The output current also decreases with the lower load impedance, which should not happen with a positive output impedance. That relation only happens with a negative output impedance whose absolute value is greater than the load impedance. There's actually a polarity flip too [the implied output voltage, prior to the output impedance, is -2.4V]. If the output impedance were positive 23.25, then the results should be more like 1.03V at 9.4R for 0.6V at 4.7R.
Hi Elvee, not shure if i well understood your method for out.imp.measurement.
Please can you post a rough sketch?
Thanks.
The subject has already been addressed, here for example:
https://www.diyaudio.com/community/...ut-impedance-with-ltspice.199462/post-5864138
This setup presupposes that test amplifier is good, a lab amplifier for example. If you use the second channel of a tube amplifier as auxiliary amplifier, V1 should be measured across R1.
If you chose convenient load and excitation voltage values, for example 8.2ohm and 8.2V, V2 will indicate directly the value in ohm, without further calculations
https://www.diyaudio.com/community/...ut-impedance-with-ltspice.199462/post-5864138
This setup presupposes that test amplifier is good, a lab amplifier for example. If you use the second channel of a tube amplifier as auxiliary amplifier, V1 should be measured across R1.
If you chose convenient load and excitation voltage values, for example 8.2ohm and 8.2V, V2 will indicate directly the value in ohm, without further calculations
What about input signal? For my KT88 SE amp, someone recommended using a 1 kHz sine wave signal and set voltage to around 2V or 3V (on the 8 ohm tap) if possible. Pink noise fluctuates too much.
For my particular example, I measured 1.82V with 6.04 ohms, and 2.37V with 46.99 ohms. Calculated output impedance for my amp is 2.18 ohms on the 8 ohm tap.
For my particular example, I measured 1.82V with 6.04 ohms, and 2.37V with 46.99 ohms. Calculated output impedance for my amp is 2.18 ohms on the 8 ohm tap.
Last edited:
I agree, re-taking measurements as first step. If results are similar, I wonder if it relates to what @Depanatoru states above, with the gain highly load dependent, so the method @Elvee provided may give different results at each load. I guess I am curious what that means for the amp when it drives a real speaker with varying impedance.
3Vrms is a little over 1W@8ohm, probably a little weak for a regular amplifier having a 5~30W output capacity. Half the rated power seems more reasonable.
You can do the measurement for various power levels, it will give you an idea of the linearity of the output impedance; looking at the residue will also give you clues about that
When gain drops with load that means there's a high Zout. Is how you look at it.
Picture the amp with a constant gain but high Zout and that's what you see.
That Zout of course isn't a physical resistor, it's the result of a drop in gain with load.
Mostly; in some cases there are additional effects but mostly it is this.
Jan
With solid state, I'm familiar with that, but I don't have any tube experience to know how they behave with different loads (i.e. if any of those secondary effects come into play). From what you're saying, I gather it's not common.
FWIW, I compared measurement results from both methods. Because of different results while measuring an amplifier with very high DF, I made a control measurement on amplifier with moderate output impedance. Results from both methods match very well.
However, with amplifier that has large NFB and extremely low output impedance, results differ at higher frequencies. All measurement were made with better than 1% precision for AC voltages and currents with equipment good up to 100 kHz.
Amplifier in question has some 110 dB NFB at 20 Hz and about 60 dB at 20 kHz. This is enough to push active circuit impedance to single digit mΩ range at any audio frequency and make output impedance to depend mainly on PCB tracks and wires impedance, as two resistors method confirms.
Either, there is some error with my measurement setup or impedance of amplifier with NFB is not the same for sourcing and sinking current or for acting as voltage and current source vs. acting as load or current sink.
It would be helpful if someone else can perform and compare measurements.
However, with amplifier that has large NFB and extremely low output impedance, results differ at higher frequencies. All measurement were made with better than 1% precision for AC voltages and currents with equipment good up to 100 kHz.
Amplifier in question has some 110 dB NFB at 20 Hz and about 60 dB at 20 kHz. This is enough to push active circuit impedance to single digit mΩ range at any audio frequency and make output impedance to depend mainly on PCB tracks and wires impedance, as two resistors method confirms.
Either, there is some error with my measurement setup or impedance of amplifier with NFB is not the same for sourcing and sinking current or for acting as voltage and current source vs. acting as load or current sink.
It would be helpful if someone else can perform and compare measurements.
I tried this method yesterday using LIMP, I thought it would be a quick and easy way to get output impedance vs frequency, but had some problems with the measurements. Interesting test though. Posted about it here LINK, suggestions welcome.
Measured difference in my post above is likely from wire inductance. Injection method requires more wires and involves more interconnections. @Rallyfinnen had the same behavior and this has put me on the right track.
I’ve made impedance measurement of amplifier output shorted with 10 cm long wire. Measured impedance ratio from 1 kHz to 20 kHz was 4x. Ideally it should be 1. So, we can’t even make a proper short circuit for the whole audio band when several mΩ is measurement target.
That 10 cm wire made 2.7 mΩ short circuit at 1 kHz and 10 mΩ at 20 kHz.
I’ve made impedance measurement of amplifier output shorted with 10 cm long wire. Measured impedance ratio from 1 kHz to 20 kHz was 4x. Ideally it should be 1. So, we can’t even make a proper short circuit for the whole audio band when several mΩ is measurement target.
That 10 cm wire made 2.7 mΩ short circuit at 1 kHz and 10 mΩ at 20 kHz.
Actually, I think an amp output would normally look at least slightly inductive since there is usually an output inductor bypassed with a resistor (solid state amp) and loop gain in the feedback drops with frequency (even in the audio band with many amps), and that would cause the output impedance to rise in the treble, since there is less feedback.
A ratio of 1 to 4 at 20kHz shouldn't be possible: for vector errors it should be √2 maximum if I am not mistaken (help Marcel!), and non-linearity should only arise if the excitation level is very different
tombo56's discrepancy is easy to explain.
With the current injection method, the AUT's output stage stays in the dead zone where the impedance is high until corrected by NFB. The last three points show the impedance doubling with frequency as expected.
The two resistor method runs the output stage in the conduction zone where the impedance stays low.
This is one argument for designing a NFB amplifier with a dominant pole > 20KHz.
Ed
With the current injection method, the AUT's output stage stays in the dead zone where the impedance is high until corrected by NFB. The last three points show the impedance doubling with frequency as expected.
The two resistor method runs the output stage in the conduction zone where the impedance stays low.
This is one argument for designing a NFB amplifier with a dominant pole > 20KHz.
Ed
That 10 cm wire, assuming 50 nH inductance has +6 mΩ from 1 to 20 kHz. That is aligned with short circuit measurement result change over frequency.
Inside amplifier is another 30 cm of thick wire for output, consisting of several hundred fine strands. That would explain impedance rise with frequency.
Important question is why method with 2 resistors doesn’t have same results as injection method. First resistor was 10 Ω high power wire wound (checked to draw same current at 1&10 kHz) and second resistor was noninductive 2 Ω film resistor. At 6 Vrms output, voltage drop with second resistor added was practically flat 22 mV at any frequency, measured with true RMS voltmeter good up to 100 khz. Using 4 Ω in parallel, resulted with some 10 - 12 mV drop over frequency.
I see now injection method as correct, but 2 resistor method poses a question why there in no change with frequency.
Inside amplifier is another 30 cm of thick wire for output, consisting of several hundred fine strands. That would explain impedance rise with frequency.
Important question is why method with 2 resistors doesn’t have same results as injection method. First resistor was 10 Ω high power wire wound (checked to draw same current at 1&10 kHz) and second resistor was noninductive 2 Ω film resistor. At 6 Vrms output, voltage drop with second resistor added was practically flat 22 mV at any frequency, measured with true RMS voltmeter good up to 100 khz. Using 4 Ω in parallel, resulted with some 10 - 12 mV drop over frequency.
I see now injection method as correct, but 2 resistor method poses a question why there in no change with frequency.
Amplifier in question is class A, so always conducting and frequency response is – 3 dB at 1.3 Mhz.
Bias current is 1A and test was made with injecting exactly 1 Arms, so never pushing output stage out of class A.
Still scratching my head.
THAT is a zobel network - which is a high frequency load. In a solid state amp it’s for local stability of the output stage - but is just a high frequency load. The problem voltage spikes generated by the OPTs inductance have mostly high frequency components to them. Damp out ONLY the above-audio highs and you reduce the amplitude of the spikes to safe levels.