power dissipation in LT Spice

[John Larkin]

If I have a mosfet with grounded source that's driving some load to V+,
switching the load low side to ground pulsewise, is there some simple way to
calculate the fet's power dissipation?

I could measure the voltage and current and multiply-integrate with "analog"
components, or do the equivalent with math expressions, but I was wondering if
anything like that was built in. Looking at the LT Spice HELP stuff, I don't see
anything obvious.


--

John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

[Joe Chisolm]

On Fri, 12 Apr 2013 22:16:22 -0700, John Larkin wrote:

> If I have a mosfet with grounded source that's driving some load to V+,
> switching the load low side to ground pulsewise, is there some simple
> way to calculate the fet's power dissipation?
>
> I could measure the voltage and current and multiply-integrate with
> "analog" components, or do the equivalent with math expressions, but I
> was wondering if anything like that was built in. Looking at the LT
> Spice HELP stuff, I don't see anything obvious.

Mouse over the fet to where you get the current probe, then
press the ALT key. The probe will change to a little
thermometer - click. You will get the power plot. Up at the
top will be the nodes (V*I). click on that.

--
Chisolm
Republic of Texas

[P E Schoen]

"Joe Chisolm" wrote in message
news:ioqdnXoTnL9sbfXM...@earthlink.com...

> Mouse over the fet to where you get the current probe, then
> press the ALT key. The probe will change to a little
> thermometer - click. You will get the power plot. Up at the
> top will be the nodes (V*I). click on that.

Good advice. I have found the help files in LTspice to be rather deficient,
but there are many tutorials and references on-line that I have found when
searching for ways to do things. Here are a few:

http://www.simonbramble.co.uk/lt_spice/ltspice_lt_spice_tutorial_1.htm

http://denethor.wlu.ca/ltspice/

http://ltwiki.org/?title=SPICE_and_LTspice_Courseware_and_Tutorials

Something else to remember when calculating power is to choose a period of
time long enough to average out, and for repetitive waveforms you should
start and end the sample at zero crossings or peaks. Also let the simulation
run long enough for transients to settle, especially when starting with
first application of supply voltage from zero. Reactive elements can throw
off the measurement. And unless you want instantaneous power at the cursor,
use Ctrl-click to get the average and RMS value (or integral) of what you
are measuring in the time window you have selected.

Paul

[Jamie]

John Larkin wrote:

> If I have a mosfet with grounded source that's driving some load to V+,
> switching the load low side to ground pulsewise, is there some simple way to
> calculate the fet's power dissipation?
>
> I could measure the voltage and current and multiply-integrate with "analog"
> components, or do the equivalent with math expressions, but I was wondering if
> anything like that was built in. Looking at the LT Spice HELP stuff, I don't see
> anything obvious.
>
>

I find it hard to believe you would ask such a question John.

THere must be some other factor you are thinking about that you're
not revealing here.

Jamie

[John Fields]

On Sat, 13 Apr 2013 09:31:24 -0400, Jamie
<jamie_ka1lpa_not_v...@charter.net> wrote:

>John Larkin wrote:
>
>> If I have a mosfet with grounded source that's driving some load to V+,
>> switching the load low side to ground pulsewise, is there some simple way to
>> calculate the fet's power dissipation?
>>
>> I could measure the voltage and current and multiply-integrate with "analog"
>> components, or do the equivalent with math expressions, but I was wondering if
>> anything like that was built in. Looking at the LT Spice HELP stuff, I don't see
>> anything obvious.
>>
>>
>I find it hard to believe you would ask such a question John.

---
Your idols don't have feet of clay?
---

>
> THere must be some other factor you are thinking about that you're
>not revealing here.

---
He's too lazy to do his own legwork and RTFM?


--
JF

[John Larkin]

On Sat, 13 Apr 2013 09:31:24 -0400, Jamie
<jamie_ka1lpa_not_v...@charter.net> wrote:

>John Larkin wrote:
>
>> If I have a mosfet with grounded source that's driving some load to V+,
>> switching the load low side to ground pulsewise, is there some simple way to
>> calculate the fet's power dissipation?
>>
>> I could measure the voltage and current and multiply-integrate with "analog"
>> components, or do the equivalent with math expressions, but I was wondering if
>> anything like that was built in. Looking at the LT Spice HELP stuff, I don't see
>> anything obvious.
>>
>>
>I find it hard to believe you would ask such a question John.

Because I want to estimate the power dissipation of a mosfet.

>
> THere must be some other factor you are thinking about that you're
>not revealing here.

Just how hot the fet will get. I don't use Spice very often, and I didn't know
about the trick that a couple of people have pointed out. Power-vs-time is
interesting, but it's still not average power.

The last time I actually did real calculus was to get the equation for fet power
dissipation in a situation with a linear load. Now I'm driving time-varying
nonlinear loads.

I can probably get close enough with just SWAG arithmetic for this application,
but I was wondering if there was some simple way to do it in LT Spice.

I might still build a "wattmeter" in LT Spice. A shunt resistor, a couple of
VCVSs, a multiplier, an integrator.

[Joerg]

Ok, Joe has explain the thermometer thingamagic via ATL-LeftClick in the
schematic. Now you have a plot power-versus-time. Mouse over its
expression on the top of the waveform window -> CTRL-LeftClick and ...
voila. A little text window open that contains average power and so
other information.

You have to wait until the simulation stopped, such integration will not
working while running.

--
Regards, Joerg

http://www.analogconsultants.com/

[Jamie]

I've learn to use a calculator years before I even touched on the use of
spice.

I know spice is a nice tool to use and at times, it could cause you
brain damage, or lack of I should say. If you don't exercise it a bit
with the connectivity from the neurons down to the tips of your fingers
via the path ways you were born with, pushing those little keys, you'll
end up like slow-Man, all talk and no walk!

Jamie

[Joe Chisolm]

--^^^^^^^^^

This should be ctrl-click for the average power.

It was too early in the morning (or is that late in the work day)
to be posting

[John Larkin]

On Sat, 13 Apr 2013 06:30:27 -0700, Fred Abse <excret...@invalid.invalid>

wrote:

>On Fri, 12 Apr 2013 22:16:22 -0700, John Larkin wrote:
>

>> If I have a mosfet with grounded source that's driving some load to V+,
>> switching the load low side to ground pulsewise, is there some simple
>> way to calculate the fet's power dissipation?
>>
>> I could measure the voltage and current and multiply-integrate with
>> "analog" components, or do the equivalent with math expressions, but I
>> was wondering if anything like that was built in. Looking at the LT
>> Spice HELP stuff, I don't see anything obvious.
>

>From LTSpice Help:
>
>"Yet another schematic probing technique is to plot the instantaneous
>power dissipation of a component. To do this, hold down the Alt key and
>click on the body of the symbol of the component. The instantaneous power
>dissipation will be plotted as an expression of voltages and currents. It
>will be plotted on its own scale with the units of Watts. The mouse cursor
>turns into an icon that looks like a thermometer when it's pointing at a
>dissipation that can be plotted. You can find the average power
>dissipation by control-clicking the trace label."
>
>Under wine in linux, it may need <ctrl-alt-left click>. At least, with
>the window manager I use, it does. (my comment).

OK, all that works:

https://dl.dropboxusercontent.com/u/53724080/Circuits/Power/Fet_Power.jpg

Thanks.



>
>"Compute the average or RMS of a trace. The waveform viewer can integrate
>a trace to obtain the average and RMS value over the displayed region.
>First zoom the waveform to the region of interest, then move the mouse to
>the label of the trace, hold down the control key and left mouse click."
>
>RTFM:-)


I did, but it turns out to be buried in "Trace Selection." It's easier to ask
here.

[John Larkin]

Fred's advice got me there. Cool.

In the pic I posted, note that I simulated for 51 us. If I simulate for 50 us,
it takes minutes to run. At 51, it finishes instantly!

[John Larkin]

I was working WBOC this morning, Without Benefit of Coffee. I'm full of Peets
now and things are clear.

The target is to make some fast 7.5 KW pulses. Soon.

[Phil Hobbs]

So it works when it's one card short of a full deck, just not two cards. ;)

Cheers

Phil Hobbs

Cheers

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510 USA
+1 845 480 2058

hobbs at electrooptical dot net
http://electrooptical.net

[John Larkin]

On Sat, 13 Apr 2013 09:08:33 -0700, Fred Abse <excret...@invalid.invalid>
wrote:

>On Sat, 13 Apr 2013 08:39:30 -0700, John Larkin wrote:
>
>> In the pic I posted, note that I simulated for 51 us. If I simulate for 50 us,
>> it takes minutes to run. At 51, it finishes instantly!
>

>Yup. Same here.

Takes longer to simulate less time!

>
>Specifying max timestep 10ns, it runs fast, 1ns, it starts at 6us/s, slows down to
>about 3fs/s, then speeds up again. I've notice similar slowdowns around fast edges
>before.

I wonder why it doesn't have a box for min timestep. I've used (and written)
other simulators that did, allowed you to deliberately trade accuracy against
sim speed.

[John Larkin]

There's always a Joker.

[Joerg]

John Larkin wrote:
> On Sat, 13 Apr 2013 10:16:02 -0500, Joe Chisolm <jchi...@earthlink.net> wrote:
>
>> On Sat, 13 Apr 2013 00:45:21 -0500, Joe Chisolm wrote:
>>
>>> On Fri, 12 Apr 2013 22:16:22 -0700, John Larkin wrote:
>>>
>>>> If I have a mosfet with grounded source that's driving some load to V+,
>>>> switching the load low side to ground pulsewise, is there some simple
>>>> way to calculate the fet's power dissipation?
>>>>
>>>> I could measure the voltage and current and multiply-integrate with
>>>> "analog" components, or do the equivalent with math expressions, but I
>>>> was wondering if anything like that was built in. Looking at the LT
>>>> Spice HELP stuff, I don't see anything obvious.
>>> Mouse over the fet to where you get the current probe, then press the
>>> ALT key. The probe will change to a little thermometer - click. You
>>> will get the power plot. Up at the top will be the nodes (V*I). click
>>> on that.
>> --^^^^^^^^^
>>
>> This should be ctrl-click for the average power.
>>
>> It was too early in the morning (or is that late in the work day)
>> to be posting
>
> I was working WBOC this morning, Without Benefit of Coffee. I'm full of Peets
> now and things are clear.
>
> The target is to make some fast 7.5 KW pulses. Soon.
>

In that case I'd consider negative gate drive for the turn off, and a
drive with some serious gusto, and maybe +12V and -10V. Maybe even a
peaker. Plus a blast shield for the FETs.

[John Larkin]

I want to make narrow pulses at low duty cycles, load powers well below 100
watts average, so fragmentation won't be a hazard. I'm figuring dpak or SO8 type
surface-mount fets and *lots* of capacitors, low 10s of joules maybe.

It's at the idea stage now, just thinking about possibilities.

[Tim Williams]

"Joerg" <inv...@invalid.invalid> wrote in message
news:astf0n...@mid.individual.net...

>> The target is to make some fast 7.5 KW pulses. Soon.
>>
>
> In that case I'd consider negative gate drive for the turn off, and a
> drive with some serious gusto, and maybe +12V and -10V. Maybe even a
> peaker. Plus a blast shield for the FETs.

Gnaw, with transistors these days, 7.5kW continuous is easy.
Theoretically, 4 x TO-247 superjunction FETs will do that at up to a MHz.
Pulses? That's childs play! ;-)

Tim

--
Deep Friar: a very philosophical monk.
Website: http://seventransistorlabs.com

[John Larkin]

On Sat, 13 Apr 2013 12:36:48 -0500, "Tim Williams" <tmor...@charter.net>
wrote:

>"Joerg" <inv...@invalid.invalid> wrote in message
>news:astf0n...@mid.individual.net...
>>> The target is to make some fast 7.5 KW pulses. Soon.
>>>
>>
>> In that case I'd consider negative gate drive for the turn off, and a
>> drive with some serious gusto, and maybe +12V and -10V. Maybe even a
>> peaker. Plus a blast shield for the FETs.
>
>Gnaw, with transistors these days, 7.5kW continuous is easy.
>Theoretically, 4 x TO-247 superjunction FETs will do that at up to a MHz.
>Pulses? That's childs play! ;-)
>
>Tim

Pretty much. The issues become capacitors, packaging, and some way to get the
drive to the load.

[Joerg]

Tim Williams wrote:
> "Joerg" <inv...@invalid.invalid> wrote in message
> news:astf0n...@mid.individual.net...
>>> The target is to make some fast 7.5 KW pulses. Soon.
>>>
>> In that case I'd consider negative gate drive for the turn off, and a
>> drive with some serious gusto, and maybe +12V and -10V. Maybe even a
>> peaker. Plus a blast shield for the FETs.
>
> Gnaw, with transistors these days, 7.5kW continuous is easy.
> Theoretically, 4 x TO-247 superjunction FETs will do that at up to a MHz.
> Pulses? That's childs play! ;-)
>

That's what one guy thought as well. All my recommendations regarding
steep drive and UVLO and all that were brushed aside. One fine day his
drive supply must have cut out, slowly since there were electrolytics,
and ... *PHOOMP*

[Joerg]

John Larkin wrote:
> On Sat, 13 Apr 2013 12:36:48 -0500, "Tim Williams" <tmor...@charter.net>
> wrote:
>
>> "Joerg" <inv...@invalid.invalid> wrote in message
>> news:astf0n...@mid.individual.net...
>>>> The target is to make some fast 7.5 KW pulses. Soon.
>>>>
>>> In that case I'd consider negative gate drive for the turn off, and a
>>> drive with some serious gusto, and maybe +12V and -10V. Maybe even a
>>> peaker. Plus a blast shield for the FETs.
>> Gnaw, with transistors these days, 7.5kW continuous is easy.
>> Theoretically, 4 x TO-247 superjunction FETs will do that at up to a MHz.
>> Pulses? That's childs play! ;-)
>>
>> Tim
>
> Pretty much. The issues become capacitors, packaging, and some way to get the
> drive to the load.
>

Don't forget an UVLO on the gate drive. If you keep pulsing while that
sags away a few hundred watts can migrate from the load into the FETs.

[John Larkin]

One has to do something to avoid a division-by-zero error in the MTBF
calculation.

[josephkk]

On Sat, 13 Apr 2013 09:28:44 -0700, Fred Abse
<excret...@invalid.invalid> wrote:

>On Sat, 13 Apr 2013 09:14:44 -0700, John Larkin wrote:
>
>> I wonder why it doesn't have a box for min timestep. I've used (and written)
>> other simulators that did, allowed you to deliberately trade accuracy against
>> sim speed.
>

>Well, it was originally written to speed up SMPS simulations, and appears
>to work hard around edges, and speed up on flat bits. Maybe there's some
>lookahead going on.
>
>Being able to specify min timestep would be nice.
>
>OTOH, I rarely, if ever simulate SMPS, so I'm not the target audience.
>
>What I really would like is polar/Smith plots.

Maximum time step is on the .options card.

?-)

[P E Schoen]

"Fred Abse" wrote in message
news:pan.2013.04.16....@invalid.invalid...

> On Tue, 16 Apr 2013 00:57:42 -0700, josephkk wrote:

>> Maximum time step is on the .options card.

> I knew that, it was *minimum* timestep we were discussing.

I thought that the Max Timestep may be a misnomer, and actually means
minimum timestep. The actual directive in LTSpice is as follows:

.tran 0 400m 10m 5u startup uic

Which is defined as:

.tran <Tprint> <Tstop> <Tstart> <Tmaxstep> <start from zero> <skip initial
operating point>

There is a parameter in the transient analysis directive for timestep from
http://ltwiki.org/index.php5?title=Simulation_Command:

.tran <Tstep> <Tstop> [Tstart [dTmax]] [modifiers]
or
.tran <Tstop> [modifiers]

It says:
"Tstep is the plotting increment for the waveforms but is also used as an
initial step-size guess. LTspice uses waveform compression, so this
parameter is of little value and can be omitted or set to zero."

Here is something I found in http://www.intusoft.com/articles/converg.pdf:
1. Gminsteps (DC Convergence)
Example: .OPTIONS GMINSTEPS=200
The Gminsteps option adjusts the number of Gmin increments
that will be used during the DC analysis. Gmin stepping is
invoked automatically when there is a convergence problem.
Gmin stepping is a new algorithm in SPICE 3 that greatly
improves DC convergence.

I tried various values of GMINSTEPS in LTspice but there seemed to be no
effect on simulation time.

Some other discussions of timestep:
http://www.aboutspice.com/details-208

I have confirmed that the simulation can be slowed down and made more
accurate by reducing the maxstimestep, while setting the Tstep or Tprint
value seems to have little or no effect.

Paul

[Jim Thompson]

On Tue, 16 Apr 2013 12:03:02 -0700, Fred Abse
<excret...@invalid.invalid> wrote:

>I knew that, it was *minimum* timestep we were discussing.

I don't think I've ever run across a setting for _minimum_ time step,
only maximum. Setting a minimum would be asking for trouble... it
could miss fast events... as LTspice has been demonstrated to do when
operated in "fast" mode.

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I love to cook with wine. Sometimes I even put it in the food.

[josephkk]

On Tue, 16 Apr 2013 12:03:02 -0700, Fred Abse
<excret...@invalid.invalid> wrote:

>I knew that, it was *minimum* timestep we were discussing.

Oops. IIRC it is on the options card as well.

?-/