A solar supercap powered Moteino (15Farad charged by BQ25504)

[WhiteHare]

... low ESR, low leakage and linear voltage drop to very low voltage with constant current, and these I think are the features people need to look for in this type of application.

Is low leakage predictive of low self-discharge rate, or are they two different animals?  Datasheets report leakage far more than they do self-discharge rates.

[perky]

Is low leakage predictive of low self-discharge rate, or are they two different animals?  Datasheets report leakage far more than they do self-discharge rates.

A more detailed question than what you might think

To a first approximation the leakage current at rated voltage can be used to calaculate a leakage resistance, which is in effect in parallel with the capacitance, so you can uses that to calculate the RC constant for self discharge (see Randles model in the below link). It's complicated a little in that the leakage resistance actually has a time domain component and varies, but appears from the link below to increase over time rather than decrease, so I think the simplistic approach would yield a conservative estimate:

https://www.gamry.com/application-notes/battery-research/testing-electrochemical-capacitors-cyclic-voltammetry-leakage-current/

Edit: Note that this says that some manufacturers actually quote the leakage current after a certain number of hours after initial charging, specifically because the leakage resistance goes up over time...

Mark.

[WhiteHare]

That's really good to know.

By the way, the test finally completed.   The RFM69HW Moteino powered by Perky's cap sent a total of 258,210 packets!  The loaded voltage at the time of failure (i.e. no more packets being sent) was 1.68v.

In theory, switching to an RFM69W would yield 2-3x that number on the same test, if it were to transmit at 16ma instead of the ~40ma during Tx used by the RFM69HW in this test. 

Thanks, Perky!

[ChemE]

In theory, switching to an RFM69W would yield 2-3x that number on the same test, if it were to transmit at 16ma instead of the ~40ma during Tx used by the RFM69HW in this test. 

I have a RFM69CW on a breakout board I could send you to test out if you'd like.  In a perfect world you'd send back an AB1815 BoB with the chip soldered on but that is up to you.

[WhiteHare]

Thanks for the offer, but yesterday after I couldn't get Joe's code to work I decided to simply order some RFM69W's.   I opted for RFM69W's because they have the same landing pattern and pinout as  the RFM69HW's.  So, it's a simpler retrofit.

Also, in more regular use, with a far more conservative transmit cycle (say, once every 5 minutes instead of once every 262ms, I suspect the difference between the RFM69HW and RFM69W on current consumption will not be so significant.  That's because the sleep current should be more dominant in that scenario.

Also, with solar, it may not matter much anyway.  Why?  I just now switched the Moteino to report the loaded voltage once every 63 seconds, and then I hooked it up to the solar charger, which had already been charging Perky's cap for a while from just ambient sunlight in the room.  i.e. the panel is not in direct sunlight.  Under those conditions, here's how quickly it's charging from a 60x150mm 3.5v solar panel that's pointed toward the window (which it's about 5 feet away from):

Code Select Expand

1,[Sender:2],000,[RX_RSSI:-39]
2,[Sender:2],204,[RX_RSSI:-38]
3,[Sender:2],207,[RX_RSSI:-39]
4,[Sender:2],210,[RX_RSSI:-38]
5,[Sender:2],211,[RX_RSSI:-39]
6,[Sender:2],214,[RX_RSSI:-38]
7,[Sender:2],216,[RX_RSSI:-39]
8,[Sender:2],219,[RX_RSSI:-39]
9,[Sender:2],222,[RX_RSSI:-39]
10,[Sender:2],223,[RX_RSSI:-38]
11,[Sender:2],226,[RX_RSSI:-39]
12,[Sender:2],228,[RX_RSSI:-38]
13,[Sender:2],231,[RX_RSSI:-40]
14,[Sender:2],233,[RX_RSSI:-37]
15,[Sender:2],234,[RX_RSSI:-36]
16,[Sender:2],236,[RX_RSSI:-37]
17,[Sender:2],237,[RX_RSSI:-37]
18,[Sender:2],239,[RX_RSSI:-37]
19,[Sender:2],243,[RX_RSSI:-38]
20,[Sender:2],245,[RX_RSSI:-38]
21,[Sender:2],247,[RX_RSSI:-37]
22,[Sender:2],249,[RX_RSSI:-37]
23,[Sender:2],251,[RX_RSSI:-37]
24,[Sender:2],253,[RX_RSSI:-38]
25,[Sender:2],256,[RX_RSSI:-38]
26,[Sender:2],257,[RX_RSSI:-38]
27,[Sender:2],260,[RX_RSSI:-37]
28,[Sender:2],262,[RX_RSSI:-38]
29,[Sender:2],264,[RX_RSSI:-38]
30,[Sender:2],266,[RX_RSSI:-36]
31,[Sender:2],268,[RX_RSSI:-38]
32,[Sender:2],271,[RX_RSSI:-36]
33,[Sender:2],272,[RX_RSSI:-37]
34,[Sender:2],274,[RX_RSSI:-37]

Granted, it's a clear, sunny, afternoon day here, but I'm sure you get the idea.

[perky]

That's really good to know.

By the way, the test finally completed.   The RFM69HW Moteino powered by Perky's cap sent a total of 258,210 packets!  The loaded voltage at the time of failure (i.e. no more packets being sent) was 1.68v.

In theory, switching to an RFM69W would yield 2-3x that number on the same test, if it were to transmit at 16ma instead of the ~40ma during Tx used by the RFM69HW in this test. 

Thanks, Perky!

Great result! I thought it might be longer, but that's excellent

BTW, assuming you transmit for 1ms roughtly at a time, then from dV =Idt/C you should get about 5.3uV drop per transfer, and that yields ~362000 transfers for 3.6V to 1.68V. Take into account the MCU active current, capacitor self discharge and MCU/radio sleep currents and you're probably not far off the calculated value.

Mark.

[WhiteHare]

Now that we have some solid data to work from, I suspect that in many cases the fancy solar charger won't be needed.  A simple solar cell, Perky's cap, and a blocking diode (to prevent discharge through the panel when there's insufficient light) may turn out to be all that's needed.

[WhiteHare]

I haven't dissected the Perky cap, but from the looks of it, it is most likely constructed from two supercaps in series.  That makes sense, since many (all?) of this genre of supercap have a max voltage of 2.7v.  Furthermore,  the Perky cap, which is rated at 5v, has a sister 7.5F supercap with a 5.4v rating.

So, generalizing, I suspect the optimal arrangement will be two in-series supercaps of the same Farad value, whether they come already packaged that way (as in the Perky cap example) or whether they get connected that way in-circuit.  The benefit of that, as compared to a single 2.7v supercap with the same Farad rating, is that more of the most desirable voltage (which, practically speaking, means 1.8v-3.6v if we are to stay in spec for a non-LDO Moteino) is available to power the Moteino.

By the way, it turns out I just now completed a drain test on a 25F 2.7v supercap (this one: https://www.digikey.com/product-detail/en/nesscap-co-ltd/ESHSR-0025C0-002R7/589-1003-ND/946803) , connected to the Moteino.  Its advertised ESR is 21mOhms, which is less than the Perky cap.  However, running from 2.7v down to 1.8v produced only about 445K transmissions.  The only difference in test procedure as compared to the Perky cap test was that I reduced the interval between transmissions to about 102ms.  I didn't reduce the rest interval further than that  so as to keep the RFM69HW within its datasheet recommended 1% max duty cycle. 

[WhiteHare]

Applying the new test procedure to the 15F supercap from the OP, so as to yield an apples-to-apples comparison, yielded a surprise:  it only transmitted 2,452 packets before cratering (see attached log).  So, I guess that particular supercap really does need longer rest intervals between transmissions....  Yeah, that potentially does cramp its usefulness compared to the newer supercaps.

Well, out with the old and in with the new!   

[perky]

However, running from 2.7v down to 1.8v produced only about 445K transmissions.

In the 7.5F case you got ~280k transmissions with a 3.6V -1.65V = 1.95V drop, and the 25F cap you got ~445k transmissions with 2.7V - 1.8V = 0.9V drop.

So is that consistent? Well both have linear discharge curves, so you should have got 280k * (0.95/1.95) * (25/7.5) = 454k. Pretty damn close to what you got. So these figures are entirely consistent.

As for the 15F cap, you're starting at 2.7V which is way into the vertical part of the discharge curve for that cap, so not surprising you got a very low figure

Mark.

[WhiteHare]

So, which do you think would perform better?  Two of https://www.digikey.com/product-detail/en/illinois-capacitor/106DCN2R7M/1572-1287-ND/5410638 in series, going from 3.6v-1.8v, or just one of https://www.digikey.com/product-detail/en/vishay-bc-components/MAL222091003E3/4830PHBK-ND/5809949 by itself, going from 2.7v-1.8v?  Both models of supercap are of type ELDC.  The BOM price would be nearly a wash.  The 20F, though, has an ESR of just 18mOhm, versus a relatively higher ESR for each of the two 10F's in series.  Each 10F has a higher ESR by itself than the 20F, but I guess putting them in series would  further double (?) their effective ESR as well?  Not sure if that's how the ESR math works. 

On the other hand, unlike the single 20F, the two 10F's in series would not be operating near the limit of their rated voltage, so perhaps that might confer a lower self discharge rate if holding a full charge for a long time?

[perky]

Slight problem with your maths here, putting two capacitors of the same value in series halves the overall capacitance

So you're asking whether a 5F capacitor from 3.6V to 1.8V is better than a 20F from 2.7V to 1.8V.  The latter is better because even though the voltage range is halved, it has 4 times the capacitance.

If you had asked whether two 20F caps in series from 3.6V to 1.8V is better than one 20F cap of the same type from 2.7V to 1.8V then the answer is they are the same (you've halved the voltage range, but it has double the capacitance), it then comes down to leakage and ESR. I think that it's probably better for the latter as the ESR would be half and assuming the leakage resistance is the same for both and doesn't change with voltage.

If you use the Randles model for the capacitor (i.e. an ideal capacitance in parallel with the leakage resistance, and then an ESR resistance in series) then putting two capacitors in series will add their respective ESRs.

BTW I think your original suggestion of just charging the cap with a solar panel with a blocking diode is good, although I would replace that diode with an ideal diode circuit to limit the voltage drop to a few 10s of millivolts (see attached)

Mark.

[WhiteHare]

Thanks!  That's a very helpful explanation.   

As it turns out, I have both the caps I referenced in my prior post, and at this very moment I'm performing a drain test on the 10F supercap.  I'll post the results once it completes.  Probably its main virtue is that, compared to other 2.7v 10F's on digikey, it's relatively inexpensive (< $2).

Anyhow, I'll start to focus on leakage rates now and try to find some supercaps that are especially good in that department.

[perky]

I've just updated my post with an ideal diode circuit that I use for other things. It's got about 1uA leakage and about 30mV drop, that's far better than a 300mV schottky

Mark.

[WhiteHare]

BTW I think your original suggestion of just charging the cap with a solar panel with a blocking diode is good, although I would replace that diode with an ideal diode circuit to limit the voltage drop to a few 10s of millivolts (see attached)

In your schematic, what does DNF refer to?  Its schematic symbol seems to indicate it's some kind of capacitor?  If so, what value?