Originally Posted by alan0354
I want to show another case of how critical the connector is. This test is for power amp stability test. I connect different values of capacitors directly to the output of the amp to test for stability. As shown in the picture, I put capacitors onto a banana plugs, 3,300pF, 6,600pF, 9,900pF, 22,000pF and 0.1uF. I test the stability of the amp by plugging directly onto the output connector of the amp for testing whether the amp can be stable under the capacitor load.
This test doesn't necessary to have a large output signal, so there is not necessary to have high current going through the connector to create hot spot. You would think this is ok for the application.......WRONG. I have to constantly ply the spring of the connectors out to get a tighter fit, I have to jiggle the connector to test. I can see if I don't press the banana connector tight, the amp can falsely look stable, but when I push it tight, it started oscillates. That little extra resistance can make a day and night difference at the output of the amp.
On the side note, there goes to tell you those so called experts and so called EE doesn't guaranty they know what they are talking. You look at those service manual from amp companies like Marantz and others, they do test for stability. The problem is when you read their test procedures, they use a speaker cable to connect the capacitor to the amp for testing. They test to like 1uF or more and claim it's stable. That is completely WRONG. That's why amplifier oscillation or instability is so common. Not only they put the capacitor at the other end of the cable, they put a 8ohm resistor in parallel with the capacitor to simulate the speaker load. This doesn't work for stability test.
That's the reason a lot of amps are NOT stable with the knitted cables when instead of using a single 12 gauge or 10gauge for speaker cable, they use many small cables like 18 or even 20 gauge cables and knit them together to decrease the inductance. Problem with this closed coupled type of speaker cables is the capacitance is very high, it can be over 6000pF. A lot of so called big name brand power amps will oscillate using this kind of cable. You need to test amps with capacitors directly connect to the output of the amp.
Also, it's important NOT to have an 8ohm resistor parallel with the capacitor as in real life, at the frequency of instability, speakers are very high impedance ( it's an inductor!!) and is out of the picture.
As an example, the Nakamichi PA-7 designed by famous Nelson Pass, it's the improved version of his famous Threshold S-300 power amp.........Sang like a canary with only 1500pF capacitor directly across the output of the amp......only 1500pF and oscillate like crazy. I tested all my amps to 33,000pF. The PA-7 is NOT stable with my custom knitted cable. I had to turn it on and off to get it stable one time to use it. This is because the capacitance of the speaker cable is connected directly to the output of the amp, not through a 12 gauge cable. The little contact resistance created by the hot spot of the banana connector in my test will give a false result. I have to push it hard. I get different values of capacitance by plugging multiple capacitors combination to cover the full range of capacitance. Amps can be stable with 50,000pF and 100,000pF, but that does not guaranty it will be stable at 75,000pF. This is electronics for you, this can be explain in Bode Plot or Nyquist stability theory in electronics.
But of cause, I don't have an EE degree, what do I know.
Haven't really been following this but interesting.
There are specialized models for high-power circuit simulation that model things like quasi-saturation and self-heating effects. Standard G-P is not sufficient; VBIC-xx (don't remember which version) added them. I'm not really up on the latest CMOS models since for the past few years I've been working at a higher level. A few flavors of BSIM (an empirical model) and another one I've forgotten did a reasonable job. I don't really "do" high-power but the problem is the same with deep submicron devices since the active area is so tiny. My problem was usually that parasitics start dominating and they are not always well-modeled for small devices and with so many metal layers and such.
Be careful doing that test with a tube amplifier; the output transformer can be damaged.
Root locus plots are always fun.
The PowerCube is a commercial version: https://www.audiograph.se/
Really good capacitors like you might use for PA (power amp) testing tend to have low ESR and high self-resonance so a little bit of excess real (contact or otherwise) resistance is sometimes enough to quiet the amp. I tended to include a high-value (perhaps 100-ohm) resistor parallel when testing stability with a capacitive load way back when I was doing more audio testing. Plenty high enough to show stability problems with a step input but less likely to send the amp into oscillation. Not that some won't still sing; "less likely" is as good as it gets. And you still have to worry about ESLs and some other speakers with very low HF impedance (the ESL's diaphragm is one big capacitor, but the impedance is not necessarily capacitive due to the transformer that couples amp to panel).
Speakers' input and amplifiers' output impedance change with power due to various effects, usually negligibly until power compression or clipping is approached, but thermal issues in both (among other things) can happen well before they near their limits.
I've never liked banana plugs; they are convenient but little else. I tend to just use bare wire or install spade lugs if I need to swap more often. Using them for testing high-power amps is gutsy, not surprised you are having such problems. Back in the day I had a special rig that was essentially heavy solid copper bus bars in a jig that would connect to a standard pair of banana plugs and then to my dummy load (non-inductive resistor in an oil can). I had a range of RLC networks for testing and some great big LC decade boxes used for transmitter testing that served double duty for adding reactance to the audio amplifier tests.
I keep thinking I'll get back into the hobby aspect; maybe when I retire. I got rid of my 'scope ages ago, just borrowed from work, but anymore I'm nervous about bringing home a DSO worth more than my house. And our spectrum analyzer doesn't go low enough for audio (but handles the Ka radar band just fine).