Low-voltage shunt references

It's possible to use ANY nonlinear device in a bridge (the left side of the bridge being R1 and R2, the right side R3 and a square-law FET, for instance) and adjust R1/R2 so that the balance occurs at any voltage you want.

The problem is, there's a temperature coefficient for all the candidate nonlinear devices. Bandgap references use the diode equation for the zeroing of tempco, which (in silicon) happens at multiples of 1.25 volts.

The only reason a forward-bias Vbe is important is that it supports an error amplifier of forward-biased bipolar transistors in Si.

I'm OK on giving up ever-lower voltage references if I can keep Si bipolar analog circuitry. The reference is so much better if the bridge balance is maintained and buffered by a good op amp...

It would take a large multiple of a few pennies to pay for an hour of engineering.

I'm sure glad that I'm not in a business where pennies matter.

I am doing a design right now with very high volume, so I am doing almost nothing else than looking for pennies

It's fun. Anyone can do a design where you just select from whatever you like. When options are limited, the fun starts

Cheers

Klaus

Some of us enjoy the design challenge to produce superb performance at very low cost... often with the added teaser... very low power. ...Jim Thompson

Let the fun begin... I'm almost there ;-) ...Jim Thompson

The greater the cost advantage, the easier the licensing negotiation, usual ly. I've been doing a number of design-outs of expensive commercial modules (lasers, detectors, and signal processing), and showing that a $700 gizmo can be replaced by something with a $28 BOM cost and better performance is relatively compelling. (That's a recent real example.)

Besides, doing good stuff with minimal apparatus is aesthetically very plea sing.

Cheers

Phil Hobbs

Indeed! ...Jim Thompson

I'm at the other end of the spectrum: extreme performance without much cost sensitivity. That's also fun.

I think the most expensive PCB-mount part that we buy is a Hittite distributed amp for $299 each.

I don't sell in high enough volume to make a doller or so matter, not at the cost of added hours of engineering, or the hassle of qualifying and stocking another part. Power usually doesn't matter either, unless component cooling becomes an issue.

If better performance might increase sales (and who knows if it will?) we may as well go for it. Selling more units will overwhelm a modest difference in parts cost.

Saving a dollar matters plenty in the low volume, moderate margin area that I'm most familiar with, too, but I get the impression you're in the low volume, LUDICROUS margin area!

When the bare boards cost $381 each, and the layout cost $10K or so, the opamps don't matter so much.

PCBs are expensive, and the difference between 6, 8, and 10 layers kinda blows away the small stuff.

Different business.

I usually have no financial interest in the finished product... only the satisfaction of having met the client's performance and cost needs... he'll be back for more designs... virtually all of my business is repeat or referrals.

My designs increase the profitability of the client's product, which is why they come back... track record is what it is all about in my business.

...Jim Thompson

Sort of. Maybe half of the things that we design are losers, because it was a bad product idea or because our OEM customer had a bad idea.

Only a few products have exceeded 1000 pieces over the years. We tend to do bleeding-edge stuff and avoid competition, so we are selling IP more than parts in boxes. Our electronics parts cost runs very roughly something like 15% of selling price. Even then, some products are marginal, or losers.

Labor is expensive to design/code/document/build/test high-end stuff. Much more expensive than parts. So a good design minimizes assembly and test hassle, which is harder to do than cruising Digikey for cheap capacitors.

The iphone is an ideal product: sold in the tens of millions WITH a ludicrous margin.

What about a ratio of schottky junctions? Schottky diodes are affordable in many processes (I guess).

Maybe not amenable to low currents or wide temp ranges...

Tim

Schottky bandgap?

I'm only 1/2 visualizing the circuits you guys are talking about. But I think Phil said he was just setting current limit's, and may not need too much accuracy as you say.

So my (new) fave current limits are back to back depletion fets. And there are LED current sources. (though unipolar) I'm wondering if you could make one look like ~1 ohm till 1 A (1V) One thousand lnd150's in parallel. (well 2k) but that will be spendy. :^)

George H.

I get the impression that as in independent in 2k17 it doesn't make much sense to hope for anything more than a fixed-price sale of IP; multinationals would rather eat broken glass than deal with not-designed-heres taking repeat cuts. Try to charge a lot

I think that's right, I look for $1-10+ savings and not pennies, still having some cost constraint, is the fun/ challenge. I can buy opamps from whomever I want, I have to shop more carefully for optics.

George H.

You didn't mention customer support. That takes ~10% of my time.

George H.