Transient measurements in HSPICE

[rmanalo]

Hello everyone,

I recently designed a high PSRR voltage reference circuit with a self-regulating feedback network. I then derived the transfer function from supply to output and I got my calculations to be 146 dB. I simulated this in AC analysis and is indeed around 140 dB at 100 Hz. I then added a sinusoidal wave with 18 mVpp 100Hz at the supply and measured the corresponding output using the .meas command was 82.3 nVpp. In the simulation graph shows only a traight line. The measured PSRR [20*log10(Voutpp/Vsupplypp)] was only -106 dB.

I again added the same sinusoidal wave but with 10 MHz and measured the output. Now at 10 MHz the PSRR from AC simulation and Transient Simulation did match. I'm concerned regarding the PSRR at 100 Hz. Do you think this is a limitation of the simulator (like the minimum voltage steps or something?). Here are the simulation plots.

 

[pancho_hideboo]

Tighten simulation accuracy.
Show me your .option

 

[rmanalo]

Only .option post

I did looked into increasing accuracy from the HSPICE Manual and added a few additional commands.

.option ABSVDC=1e-12
.option ABSTOL=1e-12
.option ABSI=1e-12

The measured output was 46.6 nVpp which is equivalent to -111.7 dB PSRR. Not close to 140 dB in AC analysis. What should I change?

 

[FvM]

There are more options relevant for transient analysis accuracy, e.g. ABSV. But what is your simulation good for?

 

[pancho_hideboo]

Set .option runlvl=6 RELTOL=1e-6

Observe signal via VCVS with -244.5mvolts.
It could relax rquirement of relative accuracy.

 

[rmanalo]

But what is your simulation good for?

It's for verification. I was tasked to show that AC analysis and transient simulations matched with regards to PSRR.

- - - Updated - - -

Observe signal via VCVS with -244.5mvolts.

Thank you for your input. If my understanding is correct, the purpose of the VCVS is to amplify the output?

 

[pancho_hideboo]

If my understanding is correct, the purpose of the VCVS is to amplify the output?

No.
Subtract 244.5mvolts.

Vout=244.5mvolt + very_small_sinusoid
Observe Vout - 244.5mvolt via VCVS.

See https://designers-guide.org/forum/YaBB.pl?num=1283902372/15#15

 

[rmanalo]

Observe Vout - 244.5mvolt via VCVS.

I don't know how to implement 244.5 mV via VCVS. Is it something like:

E1 Vout gnd Vr gnd -1   $where Vr is 244.5 mV? because if so, then why not use Vr directly?

I did however made an alternative. I put the entire circuit into a subcircuit and called each with one having a dc with superimposed sinusoidal voltage at the supply and the other without. Then plotted the result.

X1 Vdd1 Vref1 voltref
X2 Vdd2 Vref2 voltref

Vdd1	Vddb	gnd	0.95
Vddb	Vdd1	Vddb	0 Sin(0 9m freq)
Vdd2	Vdd2 gnd 0.95

.option post
.option RUNLVL=6 RELTOL=1e-6

Thoughts on this method? The result however is still the same and still did not match having only -106 dB.

 

[pancho_hideboo]

vaho n1 0 dc=244.5e-3
eaho n2 0 out n1 vcvs 1.0

Observe n2 instead of out.

 

[rmanalo]

Observe n2 instead of out.

Thank you for your help.

here is the output.


Interestingly it is similar to that of the previous. The accuracy however did not change and still got around 82 nVpp corresponding to a PSRR of 106 dB. I got the hint with the reltol and absv. I'll explore other option commands with this.

 

[pancho_hideboo]

Waveform is not sinusoid.
It is due to lack of accuracy.

 

[pancho_hideboo]

Try .option delmax=5e-9