I am surprised that none of the component manufacturers have developed an analog computer as a amp-array (as opposed to a gate array). They could have a hundred op-amps with programmable resistors, capacitors and a load of cross-point switches. There are many things that I can think of which could be handled a lot easier in analog than programming a PIC (or similar). Validation would be a lot easier too.

 

I am surprised that none of the component manufacturers have developed an analog computer as a amp-array (as opposed to a gate array). They could have a hundred op-amps with programmable resistors, capacitors and a load of cross-point switches. There are many things that I can think of which could be handled a lot easier in analog than programming a PIC (or similar). Validation would be a lot easier too.

There have been attempts made at doing that (and there still may be some available) but the it's very difficult to design such an array that can do the many unique functions that analog circuits typically do, and still make the array more economical than using op amps and discrete parts.
For example such circuits often use other devices, such as diodes, in their design which would also have to somehow be incorporated into the array.
Here's the Wiki blurb on that.

So what are these "many things" that you think could be better handled in an analog array?

 

There have been attempts made at doing that (and there still may be some available) but the it's very difficult to design such an array that can do the many unique functions that analog circuits typically do, and still make the array more economical than using op amps and discrete parts.
For example such circuits often use other devices, such as diodes, in their design which would also have to somehow be incorporated into the array.
Here's the Wiki blurb on that.

So what are these "many things" that you think could be better handled in an analog array?

Intelligent battery charging, radar (marine and road vehicle, front ends at least), sonar, environmental controls, dosing control.

 

I don't know that validation would be easier -- just consider the variability in results you are likely to get with temperature and supply voltage, let alone process variation and mismatch.

Most, if not all, of the applications you describe are industrial and/or consumer-scale in size. You generally aren't going to use a gate array for something like that -- too expensive. You are going to use a cheap MCU or a cheap ASIC. There certainly would be exceptions, but are they numerous enough to justify a market for devices that would need to be sold in huge numbers (tens or hundreds of millions??) annually to be economically viable?

It would be a huge uphill battle to compete with the kind of processing power you get for very little power consumption with today's microcontrollers.

 

I am surprised that none of the component manufacturers have developed an analog computer as a amp-array (as opposed to a gate array). They could have a hundred op-amps with programmable resistors, capacitors and a load of cross-point switches. There are many things that I can think of which could be handled a lot easier in analog than programming a PIC (or similar). Validation would be a lot easier too.

Hello there,

They already do, it's called the Quad Op Amp
They also have the Quad Transistor Array
Quad or hex diode array
7 or 8 resistor dip pack, individual resistors or with a common


There are actually things built into devices that are borderline analog digital, like the gain controlled amps we are finding in ADC's these days. They allow you to program the ADC front end amplification by a few set factors like 1/2, 1, 2, 5, 10.


The analog applications are more specific than the digital ones so it's harder to think ahead. A lot more variables and kinds of components that might be incorporated, so it's hard to design a one-fits-all circuit.

Also there is the sheer complexity. Think of what it would take to provide an op amp with programmable resistors of 1k, 2k, 3k, 4k, 5k, etc. up to maybe 100k as feedback, and the same as input resistor. Just that alone would require some doing and that's just one op amp circuit that only proivdes a fixed gain. Now introduce a couple capacitors 1nf, 2nf, 3nf, 4nf, etc. and you have a lot of components to be switched in or out just for one single amplifier. I suppose you could do the components in binary, 1k, 2k, 4k, 8k, up to maybe 80k, and switch them in allowing more than one at a time in parallel or series, but that's still a lot of stuff just for one op amp circuit. So it would be a big chip with a lot of wasted silicon area, and we have more components yet to add.

There have been analog computers built in the past, but they are usually just a one off thing and made from discrete parts.

 

There might be a niche for such devices as a means of rapid prototyping. There cost is less of an issue than with commercial production.

It seems that such a device is already available:
Field Programmable Analog Array

 

It seems that such a device is already available:
Field Programmable Analog Array

Arrrrg! Oh, the humanity!

 

Hi,

Yes apparently they didnt mind going the full mile.

Interesting devices, and they are quoting some interesting advantages like the ability to keep only a small inventory of parts rather than a whole big stockroom. Parts inventory is one of the problems we faced back in the 80's when we had to do designs that used parts that were already in stock for the most part, and had to have a really good reason for introducing a new part of some kind. It was only the really necessary things that got added.

The 5v dev system at end of life costs 129 dollars USD.

Apparently they use analog switches to switch things in or out, and ADC's and DAC's for various things that process signals.

 

Motorola had one. Lattice still might.

The problem is that whatever is used to connect the internal components together introduces something into the signal path; resistance, parasitic capacitance, an offset voltage or voltage drop, etc. This limits the operating range of whatever the circuit is to signals that are greater than these distortions.

ak

 

There have been a number of these kind of devices over the years. I have no idea what the present market presence is like, but unlike FPGAs for rapid prototyping, these analog array blocks have some pretty serious shortcoming. Migrating a digital design prototyped on an FPGA to a production ASIC is relatively straightforward. But doing so for an analog design is a lot more involved and often involves major changes to the design because of the differences in behavior that express themselves in the final performance.

 

Raytheon, in the early 90s also offered analog arrays, which were mask-configured if I remember properly, to produce small quantities or test prototypes before committing to a full custom analog ASIC.

Being fully bipolar, I remember they consumed significant amount of power.

 

There was the GAP-01 which was an analog IC with op amps, etc. that was dropped because it never was good for much. There have been several analog arrays over the years where you do the functional design and the mfg. builds it. Better than the GAP-01 but still died because of limitations. Think of analog; microvolts to kilovolts, dc to MHz, feedback requirements, etc. and it is easy to understand why there is no simple one chip does it all solution.