How sources and components are treated in DC analysis?

Not sure if this is in the documentation (if not, I'm sure it's covered well in The Designer's Guide to SPICE and Spectre). Anyway, it's quite straightforward (and actually fairly obvious if you think about it).

If the source has the dc parameter set, then that value is used as the value of the source for the DC analysis. Otherwise the time-zero value of the time-varying source is used (and of course the source is not time-varying during the analysis).

If you're doing the DC operating point that is computed at the beginning of a transient, it always uses the time-zero value.

Kind Regards,

Andrew.

Thank you.

I will look at the book and hope it is covered well. I know how DC analysis is done in general (caps are open, inductors are short and ac sources are short too). However, Cadence has a lot of voltage sources with many parameters (DC voltage, initial voltage, offset voltage,...). 

I am not sure if there kinds of parameters are considered in DC analysis. Anyway, I would like to read a thorough treatment about this.

Not sure what you mean by "Cadence" (since Cadence is the company name, not the software), but spectre only has three independent sources: vsource, isource, and port. Various aspects of these are made available in analogLib in Virtuoso by other components, but they all netlist to one of these sources.

As I said, it's very simple. It uses the "dc" value of the source (that's the dc parameter of any of these sources) if set, and if not, it uses the time-zero value of any time-varying source.

Ken's book (the one I mentioned) simply says that the DC does this first:

1. Configure all independent sources to be constant valued.
2. Replace all capacitors with open circuits.
3. Replace all inductors with short circuits.

It's not a matter of "shorting" ac sources (current sources do not short at all). It's just stopping the independent sources from being time-varying and picking a value for that source (which I described how that works). I think you're probably over-complicating matters!

BTW, dependent sources (e.g. vcvs etc) are used as-is - they are not treated differently during the DC analysis. It's only independent sources that have to remove their time-varying nature. This is hardly surprising, since there's no time in a DC analysis!

Regards,

Andrew.

Hello again,

I have just got a copy of the book now. It is night here now so I will read it in an hour.

For Cadence, I meant the software. I think I should call it spectre or hspiceD or something similar!

...spectre only has three independent sources: vsource, isource, and port. Various aspects of these are made available in analogLib in Virtuoso by other components, but they all netlist to one of these sources.

Yes. I see a lot of independent sources in analogLib and they are really confusing.

Various aspects of these are made available in analogLib in Virtuoso by other components, but they all netlist to one of these sources.

Yeah, I will take a look at it.

As I said, it's very simple. It uses the "dc" value of the source (that's the dc parameter of any of these sources) if set, and if not, it uses the time-zero value of any time-varying source.

How about the DC offset? For example in vsource, source pwl, there are two DC components (DC voltage and DC offset). Is DC offset also used in DC analysis?

For time-zero value, could you explain what does it mean? Does this mean the value of the source at the time equal to zero? Sorry English is not my native and I can't decipher that.

I'm not really sure why the analogLib components are confusing. Part of the reason they exist is for other simulators and for historical reasons. With spectre you can just always use the analogLib vsource, isource, and port components - these have a parameter which defines the type of the source and fields which change dynamically.

The dc offset (which corresponds to the offset parameter on the source) is not affecting the DC operating point. It just adds that offset to all the points defined in the PWL vector during the transient.

For any time-varying source, the time-zero value is whatever value that time-varying waveform would have had at time=0s. So if you have a sine source with a sinephase of 90 degrees, it will be the value that the sine wave has at a phase of 90 degrees (since that's the initial phase of the sine wave). 

With a real example, if you do:

v1 (n1 0) vsource type=pwl dc=1 offset=0.2 wave=[0 0.3 1m 0.7 2m -0.4]

Then n1 will be 1V for the DC analysis. If you do:

v1 (n1 0) vsource type=pwl offset=0.2 wave=[0 0.3 1m 0.7 2m -0.4]

Then n1 will be 0.5V for the DC analysis (because the value at t=0s would be 0.3+0.2).

Regards,

Andrew.

Thank you.

I have just tested these waveforms in DC and transient analysis. The result confirms what you said. I also notice that the transient analysis does use DC voltage at all.

These detailed knowledge I can't find anywhere (something like DC analysis use time-zero value if,...). I think I wouldn't believe in myself if I I don't read your post or read it from an official document!

Note that a transient simulation does use the dc voltage in some situations. For example, if there's no time-varying parameters set up at all (e.g. it's just a dc source). Or if it's a sine source and you don't have the sinedc parameter set (then it sets the mid-point for the sine wave).

I just can't remember if it's in the documentation anywhere (I don't have time to read it through to check) - sorry.

Regards,

Andrew.