Please help can't figure this out, been trying for ages

I knew what a comparator was but i didn't know what a window comparator was. I also didn't know that resistors could get down to 0.01% tolerance until I checked.

Electron wrote

In some lab I worked they used 10 turn trimpots (Bourns) for everything. You will possibly also need that for opamp offset compensation etc.

oltage is between 1.5959 and 1.62035 volts. The output voltage should be ze ro when the voltage is higher than 1.62035 or lower than 1.5959. How do I i mplement this using OP-AMPS, transistors and passive devices? Is this even possible?

No

When you said, "3 decimals so 1.620V", that is actually 4 digits of precision.

He ought to. Resistor tolerance has a direct effect on price, and 0.1% resistors, while widely available, are more expensive than their 1% counterparts, and tend to be available in a wider range of values (typically the E96 range, where 1% parts tend to be offered in E24 values).

He probably does. Most people do.

how much accuracy can one get by selecting resistors? When young I used to select from 5%ers to improve accuracy, and while ppms of drift limit what can be done it did get an improvement.

If your resistor divider pairs are thermally coupled, both drift similarly and the division ratio doesn't change as much.

NT

It's a mugs game. For some resistors the manufacturers did it, and what you got shipped as 2% resistors didn't include any that could be shipped as 1% resistors, and a surprisingly higher proportion of parts that were right up against the 2% limits.

If you buy thin film resistor arrays, you can get 0.1% resistance tolerance, and tighter - guaranteed - matching on resistance ratio versus temperature. I first took advantage of that in 1979.

AFAIK, that might've happened back in the day, with carbon comps that were terrible to manufacture any more than vague values.

Modern processes laser etch parts to spec individually, so error is proportional to production time (more or less), making precision parts expensive and loose parts cheap. This is still mostly swamped by overall production costs and especially distribution, but the difference is noticeable in quantities of millions.

Tim

some does not equal all. I got better values, so not a mug's game. More a youngster's game, now we just buy the accuracy we need.

NT

I assume basic carbon films aren't laser trimmed, just rebinned.

NT

I buy 1% resistors, on the basis it will satisfy all but my most exacting needs. Also a BOM will have one resistor part for any value. It makes inventory so much easier.

Furthermore 1% resistors are so cheap and so much less than the build cost to put the component on the PCB, their purchase cost ends up in the noise.

Obviously, and that's why I put the word in.

You were lucky. The assumption that a batch of resistors that you buy is going to have some kind of helpful distribution of values isn't always true.

Selection is time consuming, unless you can automate it, which tends to be expensive, and buying the value and tolerance you need does indeed make sense.

True, they might be hand-filed by Chinese child labor... :^)

I haven't seen any studies that showed other than a windowed normal distribution for even the cheapest axials and chips.

Even the smallest Chinese factory can afford laser trimming machines, they're not at all the industrial enterprise equipment they used to be. You can churn out many thousands of resistors per hour, if you don't mind the poor tolerances you get from the limited settling time.

Diode lasers for cutting and engraving are very cheap, as commercial tools are concerned. You can buy one yourself for low kilobux. Great for rapid prototyping in plywood or plastic.

Tim

General response here : Maybe it would be best if the thing was adjustable. Either ten turn pots or maybe just good design to set the range of the adj ustments, like not making something that wants 130 volts adjustable to 300 volts. Design, that's the word.

But then to the pots drift ? I remember some things, like the HV in a TV be ing set by cutting resistors, maybe something like that with the proper val ues.

The circuit is easy, it hits the voltage and up goes the output, then the n ext threshold slams it down. If regular BJTs had a really stable Vbe, you c ould do it with them. In fact the overvoltage sensing could be done that wa y I think. If the comparator has an open collector output, you just add ano ther open collector output.

e. Either ten turn pots or maybe just good design to set the range of the a djustments, like not making something that wants 130 volts adjustable to 30

being set by cutting resistors, maybe something like that with the proper v alues.

Pots get twiddled by people who don't know what they are doing, sometimes i n final test.

One fertile source of modification requests at Cambridge Instruments was fr om production, who'd want a wider range pot because they couldn't get the o utput they wanted.

Most of the time this represented them working out their own calibration pr ocedure, which was quicker than the official routine, but set up up the pot s in the wrong order.

Once the machine is out in the field pots get twiddled by people who really don't know what they are doing - graduate students are a particularly dang erous group.

Any pot you can design out represents design time well spent.

The idea with trimpots (10 turns) is to make an accurate voltage divider to get close with say 1% resistors, and then move the output voltage around a bit with the trimpot. For example:

• | [ ] 10 | |-----------------> | | | [ ] >100 + [ ] 5 | | | ------> [ ] 10 | | /// ///

Values are relative resistor values, the '100' from the trimpot reduces its effect, and also its drift effect.

No I wasn't.

maybe that's why no-one made such an assumption.

yes, hence it's a youngster's game.

NT

not adjusting their values at all would make a lot more sense. Whatever value they are they're saleable as is.

NT

I saw a youtube review of one of the cheap ones. Plug it in and it burns a hole through the table. The laser comes on at full power and doesn't turn off until the CPU boots.

Even I can design a driver that would prevent that, but it made me wonder how a truly reliable one would be designed so that one or more components could fail and it would never do that.