Half power rail reference

Unless you need a variable voltage then what's wrong with two resistors and an op-amp voltage follower?

If you want a power supply reference, then why not just use a resistor divider? Or do you want a rail splitter power supply? (In which case, what's the current/ voltage?)

George H.

I don't think COG ceramics are microphonic to a significant extent.

John

Yeah, if you can work with a small enough value. C0G aren't available much above what, 1000 pF or so?

you can get a few 100 nF for a couple of $, above that they get really expensive

You mean "half voltage".

generating an accurate voltage at half the supply rail by switching an outp ut at a 50/50 duty cycle to feed an RC. I can't seem to find that thread. I'm wondering what the limitations might be on the accuracy. Obviously th e difference in drive capability is a factor, but that can be mitigated by the external resistor.

within the digital device. That can be mitigated by using a slower cycle.

aveform be 0.5 MHz.

ntegrating capacitor might be important as well. Also ceramic caps can be microphonic, so film caps will be needed. In fact I need to specify that a ll through the sensor power supply.

Thanks, but trying that both with and without the quotes didn't turn up the thread going all the way back to December. It couldn't have been that lon g ago could it? Seems like maybe three for four months back.

I'm not sure how bad it would be to simply use a pair of 0.1% resistors. T hat would add error somewhere between 0.1% and 0.2% (not thinking about it too hard). There will be other sources of error from this same effect else where (dividing the 5V signal source to 3.3 volts... or maybe not. Perhaps that can be mitigated by not using equal value resistors in the input and feedback paths.

It's funny that the other guys on the project don't seem to be able to thin k about this stuff. I guess they are mostly used to just plopping down chi ps. Heck, a new FPGA guy wanted to use a $20 Xilinx part instead of a $4-5 Gowin part!

I was happy to find I can get 10 nF C0G for under a nickle and 100 nF for a round a dime. I thought they didn't come that large. They do get large!

On 2020-11-04 22:00, Rickster C wrote:

Maybe something along the lines of this?

If you want it accurate, use feedback to adjust the duty cycle.

Jeroen Belleman

Version 4 SHEET 1 880 680 WIRE 224 -16 -208 -16 WIRE 352 -16 224 -16 WIRE 400 -16 352 -16 WIRE 512 -16 400 -16 WIRE 400 16 400 -16 WIRE 224 32 224 -16 WIRE 512 32 512 -16 WIRE 176 48 -48 48 WIRE 176 96 96 96 WIRE -208 112 -208 -16 WIRE -48 160 -48 48 WIRE -16 160 -48 160 WIRE 96 160 96 96 WIRE 96 160 64 160 WIRE 128 160 96 160 WIRE 224 160 224 112 WIRE 224 160 208 160 WIRE 288 160 224 160 WIRE 400 160 400 80 WIRE 400 160 368 160 WIRE 512 160 512 112 WIRE 512 160 400 160 WIRE 608 160 512 160 WIRE 224 224 224 160 WIRE 512 224 512 160 WIRE 96 240 96 160 WIRE 176 240 96 240 WIRE 400 272 400 160 WIRE -48 288 -48 160 WIRE 176 288 -48 288 WIRE -208 368 -208 192 WIRE 224 368 224 304 WIRE 224 368 -208 368 WIRE 352 368 224 368 WIRE 400 368 400 336 WIRE 400 368 352 368 WIRE 512 368 512 304 WIRE 512 368 400 368 FLAG 608 160 0 FLAG 352 -16 Vp FLAG 352 368 Vm SYMBOL sw 224 16 R0 SYMATTR InstName S1WINDOW 3 243 28 VTop 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value PULSE(-1 1 10u 10n 10n 10u 20u) SYMBOL res 224 144 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 1k SYMBOL cap 384 16 R0 SYMATTR InstName C1 SYMATTR Value 10u SYMBOL cap 384 272 R0 SYMATTR InstName C2 SYMATTR Value 10u SYMBOL res 496 16 R0 SYMATTR InstName R2 SYMATTR Value 10 SYMBOL res 496 208 R0 SYMATTR InstName R3 SYMATTR Value 100 TEXT -32 -96 Left 2 !.model switch sw(Ron=1 Roff=100k vt=0) TEXT 568 -88 Left 2 !.tran 20m

SYMATTR Value switch SYMBOL sw 224 208 R0 SYMATTR InstName S2 SYMATTR Value switch SYMBOL voltage -208 96 R0 SYMATTR InstName V1 SYMATTR Value 30V SYMBOL ind 272 176 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value 100u SYMBOL voltage 80 160 R90 WINDOW 0 -32 56 VBottom 2

On 2020-11-04 22:56, Jeroen Belleman wrote:

Sorry, I seem to have messed that up. Let's try again...

Jeroen Belleman

Version 4 SHEET 1 880 680 WIRE 224 -16 -208 -16 WIRE 352 -16 224 -16 WIRE 400 -16 352 -16 WIRE 512 -16 400 -16 WIRE 400 16 400 -16 WIRE 224 32 224 -16 WIRE 512 32 512 -16 WIRE 176 48 -48 48 WIRE 176 96 96 96 WIRE -208 112 -208 -16 WIRE -48 160 -48 48 WIRE -16 160 -48 160 WIRE 96 160 96 96 WIRE 96 160 64 160 WIRE 128 160 96 160 WIRE 224 160 224 112 WIRE 224 160 208 160 WIRE 288 160 224 160 WIRE 400 160 400 80 WIRE 400 160 368 160 WIRE 512 160 512 112 WIRE 512 160 400 160 WIRE 608 160 512 160 WIRE 224 224 224 160 WIRE 512 224 512 160 WIRE 96 240 96 160 WIRE 176 240 96 240 WIRE 400 272 400 160 WIRE -48 288 -48 160 WIRE 176 288 -48 288 WIRE -208 368 -208 192 WIRE 224 368 224 304 WIRE 224 368 -208 368 WIRE 352 368 224 368 WIRE 400 368 400 336 WIRE 400 368 352 368 WIRE 512 368 512 304 WIRE 512 368 400 368 FLAG 608 160 0 FLAG 352 -16 Vp FLAG 352 368 Vm SYMBOL sw 224 16 R0 SYMATTR InstName S1 SYMATTR Value switch SYMBOL sw 224 208 R0 SYMATTR InstName S2 SYMATTR Value switch SYMBOL voltage -208 96 R0 SYMATTR InstName V1 SYMATTR Value 30V SYMBOL ind 272 176 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value 100u SYMBOL voltage 80 160 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 243 28 VTop 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V2 SYMATTR Value PULSE(-1 1 10u 10n 10n 10u 20u) SYMBOL res 224 144 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R1 SYMATTR Value 1k SYMBOL cap 384 16 R0 SYMATTR InstName C1 SYMATTR Value 10u SYMBOL cap 384 272 R0 SYMATTR InstName C2 SYMATTR Value 10u SYMBOL res 496 16 R0 SYMATTR InstName R2 SYMATTR Value 10 SYMBOL res 496 208 R0 SYMATTR InstName R3 SYMATTR Value 100 TEXT -32 -96 Left 2 !.model switch sw(Ron=1 Roff=100k vt=0) TEXT 568 -88 Left 2 !.tran 20m

Could that be because they never get to know exactly what you're trying to achieve or why you're trying to achieve it?

Sorry, I don't understand. The inductor is grounded on the end that looks like it should be the output. The waveforms are all sloped square waves. What am I missing?

Not sure how I would determine how to apply feedback. What would be the me asurement to use to set the feedback? I don't have a way to measure the vo ltage. That's the point, to provide a reference voltage at Vcc/2 without a ctive monitoring or control. As someone said, a pair of resistors would do the job, but with ~0.2% accuracy using 0.1% resistors. Using a 50/50 outp ut with a single 0.1% resistor should provide ~0.1% accuracy which should b e pretty good.

This is the reference voltage for delta-sigma ADCs. The input signal is be ing stepped down from a 0-5V signal to a 3.3V input with a pair of 0.1% res istors. I wonder if it might make sense to drop the input signal in the sa me manner with a single resistor and a 2:1 timing cycle. So 5V -> 3.33V gi ving 0.1% on the dropping circuit. The accuracy of the sensor is ballpark

That makes me wonder what the sensor is sensing, how it is interfaced, how quickly it needs to sense, etc. If you want help then I think a schematic or part of a schematic would be much better than trying to describe the system in words.

Here is a link to the entire project, every bit of anything we have generat ed other than the latest schematics and software which recently was hosted on github. OpenVentBristol

When I buy an ADC chip I'm pretty sure they didn't know my requirements whe n they designed it. Yes, I just checked and we have not given access to an y semiconductor makers.

I appreciate any advice, but I don't have a full set of design documents to share. This is early days on the ADC design. I'm happy to share a verbal description with numbers, but it ain't that complicated.

The stuff being sensed is slow. The accuracy requirement is 3% if I recall . As I said, I'm looking to design the very tiny analog portion of this ci rcuit to preserve as much accuracy as I can. I really don't get why you wo uld need to understand every aspect of the sensors. One of them is a diffe rential pressure sensor for an orifice based flow rate sensor with a 0 to 6 kPa range with a 5 volt Vdd and an output scaled to Vdd. The signal being sensed is the human breath with a cycle time of 2 seconds minimum. Not su re of the upper frequency limit, but the initial edges start off rater stee ply, so perhaps 100 Hz.

Does any of that help?

Was that an attempt at sarcasm?

If it ain't complicated then a schematic drawn on the back of an enveleope and put in a scanner is the way I would likely convey the information. You may prefer to use a schematic capture tool which is fine. A simplified schematic showing the intended ADC interface circuitry is all that's needed.

Well if I were designing an interface from a sensor to an ADC then some of the things which might go through my mind are:

What is the source impedance and voltage or current range coming in from the sensor? Single ended or differential? What bandwidth do I need? What is the input impedance of the ADC and what voltage range is needed?

If you already have an intended circuit for that then it should easily fit on the back of an envelope.

Ok well 100 Hz is not the hardest frequency to deal with so do you already have an intended interface schematic from this sensor to your ADC?

?Does any of that help?

A little. A schematic would help a lot more than a Google drive with so many documents that there's no way I would know which, if any, is relevant.

be

as

?usp=sharing

to

to

al

the

ed

f 2

any

The 3.3V for Vcco is ratiod from the 5V sensor supply as the output value i s a fraction of the sensor supply. Vref at Diff + input is half supply vol tage. This may not be of value though. I'm pretty tired and I have a hund red other things I'm looking at. I don't think I work this hard when I get paid to do this stuff. But at least I get to do it on my own without havi ng to get consensus from a dozen other people some of who don't even do har dware.

I'm not sure if I can eliminate the voltage divider on the output of the 5V sensor if I make the two Ra resistors different values. That would let me eliminate the op amp as well. Yeah, I'm pretty sure I can. Put two Ra in series for the input and a single Ra from the D-S output and the input ran ge should be 0 to 5V. The ADC output sample rate is 200 SPS with a clock o f 33.556632 MHz. MEMS clocks are custom programmed at Digikey from blanks. Just as rapid delivery as stock crystals and they seem just as good, 25 p pm. I guess they laser a PN on them when they program them. Some table to p machine I'm sure. Keep the hopper filled and they can ship thousands per day.

There will be four or more ADC like this and they can all share the Vref si gnal. But I don't see that this is really much better than just a simple d ivider. Slightly better accuracy from the single resistor instead of two.

ASCII art was more fun than banging VHDL.

Ok well if I wanted a 1.65V reference for a sensitive ADC circuit I'd avoid square waves unless there was some good reason why they couldn't be avoided. Noise generated by fast transitions can easily end up in unexpected places.

I'd be more likely to have a look here if an external reference was needed:

Assuming it wasn't sufficient to just use two resistors, or two resistors and an op amp.

I'm tired too and will be away for the next 24 hours.

What is D-S Out?

Do you see the circuit? Oh, you are asking about D-S out, so I guess not. Searching I see I didn't mention this is a sigma-delta ADC input. As I said, I'm very tired at the moment.

Are you familiar with those? You should also note that I mentioned the bandwidth of the signal can be limited to 100 Hz with no ill effects which the capacitors should accomplish just fine.

Heck, the O2 sensor has a 90% rise time of 5 seconds. We'll run a low pass filter on the 200 Hz measurement to get lots of precision and low noise. The pressure sensors are the sort of signal you see on hospital TV shows.

I'm off to have a bit of ice cream and bed. Tired am I.

You wanted this to provide a reference??? Sorry, I assumed you wanted a power rail splitter.

Forget it then. You need a simple resistor divider, maybe buffered with a unit gain amp.

Jeroen Belleman

The resistor divider has tolerance related error while the 50/50 thing powe red from the same supply as the signal source canceling variations has rath er near zero error. Advice is fine, but I was asking if anyone remembered the thread where it was already discussed in detail.

There must be someone who was involved in the thread who remembers the subj ect line.