CA3140 model. Again!

Lostgallifreyan wrote in news:Xns9FE1BE34E7BC6zoodlewurdle@216.196.109.145:

First post had broken formatting of the model (web page copy), here's the copy I originally found and used. (Also, crossposted to s.e.d this time, and corrected some typos above..) Maybe this model is still borked in copying, or maybe it has fundamental flaws. Please help me find out...

• CONNECTIONS: NON-INVERTING INPUT
• | INVERTING INPUT
• | | POSITIVE POWER SUPPLY
• | | | NEGATIVE POWER SUPPLY
• | | | | OUTPUT
• | | | | |

.subckt CA3140 1 36 15 20 12

• +IN -IN OUT +VSS -VSS

• SPECTRE: + IMAX=1000 DD2 24 16 MD2 G3 31 6 16 31 -1.07249255E-06 R2 31 6 100000 C2 11 6 1.2E-11 RP1 31 20 3750 RP2 31 12 3750

• SPECTRE: + IMAX=1000 DD4 9 11 MD4 .MODEL MD4 D IS=10.0F XTI=1.0P N= 8.028126E-01
• SPECTRE: + IMAX=1000 EU6 9 31 2 31 1 RO1 11 26 20 FF1 31 28 VFF1 1 VFF1 26 2 0.0 FF2 31 20 VFF2 -1 VFF2 33 31 0.0 FF3 12 31 VFF3 -1 VFF3 31 27 0.0 DD8 27 28 MID DD7 28 33 MID .MODEL MID D XTI=1.000000F N=1 IS=10.000000F
• SPECTRE: + IMAX=1000 VP 20 22 -29.2613 VM 21 12 -29.3113 DD5 25 22 MID DD6 21 17 MID VP1 20 30 2.7501 VM1 32 12 .859456 DD9 2 30 MD9 DD10 32 15 MD9 .MODEL MD9 D XTI=1.000000F N=1 IS=10.000000F
• SPECTRE: + IMAX=1000 HH1 25 2 POLY(2) VIC2 VIC1 0 1960 0 1540 0 0 0 0 0 0 HH2 2 17 POLY(2) VIC3 VIC1 0 -1960 0 -1540 0 0 0 0 0 0 VIC1 37 3 0.0 VIC2 2 14 0.0 VIC3 14 15 0.0 VPP 37 0 1 RPP 3 0 100.0K RO3 15 20 200.0MEG RO2 15 12 200.0MEG .ends

Lostgallifreyan wrote in news:Xns9FE1C0DDF59A8zoodlewurdle@216.196.109.145:

Note that line wrapped; all others are intact in that post.

Lostgallifreyan wrote in news:Xns9FE1C0DDF59A8zoodlewurdle@216.196.109.145:

Please help. I'm sure plenty of people here know Spice, and use the CA3240, and would have no trouble trying this subcircuit (below) as a quick substitution just to see if it works.

The description pin order does not match with the

• CONNECTIONS: NON-INVERTING INPUT

• | INVERTING INPUT
• | | POSITIVE POWER SUPPLY
• | | | NEGATIVE POWER SUPPLY
• | | | | OUTPUT
• | | | | | .subckt CA3140 1 36 15 20 12

• +IN -IN OUT +VSS -VSS

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Amazing. Thanks. :) I never noticed that because I assumed the post I found it in was right because all the other amp models use the first order stated in comments, not the second (so I didn't notice the conflict at all..)

I corrected that and saw traces graphed, at last... I still get some errors though, for example: (in LTspice) Analysis: Time step too small: time = 0.00138, timestep = 1.25001e-018: trouble with node "n010"

Things like this are why I didn't notice the obvious error, I was confronted with many more subtle ones like this one which persist, there really does seem to be a flaw in the model even after the pin correction.

Node n010 in my circuit is the noninverting pin on a differential amp, with 3K3 in, and 680R to ground, nothing unusual there, and exactly what LT1215 wants for good fast and accurate performance. No other amp model I tried has trouble with it, just this one...

I tried changing the resistors to ten times the low values LT1215 likes, to something more usual for CA3240, and removing the compensation caps that LT1215 needs and the CA3240 doesn't, and I get a different error: Analysis: Time step too small: time = 0.001368, timestep = 1.25001e-018: trouble with U2:mid-instance d:u2:d6

Sorry if this is not helpful info, I'm doing the best I can short of posting a huge post with the entire context, and even from this is should show that there is something wrong with model internally, beyond the original pin order error, and where one major flaw exists, there may well be others too. Please can you tell me if this model is in any way redeemable. The CA3240 is a fine and venerable old IC and I'd really like to see a good working model for it.

Looking at the Harris and Intersil datasheets for CA3140 I see an identical schematic. I don't know much about Spice notations in subcircuits, but when I did this in the past for some other amp (possibly LF412 or more likely LM317) I saw a close match with transistor and resistor numbers that helped me to understand that I was at least looking at a real model of a real device. :)

In this case, things are not so clear! To start with, both input pins should connect only to an FET gate and a diode, yet the model appears to show at least 4 internal connections to the inverting input pin. Is it a fine model of exacting conditions beyond anythign the datasheet sdescribes, or is it just a broken mess? I really have no way to know without help, because no matter what I do I never see it working so I can't learn from breaking it.

news: snipped-for-privacy@4ax.com:

in

comments,

I caught it because I often change subcircuit node order to match my home-made symbols.

though,

with node "n010"

with

in,

good

just

something

the

trouble with U2:mid-instance d:u2:d6

huge

me

IC

Back up to some simple circuit configuration and make sure that it behaves as an OpAmp... verifying there isn't some other node out of proper order.

I'm not familiar with that CA3140 model, but it seems to have a lot of behavioral components, so add some capacitance to slow your external nodes, then see if it converges. ...Jim Thompson

Further perusing that subcircuit, it seems I was the original poster of that netlist, having found it in some Analog Workbench materials.

But I've never used it myself.

Looking it over... it's crap of the finest level :-)

Behavioral modeling to an extreme.

(Though that "Dynamic Current Sink" looks like a disaster waiting to happen.)

Amusingly the schematic bears a strong resemblance to my Master's Thesis of 1968, except I used JFET's, the only kind I could make back then :-)

I'll see if I can't write a better subcircuit. ...Jim Thompson

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

I don't think it does. :) Try this, a simple voltage follower, the kind often used to generate a split rail from a single supply. I used 10K resistors to halve a 15V single-rail supply, and to add some capacitance to slow the transition on the +input I added a 10µF cap across the ground-side resistor.

If I use an LF412 model I get the exact result expected, as with LT1215. If I omit the 'startup' part of a .tran directive I see the 300 ms or so of cap charging curve omitted from the graph, as expected. In either case with either model, the output is a clean 7.5V.

If I do this with the CA3140 model, it works if the startup bit is added, but if not, it oscillates at around 7.5V. The datasheet suggests a 3K9 resistor for feedback instead of the direct link used for voltage followers, but if I add that the output, while not oscillating, is only 2V! I know that a REAL CA3140 doesn't behave this way as a voltage follower because I've done it.

it gets weirder... Keeping the single 15VDC supply, add a new 5V supply to feed the 10K+10K divider. Output should be 2.5V, with or without the cap charge curve depending on use (or not) of 'startup' in the .tran directive. It isn't. With 'startup' the output is a millivolt below 15V (that amp cannot swing that far!), and there is no sign of the cap charge curve at the start. Without 'startup, it oscillates at about 125 KHz between 45.5 kilovolts and

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Well, thanks for looking at it seriously now. Better late than never. I knew it couldn't just be me it did strange things to. What amazes me that in TEN YEARS no-one else seems to have noticed and posted about it anywhere. It's all over the net now, I'd have thought someone might have mentioned it, but no...

That will be awesome. I think all who love the CA3140 will welcome it. :) See my other post, it shows some very strange stuff...

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Here's something I can do that may help... It's as exact a rendering of the CA3140 schematic as I can make from the Intersil datasheet. Resistors have correct values, and numbering of all parts is exact (though the MOSFETS get an M where bipolars get a Q). I have repeatedly double-checked the wiring so the nodes should be right (though arbitrarily numbered in the netlist). I made (far simpler) working subcircuits for SPST and SPDT switches yesterday (things strangely lacking from LTspice as supplied!) based on netlists, and applied my methods to this one, below, so it should be intact. I don't know nearly enough to get the modelling right for the CA3140 diodes, bipolar and FET transistors, and what appear to be Schottky diodes, let alone add anything else vital like temperature related behaviour, but this starting framework might save you some work if you're willing to add the vital details. It likely also needs a reduction to 5 pins from 8 to fit a standard op-amp model.

• 8, Strb -------------------------|
• 7, V+ -----------------------| |
• 6, Out --------------------| | |
• 5, null -----------------| | | |
• 4, Gnd --------------| | | | |
• 3, +In -----------| | | | | |
• 2, -In --------| | | | | | |
• 1, null -----| | | | | | | |
• | | | | | | | | .SUBCKT CA3140 31 21 23 15 30 16 1 12 D1 N001 N002 Diode Q1 N005 N002 N001 0 PNP Q2 N007 N002 N001 0 PNP Q3 N006 N002 N001 0 PNP Q4 N012 N005 N006 0 PNP Q5 N024 N005 N007 0 PNP Q6 N011 N005 N002 0 PNP Q7 N005 N011 N017 0 NPN R1 N017 N015 8K M8 N020 N015 N011 N011 PMOS D2 N020 N015 Diode D3 N022 N021 Schottky D4 N022 N023 Schottky D5 N022 N024 Schottky M9 N025 N021 N024 N024 PMOS M10 N026 N023 N024 N024 PMOS R2 N025 N029 500R R3 N026 N027 500R Q11 N029 N025 N030 0 NPN Q12 N027 N025 N031 0 NPN R4 N030 N015 500R R5 N031 N015 500R Q13 N012 N027 N015 0 NPN C1 N027 N012 12pF Q14 N018 N020 N032 0 NPN Q15 N016 N020 N015 0 NPN Q16 N016 N028 N015 0 NPN D6 N028 N033 Diode R6 N032 N015 50R R7 N033 N015 30R Q17 N001 N012 N018 0 NPN R8 N018 N019 1K Q18 N013 N019 N016 0 NPN Q19 N006 N009 N013 0 NPN Q20 N001 N003 N008 0 NPN R9 N004 N010 50R R10 N009 N010 1K R11 N010 N013 20R R12 N008 N014 12K R13 N001 N003 5K R14 P001 N015 20K M21 N028 N016 N014 N014 PMOS D7 N001 N004 Diode D8 P001 N003 Schottky

Wow! That's extraordinarily helpful. Now all I have to do is search my libraries for appropriate models. (I believe the diodes you call Schottky are actually zeners... see D8 for instance... it's reverse biased at all times.) ...Jim Thompson

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Now I feel better. :) I understand a need to do what I can for myself. I just wish I'd known that makign a SUBCKT from a netlist was this easy, just takes time and care. But the real core detail is beyond me, this is where I really do need help.

I'll see if I can figure out the reduction needed for the 5-pin amp model but there seems to be a conflict of interest because some might like that offset network modelled in full. Working out what to include was easy, workign out what can safely be omitted might be more than I should try to do. This model needs to help all those who encountered that other one. :)

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Ooops. :) So true, I don't know why I did that, my last circuit had enough zeners. I think I saw the Schottky symbol in a moment of uncertainty. Also, no zener voltages are specified, so that's another detail I don't know enough to solve here.

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

I tried something of my own, so far with no luck... Googling for 'PMOS Bipolar process', and 'BiMOS Operational Amplifier', hoping to find some IC that also had a good working Spice subckt that I could raid models from. I learned that a CA3260 exists, but could not find a model for it. NTE7144 may be another source of appropriate models IF there's a subckt for it, but again, I can't find one. I'm fairly sure I don't know enough to decide if it's usable even if I do.

Incidentally, that old CA3140 model appears to have no transistor models, and a very wrong diode model count. I'm no judge of these things, but when I look at it I can imagine how Mulder or Scully feel when confronted with a human form that has no clearly discernable anatomy.

I found this:

...but I have no way to get a look at it.

I do all my searches -IEEE :-)

They're invariably worthless anyway. ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | 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

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Ok :) I figure that if they need to hide them that much, maybe they are like the Emperor's New Clothes, and they don't want too many people to notice.

How about these?

news: snipped-for-privacy@4ax.com:

I found some high-voltage MOS models dating back to the dark ages, and am now trying to remember how they work ;-) (They're subcircuits that include the parasitics found in BiCMOS.) ...Jim Thompson