Square root circuit

I really like to do technical management, brainstorming and architecting and reviewing designs. I don't like the scheduling and HR bits.

I've been to too many meetings with those folks, including the founder/CEO. They all think in spreadsheets and communicate in PowerPoint. We get visited and audited by quality and logistics and management people who have never actually seen the products that we suppy to them; I have to bring products up to the conference room and show them. The QC people have no idea of how many of our units are deployed or what the field failure rates are. We've shipped a couple thousand boxes with zero field failure rate, but they scored us "F" on quality. We really don't care.

I just bailed on a conference in europe. The agenda was breakfast, Powerpoint, lunch, Powerpoint, dinner. Golf the next day. Is golf a required course in business school?

Only a month late? That's not bad. Usually the customer has no alternative in that time frame.

He could be a participant here if he wanted to. I guess he prefers to be universally ridiculed. Whatever turns you on, I suppose.

He should invent a new nym and start over. But I think he's done that a few times already.

LTSpice will do the integration for you if you ask it right.

It may be, but it's a lot harder to get it accurate.

His customer was BT and in the meantime his startup was bought by another startup. The other startup had a full size PC-AT board that would do what my PC-104 board did. They bit the bullet and made a case several times larger than the one they were going to put my boards in.

I think you don't miss schedules with British Telecom. Well, maybe once.

A job I consulted on was for a satellite FXS-FXO board for Alcatel. It was a nice piece of work, but the former salesman turned entrepreneur couldn't manage his way out of a paper bag. He gave Alcatel lame excuses for several months until they flew six of theirs over for a design review. The head guy didn't speak English so he got some translations during our design review, but not as much as you might expect... at the end they revealed their hand and the head guy *did* speak English and told us that for the most part they were happy with the technical members and asked our fearless leader to talk to them in private. When he came out some minutes later he was nearly white. I don't know if they berated, bullied, threatened or beat him, but he was clearly shaken. Just desserts in many ways.

We got the board finished, but to the best of my knowledge they never used it. This was around 2000-2001 and I think Alcatel had to lay off a lot of people about then. The product was aimed to satcomms for "remote" locations and I think they just stopped worrying about that part of the market. If we had delivered on time, they may have gotten it into production before the big meltdown. I don't remember the reasons for the delay, but it wasn't all technical. Much of it was our guy just not doing what he was supposed to do because it would put a few more pennies in his pocket.

I have three U6 Pros that I like a lot. In the microscope project that I'm working on, I'm using one to control two VCOs for X and Y acousto-optic deflectors, and to run an LTC2602 dual 16-bit DAC over SPI. (The LabJack uses PWM for its DACs, which is cheating.)

The box with the DACs also has four 74LVC74s running as /4 Johnson counters (/16 on each channel), which takes the 50-110 MHz VCO range down to something the LabJack can handle (8 MHz max).

One nice thing about the LabJack is that you can set it up to do lists of things with vaguely decent hardware timing, in what they call "stream mode". That makes the internal timers useful for something, which they wouldn't be otherwise. Interleaved control and data acq is usually the Achilles' heel of lab automation setups.

The software takes 15 or 20 measurements across the code space, fits a cubic spline for code in terms of frequency, i.e. inverse interpolation, and thereby gives me a nice evenly spaced square raster.

(Yes, the real version is going to use DDSes.)

The documentation is pretty sketchy, but there's a reasonable amount of sample code and the developers are very responsive to forum posts and email.

Cheers

Phil Hobbs

The +12 is bootstrapped from the LM399, so it's almost as good.

It would work fine, better than what they did. What's stupid about that circuit is that the influence range of the "fine" pot depends on the rotation of the coarse pot. When the coarse pot is at max, the fine pot has zero effect.

Set the coarse pot near max and do that again.

Two pots from V+ to ground. Sum the wipers with two resistors, asymmetrically, any coarse/fine ratio you like. That will keep the fine adjust range independent of the coarse setting, and the TC will be lower. Add one more resistor if you want the overall range to be limited, as on that schematic.

And increase the hairball factor!

Hardly.

Actually, you could put the pots in series and still sum the wipers with two resistors. Make the upper summing resistor small and the lower one big, and that maintains the existing voltage range.

A piecewise linear approximation using single OA and diode+resistor gain setting branches may be good enough. It's quite straightforward.

Two pots in series. Each wiper goes to a resistor, and the other end of those resistors join to be the output. You need a schamatic?

Whoever designed that was an idiot. It's a mess.

Hah, that's what JL said too. (thanks for the advice.) I was looking at the price of quad transistors.. I might as well use the analog multiplier from AD. (we use them in another circuit.)

The offset voltage stinks (AD633)

So with a few more diodes and R's.....

Some weird kinda of weird math fun.. (Ok someone must have already done a multi diode-resistor opamp- break point thing, that approximates a sqrt.)

George H.

You don't need a quad transistor, you just have to run it inside a feedback loop.

Cheers

Phil Hobbs

OK, I wanted to prototype it today, but I ran out of time. (no lunch, but I left a bit early...)

A break point thing with diodes and R's should work.

As you say it goes in a control loop, which covers up a lot of sins.

George H.

The usual model of a MOSFET is that the drain current is proportional to the square of the G-S voltage, so that's a square-root candidate. The law only works near the threshold voltage, you want a low-transconductance MOSFET to get any range from this. BSS123 looks promising.

A second possibility is the temperature rise of a thermally insulated resistor (square of input current); either use a thermistor with a controlled current (and do pulsed temperature-measure cycles), or couple a thermometer to a resistor. I envision a finger milled into PCB material, with a sock over it. Either way, measure difference from ambient temperature. Delta-T is the input, resistor forward voltage is proportional to sqrt(delta-T).

On Wednesday, July 1, 2015 at 11:28:18 AM UTC-7, George Herold wrote:

in

thermostat

I did the circuit using 3904 transistors and changed the reference and pick

-off point to make the graph easier to read. It looks SQRT-ish. You can t ake one of parallel transistors out and double the value of the Ref resisto r and get a very similar result.

- The typo of "80 uA TO 1.7A" may have been intended as "80 uA @ 1.7A"

- The "12" of the reference resistor should read 12V.

Version 4 SHEET 1 8124 1208 WIRE 2288 0 1440 0 WIRE 2368 0 2288 0 WIRE 2288 32 2288 0 WIRE 2288 144 2288 112 WIRE 2384 144 2288 144 WIRE 1440 176 1440 0 WIRE 1600 176 1440 176 WIRE 1888 176 1680 176 WIRE 1968 176 1888 176 WIRE 2208 176 1968 176 WIRE 2384 176 2384 144 WIRE 1968 192 1968 176 WIRE 2288 192 2288 144 WIRE 1888 240 1888 176 WIRE 1904 240 1888 240 WIRE 2208 240 2208 176 WIRE 2224 240 2208 240 WIRE 1440 288 1440 176 WIRE 1744 352 1648 352 WIRE 1888 352 1744 352 WIRE 1968 352 1968 288 WIRE 1968 352 1888 352 WIRE 2288 352 2288 288 WIRE 2288 352 2208 352 WIRE 1744 368 1744 352 WIRE 1968 368 1968 352 WIRE 2288 368 2288 352 WIRE 1440 416 1440 368 WIRE 1536 416 1440 416 WIRE 1648 416 1648 352 WIRE 1648 416 1616 416 WIRE 1680 416 1648 416 WIRE 1888 416 1888 352 WIRE 1904 416 1888 416 WIRE 2208 416 2208 352 WIRE 2224 416 2208 416 WIRE 1440 448 1440 416 WIRE 1440 560 1440 528 WIRE 1744 560 1744 464 WIRE 1744 560 1440 560 WIRE 1968 560 1968 464 WIRE 1968 560 1744 560 WIRE 2288 560 2288 464 WIRE 2288 560 1968 560 WIRE 2448 560 2288 560 FLAG 2384 176 0 FLAG 2368 0 out1 IOPIN 2368 0 Out SYMBOL voltage 1440 432 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value 12 SYMBOL res 1632 400 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 11.3k SYMBOL res 2272 16 R0 SYMATTR InstName R2 SYMATTR Value 8.45k SYMBOL voltage 1440 272 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value 12 SYMBOL current 1600 176 M90 WINDOW 0 65 40 VBottom 2 WINDOW 3 -64 144 VTop 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName I2 SYMATTR Value PULSE(0 100e-6 0 20e-3 20e-3 1e-6 40e-3 1) SYMBOL npn 1680 368 R0 SYMATTR InstName Q6 SYMATTR Value 2N3904 SYMBOL npn 1904 368 R0 SYMATTR InstName Q7 SYMATTR Value 2N3904 SYMBOL npn 1904 192 R0 SYMATTR InstName Q8 SYMATTR Value 2N3904 SYMBOL npn 2224 368 R0 SYMATTR InstName Q9 SYMATTR Value 2N3904 SYMBOL npn 2224 192 R0 SYMATTR InstName Q10 SYMATTR Value 2N3904 TEXT 1488 48 Left 2 !.tran 20m

So look at the AD635.

Sure. Everybody has ....