Precision Voltage Source from LI-Batteries

[Noobster]

Hi,
my problem might be too specific but i did not find an adequate solution in my search. I am looking for the following:
-Stable voltage source from rechargeable batteries (got those) for measuring applications
- fixed voltage in the common Range 2.5-5V would be ok, adjustable better
-output 50-200mA
-short term (<2 hours) low drift (temperature range should be +-2°C during that time)
-budget < 5 Dollar
Sadly i can´t give an answer on what precision i want to achieve, it is more like this design challenge has come up a few times and i would like to have a go to solution that is not too complicated. The Idea is as precise as possible so i have one thing less to worry about.

[DavidKo]

What means stable? First what came into my mind was TL431 (or equivalent). See datasheet for recommended connections.

[opa627bm]

Use Ref5050 will be a good starting point.

[dobsonr741]

How about an AD584 module from eBay?
Or if you can afford 10x times better goodness at 10x price then ADR1399H-EBZ.

[MasterT]

1. High current OPA :
   TCA0372 or LA6500
2. LDO:
   AMS1117, LD1117

[Noobster]

Hi,
thanks for all the suggestions. What would be the best way to buffer the voltage reference (opa as voltage follower ?). Does it (for thermal and load behavior) make more sense to use voltage reference plus follower or a linear regulator ? Also i am concerned with the voltage drifting as the battery voltage decreases.

[Conrad Hoffman]

Cheapest thing I can think of would be some voltage reference that hits whatever sweet spot between performance and price that you like. Follow that with a voltage divider made from inexpensive 1% metal film resistors. Switches are expensive, so use a header with jumpers to set the voltage. Follow that with an opamp that, again, hits whatever sweet spot you like, as a buffer. Put a simple NPN transistor at the output, in the feedback loop, so you get enough current and not heat up the upstream circuits.

[RoGeorge]

For <$5 budget you can try a LM317 with a resistive voltage divider, something like this:
https://www.st.com/resource/en/datasheet/lm317.pdf

[magic]

LM317 could work as a power buffer, but I'm not sure if it will perform well as a reference combined with a buffer.

Staying in the realm of ultra low cost solutions, it's very easy to put LM317 in the feedback loop of TL431 so that the latter acts as the reference. You pretty much need just the two ICs and one resistor (from LM317 output to adjust, to supply TL431 bias, 1kΩ will do). Minimum output voltage is 3.75V.

There are some reference ICs with output force and sense pins, perhaps these could drive an external buffer to eliminate a separate opamp from the BOM.


Important question: what supply voltage is available?

[Kleinstein]

5$ is a very tight budget and would not really allow for much precision.
A voltage regulator would be cheap, but when loaded the regulator heats up and this hampers accuracy. So one should have a separate reference / regulator and power transistor of some kind.
A cheap version may be with a TL431 and power transistor.

[Noobster]

Thanks again for the suggestions.
@magic: I have multiple packs with 3 cells so 11-12.6V also typical 18v batteries from power tools.
@Conrad Hoffman: that sounds very good

Here is also where my rudimentary knowledge stops, what to look for in the single components: Are all OPAs precise buffers? etc. Concrete component suggestions would also be helpful (i doubt i would find the best fit).
Generally it probably makes sense to spacially (thermally) divide the "power stage" from the reference (and the buffer?).

[magic]

This is what I suggested. I have never characterized this circuit for accuracy because I have only used a variant of it to build a CC/CV float battery charger (LM317 was doing CC part, not included here), but it should be much better than LM317 alone.

With such high input voltage LM317 works fine.

TL431 could be replaced with something better. It's certainly possible to have separate reference and opamp chips, but perhaps some integrated reference with force/sense capability could suffice. Alternatively, a 5V ref and opamp at unity gain, then no particularly accurate resistors are needed. I wonder if those AD584 from eBay that somebody mentioned above would work, as this covers everything - reference, opamp, feedback divider.

[iMo]

An uA723 oldie again? With a PNP power transistor for those 200mA?
Not sure it fits <$5 today..
The TC will be rather low, perhaps comparable with 317 or 431..

[magic]

Well, of course.

Not sure how you would use a PNP, but NPN emitter follower (or LM317 - yes, I know I'm boring ) is an obvious option. Current limiting is possible using the 723's facilities.

I recall a whole thread about getting the most out of that chip.
Also, National's version was the winner for long term stability.

[RoGeorge]

IIRC 431 is better in thermal stability, and 723 was tricky for lower voltages.

TI has many SMD voltage references, some might be slightly better than TL431 and still cheap, e.g. REF35 $1 in kpcs https://www.ti.com/power-management/voltage-reference/overview.html thought that's a little more than $3/1pcs at mouser.

[Conrad Hoffman]

Not sure how good the performance might be, but you could make a discrete bandgap reference with a couple transistors glued together to keep the temperature the same. Cost would be pennies. Google bandgap references.

[iMo]

Well, of course.

Not sure how you would use a PNP, but NPN emitter follower (or LM317 - yes, I know I'm boring ) is an obvious option. Current limiting is possible using the 723's facilities.

I recall a whole thread about getting the most out of that chip.
Also, National's version was the winner for long term stability.

Yep, there is NPN or PNP possible with 723.. I've been using in my floating +/-15V source the PNPs there, as well as I added the foldback current limiting which works nice too with 723..

[magic]

Not sure how good the performance might be, but you could make a discrete bandgap reference with a couple transistors glued together to keep the temperature the same. Cost would be pennies. Google bandgap references.

Unlikely to be competitive with IC references, because for any 1°C of ΔT you get 2mV of ΔVbe and this is amplified over 10x by typical bandgap circuits by the time you reach 1.2V, and OP wanted 2.5~5V.

However, you can select two TL431 such that the sum of their drift is close to zero at the temperature of interest, glue them together, and build my circuit using the second TL431 in place of R2.
 

[iMo]

I think the OP should have refined his/her requirements a bit, as we may suggest almost anything here in this section 

[RoGeorge]

For discrete components, maybe a low noise, LED based reference then?
https://www.eevblog.com/forum/chat/leds-as-a-low-noise-shunt-regulator/
https://duckduckgo.com/?q=walt+jung+led+regulator

Not mine, saved from diyaudio forum IIRC:

[Noobster]

"I think the OP should have refined his/her requirements a bit, as we may suggest almost anything here in this section". Yes, and i love the suggestions 
Let me try to narrow it down, i think most important is
-temperature drift
-drift from input voltage decreasing

has anyone tries simple ways of temperature "control"? like a ptc resistor and make one expoxy block out of the circuit, put in styrofoam ?

[iMo]

..Let me try to narrow it down, i think most important is
-temperature drift
-drift from input voltage decreasing

has anyone tries simple ways of temperature "control"? like a ptc resistor and make one expoxy block out of the circuit, put in styrofoam ?

You have to make the specification like this, for example:
0. input voltage 8V to 15V
1. output voltage 5V
2. output current max 200mA
3. output voltage temp drift <20ppm/C
4. long term drift <100ppm/1kHr
5. line regulation <40ppm/V
6. load regulation (10mA to 100mA) <40ppm
7. output noise (1Hz to 1kHz) <10uV rms
etc.

[Noobster]

0-2 should be answered. For the rest: 4 doesn´t matter, the rest as low as possible for the budget with special emphasis on thermal drift and line regulation. One example i use this on is distance measurement with hall effect sensors where i measure <0.5mV (this sucks with just a cheapo linear regulator). I am a mechanical engineer so i don´t enjoy building a new circuit everytime to fit my needs exactly, therefore i want something that works everytime. For example for mosfets i buy just one type that is overkill for most of my applications but i have them here and don´t need to order new ones/check datasheets each time. I know this seems strange, sadly for me electronics is something i have to deal with but don´t enjoy. Thanks for all the suggestions, to me the solution voltage reference->opa +transistor seems the most versatile. What do i look for in opa and transistor selection?

[lyxmoo]

LT1372 typical application Li-Ion Cell to 5V SEPIC Converter.

take a look.

[Noobster]

I did some more searching and stumbled across the LT6658 which would fit the budget. What do you guys think of it ? I doubt it is easy to achieve a better result with a diy circuit (at least for me  )?
@lyxmoo looks good for high loads but seems noisy to me (again maybe im wrong)