开云体育

It's not about convergence really; LTspice cannot find the DC operating point easily.? If you add the Spice directive .option gminsteps=0, the simulation runs and gives credible results for the AC sweep.

By the way, asking for 10 thousand points per decade in your AC sweep is huge overkill, especially as you are sweeping over many decades.

Best wishes
John Woodgate OOO-Own Opinions Only
J M Woodgate and Associates 
Rayleigh, Essex UK

开云体育

Jan wrote:

"I'm trying to simulate a Sallen Key filter with a cutoff frequency of around 40kHz using a unipolar 5V source (See TEMP section). The initial operating point solution does not converge.

I have added series resistances everywhere so every point in the circuit has a connection to GND as described in

Since I cannot find any problem in my circuit, I'm out of ideas on where LTSpice might fail.

Can someone hint me to where the problem might be?"

Your circuit converges fine and runs with .options gminsteps=0.

It make no sense to have 10000 points per decade in an .AC analysis, unless you have a circuit with a Q of 1000s - I changed it to 24. It just takes longer and you produce a huge .RAW file.

Regards,
Tony

开云体育

开云体育

In order to copy a schematic or part of one to another schematic, you have to have both schematics open in the same instance of LTspice. Open them both before you try to copy.

Best wishes
John Woodgate OOO-Own Opinions Only
J M Woodgate and Associates 
Rayleigh, Essex UK

开云体育

I name all nodes that might be "interesting" so I know where everything is.

Donald.
--
*Plain Text* email -- it's an accessibility issue
() no proprietary attachments; no html mail
/&#92; <>

Hello analogspiceman

I recently asked Mike if would be possible to add a new feature to .savebias
and .loadbias invoked by the keyword “eachstep” whereby in a stepped
simulation, each succeeding step would load the bias saved by the preceding
step. Among other uses, this could allow eliminating the startup transient
between steps when performing a time domain frequency response analysis.

What was his response? I'm also curious because, by chance, I saw this
topic come up on electronics.stackexchange a few days ago.

--

Vlad
______________________
ltspicegoodies.ltwiki.org -- holding, among others:
a universal analog/digital filter, block-level models
for power electronics (and not only), math blocks
with a more stream-lined approach, some digital
ADC, DAC, (synchronous-)counter, JKflop, etc.

I recently asked Mike if would be possible to add a new feature to .savebias and .loadbias invoked by the keyword “eachstep” whereby in a stepped simulation, each succeeding step would load the bias saved by the preceding step.? Among other uses, this could allow eliminating the startup transient between steps when performing a time domain frequency response analysis.

Currently bias data may be saved from individual steps, but the syntax is cumbersome as each step must be specified:

.savebias bias1.txt time=1m step=0
.savebias bias2.txt time=1m step=1
.savebias bias3.txt time=1m step=2

Unfortunately, there is no way to load different bias data for steps, which is the reason for my request to Mike.

Yes, I realize that .savebias type commands do not save branch currents (inductor states), but that may be neatly circumvented by creating a simple inductor subciruit that converts a capacitor into its dual via controlled current sources. (I posted this a few days ago.)

.subckt L 1 2

.param L=1 Rser=1m Rpar=1/Gmin

C1 3 0 {L} Rser={1/Rpar} Rpar={1/Rser}

G1 1 2 3 0 1

G2 3 0 2 1 1

.ends L

?

Merry Christmas and a Happy New Year

thanks all

When I run simulations, I get a messages in the Spice error log stating

Hi, friends.