Hopefully it’s time to do some more dedicated evenings ?
(unfortunately I broke three cars commuting since october - don’t ask why)
I measured some of the LDOs on hand, some that I brought back from today’s meeting.
All LEDs were removed(*) and the source was 5v from USB.
One 1117 that I had for a while seeped 2.3mA (without load).
Another cheap pair of devices, each has a 1117 attached to a USB dongle (no uart lines) to extract 3.3v. One uses 115uA and the other 83uA. I am surprised - these are good numbers.
The 1117 board from @clixx_io used only 45uA. Again this was unexpected (David, what is the source for these?). An older @clixx_io board gulped 4.3mA… talk about variability of the 1117s.
The 1825S from @Harvs were not all the same. One took 125uA, two others 115uA, and the fourth I forget but I think it was also 115uA or less. The esp when in deep sleep uses 150uA with this one (it is now set up and running as a test).
Naturally the voltage drop is important here and this I did not yet test.
(*) About LEDs. I had great difficulty desoldering these tiny smd LEDs, especially when they are in a tight location. Then an accident showed me that with a little push the LED will just pop out, they seem to be very brittle. Nice and easy.
Variability in a spec like this is very common, you’ll note none of them are out of spec WRT the datasheet. The ones I gave you were spec’d at max 240uA leakage, with a 120uA typical. This is also a very temperature dependant spec.
It’s also common among low volume precision equipment manufacturers to test critical components to tighter tolerances than spec’d in the datasheet and pick the best ones. This is an expensive process though for manufacture.
However I see no reason to take that gamble here. The $0.16 regulator I linked above has a lower dropout voltage than the 1117, and are actually spec’d with a typical leakage of 25uA and max of 50uA.
Then we should test the RT9020 in a typical setup, and I have a few spare ESPs I can use now.
I am now trying to decide if I should standardize on 3xAA or 4xAA (or 1x18650 which @clixx_io offered a bucket of).
Using 4xAA will provide proper voltage throughout the life of the battery but the other two options that provide 4+v initially and drop to 3-v will need to be replaced before being depleted as some ESPs require over 3v. I am keeping an eye on the ones under test to see if the same ones are consistently disturbed by the low voltage.
It really did take me some time to locate these:
If you need any I have some spare. I had to buy more than a dozen to get a good price and so if anybody wants some I can bring them down next time.
With a dropout of 1v it is not suitable for 1x18650 but with 4xAA (5.4-4.?v) it may be acceptable. I am about to start testing it in this setup which should take a few weeks to complete.
Sure. Well it was just designed as a handy device for turning things on/off from 5-20V.
To do a special battery board would probably need a different type of regulator like a buck-boost type.
Something like this for example:
Maybe, but such devices are far less efficient than a dedicated LDO, and efficiency is a primary issue for us.
So keep in mind “drop out voltage” is a spec relative to max current. i.e. 1V dropout literally means at max output current, the output voltage will fall out of spec when the input voltage is less than 1V higher than it.
But still, I think the 1117 isn’t the best choice for this application.
I’m not convinced a switching reg is best for this app either. There may be some that would be ok, but they have their downsides. e.g. the one you’ve linked has too low an output current, and being a switched cap converter (not a buck boost) it will generally have fairly horrendous EMI. That’'s ok for some applications, but not for sensor nodes.
Quick note (have to rush out). Tried the nice 1117 from @clixx_io and the esp-201 did not come up. It delivers only 3.15v (from 4xAA delivering 5.4v). I will test another LDO.
[later] I am now using another 1117 that tested at 85mA when idle. On the ESP it now uses 120mA in deep sleep so not bad. The voltage I get is just under 3.5v, I will see how it handles when the batteries are weaker. I also set a static IP and this shaved about 1s from the active phase (which is still too long, WiFi up 3s after wakeup). This is an esp-201.
Hi Derryn,
have read through the thread and have probably missed the point but I’ll ask anyway.
Is there some particular reason the low power side, presumably a cpu and temp sensor, needs a regulated supply? Can’t this just be driven directly from battery?
Also, looking at Eyals battery voltage graph. If the battery pack is depleted in 5/6 days then by my calculations the average current draw is 16mA.
Cheers,
Steve
Thanks Eyal + Derryn for those comments.
Of course I can make up a different regulator board. I’m very sure that Texas-Instruments would have some IC that does everything battery management wise.
By the way, I have some tiny Solar Cells that generate high voltage at low currents. I’ll bring those along next time along with my bucket of 18650’s that I have left.
G’day Steve,
Depends on which iteration of this conversation we’re talking about 
With some maths in the other thread, we’re back on just using an ESP with no additional cpu. Essentially it came to do, if the time between sending readings < 10min, the sleep current was significantly less than the active current, so there wasn’t much point optimising it.
On the other hand, you’re right. It depends on which microcontroller one uses and its voltage range. The MSP430’s high end is only 3.6V, so that’s why I included a low power reg, but a different micro may have been a better choice anyway. We never really looked into it as far as we could have, as when we did, we came to the conclusion we didn’t need it
Fair enough. Did you work out what was causing the high average current draw? If you’re cycling through wake/sleep, averaging 16mA would suggest not sleeping long enough or something very thirsty in the wake cycle.
The issue is this. When fully charged, a set of 3xAA or 1x18650 deliver about 4v. This is rather high for the esp, which will then drain 600-800uA in deep sleep (instead of 20-30uA). If it only wakes up every few minutes then this current becomes significant.
This depends on what the battery pack is and which esp is used. I now improved my app to complete all processing in less than one second on an esp-07. The -01 is also good but the esp-12 (and probably others with 4MB flash) takes over 4 seconds.
With a regulator I can use 4xAA which directly deliver over 5.5v that is too high for the esp. I am not yet decided on what setup will work best for my case.
If I read once every 5 minutes then I am active for 1s and asleep for 299s. A sleep current of 600uA will use up the battery twice as fast as 100uA, given that the 1s of activity uses about 100mAsec. I expect 3xAA to last a few weeks now, the original tests used a very hungry regulator (1-2mA Iq). I have a few tests running now.
At this early stage of the project there are too many variables that were not closely observed yet, and we are studying this situation with an open mind.
Hi all,
I went away and found this IC specifically made for generating 3.3V at 91% efficiency from Batteries with continuous discharge. It’s made by the same maker as the other VREGs I was using:
It says that the quiescent current is 12uA with input ranges 0.85V to 4.5V (with step up) to 3.3V
Let me know if it’s worth getting some.
Interesting, very low Iq.
4xAA deliver much more than 4.5v so the usual 5.5v max most LDOs expect is more like it.
Is 91% good? I do not know what is expected from a good quality device.
Still, 3xAA will be good if it can boost to 3.3v when the batteries head to 3v and below.
We need to study the data sheet in detail first.
Nice device.
There’s quite a few tricks to reading datasheets for switching converters. Very rare you will ever get close to their headline specs.
Perhaps we could go over the datasheet tomorrow night if you like.
Good idea.