Micro Moteino?

[joelucid]

Only unfortunately the core idea is based on a misunderstanding of ohms law:

https://lowpowerlab.com/forum/index.php/topic,1269.msg8530.html#msg8530

[kobuki]

Hmm, but he does use a cap for storing power... See here.

[joelucid]

Hmm, but he does use a cap for storing power... See here.

The cap is only there for sleep, it's pretty small. I think it's a combination of a wake-up device plus sleep power provider. The piece I would implement in Tino by using a ultra low power HW watchdog timer.

Some more thoughts on the effectiveness of the booster:

Obviously at best Vout * Iout = Vin * Iin because of conservation of energy. Therefore Vout/Vin = Iin/Iout and every increase in Vout needs to be paid for by increasing Iin. Now when the internal resistance RI of the battery is low that's no problem. However when it is large the increased current will reduce Vin, working against the booster. So for a given RI where's the sweet spot? Let's call the coin cell open voltage Vbat and equivalent input resistance of the booster Rin.

Then Vin = Vbat * Rin / (RI + Rin) and Iin = Vbat / (RI + Rin). For a given Rin we extract a power of Vin * Iin from the battery which becomes:

Vbat * Rin / (RI + Rin) * Vbat / (RI + Rin) =  Vbat^2 * Rin / (RI + Rin)^2

The derivative is

- Vbat^2 ( Rin - RI ) / (Rin + RI)^3.

It has a root at Rin = RI. Therefore you extract the maximum power from the coin cell at an Rin that's equal to the internal resistance. You can then decide how to transform that power - btw, converting it to 3.6V is hardly going to prolong the lifetime because 328p/radio use more power than at lower voltages.

At first that seems to imply that as long as Vin doesn't break down to around 1.5V without the booster you could still get some gains using it. But graph the extracted power at http://www.derivative-calculator.net/#expr=b%5E2x%2F%28a%2Bx%29%2A1%2F%28a%2Bx%29 and you'll see how flat it is around it's maximum. I don't see this even compensating for the booster efficiency losses.

All this is consistent with the fact that ultra low average pulsed loads can extract close to, if not more than, nominal capacity from a coin cell (see the listen mode thread for sources).

In the end you want to provide a platform with as little conversion losses as possible (including booster and cap charging losses), run it at as low a voltage as possible and support developers in writing code that creates the shortest possible pulses and the lowest average loads.

Of course a booster does make sense for certain scenarios (I want to include one using the 328p). I'm looking at 1) providing a larger voltage at low power (scenario PIR sensor) and 2) providing high power pulses by precharging a cap (sprinkler controller). But powering the radio and the 328p using it is a step back in efficiency.

Joe

[kobuki]

The author of the blog has already spent a year refining his designs. While your calculations are probably correct, I doubt he's some fool experimenting for naught for a year. I especially like the idea of being able to use a single, common AA cell for operation. You both have different design ideas but only time will tell which one works better or at all. The solution in the blog can be further refined or fixed. Such as your own designs with Tino. Neither one is a finished, well tested, available product (vs. the Moteino, which already is). There are other projects that follow a similar design (using a boost converter), albeit without a cap AND have a pretty long battery life. Even many common battery-operated consumer devices do. They can suck a 1.5V cell down to 0.8V or less until they demand another set of batteries.

[kobuki]

@Joe: reflecting to your latest edit. As a consumer, I only care about battery life (and price). If I'm not bothered with swapping batteries for like 1-1.5-2 years, then I think it's fine Whatever solution lies inside.

[joelucid]

@kobuki,

I'm just trying to learn from these projects and incorporate the best ideas in my design. ULPNode has already convinced me to put a ultra low power HW watchdog on the Tino to at least match its sleep power consumption. Looking critically at all aspects from a coin cell perspective. Necessary for me, you can just pick up the final product as you say :-) - at least the ones that ship at some point.

Now this is also a sport - so 2 years battery life are unacceptable to me :-)

Oh yeah, btw, boost converters are great for single non coin cell operation.

[kobuki]

Yeah, as a consumer, my needs are different. As an engineer, my interests are again different, although I'm no EE design expert guru at all (I have an SW engineer background instead), only a hobbyist wanting to learn. You might not be satisfied with 2 years "only", but is it worth the effort to achieve a lot more? I think in this case it's worth thinking about the "what is good enough?" approach You might be able to produce something that only last 2 years on this coin cell... But you already have a product then, ready to be refined. And the world doesn't stop, so we might see some new type of battery within 2 years, existing ones increasing in capacity while maintaining the same size, etc - so I'm not saying it will, but it at some point might become pointless, trying to achieve a significantly higher goal in the short term. Please just think of it as me thinking out loud, I'm not trying to argue with your aims. But I also expect the IoT world and thus battery-operated technological advances explode in the following years (well, explosion has already begun, in fact...)

BTW, should you ever start selling Tinos, expect me to be one of your first customers

[joelucid]

You might not be satisfied with 2 years "only", but is it worth the effort to achieve a lot more? I think in this case it's worth thinking about the "what is good enough?" approach

Yeah - if it were only so easy once you've been bitten by the bug. I blame Tom for turning me on to this  - but hey, I'm pretty sure I could stop if I wanted to ...

[TomWS]

You might not be satisfied with 2 years "only", but is it worth the effort to achieve a lot more? I think in this case it's worth thinking about the "what i
Yeah - if it were only so easy once you've been bitten by the bug. I blame Tom for turning me on to this  - but hey, I'm pretty sure I could stop if I wanted to ...

In the words of Etta James:

It's a world full of trouble and a world full of pain,
I'll take the problem, but I won't take no blame.

Tom

[joelucid]

I doubt he's some fool experimenting for naught for a year.

I came back to this topic today as I wanted to check if an ultra low power voltage supervisor might be superior to the ULP watchdog timer I'm planning to use on my TinoV2. After all being able to get an interrupt on low voltage would be a nice feature on top of providing a regular wake signal.

After I couldn't find anything remotely interesting I checked what the ULPnode uses. Turns out it's the http://www.onsemi.com/PowerSolutions/product.do?id=NCP302 with a quiescent current of 500nA!  That's more than my total energy budget for TinoV2 including 328p sleep current, WDT, measurement, radio sleep and TX.