F4 Beast Builders

I forget to specify one additional measurement:

.measure ac MagVout find mag(V(out)) at=1khz

I haven't figured out how to restrict the frequency range of the measurements, so you must restrict the range to be above 1kHz in the AC Analysis Menu.

I do not understand why dB units are shown in:

phase_margin: atan(im(v(out))/re(v(out)))=(33.7472dB,0°) at 1.75667e+06
gain_margin: -20*log10(mag(v(out)))=(26.0499dB,180°) at 33.9781
 
Member
Joined 2009
Paid Member
Thank you Lynn!

that is the new answer

magvout: mag(v(out))=(20.0676dB,0°) at 1000
f3db: mag(v(out))=magvout/sqrt(2) AT 210135
phase_margin: atan(im(v(out))/re(v(out)))=(33.7472dB,0°) at 1.75667e+06
gain_margin: -20*log10(mag(v(out)))=(26.0499dB,180°) at 33.9781

stable?
 
Generg, does the bandwidth go up if you lower the toshiba gate resistor values on that circuit? I just tried a simulation of a slightly modified version of Nelsons headphone opto bias circuit with this amp. I broke a rule by using resistors on the output transistors but it sets both the bias current and offset. When the toshiba gate resistors are at 1000 ohms the bandwidth is 330khz and with 100 ohms it is at 1Mhz.
 

Attachments

  • Distortion.png
    Distortion.png
    18.3 KB · Views: 333
Thank you Lynn!

that is the new answer

magvout: mag(v(out))=(20.0676dB,0°) at 1000
f3db: mag(v(out))=magvout/sqrt(2) AT 210135
phase_margin: atan(im(v(out))/re(v(out)))=(33.7472dB,0°) at 1.75667e+06
gain_margin: -20*log10(mag(v(out)))=(26.0499dB,180°) at 33.9781

stable?

The gain margin frequency of 33.9781 tells me you are doing the AC Analysis starting at a frequency below 1kHz.
 
The phase margin of 33.7deg is ok. But are these measurements with the AC stimulus at the input or in the feedback path? Instability and oscillation occur when the phase shift in the feedback path turns into positive feedback and the input signal becomes amplified by the feedback.
 
Member
Joined 2009
Paid Member
Generg, does the bandwidth go up if you lower the toshiba gate resistor values on that circuit? I just tried a simulation of a slightly modified version of Nelsons headphone opto bias circuit with this amp. I broke a rule by using resistors on the output transistors but it sets both the bias current and offset. When the toshiba gate resistors are at 1000 ohms the bandwidth is 330khz and with 100 ohms it is at 1Mhz.

Hi Rob,

I have the same results...:))
 
Member
Joined 2009
Paid Member
The phase margin of 33.7deg is ok. But are these measurements with the AC stimulus at the input or in the feedback path? Instability and oscillation occur when the phase shift in the feedback path turns into positive feedback and the input signal becomes amplified by the feedback.

Did I do it right? for the first data I set the AC of the input source AC=1

magvout: mag(v(out))=(20.0676dB,0°) at 1000
f3db: mag(v(out))=magvout/sqrt(2) AT 210135
phase_margin: atan(im(v(out))/re(v(out)))=(33.7472dB,0°) at 1.75667e+06
gain_margin: -20*log10(mag(v(out)))=(26.3923dB,0°) at 1.04701e+07

and of this data I deactivated the input source and activated the feedback source with AC=1

magvout: mag(v(out))=(20.0771dB,0°) at 1000
f3db: mag(v(out))=magvout/sqrt(2) AT 209245
phase_margin: atan(im(v(out))/re(v(out)))=(35.4522dB,0°) at 1.81029e+06
gain_margin: -20*log10(mag(v(out)))=(27.3892dB,0°) at 1.42114e+07
 
Both sets of measurements look plausible, but I am surprised that the -3dB frequency is so high with the input and actually lower for the feedback loop. Is this with a CFA input stage or something else? In my simulations with a jfet gate stopper of 33K is get -3dB at around 105kHz from the input, and about 800kHz for the feedback loop.
 
If you put the circuit above on the second stage of a three stage amp to drive another source follower you could fun the feedback for the offset from the final output stage back to the op amp. Then it should keep the output offset at zero and the second stage bias constant. You would need to have the voltage gain in the first stage though.
 
Member
Joined 2009
Paid Member
Indeed Lynn the first data had considerable CFB, now with out only global FB

input AC

magvout: mag(v(out))=(21.1412dB,0°) at 1000
f3db: mag(v(out))=magvout/sqrt(2) AT 188336
phase_margin: atan(im(v(out))/re(v(out)))=(33.6045dB,0°) at 1.76034e+06
gain_margin: -20*log10(mag(v(out)))=(26.3862dB,0°) at 1.04495e+07

feedback loop AC

magvout: mag(v(out))=(21.1507dB,0°) at 1000
f3db: mag(v(out))=magvout/sqrt(2) AT 187676
phase_margin: atan(im(v(out))/re(v(out)))=(35.353dB,0°) at 1.81407e+06
gain_margin: -20*log10(mag(v(out)))=(27.3796dB,0°) at 1.42083e+07

both -3dB are very similar