Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
- LTspice
- Messages
Search
|
Re: Frequency
Ganesan
Thanks
toggle quoted message
Show quoted text
On 9/20/2011 3:00 PM, John Woodgate wrote:
|
|
Re: Frequency
Tony Casey
<snip>
--- In LTspice@..., Ganesan <dg1@...> wrote: </snip> I think it explains everything once you consider that SPICE's automatic timestep algorithm is not exposed to the user, unless you override it. Perhaps you can see why I questioned your use of SPICE in this context. The "better way" is to accept that SPICE is the wrong tool for this job. Regards, tony |
|
Re: Frequency
Ganesan
I think measuring the period over a thousand cycles and running it
toggle quoted message
Show quoted text
through a software delta- sigma might be a better choice.. I just dont know how to implement this in LTspice parlance.. Cheers AG On 9/20/2011 2:47 PM, Ganesan wrote:
|
|
Re: Frequency
John Woodgate
In message <4E78EC92.4080505@...>, dated Tue, 20 Sep 2011, Ganesan <dg1@...> writes:
femptosecondsPlease remove the p. -- OOO - Own Opinions Only. Try www.jmwa.demon.co.uk and www.isce.org.uk John Woodgate, J M Woodgate and Associates, Rayleigh, Essex UK When I point to a star, please look at the star, not my finger. The star will be more interesting. |
|
Re: Frequency
Ganesan
I tried the numdgt=15. As I suspected it improves the "Precision", but
toggle quoted message
Show quoted text
not the "accuracy" cheers AG On 9/20/2011 2:42 PM, Ganesan wrote:
|
|
Re: New Member and Updated Help File
Lewis
Here's an update on using the provided help from LTspice as a basis for
an expanded and annotated help in LTwiki <> : - Mike graciously allowed use of the existing help as a basis for an expanded and annotated help in LTwiki, specifically the LTspiceHelp.chm not the scad3.pdf. - In doing so, we must make it clear what is the original versus the annotated. - I've succeeded in converting the Help to an HTML series of files. This conversion can be seen at: LTspice Help <> -I'm now working on my test server LTwiki to (more or less) seamlessly incorporate this help as a stub for a section titled 'Annotated and Expanded LTspice Help'. The approach I'm trying is Iframe Widgets into the MediaWiki engine that you know of as LTwiki.org - The end result is to have this Help on the wiki, but the annotating authors would then further explain, illustrate and annotate a given stub as they see fit. best regards, Lewis --- In LTspice@..., "Helmut" <helmutsennewald@...> wrote: level of traffic on this list. Cadence tools (Virtuoso, etc), but have used Mentor Graphics in times past as well as Daisy systems in the very beginning. I now need to get a lot of BSIM3V3 models brought in so I can "play" on LTSpice. has the latest and greatest, but, the help seems very outdated. Case in point: Special Functions. The help says that the standard logic gates are available plus a VARISTOR and a MODULATE. However, when I go to place a symbol in the schematic, it shows that there are SRFLOP, PHIDET, BUF, and BUF1 available too. the help has been updated to at least mention these newer components? I'm imagining that they have been there for awhile now and didn't recently appear after I ran the FUT.
|
|
Re: Frequency
Ganesan
Thanks ..Explains the RC case but not the RRcase.. Further 1ppm of 10k
toggle quoted message
Show quoted text
is 100psec. My real frequency in more complex circuits are 100s of megahertz.. That puts the steps size in the 10 femptoseconds.. Yikes... There must be a better way.. On 9/20/2011 2:13 PM, Tony Casey wrote:
|
|
Re: Frequency
Tony Casey
<snip>
--- In LTspice@..., Ganesan <dg1@...> wrote: </snip> Hello AG, The first thing you can do is try all the usual things that improve accuracy: 1. Add a minimum timestep (I realise you have a philosophical objection to this). Adding a minimum step of 10n completely fixes your measurement issue in this case - 0ppm error. 2. When necessary, you can invoke double precision data storage by adding .option numdgt 15. Read the .option section of the Help. 3. Add a redundant slave B source with the tripdt and tripdv clauses. these can often be more efficient than a minimum timestep. I suspect trying to accurately count the frequency of high stability sources in SPICE might be a complete waste of time, although that might depend exactly what you mean by high stability. It is in principle possible to model a Caesium standard with SPICE, although with stabilities of the order of 1E-14, the required simulation time might be weeks or months. Just a thought. Regards, Tony |
|
Re: Frequency " How do you measure frequency of an oscillator accurately?"
Ganesan
It drops to 12 ppm when I replace the capacitor with a 10K resistor. A
toggle quoted message
Show quoted text
pure resistor divider should give better accuracy than that.. Cheers AG On 9/20/2011 1:40 PM, Ganesan wrote:
|
|
Re: Frequency
Ganesan
I uploaded the file
toggle quoted message
Show quoted text
File-> Temp ->RC_meas_freq.asc <> I applied your recommended technique. I get an accuracy of 40 ppm.. I am afraid the result may be worse in very complex circuits with non-linear caps.. I would like to get an accuracy of 0.1ppm to 1ppm (Xtal oscillators and time references). " How do you measure frequency of an oscillator accurately?" cheers AG On 9/20/2011 2:42 AM, Helmut wrote:
|
|
Re: Regardinfd Delay Element simulation
Other than that, all are fine??please confirm..Well, try it and see! shall i give clock signal for the nodes VIN_P, VIN_N, and VBIAS_P??You can give whatever signals you want to, to those nodes. It is your circuit to simulate. Just don't leave them floating or LTspice will complain. My guess is that VBIAS_P should probably be a DC voltage, and then VIN_P and VIN_N would be some sort of differential signal or common-mode signal or some combination of the two. Whether it is a single pulse or a clock signal or a semi-random data signal, is entirely up to you and what you want to simulate. If you don't know what signals to use, please ask your co-worker or your teacher what to do. Andy |
|
Re: Data not saved in a raw file?
--- In LTspice@..., Walking Through <walkingthrough@...> wrote:
Hello Mark, You could additionally add the following options. .options gmin=1e-10 abstol=1e-10 reltol=0.003 Maybe you can send me your files if you don't want upload the files to the group's Files/TEMP. Best regards, Helmut |
|
Re: Behavioral source generates an error message?
--- In LTspice@..., Walking Through <walkingthrough@...> wrote:
Hello Mark, The simulation time variable is always 0 or greater 0. The following equation works without warnings. B1 VControl 0 V=2.39495e-005*if(time < 2e-009, 1.0 -exp(-time/2e-009),0.632121*exp(-(time - 2e-009)/2e-008)) Best regards, Helmut |
|
Behavioral source generates an error message?
Walking Through
A netlist I have contains (apologies for the line wrap):
B1 VControl 0 V=2.39495e-005 * if (0 <= time & time < 2e-009, 1.0 - exp(-time/2e-009),0.632121*exp(-(time - 2e-009)/2e-008)) This generates a message in the log file: Questionable use of curly braces in "b1 vcontrol 0 v=2.39495e-005 * if (0 <= {time} & time < 2e-009, 1.0 - exp(-time/2e-009),0.632121*exp(-(time - 2e-009)/2e-008))" Error: undefined symbol in: "[time]" The netlist doesn't have "{time}" anywhere. Nor does it have "[time]". Mark |
|
Re: Data not saved in a raw file?
Walking Through
Hello Helmut -
You wrote: Here is excerpt from the netlist: * ======= Simulation Options ======= * .options noopiter .options topologycheck=0 .options gminsteps=0 .options srcsteps=0 * These go a LONG way toward reducing the simulation time. If these weren't in place, just finding the initial operating point would take on the order of 30 minutes. --Mark P.S. Just found something odd in the log file and will post about that separately. |
|
Re: NAND .model
--- In LTspice@..., "s35148" <s35148@...> wrote:
Hi I have some simulations (one I am fighting with right this second) which use the 4093 NAND. My simulation uses a part from the CD4000.lib file that is in this group. (Look here: ) It's not a bad model, although the input thresholds and output levels are defined, but fixed (not changing with supply rails). I have added a behavioral source on the output for one of my simulations so that that it doesn't output more than the supply level, and I may have to put something on the input to similarly scale the thresholds for start-up simulations (or just do it the right way and build my own model!) Hope that helps. A. |
|
Re: Frequency
--- In LTspice@..., "Tony Casey" <tony@...> wrote:
Hi, If there is a section in LTwiki to show examples for specific calculations, this would be a good example to be added under the title "Frequency and duty cycle". I admit that for most of us this example is rather obvious but not for scared/confused users running a simulator for the first time :) For the more complicated situations, I used to archive every solution I found adequate under an appropriate title that I can find easily by searching anytime later. Kerim |
|
Re: Data not saved in a raw file?
--- In LTspice@..., Walking Through <walkingthrough@...> wrote:
Hello Mark, I wonder about the slow simulation speed. You should help LTspice with some settings(.options) to improve convergence. This should reduce the simulation time by factors if not decades. Best regards, Helmut |
|
Data not saved in a raw file?
Walking Through
Hello -
Last Thursday a simulation consisting of a 100 x 100 grid of resistors and capacitors was started.? I knew this was going to take a long time (transient response) and arranged for use of a dual-Xeon computer. Simulation time was from 0 to 1 microsecond. Returned Monday to find the simulation done and plots on the display.? In the past, LTspice saved the raw file and I could use ltsputil to retrieve data. Read: No, I didn't export the data before closing LTspice. LTsputil says the raw file contained data to 27 nsec. Windows XP shows the raw file as 19K long.? Where did the other 973 nsec go? Also.... How often do displayed waveforms get refreshed? I remember checking Thursday afternoon and seeing that 27 nsec... but while the display showed 1 usec on Monday, the rawfile only contained data to 27 nsec. --Mark |
|
Re: Regardinfd Delay Element simulation
Other than that, all are fine??please confirm..
toggle quoted message
Show quoted text
shall i give clock signal for the nodes VIN_P, VIN_N, and VBIAS_P?? please confirm. Thank you --- In LTspice@..., Andy <Andrew.Ingraham@...> wrote:
Please help me on this and do the needful.You left unconnected signal pins. Look at nodes VIN_P, VIN_N, and |
to navigate to use esc to dismiss