SPICE (or other) models for data converters

All I would be interested in would be the analog inputs.  I can't imagine doing anything with digital interface or trying to look at linearity etc. of the conversion, I'd leave that to real world testing.

The one thing I would be interested in testing is the op-amp buffer interaction with the input for certain circuits.  Maybe an extra input could be added to the spice model to trigger the selection of the analog input which would simulate the internal mux of the adc.  Especially for SAR types, this would be useful to simulate the buffer amp stability when charging the internal cap and help to select RC buffer components.

Also.. for converters with built in amps, it would be nice if that was included in the model with a pin showing the output of the internal amp.  I guess there would need to be some way to set the gain for those which are set through a digital configuration (maybe extra input pins on the model?) but it would be nice to see the signal feeding the actual ADC circuit.

Thanks - that's been my thought for some time - if we just did the analog input (and reference input), that would help with designing the external circuitry. 

Good thoughts on the other topics - PGAs, etc.

I appreciate the feedback. Look forward to hearing from some more of you folks out there!

I would absolutely love having a spice or ibis model to simulate the analog aspects of the input and output pins. It would take much out of the guesswork of so many things - termination, layout, EM simulation, signal integrity...the list goes on. Really would be fantastic. In fact I made a post on this not long ago, here: http://e2e.ti.com/support/data_converters/audio_converters/f/64/t/31731.aspx

I think it would be fine to provide only the analog section, such as the output of a current DAC.  But it would be very useful to have as much of the linearity, noise, and other performance aspects.  It would understandably be difficult to model any noise which bleeds through from the digital section if the digital section is not in the model, so that would be the really difficult part.

As for the digital part, don't people use Verilog for that?  I must confess that I don't know what most engineers use for simulation of analog plus digital designs, but the ones I've talked to use both Verilog and SPICE, although I don't know how they divided up the task or how they combine the results.

Why not provide a partial model which covers the analog section, and expand on that later if possible?  I tend to deal with the analog and digital sections of the design separately, so it would be fine for my workflow.

I use the MDAC's for programmable gains, etc and it would be very useful to have a model

allowing some kind of relative input and use the parameters at Vref, feedback and output

in my model circuit. I have tried to simulate my own with limited success and lots of variables.

Good suggestions on the MDACs - certainly those devices could be modeled to some extent to show the analog behavior, as magic don suggested.

@Kurt G - I hope you have found our IBIS models - we have a growing number of them available for TI data converters. If you haven't found them, simply search for "IBIS models" in the keyword search and you should be able to find the repository of models that is growing.

- Rick

There are two parameters that I'm worrried about at the end of the day...

1. Freqency

It would be nice to be able to include models of the A/D converters that simulate the filtering that goes on inside them.

2. Noise

How much noise is the A/D converter adding to my signal?

Steven,

So, how would you want to see this? In the analog domain, or digital? Thinking about this from a SPICE perspective, modeling the analog front end of the converter could get you the frequency stuff, and we could simulate the digital filters in an analog domain (I think). Same with the noise to some extent. The question is, would you want an "effective analog output" to use in your modeling, or would you want to get the digital bits out and then try to determine something from that? This is where it gets tricky, since the mixed-signal uses of SPICE are pretty variable across simulators and the simulation time would likely be slower.

Thoughts?

- Rick

Simulating the digital effects in analog would be great. So you have the front end analog ports, you plug in the A/D converter but instead of it spiting out a digital output, it could be analog for the output with the noise of the A/D converter and the frequency 'filtering'. Then if you so desired you could dump the data file into a programming language of your choice and simulate the digital side with not to much coding. (You could even build some kind of tool like this into TINA)  You wouldn't really need to do mixed signal stuff, emulating an A/D converter would be take too long to simulate. From an analog design perspective, you only need to worry about how this is going to modify my signal. -Steve

Oh, one more thing it would be nice to have the limits of the op amp built in, so if you tried to drive 10V into a 5V converter it would hit the 'rail'. Also it would be nice to have the differential inputs simulated correctly for the appropriate op amps

The best way to use simulation model for A/D converter is for early design of the front-end (i.e opamps) that will be used for the said A/D. And especially when there are many needed opamps around the A/D especially again for instrumentations.

An analog front and with the PGA is great but you need the frequency effects of the delta sigma converters.

Dear Friends,

            I have been looking for a 'SPICE Model'  for converters (A/D's & D/A's). As I work with MATLAB & Simulink and I try to include a SPICE models to simulatein my circuts as it takes me more closer to real world simulation. If I try to simulate components shipped with simulink is good too but they dont represent real components as most of them are Ideal.

          As far as my simulations are concerned I would say a SPICE model should containi all necessary parameters required for a basic A/D's & D/A's and unfortunately 'Noise' parameter is not possible to include in my simulations as my simulation will completely ignore this parameter  automatically when I convert a 'SPICE model file' into a simulink block.

        I am eager to know when can I expect a SPICE model for A/D' & D/A's to be available from Texas Instrument?

Any suggestions will be appriciated.

With best regards :)

Well, you just taught me something - I had no idea that SPICE modes could be pulled into a Matlab environment (I don't use Matlab or Simulink). Very interesting. That opens up a whole different set of things to think about.

As for when we might provide SPICE models for data converters, I'd have to say we really don't know - as you can tell by this thread, I'm exploring what we might want to do. My initial thoughts on this have been mostly confirmed by the posters here - that what we should start with, at the very least, is a model of the analog input and the reference input, to help people design the external analog circuitry. We actually have done a little work in this area, but it's far from ready for release.

Adding on to this the idea of actually modeling more of the frequency or noise responses, all in an analog mode, is intriguing to me. Not sure just how difficult that would be, but it's something I'd like to explore.

To be clear, I'm only talking here about precision data converters. I know that folks that use the higher speed converters for communication applications often do simulate in Matlab and such, and perhaps the high-speed data converter group is doing or considering something else - I can't speak for them (but I am going to ask them to see what they're doing in this area).

To all who have contributed on this thread - THANKS. You've been a big help. I'll let you know when plans firm up.

Hi Rick,

I know this thread has grown a bit stale, but I just found it searching and wanted to chime in my interest as well.

Right now we're working on a new design, part of which uses R-2R MDACs (the DAC8801 / DAC8803 specifically) in a digitally-controlled state variable filter.  It would be hugely helpful to have a SPICE model for the analog behavior, from VREF in to IOUT1 & IOUT2 outputs.  The DAC word can be set as a parameter, or as separate bit inputs, or whatever.  What we need to model is stuff like code-dependent output capacitance and code-dependent multiplying bandwidth (think of a model + test circuit that can reproduce Figure 8: Reference Bandwidth from the DAC8801 datasheet).

We're seriously considering slogging through trying to construct our own SPICE model to do this in-house, but it be a lot better (and probably higher fidelity) if you folks had one available.

Thanks in advance,

--Matt

Hi Matt,

You are the first I have seen that has requested a SPICE model for our R-2R MDACs. It is funny that you bring this up as it is something I have been recently thinking about possibly assembling our TINA spice. I still need to talk it over with some of the DAC designers to make sure that it can be done but offhand I do not see any problems. So, to answer your question, right now we do not have anything we can provide but this is something I may consider looking into in the near future.

Regards,

Tony Calabria