Graham Maynard said:
Graham:
We are aware this is a issue that particularly worries you, yet it is not the crux of the matter.
Remember these plots depict the error signal, i.e. the amount of correction NFB thinks must be inserted to strighten things. The real issue is how the actual output voltage follows input and controls cone movement, for this is what actually gets listened.
This nonwithstanding, a much better high frequency response is readily achievable with modern devices, meaning negligible error signal phase shift can be attained up to and above 20 KHz.
Rodolfo
This is the kind of output test I would do, though I have never actually used a current source before to do it.
The error is greatest at maximum amplitude and is close to zero again at minimum - though not zero due to the fractional phase shift. Similar to the more efficient Symasym.
Now this error of 3mV at 2A = 16V @ 8R is greater than the steady sine measured THD for the circuit, BUT quite close to the figure for the amplifier distortion figure for first cycle distortion.
This back-EMF induced error is greater than THD with most high NFB amplifiers, but will not be demonstrated without either amplitude nulling the output when driving a dynamic loudspeaker load, or reverse driving an output stage.
Of course non feedback designs do not actively produce any damping and the voltage cannot keep from drifting under the influence of back-EMF beyond the passive bias loading of the loudspeaker, but at least the amplifier does not itself then generate a phase shifted error potential, which is likely why some of Nelson's designs have earned a good reputation.
PS Could someone at diyAudio PLEASE change the upper attachment limit to 1024. What a nuisance it is having to convert image sizes - the detail gets lost !.
The error is greatest at maximum amplitude and is close to zero again at minimum - though not zero due to the fractional phase shift. Similar to the more efficient Symasym.
Now this error of 3mV at 2A = 16V @ 8R is greater than the steady sine measured THD for the circuit, BUT quite close to the figure for the amplifier distortion figure for first cycle distortion.
This back-EMF induced error is greater than THD with most high NFB amplifiers, but will not be demonstrated without either amplitude nulling the output when driving a dynamic loudspeaker load, or reverse driving an output stage.
Of course non feedback designs do not actively produce any damping and the voltage cannot keep from drifting under the influence of back-EMF beyond the passive bias loading of the loudspeaker, but at least the amplifier does not itself then generate a phase shifted error potential, which is likely why some of Nelson's designs have earned a good reputation.
PS Could someone at diyAudio PLEASE change the upper attachment limit to 1024. What a nuisance it is having to convert image sizes - the detail gets lost !.
Attachments
Hi Rodolfo,
I am freely passing on my viewpoint of entries presented here, and as back-up showing some of my own results, which I do not claim to be perfect, but which satisfy me.
Forward and reverse control phase shifts being different means that topology has the most significant effect, even with high speed devices.
If the error signal was being illustrated I was not aware of this, and it was NOT stated. I said I could understand how those those plots had been made and asked for illustration.
PMA are those plots output terminal potential error, or an internal voltage error ?
As far as controlling the loudspeaker - I suggested that forward amplifier testing be completed using a realistic load with fundamental nulling, for that includes the imaginary back-EMF current.
When an amplifier is virtually immune to back-EMF its forward measured response can be better relied upon, but I must agree and have written elsewhere that I do not assume that a perfect voltage amplifier is the best way to drive a loudspeaker.
Cheers ........... Graham.
I am freely passing on my viewpoint of entries presented here, and as back-up showing some of my own results, which I do not claim to be perfect, but which satisfy me.
Forward and reverse control phase shifts being different means that topology has the most significant effect, even with high speed devices.
If the error signal was being illustrated I was not aware of this, and it was NOT stated. I said I could understand how those those plots had been made and asked for illustration.
PMA are those plots output terminal potential error, or an internal voltage error ?
As far as controlling the loudspeaker - I suggested that forward amplifier testing be completed using a realistic load with fundamental nulling, for that includes the imaginary back-EMF current.
When an amplifier is virtually immune to back-EMF its forward measured response can be better relied upon, but I must agree and have written elsewhere that I do not assume that a perfect voltage amplifier is the best way to drive a loudspeaker.
Cheers ........... Graham.
Re: loop gain
Yes, and also this paper. It is a subcircuit you can drop onto your schematic. The STEP command is used to turn the probe's voltage and current sources on and off, and a plot command which uses some odd-looking syntax to refer to voltages and currents in each state of the STEP is used to compute loop gain.
This circuit came from the Yahoo LTSpice users' group.
estuart said:
Yes, and also this paper. It is a subcircuit you can drop onto your schematic. The STEP command is used to turn the probe's voltage and current sources on and off, and a plot command which uses some odd-looking syntax to refer to voltages and currents in each state of the STEP is used to compute loop gain.
This circuit came from the Yahoo LTSpice users' group.
Graham,
I have already described it, you have not been patient enough
The amplifier is driven to its output with input grounded. The plots show output pin voltage vs. current forced into output. X axis is current to output, Y axis is voltage at the output pin created by forced current, as a voltage drop on output impedance. This output voltage should be ideally zero (for infinite damping factor, not dependent on frequency and amplitude). I create it in simulation with X axis current, Y axis voltage.
I am going to measure it by conventional scope as X-Y plot.
I have already described it, you have not been patient enough
The amplifier is driven to its output with input grounded. The plots show output pin voltage vs. current forced into output. X axis is current to output, Y axis is voltage at the output pin created by forced current, as a voltage drop on output impedance. This output voltage should be ideally zero (for infinite damping factor, not dependent on frequency and amplitude). I create it in simulation with X axis current, Y axis voltage.
I am going to measure it by conventional scope as X-Y plot.
janneman said:
Hi Jan and all other LTspice users,
OK LT guys, you win as I came down a peg or two (saves of lot of energy). I've reinstalled the S/W package from LTC.
Still, the data exchange isn't foolproof, because the definition models are not included in the asc files (as far as I know). Any suggestions how to tackle this source of annoyance.
And Mike,
Now that I can read your drawings (Stuart's_LG2.asc, upper diagram), I saw a loop gain probe put at a place, which I had never expected. If it was you intention to characterize the global feedback loop in this way, I'm sorry to say so, you'll get erroneous results. But maybe you had something else in mind. Anyhow, this too explains that we have been at cross purposes.
Cheers,
.
loop gain, Middlebrook
Hmm, sounds pretty good and effective, the way as it should be done.
I'll have a try to implement this I/V switching mechanism into Micro-Cap, as up till now, you have to enter the circuit diagram two times (a little bit clumsy), one for voltage and one for current.
Thanks a lot Andy and Mike
Cheers,
andy_c said:
Hmm, sounds pretty good and effective, the way as it should be done.
I'll have a try to implement this I/V switching mechanism into Micro-Cap, as up till now, you have to enter the circuit diagram two times (a little bit clumsy), one for voltage and one for current.
Thanks a lot Andy and Mike
Cheers,
estuart said:
Sure. The best way to have your files usable by others is to not modify standard.bjt, standard.mos etc. to add models. Instead, put each model in a separate text file in the same folder as your project. Then, in your project, do an "Edit, SPICE directive" (shortcut "S") and put in:
.INCLUDE mymodel.mod
for each model you want to use that's not in the standard files. This is messy I know, because you end up with lots of duplicate model files all over the place, but it's the only way I know to keep your projects portable. Then, when you post the project, use a zip file that contains the .asc project files and all the needed models. Also, if you need help in the LTSpice users' group, that's the way they do it there also.
I don't claim this to be a great thing
, but since it's free. well...
Hi Graham and Rodolfo,
by the the time I have made measurements and they are quite different from simulation.
0.25A/d, 10mV/d, 1kHz, EC AB1 amp
0.25A/d, 10mV/d, 10kHz, EC AB1 amp
0.625A/d, 50mV/d, 1kHz, symasym
0.625A/d, 50mV/d, 10kHz, symasym
The symasym (class AB) has 500mA quiescent current
by the the time I have made measurements and they are quite different from simulation.
An externally hosted image should be here but it was not working when we last tested it.
0.25A/d, 10mV/d, 1kHz, EC AB1 amp
An externally hosted image should be here but it was not working when we last tested it.
0.25A/d, 10mV/d, 10kHz, EC AB1 amp
An externally hosted image should be here but it was not working when we last tested it.
0.625A/d, 50mV/d, 1kHz, symasym
An externally hosted image should be here but it was not working when we last tested it.
0.625A/d, 50mV/d, 10kHz, symasym
The symasym (class AB) has 500mA quiescent current
mikeks said:
Hi Mike,
One of the two:
1. global (major) feedback loop and erroneous results. In this case the top your R10 (my R1) should be be connected directly to the output, instead of tied behind (left) to the loop probe.
or:
2. correct results, but targeted at a different loop, but then, what kind of loop, inner, Miller?
Please, explain (and check if we are looking at same diagram).
btw I couldn't see any load resistor (8Ohm), why?
Cheers, I'll be back in 1 hour.
