8

In super small low power devices, where space is a big premium, complicated passive networks can take up a lot of space. Is it possible to get a passive or mostly passive ASIC fabricated? If it is possible, how much would it cost? Obviously it would never approach the tiny cost of discrete passives, but when you really need that space it might be worth it.

In terms of costing for ASIC manufacture, this question was really helpful, but it's talking about more traditional ASIC's where the main thing is the active components.

I'm also wondering, maybe ASIC is the wrong word? Is there another word for this kind of thing?

BeB00
  • 5,303
  • 2
  • 18
  • 34
  • What do you mean by "passive"? Like a bunch of tiny resistors and capacitors? – Eugene Sh. Aug 16 '17 at 18:45
  • 1
    Yeah pretty much. I guess you would just use FETs for resistors. Silicon capacitors are pretty inefficient spacewise, so it's not the best idea. – BeB00 Aug 16 '17 at 18:46
  • 1
    So maybe you are talking about "analog" rather than "passive"? – Eugene Sh. Aug 16 '17 at 18:49
  • can you give us a rough idea of how much power you'll try to put through this (because that is what makes resistors large), how high your voltages are and how large your largest capacitances (because that is what makes capacitors large), how large your largest inductance is, and *just how complex* in numbers of components your circuit is? And: what's the size you currently have, what's the size you aim for? These are all necessary boundary conditions that one would have to first put down before thinking about a price – Marcus Müller Aug 16 '17 at 18:51
  • This is Low voltage signal stuff, so around 3v (maybe max 10 i guess). The application I had in mind is actually purely a resistive network, with maybe 15 resistors of 01005. I'm also curious about more complex applications. http://www.sga.se/website/files/SGA-AnalogASICCutsCost.pdf suggests that it is occasionally done, although I cant see much info on it – BeB00 Aug 16 '17 at 18:54
  • Hm, I don't really see how you could do the contacting for something build out of 15 resistors on an area smaller than that of 15x 01005? – Marcus Müller Aug 16 '17 at 18:55
  • Not every resistor needs to have it's nodes available, and if you used 0.4mm pitch bga it would be smaller by default i think – BeB00 Aug 16 '17 at 18:56
  • 3
    You go to great lengths to avoid resistors, capacitors and especially inductors in ASICs so you are in a bad spot to start with. What you do is create very weak current sources and minute small capacitors, opamp style intregrators and so on to fake the passive value you need. – winny Aug 16 '17 at 18:57
  • No, 0.4mm pitch is *larger* than the distance between two parallel 01005 resistors. Unless you need less than half of the nodes contacted (in which case, you can simplify the network!!), you don't gain *anything* by that. – Marcus Müller Aug 16 '17 at 19:04
  • 1
    If you stated what application this was, it's possible we could suggest how that application could be converted from a passive circuit into an active circuit which could then be more easily placed within an ASIC. – horta Aug 16 '17 at 20:01
  • It might be worth explaining the function of your network a little bit more. How many pins does your network have? Is current always flowing in one direction? How stable does it need to be? What power dissipation do you expect / can you tolerate? This looks like it may be a [case of the XY problem](https://meta.stackexchange.com/q/66377/212464) – Floris Aug 16 '17 at 20:43
  • Dual,Terminated networks? 1M? – Tony Stewart EE75 Aug 16 '17 at 21:33
  • When I thought of the question I did have a specific application in mind, but I was mainly just interested in whether it could be a solution generally. The application was just a series of 2M resistors to bias some inputs for an ADC. – BeB00 Aug 16 '17 at 23:07

4 Answers4

20

Not an expert answer, but some disconnected relevant facts:

One of the reasons that ICs often require external passives is that standard silicon ICs are not a good substrate for making capacitors and inductors of significant value on-chip. Thus, a conventional IC, "AS" or not, is not a space-efficient way to put large passives on a board (other than high-value resistors). It might still work out to be an improvement if it's not the value but the number of parts that is a problem (i.e. packages and interconnects account for most of the area), but this is why, for example, your microcontroller always has external decoupling capacitors.

Resistor networks and capacitor networks are an existing common form of passives-in-a-single-package. I haven't heard of any mixed passive types in a single package, though — I expect it would be difficult in the same way as putting passives in a normal silicon IC because they have very different materials and manufacturing techniques.

Resistor networks can be used:

  • To save space on the board; they were much more common and relevant in the through-hole days, I believe, where even if you placed the resistors vertically you would need an entire hole for the other pin, and a network would let you have \$N\$ resistors with only \$N+1\$ pins).
  • As a high-precision voltage divider or other configuration of matched resistors (this is often seen in high-end measuring instruments). Multiple resistors made using the same process and on the same substrate can have tighter tolerance, more temperature-stable ratios of resistances than independently manufactured resistors.

Hybrid integrated circuits, or hybrid modules, are sort of an intermediate technology between PCBs and single-semiconductor-die ICs, which can include passives either formed on the board itself or soldered on — but they are not noteworthily compact nor cheap. The value is in having a precise, pre-built, sealed module that you can plug/solder into a larger board.

Kevin Reid
  • 7,444
  • 1
  • 25
  • 44
  • Do you know which component manufacturers are big on resistor networks? – BeB00 Aug 16 '17 at 18:55
  • @BeB00 I have no experience here, but that sounds like a question to answer starting by seeing who makes the resistor networks available at your favorite component supplier ([e.g.](http://www.mouser.com/Passive-Components/Resistors/Resistor-Networks-Arrays/_/N-e89l)). – Kevin Reid Aug 16 '17 at 18:57
  • With resistor networks it looks like the smallest are basically just a line of 0201 resistors in one package, so theres no space saving. – BeB00 Aug 16 '17 at 19:02
  • May be worth mentioning that thin-film resistor value tolerance tends to be very loose, but the resistor ratios can be very precise. (Ohms per square sheet resistance is a process variable, but number of squares in a design element only depends on lithographic accuracy.) So it can make sense to build a resistor divider network, but replacing unrelated discrete resistors may not work as well. – MarkU Aug 16 '17 at 20:44
  • @MarkU That's what I was getting at by saying "matched", but I've put in a few more words on that. – Kevin Reid Aug 16 '17 at 20:52
  • @BeB00 I'd go with Vishay - I've sampled some pretty impressive precision R networks in DIP packages from them a couple of years ago; while it may not be the best choice in terms of board space, it was *great* in terms of R value tolerances & stability (I've based an analog audio synth on them, with great results). –  Aug 17 '17 at 14:40
12

ASIC means Application Specific Integrated Circuit. If that integrated circuit would contain only passive components then we could still call it an ASIC.

Making a silicon die with only passive components is very well possible but there are severe limitations. On an IC capacitors are only practical up to a few nF. Inductors can only be a few nH. Resistors can be made in almost any value.

These passive components (except for the inductors) might have parasitic diodes to the substrate so it is possible that they're not 100% passive.

Designing and producing an ASIC is never cheap. Depending on the manufacturing process, making the masks (needed to define the patterns to make the components) can be anywhere between $ 10000 to $1000000 (yes, on million US dollars).

The actual cost of producing one IC once you have the masks is lower, it can be a few cents each up to tens of dollars depending on the size (smaller size cheaper). But with a high mask cost you'd need to make many of these chips (and use/sell them) for this to be cost effective. The mask cost is shared between all the ICs you produce so the more the better.

I worked at Philips Semiconductors in the past and they had a special process for passives only. The idea was that these ICs would provide the supply decoupling capacitors and RF matching components for a separate, small IC with the active components and made in a more expensive process technology. This smaller IC would then be mounted on top of the die with the passive components.

This was a bit complicated to produce so it was only used for some niche products as far as I know. It is certainly not mainstream technology.

Bimpelrekkie
  • 80,139
  • 2
  • 93
  • 183
5

You don't have to have actives on an ASIC (although I don't know if anyone has ever done this). There are manufacturers who will do this kind of thing, it is expensive and you need large quantities.

There are better ways that would still be more costly than a PCB with components but cheaper than an ASIC. Components can be built inside the PCB.

enter image description here
Source: EDN

Or you can get integrated packages:

enter image description here

These last two would probably be cheaper than an integrated circuit where the passives are built into silicon, however embedding components in these new processes will be much more expensive than using a PCB with micro vias and stacking PCB's

Voltage Spike
  • 75,799
  • 36
  • 80
  • 208
  • 3
    *I think ASICs are more in the digital domain* Not true, I've worked on many ASICs over the years and all of these had a significant part with Analog electronics. Which I partly designed ;-) – Bimpelrekkie Aug 16 '17 at 19:43
  • Yeah, in my experience that's what I've heard. I realize its a fuzzy distinction. – Voltage Spike Aug 16 '17 at 19:52
  • Digital/analog has nothing to do with it being an ASIC. ASICs these days usually have analog front-ends and digital internals. But there's no reason you couldn't have it be entirely one or the other. They're simply orthogonal ideas. – horta Aug 16 '17 at 19:58
  • Look, in every case I've heard it ASIC had to do with digital and IC with analog. I'm telling you *what I've been told* I didn't say it was right. – Voltage Spike Aug 16 '17 at 20:03
  • 3
    I think we're just suggesting you remove that wording from your answer so that it doesn't further confuse readers of your otherwise good answer. – horta Aug 16 '17 at 20:15
  • There are PCB materials that allow to create discrete resistos within the PCB. Here's an example: [Ohmega](http://ohmega.com/overview/). – Nick Alexeev Aug 17 '17 at 14:38
2

Yes, definitely. Here at Keysight, for example, we do create some of our own basic strip-line filters for probes and other equipment. Building passive components into chips is standard.

Other clarifications. ASIC vs. IC - If you're designing a chip for a specific purpose, then the "AS" is valid. It's a potato/potahto situation.

Digital vs. Analog? - Both are common. Digital ASICs generally do processing and computations, analog ASICs generally massage a signal into a more usable form. We talk about this one on the EEs Talk Tech electrical engineering podcast here:

https://eestalktech.com/2017/07/13/fast-adc/