cheap ADC

torsdag den 20. januar 2022 kl. 17.31.18 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

more than one way to do it, one is basically successive approximation with a pwm dac another is delta-sigma with RC as "integrator" that might have an advantage in that the comparator doesn't follow the input signal it is always ~1/2 Vcc

Version 4 SHEET 1 1316 680 WIRE -96 -992 -176 -992 WIRE -176 -960 -176 -992 WIRE 496 -832 -320 -832 WIRE -96 -752 -96 -992 WIRE -96 -752 -128 -752 WIRE -128 -720 -128 -752 WIRE -96 -720 -96 -752 WIRE -320 -688 -320 -752 WIRE -304 -688 -320 -688 WIRE -144 -688 -304 -688 WIRE -16 -672 -64 -672 WIRE -144 -656 -480 -656 WIRE -16 -656 -16 -672 WIRE 192 -656 -16 -656 WIRE 192 -640 192 -656 WIRE 272 -640 192 -640 WIRE 496 -640 496 -832 WIRE 496 -640 432 -640 WIRE 752 -640 496 -640 WIRE 912 -640 832 -640 WIRE -304 -624 -304 -688 WIRE 272 -592 208 -592 WIRE -128 -576 -128 -624 WIRE -96 -576 -96 -624 WIRE -96 -576 -128 -576 WIRE 912 -576 912 -640 WIRE -304 -528 -304 -560 WIRE 208 -512 208 -592 WIRE 912 -464 912 -512 WIRE 208 -368 208 -432 WIRE -96 -224 -176 -224 WIRE -176 -192 -176 -224 WIRE 496 -64 -320 -64 WIRE -96 16 -96 -224 WIRE -96 16 -128 16 WIRE -128 48 -128 16 WIRE -96 48 -96 16 WIRE -144 80 -176 80 WIRE -16 96 -64 96 WIRE -480 112 -480 -656 WIRE -416 112 -480 112 WIRE -320 112 -320 16 WIRE -320 112 -336 112 WIRE -304 112 -320 112 WIRE -144 112 -304 112 WIRE -16 112 -16 96 WIRE 192 112 -16 112 WIRE -480 128 -480 112 WIRE 192 128 192 112 WIRE 272 128 192 128 WIRE 752 128 432 128 WIRE 912 128 832 128 WIRE -304 144 -304 112 WIRE 272 176 208 176 WIRE 496 176 496 -64 WIRE 496 176 448 176 WIRE -128 192 -128 144 WIRE -96 192 -96 144 WIRE -96 192 -128 192 WIRE 912 192 912 128 WIRE -304 240 -304 208 WIRE 208 256 208 176 WIRE -176 272 -176 80 WIRE 912 304 912 256 WIRE -480 384 -480 208 WIRE 208 400 208 336 FLAG -480 384 0 FLAG 272 224 0 FLAG 208 400 0 FLAG -304 240 0 FLAG 912 304 0 FLAG -176 -112 0 FLAG -128 192 0 FLAG -176 352 0 FLAG 272 -544 0 FLAG 208 -368 0 FLAG -304 -528 0 FLAG 912 -464 0 FLAG -176 -880 0 FLAG -128 -576 0 FLAG -480 -656 Vin FLAG 912 -640 Vout1 FLAG 912 128 Vout2 SYMBOL voltage -480 112 R0 WINDOW 3 -226 54 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR Value SINE(1.5 1.4 5000) SYMATTR InstName V1 SYMBOL Digital\\dflop 352 80 R0 WINDOW 3 8 12 Left 2 SYMATTR Value vhigh=3 SYMATTR InstName A1 SYMBOL res -320 96 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 10k SYMBOL res -336 -80 R0 SYMATTR InstName R2 SYMATTR Value 10k SYMBOL cap -288 208 R180 WINDOW 0 24 56 Left 2 WINDOW 3 24 8 Left 2 SYMATTR InstName C1 SYMATTR Value 10n SYMBOL res 848 112 R90 WINDOW 0 80 54 VBottom 2 WINDOW 3 89 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1K SYMBOL cap 896 192 R0 WINDOW 0 -58 48 Left 2 WINDOW 3 -56 80 Left 2 SYMATTR InstName C3 SYMATTR Value 8n SYMBOL Comparators\\LT1715 -112 96 R0 SYMATTR InstName U1 SYMBOL voltage -176 -208 R0 WINDOW 3 -64 56 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR Value 5 SYMATTR InstName V2 SYMBOL voltage -176 256 R0 WINDOW 3 -64 56 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR Value 1.5 SYMATTR InstName V4 SYMBOL Digital\\dflop 352 -688 R0 WINDOW 3 8 12 Left 2 SYMATTR Value vhigh=3 SYMATTR InstName A2 SYMBOL res -336 -848 R0 SYMATTR InstName R5 SYMATTR Value 1k SYMBOL cap -288 -560 R180 WINDOW 0 24 56 Left 2 WINDOW 3 24 8 Left 2 SYMATTR InstName C2 SYMATTR Value 10n SYMBOL voltage 208 -528 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V6 SYMATTR Value PULSE(0 3 0 0 0 50n 100n) SYMBOL res 848 -656 R90 WINDOW 0 80 54 VBottom 2 WINDOW 3 89 56 VTop 2 SYMATTR InstName R6 SYMATTR Value 1K SYMBOL cap 896 -576 R0 WINDOW 0 -58 48 Left 2 WINDOW 3 -56 80 Left 2 SYMATTR InstName C4 SYMATTR Value 8n SYMBOL Comparators\\LT1715 -112 -672 R0 SYMATTR InstName U2 SYMBOL voltage -176 -976 R0 WINDOW 3 -64 56 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR Value 5 SYMATTR InstName V7 SYMBOL voltage 208 240 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value PULSE(0 3 0 0 0 50n 100n) TEXT -712 488 Left 2 !.tran 5m RECTANGLE Normal 576 480 144 -128 2 RECTANGLE Normal 576 -288 144 -896 2

That's certainly better; the FPGA LVDS receivers are mediocre comparators. But the game was to minimize the number of external parts. I'll be digitizing a thermistor on a heat sink, so it's not a precision thing.

If we go with an available Lattice FPGA, it doesn't have an ADC.

Some sort of delta-sigma signal processing would be interesting.

I could use an LM71, SPI temp sensor, which is just not as interesting.

That is important in an FPGA because the differential inputs are not rail to rail. In fact, they are rated to work over a very limited range. I'm not sure how well they will serve in this application. A good comparator is rated for offset voltage. The spec on an LVDS mode input in an FPGA has a very loose spec for that. Given that the input circuit is basically a voltage divider it works to increase the significance of input offset relative to the input signal.

If you are trying to set an alarm for some condition with a loose input spec for accuracy or offset, then this can work fine. It only costs 3 pins on the FPGA. I worked on a design that used 8 such ADCs. Rather than use resistors to set the reference point I used another output with a 50/50 duty cycle to get an accurate Vcc/2 point that follows the rail without resistors removing some source of error and could be shared with all the ADCs. It also allowed you to adjust the reference voltage if there became a need to make an adjustment.

snipped-for-privacy@4ax.com a ecrit

Probiblie bEst iz too juice a FLIR IR camarade wiz Anna Log outpud ant prozes ze vidio on 1 FPGAYE pin. Ze moost rat airraids r ze hot test

Tom

fredag den 21. januar 2022 kl. 03.08.06 UTC+1 skrev snipped-for-privacy@highlandsniptechnology.com:

even simpler tmp05/tmp06 which is pwm output so all you need is a counter in the fpga

or for easy mounting on a heat sink,

Who in their right mind comes up with a device name "tmp05"? ;-)

74 cookies! A record!

That would be a good way to couple thermally to the heat sink. We have LM35 in stock, in TO220, too, but that would require an ADC and probebly an opamp, which isn't bad.

I want to simulate the mosfet junction temperatures in realtime, and the heatsink temp is part of that math.

Isn't the FPGA the place where you put arbitrarily complex logic? Why try to minimize it? The filter typically used in the ADC you indicated above is often an integrate and dump, otherwise known as a counter. The filter characteristics are such that frequency multiples of half the sample rate are on nulls. This helps with the antialias filtering, but to get a sharper response near the half sample rate requires a FIR filter or similar. In fact, best is a filter that compensates with some boost to compensate for the integrate and dump roll off. A design example I saw from Lattice included both an integrate and dump followed by a FIR filter. However, I believe the FIR coefficients were all just 1, so after decimation it was also an integrate and dump equivalent. I suppose you could add any coefficients you wished. It was only a app note after all.

Often the application does not require any significant filtering and the integrate and dump is simply the way of increasing the resolution from 1 bit to N bits by counting 2^N input samples.

why spend time to implement a SPI, use three/four pins, or logic and extra parts to fudge an ADC and calibrate that when you can get a calibrated part that only need a single pin and very simple logic

Not sure what you are comparing to what. An ADC inside the FPGA uses no external parts other than perhaps resistors and capacitors. You are talking about a temperature sensor, I was only talking about an ADC. If the cost is paramount, the passives are cheaper than the TMP05/06, only a few of them are active, in stock in small quantities at Digikey and cost dollars!!! Am I looking at the wrong parts? I've used an analog temperature sensor that was only $0.10 in quantity, the TMP236! Well, looks like they are a bit more now, $0.27. I got a penny happy group of designers to use that instead of a thermistor which was the same price.

The semiconductor shortage seems to have decimated the inventory of temperature sensors. Digikey will show 15 or more parts, but only 2 in stock and 52 week lead times once those are gone!

Yeah, resistors and capacitors with a thermistor are looking pretty good.

On a sunny day (Sat, 22 Jan 2022 19:14:32 -0800 (PST)) it happened Rick C snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:

Any diode will function as temperature sensor. Been there done that, even used the forward biased cb junction from a TO220 transistor screwed to the heatsink. Is sort of linear, and takes only one calibration point, collector at ground potential from heatsink, no isolation needed. Any old TO220 will do :-)

on pin 9 of the PIC.

What was the temperature resolution of this setup? When you mention calibration point, do you mean a one time design calibration or a calibration for each device? I know they use diode junctions on chips to measure temperature, but typically they use a special circuit to drive with two known currents, eliminating the need for calibration, no?

On a sunny day (Sun, 23 Jan 2022 04:06:46 -0800 (PST)) it happened Rick C snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:

Well in this case I wanted to stabilize the temperature, so measured the ADC steps for the temperature I wanted and the control loop then steered towards that.

the same way you can set an alarm level. There is a linear slope of about -2.1 mV / degree C for a Si diode, accurate enough for calculating heatsink temperature. If you power the junction via a resistor from some higher voltage, say 5V or 12V then the current change due to Vcb change is not really that much over a normal range.

Would you need to do the calibration for every unit built, or just once for the design? Does the voltage for your set point vary with normal device variations?

I ask this because I thought it was the slope with temperature that is pretty consistent between devices while the details of the measured voltage would vary. So to avoid having to calibrate each device measurements were required at two currents?

Or are you saying your requirements simply don't require that much accuracy?

On a sunny day (Sun, 23 Jan 2022 06:09:48 -0800 (PST)) it happened Rick C snipped-for-privacy@gmail.com wrote in snipped-for-privacy@googlegroups.com:

Driving the heatsink (in my case the alu plate that holds the photo diodes, opamp amplifier and tritium light) to a specific temperature and maintaining it from there from the ADC steps gives very very high accuracy. The system is basically a thermostat So calibrated at a specific temperature for that specific instrument once as there is only one.

If you wanted to know ADC steps for other temperatures do the mV / C thing. John wants to measure heatsink temperature over a big range I think, this would work fine with same small error. But you could just as well glue some LM35 or so to the heatsinks .. I have used those too. Or some other sensor chip with i2c or spi interface. Tom was joking about using a FLIR camera and doing the video processing in FPGA, gives you about +-2 degrees C accuracy or so.

On the other end of the spectrum: we had 'clicksons' in the past that would open at some high temperature. No idea what his specs are

sure if you need to build thousands, if you only 100 maybe not so much

Building 100 still requires 100 ICs... if you can get them. Whatever.