-
Hello all,
I have been contemplating making my own voltage reference to use both for checking my meters for similarity between them, as well as a means to check some projects that I have in mind for the coming months.
Naturally I have read the TI datasheet several times as well as the addition BB application notes.
In honesty I have mostly just copied the data sheets in my schematics, however, as expected the datasheets made reference to how important grounding, particularly star grounding, is for a reliable voltage reference.
I was hoping that those more experienced than I would take a look at my design to see if anything is amiss, or could be improved.
If I could get an extra 2 digits out of it (10.00002vdc), compared to its base accuracy (10.0025vdc), that would be the best case scenario, even if that is wishful thinking.
For what its worth, besides using it to compare commercial meters, I plan to try and make my own version of Scullcom 6 digit meter, along with an integrated constant current load, which would also allow me to use the meter for power and resistance calculations, it would included dual adc channels so I could measure both at the same time. Control via STM32/Arm Cortex M3, at least that's my thinking at this stage.
I realize that I could simply buy a reference of Ebay/Banggood/AliExpress with the AD584 for sub $20AUD, which will be a reasonable amount cheaper than what I am planing, but thought the learning experience would be worth the additional cost.
So this hopefully will be the first of my successful steps on a road to bigger and better things.
Thank you
-
The buffer behind the filter section is odd. What is called noise filter is more like a buffer that is tolerant to capacitive load, except that the resistor / cap values are a bit odd.
Fitlering with R11,C7,C9 is also not very effective.
Usually R11 and C7 / C9 would be smaller (e.g. 50-100 Ohms and 100 nF) to make a proper buffer. The extra buffer is more like making things worse as it is not very tolerant to capacitance at the output.
Filtering would be better done at the input to the "filter stage", using the 3 resistors that are allready there for averaging and add a capacitor to ground to make it some filter. Filtering a reference is anyway not very effective as the critical part are usually the low frequencies. So filtering is more something to add if it is easy as there is already a buffer anyway. The first stage filtering is allread at the noise reduction pin of the reference chips.
There is no real need to adjust all 3 references. Adjustment at one reference and than a better quality pot and resistor is likely better.
The capacitors used for noise filtering can be an issue. X7R capacitor can be a bit microphonic and may have leakage issues. One should consider film caps (e.g. Wima MKS) instead.
-
Thank you for the incites, I have taken your advice and adjusted the values of R11, C7 and C9.
Also, removed the buffer section. The only reason I included it in the first place was so that the 2nd op-amp of the OPA2277 wouldn't be floating in the breeze.
Would these be good for filtering as replacements for the ceramic filter caps?
https://datasheet.lcsc.com/szlcsc/PANASONIC-ECPU1C104MA5_C150376.pdf
In regards to the trimmer, would it be better to use a smaller value, so as to have more resolution on the 25 turns? (1k?) Or would this make it unusable.
My thoughts were that I would only install 1 trimmer at a time, and then if I was unable to trim it with one, I would install the second and so on.
If it happens that I don't need any trimming I would leave all uninstalled.
-
There is also a single OP counterpart (OPA277) - no need to use the dual part.
For the trimmer 10 K looks like a good value - 1 K may get relatively warm. If more adjustment range is needed, one could reduce the resistor to the trim input. If really needed one could add a fixed resistor part for more coarse trim. 3.3 M precision resistors may be a bit tricky to get anyway - I would consider 2-3x1 M in series.
The panasonic caps don't tell which dielectric though likely PPS which should be OK.
With only 100 nF C0G ceramic may be an alternative. For the 2.2 µF at the Ref. filters I would consider THT caps - for just a single unit and enough space this is less of a problem than soldering relatively sensitive SMD film caps. -
Main reason I was considering the OPA2277 was due to the chip shortage, it seems that they are more common and available than the single OPA277.
The Panasonic are polypropylene film, although they are also available in PEN film.
When you say the 2.2µF filter cap as a single unit, are you implying that I should tie all the filter pins together and use a single filter cap for each IC?
I was under the impression that the 3.3MΩ wouldn't need a tight tolerance, and a TCR of only 100ppm or better according to the datasheet, which matches the component in my BOM.
Although the 3.3MΩ is also available in 50ppm, if thats still not good enough, then I have two options for better TCR, change package (currently 0603) or change value.
Vishay make a 2ppm 3MΩ 2010 package, this would be the best available for the 1MΩ also, as that one is also 2ppm with a 2010 package.
Given they are $25 each, versus $0.05, in your opinion would it be worth the expense? Honestly I don't mind spending it if the result is worth while. -
There are other precision OP available than the OPA277. Given the level of accuracy, a OPA207 may be OK too.
How stable the 3.3 M resistor needs to be depends on the amout of trim used. The change in voltage depends on that resistor. With a small effect from 3.3 M a 100 pmm/K resistor should be good enough. No need to go nuts on that resistor.
For the filtering it would normally be 1-2.2 µF for each chip. I have not thought about it, but connecting the filter pins togehter should also work, though would not give an advantage as one would also need a larger cap. The only point would be also having a 7 V output.
I very much doubt the panasonic caps would be PP - this is way to heat sensitive to be soldered SMD with normal solder, let alone lead free.
PP caps would normally also be quite large volume. THT PEN or similar is the easier choice - SMD film caps are limited in capacitance and heat sensitive.
-
You may very well be right about it not being PP, as I was only relying on the vendors categories, rather than anything direct from Panasonic.
The downside of going from a double to single op amp package is that the pinout's are different, however, in taking you advice regarding the single op amp package, this allows me access to trimming of the offset voltage, which is a decent reason in itself to change, the datasheet indicates a 20k trimmer, so like the rest of my trimmers I would use ones from Bournes with a TCR of 100ppm, if its worth while.
Having said that I think having it on the PCB doesn't hurt, and one can always just leave it unpopulated.
Sorry with the filter cap, I was talking about a single cap (C12) is this ok, in the same sense, should it be ~330nF given there is 3 infeeds.
VS 
-
The higher grade OPs with an offset below some 100 µV don't really need an offset trim. The tendency is to make the TC worse and add uncertainty from the trimmer wiper. It won't hurt to have the footprint, but I would likely not populate it. The residual offset is kind of part of the reference.
For the filtering it depends on the use. A DMM is usually not very sensitive to higher frequency noise. So there is little to no difference in having 100 nF or 1 µF. The main frequency range that can make a small difference is some 25 Hz or 2.5 Hz depending on the meter (the AZ mode with some 20 ms or 200 ms sampling and than a break for zero reading gives some noise aliasing in this range). The NR pin is supposed to use 7 K resistors and is thus the more effective point for low frequency filtering. Chances are the 7 K is a reasonable compromise for added noise and filtering efficiency. It still does not hurt to have space for a larger cap (or a 2nd in parallel) for C12, just in case. -
No worries, I think I'll just leave the footprints there, but without the intention of populating them.
Rather than increasing the size of C12, I have added 2 additional caps, at the leg of each output, rather than the single C12.
I've changed these 3 caps to 1210 PEN caps.
https://content.kemet.com/datasheets/KEM_F3078_LDE.pdf
The upside of SMD rather than THT caps is that if I find that I need more than 100nF, I can stack them.
The other option would be to use 2220 1µF caps that are also PEN.