Using SPICE modelling of designs - call for info...

[Nicko]

Hi,

I'm in the middle of designing my first serious PI detector. As I work mainly with SMD components and DSPs, thats the route I'm taking...

I've been modelling the front end with SwitcherCAD III from Linear Technology (an *excellent* *FREE* windows SPICE package from a major manufacturer - see below). The design is currently based on a high-side N HEXFET rather than the normal low-side NFET or high-side PFET approach - this is mainly due to the very cheap (c. $1) cost of high-side MOSFET drivers that can fully enhance NFETs from very low battery voltages. I know that these sorts of components are only available as SMDs, but they are very simple and cheap to use. They also allow you to do nice things like current-limit the coil driver, so that you protect the driving FETs etc. against being destroyed by little "accidents" ;-)

There are various ultra-low-noise opamps available from the likes of AD, and I'm trialing those at the moment - you can't "breadboard" low-noise SMD designs very easily, so I'm having to produce PCBs for each iteration - again, modelling has enabled me to get it very close to what I want first time.

Modelling allows me to experiment quickly with the effects of different driver configurations, particularly when trying to drive the FET into full enhancement at high speed, and equally to turn it off quickly as the gate capacitance of most power FETs is very high. Tuning this can make a radical difference to the power generated and the resultant behaviour of the transmitter. Many of the designs I've looked at seem to have rather ad hoc approaches to this.

Has anyone else approached this sort of design in a quantitative way, or is most design qualitative (i.e. experimental)?

I'm assuming coils with a nominal inductance of approx. 100-120uH and a series resistance of about 1.3ohms. I would be particularly interested in hearing details of coils used in other detectors, specifically in how the shielding (if any) is carried out.

Also, are there any academic/technical papers in the public domain that anyone is aware of that compare coil configurations for PI detectors with the effect on standardised targets? I'm particularly interested in the subtle changes in inductance, again for modelling reasons.

I will make this design public, if there is interest.

Many thanks,

Nicko



SwitcherCAD III home page

[Robert Hoolko]

Re: Using SPICE modelling of designs - call for info...

Nicko

I am not doing circuit design so I am not using a SPICE program, but I definitely agree about the benefits of modeling the design before building it. I am only doing software, but I still find it beneficial to model the design in a math program (I use Mathcad) before writing it in assembly language. It is usually much easier to set up and test a model than it is to build the real thing. It can eliminate a lot of iterations from the design cycle. Or at least the iterations are done on the model rather than on the hardware. That is especially important if you have to do a new PCB for each iteration.

By the way, I liked your bat machine.

Robert

[nickf]

Re: Using SPICE modelling of designs - call for info...

One difference I've noticed in available circuits is the configuration of the bipolar transistor that turns the FET on and off. I've followed Mark Stuart's approach in my detector which is to use the transistor between the gate and negative supply with the collector resistor going to the +ve supply. In this way the gate capacitance is discharged through the transistor rather than a resistor when turning off. I've never tried the alternative though so I don't know how big a difference it makes - perhaps the SPICE model will show this so long as the model of the gate capacitance is OK.

[Nicko]

Re: Using SPICE modelling of designs - call for info...

Yup - the key is how fast can you turn the FET off. Using a real driver, the FET gate is essentially grounded through a low-impedence path leading to very rapid turn-off and a resulting high dI/dt.

ATM I'm playing with a number of FET models, including the ubiquitous IRF740/840. For some FETs I have several models (SwitcherCAD can use pretty much any PSPICE model), including those from the manufacturers of the devices. The models tend to differ in subtle areas, such as temperature coefficients, but critical parameters seem fairly consistent even though the actual physical model may be very different.

So far, my practical tests using an arbitrary function generator and a selection of power FETs with a Tek 3054 'scope seems to show that the modeling is very close to reality (withing 5%). I've added models for some Rectron transient voltage suppressors and they model quite well too - the 1.5EK440CA limits the back EMF quite nicely (FETs may be cheap, but they are a pain to replace in the field!)

Nicko

[Nicko]

Re: Using SPICE modelling of designs - call for info...

Robert,

> By the way, I liked your bat machine.

Thanks - Not quite finished yet - Its a major major project. My DSP of choice for that is the ADSP2191M, but AD have made it very power hungry, so I'm a bit miffed as I need a portable unit.

With good/excellent battery management I've managed to get 10hrs from 1800mA/H Nimh batteries, but its been very hard work indeed.

I've still got some nasty timing issues with the DSP driving the screen, and I may yet have to add a dedicated display processor... ho, hum...

Nicko

[simeon webber]

Re: Using SPICE modelling of designs - call for info...

hi

i found a usefull chip is the tc 442x
x= 0 - 9

forinstance the tc4420 is a single section fet driver
{very fast }

the others in the family are dual and very high esc and voltage breeds
by microchip
to compliment a pic when driving fast pwm
as you point out clean transitions are the key

i find if you load the pin the gnd in the arrangement in my schematic {in another post}
you can use a variable resisotor to add a top notch as a drag coificient as the voltage peeks

again
i may add these tc devices to the design i published

any views or other ideas about this ?

i would welcome for one.... interesting topic and a good read.....
thanks

example circuit http://homepage.ntlworld.com/vsm.libs/DESIGNS/water.exe

look in fet driver design sheet

{click design in the design menu in proteus lite}
you can download proteus lite nag inc at
www.labcenter.co.uk/download/prolite.exe

[simeon webber]

Re: Using SPICE modelling of designs - call for info...

just too add

if anyone has an opamp processor or other library problem with proteus {inc creating some dll's types} let me know
and ill do it and post on vsm.libs....

[simeon webber]

Re: Using SPICE modelling of designs - call for info...

first impressions of switcher cad


is it looks quite good

at selling linear tech components and supporting no others by default

one thing i learned is this

it is one thing supporting pretty much any pspice model ..{no ibis no real time models no vhdl no teccor....etc......}
it is totaly another using pspice models with any app..... accuratly

it all depends on the app..... no !

it depands sowly on what processor you use

an amd >athlone 1g has no FPU error {built into intel {and amd till now} by the american goverment}

all intel and amd after 1994 have a maths co processor error built in till amd stoped doing this {or in my opinion moved the error to another level so it is still a problem to them but not to us....

in a vien attempt to stop would be terrorists from using them for targeting missiles {read a bit youll see}


for this reason only a newbrain p60 p90 {this uses a separate fpu on controller chip..... .
nexgen controllers are intel complient but no fpu in chip

processors or any other fpu from intel have a level 5 fpu folded corner in there architecture .....



so now tell me....... about accuricy in simulations prior to 1994....
when switcher cad was first built....
then tell me what chip maker you use.... most off.....


then tell me why????? i am wrong


i am reminded that linear tech are the grubby crap basicaly custom makers of chips i used to replace heeps of in badly designed amstrad and sudo japanese {korea or china or tiawan etc....} companys parasitic plastic all i can describe as total prefabricated expanded polyurathane junk stereo hifi cb and commerical radios tv ooozing like a box in a box for 20 years of my life......


so i guess this is there way to sell heeps more.....


hum.... pll synthasis is usefull but there cad is very poor giving poor representation by way of any instruments or realtime results or charting


so..... a chip of the old block

and no real differance in it from 1984 except a work around for the error that then was built in to all apps

and isnt multithread

a good app for free quick design if you not fully aware what you need too use

looks crap when you first open it on a slow machine { i tried a 486 under win 95 it still runs full speed}


but with apps like proteus
use a fast machine and 3d card open the app to a network of servers or three machines one dual cpu set as a file server
and its a dream come true

this is what you need......... proteus V6

i will stick to spending some cash to get some flashy bits knobs whistles views and twiddly things