I am running an AC analysis of a two-stage amplifier in unity gain feedback. The input has 0.9V DC and 1mV AC. I am trying to plot the voltages and the gain in dB with db() function. But it seems that the db() function always give positive value... As shown on the plot above. The output voltage vo (6.05e-05) is below vi (0.001), but the gain, which is db(v(vo)/v(vi)) is +24.37dB. Am I missing something really simple?

@Richie Fan in AC plots xschem plots all variables as complex numbers with magnitude (

gain

) and phase (

ph(gain)

)(in degrees). If you plot

ph(gain)

you will see it has a phase of 180deg, so

gain

it is actually a negative real number. To see the correct gain just plot this:

"vout vin / db20()"

(including quotes) You can display also an alias name for the wave:

"gain; vout vin / db20()"

this expression means: create a new "gain" wave, by taking vout and vin, dividing the first by the second and applying 20*log10() (db20()).

In the ac example in xschem demos, the phase is plotted with ph(out), is that an xschem or ngspice expression?

@Matt Venn frequency plots are saved by ngspice in the raw file as complex data with real and imaginary part. So node

out

in a ngspice generated AC raw file is a complex number:

out = re + i * im

. Since we are mostly interested in magnitude and phase rather than real and imaginary parts (unless when doing nyquist plots) xschem transforms data

out=re + i * im

into magnitude:

out = √(re² + im²)

and phase:

ph(out) = atan2(im, re)*180 / π

. So, yes

ph(out)

is xschem specific, it is the phase of

out

measured in degrees. > if it's xschem, why is not "out ph()" like the gain "out db20()" ? Because as said above in xschem

out

is the magnitude of ngspice saved node

out

. You can't get a phase out of a magnitude. Xschem already has transformed the complex ngspice-saved number

out

into magnitude (

out

) and phase in degrees (

ph(out)

). The db20() function just does

20*log10(out)

to transform magnitude

out

into a value in

dB

> and how does xschem know to use the ac analysis in the plhase and gain plots (I'm looking a the ac example in the xschem demos) If there is only one simulation (only one AC sim for example) xschem will load this one and figure out it's an AC plot. If there are multiple simulations (op, dc, tran, ac) the graph can specify which simulation to load. See the Raw file and Sim type text boxes.

thank you very much Stefan!

@Matt Venn

so is ph(out) a variable name then?

Yes, from every complex variable

xxx

(two real numbers, one representing the real part, the other the imaginary component) saved by ngspice in ac raw files xschem creates

xxx

(the magnitude) and

ph(xxx)

(the phase in degrees)

Thanks!

Interesting way to visualise the nyquist plot @Stefan Schippers. Thanks for including this feature in xschem. Could I ask if the schematic of the

test_nyquist.sch

could be shared or where are the ac examples in the xschem repo ? I tried the ngspice expression for the imaginary part as

imag(vout_vector)

but it doesn't generate a nyquist plot.

the test_nyquist has been added to the xschem distribution. This example uses generic MOS models, not sky130 specific devices, so you can test it without an installed pdk. https://github.com/StefanSchippers/xschem/blob/master/xschem_library/examples/test_nyquist.sch