1 Watt waveform performance is to my eyes is not a real life test.......esp stability when the power gets going. On my high power amps (200W+) one has to be careful that the 10Khz waveform doesn´t show appreciable ringing which can reveal reactive instability on longish LS leads. The 1950´s article by John Moyle https://ia801508.us.archive.org/21/items/moyle_amp/moyle_amp_text.pdf should be in everyone´s build/diagnostic kit and this takes the skill of the subject a dimension further. Since there are so many amp designs configured around the Mullard/Philips 20-30W LTP version relevant to the thread start, this should be nice midnight reading.
To round up....In that 1950´s era it was common that tube amp stability tests should also include tube amp survival on two nasty conditions which are seldom in home setup´s;
1, Accidental disconnection of loudspeaker at full load; i.e open circuit.
2, A short circuit on the output at full load.
I am not suggesting to do any test of the sort that could destroy our beloved amps, let expensive tubes and output transformers.
Le bench baron
Well done! Now, I'm sure you know, nfb does its magic by trading in some of the open loop (withouth feedback) gain. Like, when no-nfb gain is 100, you close the loop for say gain of 20. The loss of gain is traded for better performance.
So I'd like to know what your gain is without fb?
Jan
Great articles, the best I've seen aimed at the amateur so far. Thanks for posting them. Certainly, the OP will want to test for stability, which would have been my next suggestion once he's had a look at how the transformer behaves with GNF, and been able to reduce any excess ringing. It should be able to hold up unloaded, with some capacitance across the load, and into a small purely capacitive load.
Jan,
From the measurement I believed my open voltage loop gain without GNF is between 270-280. This is way more than LTspice simulation which indicated about 215.
From the measurement I believed my open voltage loop gain without GNF is between 270-280. This is way more than LTspice simulation which indicated about 215.
Bear in mind that´s only a 20% change, quite in order for tube gain varations i.e gm/v are never guaranteed esp. when emissions start to degrade specs. One might find a tube change can re-vitalize those previous measurements closer to LT spice, but In tube life I err-on-the-side-of-caution and never take any calculation to perfection as the LT simulation suggests...(it is only a tool). The old designs that sounded great often used then the state of the art 20-50% tolerance carbon composition resistors !
Le bench baron
Le bench baron
Tanker135, hopefully squarewave testing will provide some simple waveform results from which to assess stability margins. That implies you have access to a squarewave generator with sufficiently fast rise-fall times, and a scope with similarly large bandwidth (ie. at least perhaps 5MHz) to observe the nature of any ringing induced by the square wave edge transitions ?
Initially with that testing you start with a stable amplifier and use a nominal resistive load on the speaker output. and a lowish signal input (and output) as the intent is to firstly assess the small-signal feedback performance of the GNFB loop. Testing may then show up whether the feedback can benefit from some form of additional compensation, or not, in order to provide a better margin for stability. This is as far as some go.
After that you could test for stability at no load and with capacitive only loading, as a way to bullet-proof the amp from anything its output may practically come across - that can become tricky to navigate through, as some forms of compensation tweaked for best resistor-only loading stability could make the amp more unstable for capacitor only loading.
After that you could test across the full power bandwidth, as a way to ensure there are no quirky local feedback gremlins causing sporadic hf oscillation, or as yet unexplored low frequency oscillations or peaking.
I would suggest not trying to model high frequency stability performance using simulation, as I doubt that the transformer model can sufficiently equate to what a real output transformer does out past 50-100kHz (or to that matter below 10Hz), and every output transformer is a different subtle mix of distributed inductances and capacitances so there is likely no simple generic model or other simulations to get you to a good operating point.
Initially with that testing you start with a stable amplifier and use a nominal resistive load on the speaker output. and a lowish signal input (and output) as the intent is to firstly assess the small-signal feedback performance of the GNFB loop. Testing may then show up whether the feedback can benefit from some form of additional compensation, or not, in order to provide a better margin for stability. This is as far as some go.
After that you could test for stability at no load and with capacitive only loading, as a way to bullet-proof the amp from anything its output may practically come across - that can become tricky to navigate through, as some forms of compensation tweaked for best resistor-only loading stability could make the amp more unstable for capacitor only loading.
After that you could test across the full power bandwidth, as a way to ensure there are no quirky local feedback gremlins causing sporadic hf oscillation, or as yet unexplored low frequency oscillations or peaking.
I would suggest not trying to model high frequency stability performance using simulation, as I doubt that the transformer model can sufficiently equate to what a real output transformer does out past 50-100kHz (or to that matter below 10Hz), and every output transformer is a different subtle mix of distributed inductances and capacitances so there is likely no simple generic model or other simulations to get you to a good operating point.
Nice comment from Tim. You have what looks like a sophisticated design and some nice output transformers. You should be able to dial in a fine, stable amp.