1. Attachments are working again! Check out this thread for more details and to report any other bugs.

2002 power steering shudder fix

Discussion in 'Generation 1 Prius Discussion' started by Behrens, Jan 31, 2015.

  1. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    You are at the cutting edge because you have a test article. Be careful about higher voltages as the heat, the power, increases by the square of the voltage:

    watts = (volts * volts) / resistance​

    To burn out tin whiskers typically takes small voltages. But larger voltages, a quick impulse, can vaporize the debris without burning up the part. Two alternatives come to mind: small Tesla coil and a charged cap. But we're getting into the fun stuff I like to play with. But a resistor in series to limit the current would also work and be a lot simpler.

    Also I'd like to suggest is measuring the resistance of the part and see if the variability Chapman reported shows up. This would give a way to measure or quantify the 'noise.'

    Bob Wilson
     
  2. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    I'm also intrigued by the audio speaker idea Bob mentioned earlier. I've never tried it myself (the dealer kept my old bad rack, so no playing for me).

    -Chap
     
  3. Behrens

    Behrens New Member

    Joined:
    Jan 24, 2015
    15
    0
    0
    Vehicle:
    2002 Prius
    Model:
    One
    Hi Ho, several days after application of the 18vdc pot sweep, it appears there was some improvement.No PS fails but still some shudder at the center.I soldered wires on VT1,VT2 and both gnds and attached various capacitors as a filter between each vt & gnd without success.Any suggestions in a circuit to cleanup the signal? Rich
     
  4. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    You need a low-pass filter with a long time constant, 0.05-0.1 seconds, but we don't have enough details to specify the hard requirements. So let me toss some ideas:
    1. PI filter - cap_1->choke->cap_2, measure the VT1 and VT2 impedance for the R of RC for the last cap, cap_2. Then use cap_1 and choke to impose the input pass bandwidth. But there are a lot of unknowns which makes this an exercise in empirical engineering.
      • In theory, the input signal could go through a linear amplifier to change the signal impedance. With more power/current across a greater voltage range, the output PI filter would be more effective and a simple divider network return the signal to values used by the steering ECU.
    2. De-bounce circuit - the noise is similar to a noisy switch. I've not studied or worked with de-bounce circuits and they usually are designed for ON-OFF but there may be something out there for noisy pot wipers.
    3. MSP430 (or equivalent) - digital, software filter, feed VT1 and VT2 to A-to-D converters and read the values. Then using software filtering logic, to drive output transistors to specific values with a maximum rate of change. The advantage of a software filter is you can tune the response. The electronics are not trivial but could be fun: NHW11 Prius Temperature Hack
      • The 'glue' logic would be significantly different. The interface to VT1 and VT2 would be designed to be a passive, mid-scale trim so if the microprocessor is inert, there would be no signal to VT1/VT2. I suspect the output voltage swing is such that the transistor ON voltage, ~0.6V, won't be a problem. Very advanced, use a digital pot but this would require something other than a perf-board prototype.
      • The input side would feed to A-to-D pretty much raw, running as fast as possible. Software on the input side can detect excessive dV/dt and toss out or down-filter those values. The output side can also have rate limits.
    The nice thing about the digital solution is further degradations have little to no effect. In contrast the analog, filter approach is effected by input side effects.

    Bob Wilson
     
    #24 bwilson4web, Mar 27, 2015
    Last edited: Mar 27, 2015
  5. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV


    If your pot sweep several days ago made some (what sounds like significant) improvement except for some shudder at mid-range, maybe before giving up it would be worth just doing some more.

    I'd suggest since you found 9 V to be not very effective, and 18 V to have some effect, maybe you could safely explore the territory in between to see if you find a voltage that is still effective but a bit more restrained than 18 V. Also, rather than sweeping the sensor across the whole range, you could concentrate on just the smaller region near center, as that seems to be where the trouble is (it is where the pot wipers spend most of their time).

    Remember that the sensor we're interested is a torque sensor, not an angle sensor; exercising it over its range doesn't require moving the tires on the ground through any angle at all, but just leaning the wheel to the left and right against the resistance of the steering.

    If you do get it satisfactorily cleaned up, there was a post I put up years ago (after too much chips y salsa) concerning a theoretically possible but untested approach of rearranging the pins at the connector in order to put a more constant wetting current across the pots in the hope of keeping the contact area cleaner long term. However, (a) I never tried it, and (b) it was based on assumptions I was making about the actual resistances down there, that don't seem to be consistent with what I've actually more recently measured. So I certainly can't guarantee anything about the idea (even that it won't tear the wheel out of your hands and drive you into a tree). Maybe if I get around to solving the system represented by my most recent resistance measurements (or someone beats me to it)....

    -Chap
     
  6. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    Wire them in parallel?

    I like your suggestions about repeating the 'burn out' test but with instrumentation. If possible, configure a VOM in amp mode to monitor the current flow while applying torque. Alternatively, use a coil-speaker (i.e., 8 ohm or less) in series and listen for the noise. When you hear it, do multiple sweeps to see if the level of noise decreases.

    Bob Wilson
     
    #26 bwilson4web, Mar 28, 2015
    Last edited: Mar 28, 2015
  7. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    No, because VT1 and VT2 still have to be complementary signals, but (on the assumption that the two pots down there are matched - VT1 to VT2 - and symmetrical - Vcc to ground) there was a simple permutation of the pins in the connector shell that looked promising, as in my original post.

    The fly in the ointment is, based on my later resistance measurements on a known good rack, I'm not sure the assumption holds, because I definitely don't see symmetrical resistances Vcc to VT{1,2} and VT{1,2} to ground.

    It ohms out as a voltage divider (if that's indeed all it is) that would provide > 2.5 V at neutral input, but when plugged into the ECU the neutral input seems to be right at 2.5 V where you'd expect. Perhaps this means the ECU is already built with a low enough input impedance to pull the signal down, and the pots down there are built to compensate. In that case my simple permutation would not produce an equivalent network. (It might also be moot, if they made the ECU input impedances that low precisely for the purpose of drawing a wetting current, which was what I was setting out to do.)

    -Chap
     
  8. Behrens

    Behrens New Member

    Joined:
    Jan 24, 2015
    15
    0
    0
    Vehicle:
    2002 Prius
    Model:
    One
    Hi Ho, I just completed a longer drive test and had no PS lockouts until the end of a 150mi trip.Then lockup after going over RR tracks with no wheel movement.Thanks for the reminder that this is a torque sensor and the wheels don't need to move.I'll redo the volt sweep and let you know the results.There is a debounce ckt from ELM the 401might work at Elm Electronics - ICs for the Electronics Bench do you think? Rich
     
  9. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    I can see how the phrase "rotary encoder" could have piqued your interest, but what they mean there is a digital rotary encoder, the kind of thing you see pictures of if you google for "Gray code".

    The torque sensor in your steering rack, though, is an analog device.

    -Chap
     
  10. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus


    I went back and re-read the previous thread:
    I didn't see where we verified that the wipers move in opposite directions under torque. The wipers could move in parallel or complimentary. This is what would help us find out:
    • Connect GND and Vcc to a 5-9V source
    • Connect VOM so its BLACK is on one wiper, VT1, and RED is on the other wiper, VT2
    • Apply a torque and see if signed voltage, proportional to the steering input shows up
      • Since there is noise, hold the torque constant and see if the voltage decays to zero
    • Move VOM to GND
    • Apply a constant torque (i.e., hang a weight or bungee to one side of steering wheel)
      • Measure each wiper voltage and report
    • Reverse constant torque (i.e., hang weight on other side)
      • Measure each wiper voltage and report
    This will clearly verify the wipers move in opposite directions. We had assumed this would be the case to maximize the voltage difference proportional to the torque. However, this is not required if two, A-to-D converters are used. If a single, A-to-D is used with floating references, it would require opposite signed input. However, I've fiddled with them and it would be a lot easier to use two, A-to-D converters, one for each VT1 and VT2, relative to ground. The software can handle the maths.

    Looking for the voltage wander independently, VT1 and VT2, would help us learn if only one of the two torque sensors has the noise problem. I would expect one to be having a greater problem than the other but similar magnitude. I would be surprised if one is 'clean' and only the other one is 'dirty.'

    I am working up a better resistance model. More later,
    Bob Wilson
     
  11. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    Hey Bob, did you glance again at my chart recorder strips?

    Also, there is a drawing of the sensor in the NCFM.

    I guess what "opposite directions" means could be subject to some unavoidable language ambiguity. They are both attached to a shaft that's only turning one way at a time ... but they're symmetrically placed.

    [and don't forget it was a totally retro, actual chart recorder, with a 6 mm phase offset between pens so they don't collide. :)]

    -Chap
     
  12. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    NCFM page 126 is where the drawing of the sensor is, and page 127 has the graph of VT1 and VT2 voltage showing them with positive and negative slopes. VT1 is the one that increases with driver torque to the left, VT2 increases with driver torque to the right, as I had described here once upon a time.

    -Chap
     
  13. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    Ahhhh! New Car Features ... I never got one for our NHW11.

    Any possibility of posting their sketch?

    BTW, I was able to make a functional resistance network that handled all four, VT1 and VT2 values to Vcc and GND. The values suggested the same 2k pots with the VT1 and VT2 offset. But two of the resistances did not work in a ugly way. But this is a solvable problem.

    I have six equations and six unknown resistances and they are all first order, linear. A big sheet of paper and sharp pencil and I should be able to get good values for:
    • VT1-to-GND common
    • VT1-to-Vcc common
    • VT2-to-GND common
    • VT2-to-Vcc common
    • GND common-to-GND external (*)
    • Vcc common-to-Vcc external (*)
    The "*" resistances came out badly, one having to be a negative resistance.

    I was hoping they would have used an encoder like the accelerator pedal and throttle position encoder. But these are isolated, independent variable pots that do not share a common Vcc and GND. Now the wiring harness could have tied these together but we still have a problem of some other fixed resistances. Hummmm, this gives me a horrible thought.

    If they are two, independent pots tied together, there could be two, fixed resistors on both the Vcc and GND ends of the fixed resistances. That would be an ugly circuit to model ... very ugly.

    I'll try to post a sketch of what I was thinking about tonight.

    Thanks,
    Bob Wilson
     
  14. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    Ok,

    Understand this is a 'work in progress.' I don't have all of the answers but let me share what I've done.

    RESISTOR MODEL

    pairs resistance calculated R R(name)
    1 VT1-VT2 1231 1070.9 R(GND)
    2 VT1-GND 1421 1422.7 2310 R(VT1-GND)
    3 VT2-GND 1423 1422.7 2310 R(VT2-GND)
    4 VT1-Vcc 615 616.2 697 R(VT1-Vcc)
    5 VT2-Vcc 617 616.2 697 R(VT2-Vcc)
    6 Vcc-GND 807 1503.5 R(Vcc)

    • pairs - the signals Chapman measured
    • resistance - the values found
    • calculated - the network calculated resistance values for two potentiometers (pots) tied together at the ends with independent VT1 and VT2 sweeps
    • R - resistances that would generated the calculated values
    • R(name) - same as Chapman's signals for expected values
    The good news is some standard resistances, ~2300 and ~700 ohms, gives values that match Chapman's measured value. But this model assumed no resistance between the network and Vcc and GND. When those resistances are calculated, the differences are too great.

    ACCELERATOR ENCODER

    There is a fairly common part that measures angles in the accelerator and throttle position:
    [​IMG]
    The encoder has a nylon bushing that rotates the shaft through 90 degrees. BTW, this one came 'bent' when I was rebuilding accelerators.

    Opening one up, we find:
    [​IMG]
    The disk rotates with two, pairs of brushes. On the right are six contacts for the resistance traces:
    [​IMG]
    So the outer traces are the traditional, potentiometer structure. The outer pads connect to Vcc and GND and the wiper sweeps an angle. So there is a resistance from between Vcc and GND. But what I forgot was the inner arcs that pickup the brush signals VT1 and VT2 and vary too. So what is missing from my model:
    • all of the pad to arc trace resistances
    • inner arc trace resistance
    Here are the typical resistance values I've measured with accelerator encoders:
    [​IMG]
    I used a test fixture with 21 rotations of a turnbuckle to go stop-to-stop so a half turn is the ".5" change. So these are the resistance values measured from one side of the pots to the variable trace:
    [​IMG]

    Upon reflection, I may be able to replicate Chapman's measurements with an existing, spare accelerator. But two of his measurements are a challenge. We can reduce resistance by putting them in parallel. Just I've not figured out how to get there from here, yet.

    CLEANING OR TIN WHISKER BURNOUT

    This is the difference when an encoder is mechanically cleaned:
    [​IMG]
    Here is another, more dramatic cleaning result:
    [​IMG]
    But this was when I was doing a mechanical cleaning.

    Burning out the tin-whiskers is a new approach inspired by work at NASA Goddard Space Flight Center:
    • does not require disassembly
    • no one has studied what happens after burnout (whiskers may grow back)
    • tin whiskers that have not made contact will continue to grow
    If I were doing tin whisker burn out, I would use a large choke in series and rock the encoder back and forth. The upon reaching a discontinuity (i.e., a tin whisker), the electomotive 'kick back' would send a high-voltage pulse to more than vaporize the tin whisker. GSFC pointed out that older VOM ohm meters had enough to often burn out tin whiskers, making the devilishly hard to detect.
    INITIAL THOUGHTS

    When I was testing the accelerator encoders, I wanted parallel resistance traces. The difference would be the resistance difference Vcc-to-arc and arc-to-GND. This fixed value would be ~500 ohms.

    If the VT1-VT2 and Vcc-GND values were reversed, the circuit would be trivial:
    1. Big arc 2.5-3.0k ohm - with 2.5k, in parallel there would be 1.25k resistance for Vcc-GND
    2. Small arc series 800 ohm - would make sense as the outer arcs would be in parallel to VT1-VT2
    I now see how to refine my model to reflect the internal configuration.

    I really need to know if the wheel is turned right, do VT1 and VT2 both increase or does one increase and the other decrease. This tells me exactly how the Vcc and GND are connected to the encoder.

    Bob Wilson
     
    #34 bwilson4web, Apr 3, 2015
    Last edited: Apr 3, 2015
  15. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    Hi Bob,

    Do you need to know that badly enough to read post 32? :)

    -Chap
     
  16. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    Bob,

    For your topmost calculated value VT1-VT2, I'm getting (697+697)∥(2310+2310) ≈ 1070.9 rather than 1503.5, did I do that right? I get the same values for the rest.

    Did you happen to also compare the values to the second set of measurements I took, the one where each measurement isolates exactly two of the (four most obvious) unknowns? (I haven't yet, just asking.) I took that second set because I was making algebra-steam trying to get a linear system out of the first set, but the second set made it easier.

    I also thought about modeling additional unknowns, such as resistances between Vcc and network Vcc, ⏚ and network ⏚ (but I would assume negligible values for both, probably soldered), or VT1 and VT2 wiper resistances (possibly non-negligible in an old rack but I'd hope small in a new one).

    It occurred to me that for modeling just the four basic resistances, with six equations I'd have two df and be able to put (really conservative) confidence intervals around the inferred values, but as soon as I was trying to model six resistances I'd be out of df and could only get a bare least squares solution with no confidence information. But if you have made six linear equations out of my first measurements, heck now we've got twelve and possibilities open up....

    -Chap
     
  17. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    One thing I remember from my earlier disassembly of a heat damper servo is that the printed traces can have regions that look the same to the eye but have different properties:
    [​IMG]
    In the case of the servo gear, there is only one region of one trace (between my two arrows) that has noticeable resistivity and therefore a voltage gradient. Everywhere else, though the traces look the same, they are highly conductive and equipotential (either with the inner ring or the middle ring, the two rails of the supply). If you look near my arrow on the left, there is a kind of barely visible score mark across the outer trace that indicates where the resistive area starts. That really was visible on the part. There was another one visible at the other end on the real part, but in this compressed JPEG I saved I can't see it any more (and I'm not taking the car apart again just for a better shot), so I put a dashed arrow there ... it's somewhere in the right neighborhood.

    With all that in mind, Toyota's drawing of the torque-sensor element does Escher things to my eyes if I look at it too long, but I've added a colored version representing my best guess at what's going on, with the middle and right connection point being the two supply rails and the left one being the torque signal (and the left trace being simply conductive and equipotential).
    [​IMG]

    For what it's worth, the resistor body, with the traces, is carried on the upper portion of the torque sensor, connected to the driver, facing down; the wipers are on the lower portion, connected to the road, facing up.

    -Chap
     
  18. ChapmanF

    ChapmanF Senior Member

    Joined:
    Mar 30, 2008
    23,069
    14,974
    0
    Location:
    Indiana, USA
    Vehicle:
    2010 Prius
    Model:
    IV
    Hmm, the more I look at their drawing, the more it bothers me, as they have drawn "resistor 1" and "resistor 2" oriented the same way along a circular path, instead of in a mirrored orientation as I would have pictured. That would mean that driver leftward steering effort brings both wipers nearer the contact-takeoff ends of their traces, and rightward steering effort takes them both farther from that end.

    There is no doubt that the output signals have opposite slopes: it says so on page 127, and that's exactly what my chart recorder showed. VT1 increases with driver leftward effort, VT2 increases with driver rightward effort. If the drawing is correct, that would imply that "resistor 1" must be wired with Vcc at the pad I colored blue and ⏚ at the pad I colored red, while "resistor 2" would have to have those connections reversed. And because we are not seeing resistances that are symmetrical "above" and "below" the wipers, that would tend to mean that resistor 1 and resistor 2 are not just two copies of the same animal.

    Curiouser and curiouser....

    -Chap
     
  19. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    It turns out that a single resistor makes the circuit:
    pairs resistance calculated R R(name)
    1 VT1-VT2 1231 1070.9 R(GND)
    2 VT1-GND 1421 1422.7 2310 R(VT1-GND)
    3 VT2-GND 1423 1422.7 2310 R(VT2-GND)
    4 VT1-Vcc 615 616.2 697 R(VT1-Vcc)
    5 VT2-Vcc 617 616.2 697 R(VT2-Vcc)
    6 Vcc-GND 807 808.7 1750 R(Vcc-GND)
    7 1503.5 R(Vcc)

    • R(Vcc) - calculated resistance Vcc-GND without external shunt resistor
    • R(Vcc-GND) - a resistor applied outside the dual-pots, gives the observed 807 ohms
    • VT1-VT2 error, 1231-1070.9 ~= 160 ohms - the trace VT1 and VT2 pickup spans, ~80 ohms each
    Now if we can find a SPICE expert to model it. <GRINS>
    [​IMG]
    The "80Ω" was added to resolve a 160 ohm problem with VT1-VT2 resistance calculation. We may need to adjust the 697 and 2310 ohm values to make a more accurate model. But we're getting close to 'the weeds.' We're rapidly approaching the area where SPICE could resolve the 'fine detail.'

    I can only speculate why the 1750 ohm resistor might play a role. It balances the resistance network but the function remains ... unknown. Perhaps it improves stability of the Vcc-GND voltage by keeping the net load on the power supply constant. Variable resistances could change the output voltages in bad ways. It might also reduce sensitivity to electrical noise. Regardless, without it, this simple model does not work. BTW, the 1750 ohm resistor could be two 3500 ohm resistors, one per encoder.

    Now if a single ADC is used between VT1 and VT2, the direction of the wipers is critical. But if two ADCs are used, one for VT1-GND and VT2-GND, the software can easily calculate the offset.

    Bob Wilson
     
    #39 bwilson4web, Apr 3, 2015
    Last edited: Apr 4, 2015
  20. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,067
    15,372
    0
    Location:
    Huntsville AL
    Vehicle:
    2017 Prius Prime
    Model:
    Prime Plus
    Ok, I picked up MacSpice and learned enough to get some useful data. There are other SPICE equivalents. But I am not a SPICE expert so it may take a little while longer to get it all figured out. My first example:

    - - - - - - - - - - - beginning of model - - - - - - - - -
    Resistor Network model_020.CIR

    * This models the Prius, steering encoders

    **** CIRCUIT TOPOLOGY DEFINITION SECTION

    * Rvt1 is an 80 ohm resistor between the VT1 external point and the internal
    * pickup brushes of the encoder on one side.
    Rvt1 1 2 80
    * node 2 are the pickup brushes
    * node 7 is the floating GND
    Rvt1_gnd 2 7 2310
    * node 3 is Vcc
    Rvt1_vcc 2 3 697

    Rvt2 4 5 80
    * node 5 are the pickup brushes
    * node 7 is the floating GND
    Rvt2_gnd 5 7 2310
    * node 3 is Vcc
    Rvt2_vcc 5 3 697

    * bias node to bring Vcc and GND resistance in sync
    Rvcc_gnd 3 0 1750

    * Spice converts all entities in this section to lower-case
    * so that the above line is the same as "r1 1 2 1k".

    * The ground node must always be given the identifier 0 (zero)
    * Every spice circuit must have at least one connection to node 0
    * because otherwise it is impossible to calculate the voltages.

    * VDD is the voltage source called DD. The positive node of a
    * voltage source is listed first (i.e. 1 in this case).
    * The next parameters specify that the source is DC and produces
    * 10V, but note that SPICE ignores the V, it is a comment
    * intended to make the file easier to understand.

    VDD 6 0 DC 10000V
    * 5V is assumed to be the operating voltage range
    * 10000V is a DUMB ohm meter, resistance = 10000V / current

    * The following are power supply probes so we can
    * see the current between any two points. A zero resistance
    * is converted to 1000 ohms but a tiny resistance is left alone.

    Rvcc 6 3 0.000001
    Rgnd 7 0 0.000001

    **** CONTROL STATEMENTS SECTION

    .CONTROL

    * All commands in this section *are* case-sensitive. This means
    * that "print r1" is *not* equivalent to "print R1". It was therefore
    * unwise of me to use upper case element names above.

    echo "There are no commands in this file..."
    echo " ...except the ones used to print this message."
    .ENDC

    **** OUTPUT STATEMENTS

    * The last line in the file must always be:
    .END
    - - - - - - - - - - - end of model - - - - - - - - -

    So I loaded this model into MacSpice and got:

    MacSpice 1 -> source /Users/bobwilson/steering/model_020.CIR

    Circuit: Resistor Network model_020.CIR

    There are no commands in this file...
    ...except the ones used to print this message.
    MacSpice 2 -> op
    MacSpice 3 -> print all
    v(1) = 7.68208e+03
    v(2) = 7.68208e+03
    v(3) = 1.00000e+04
    v(4) = 7.68208e+03
    v(5) = 7.68208e+03
    v(6) = 1.00000e+04
    v(7) = 6.65115e-06
    vdd#branch = -1.23654e+01
    MacSpice 4 ->​

    So if we take 10000V / 12.3654A = 808.7 ohms.

    I need to learn more about SPICE to see if there is a way to automate evaluation of the resistance between points. Then some way to switch the power supply between the different, external contacts. A little massaging of the resistor values and it will quickly resolve to a final resistance network that matches the measured resistances. <SIGH>

    I wasn't planning to become a SPICE expert <GRRRRR>.

    Bob Wilson
     
    #40 bwilson4web, Apr 4, 2015
    Last edited: Apr 4, 2015