Hi guys - I’m doing some looking into using an Imp in battery powered products. I like the ease of the platform, but naturally wifi and battery may not be the best combination for long life. Now, I recognize there are some great engineering tactics in both hardware and software to get the best out of battery life and so I’d like to learn about those.
For instance, I looked the Electric Imp Nora, I like the idea of the boost for when the wi-fi radio is needed.
Another idea is to save data locally and then do a large data to cloud when needed. This goes hand in hand with turning on the radio only when needed.
I’m looking for some real world examples of Imp and battery. What have you done to get the best battery life? What do you recommend for new designs?
Nora is a pretty good example; it minimizes sleep power and time awake, whilst staying low cost. There’s also a great guide on designing battery powered devices here https://electricimp.com/docs/hardware/power_story/ which is well worth a read.
For local data storage, there’s a SPI flash filesystem library about to come out which helps make storing lots of data locally much easier.
Many customers are shipping battery powered devices (Toymail, Edyn, Hiku, Budweiser, etc); if you’re building something commercial then make sure you get signed up on the commercial support system so that our team can help with your power story, schematic & board reviews, etc. Email email@example.com for more on that.
Thank you Hugo. I am contacting Peter and Eric and hope to hear from them soon regarding more details on Nora. the SPI flash filesystem sounds great. I really appreciate items like that since it would have been one of the first things to code.
I’d love to hear from other users and designers on their strategies too.
I thought I’d bump the topic to see if anyone has any experince to share with Imps and battery power/low power.
I’ve been reading some ideas and app notes online. One reference I enjoyed is TI’s app note for AA battery power of their chips. The power scheme here may also apply to IMP, though the IMP will cold boot, I think, each wake up. http://www.ti.com/lit/ug/tidu787/tidu787.pdf
The imp cold boots on power cycles, warm boots on wake cycles.
That TI appnote is very weak sauce in my opinion.
The CC3100/3200 seem awfully slow (4-5 seconds to wake, DHCP, and send a UDP broadcast?! The imp is more like 1-2 seconds to wake, DHCP, DNS, TCP, TLS and exchange encrypted data with a server in the cloud)
The CC chips also write to SPI flash at every boot - to a single page, which is specced (as is common) to 100k cycles. Every boot uses a cycle, which is insane; the use case noted in our Nora design (waking every 15 mins) would wear out the flash in 3 years making the device not just out of battery, but needing to be totally replaced…
They’re boosting to 3.3v which seems unnecessary as most peripherals are specced for 2.7-3.6. Boosting lower = less conversion losses and generally less power usage (radios have many LDO’ed rails).
The nora design is much simpler, lower power, and higher performance. It’s also likely cheaper
Thank you again guys.
For the TI app note, I’m not as interested in their wifi module, so no worries there. I also noted the SPI flash usage and feel like that is a bit wonky. What I did like about the note was the power section and the idea of turning off the regulation completely with a very low power wake up switch of sorts. I’m not familiar with low power design enough and this concept got me to think deeper about options for sleep state.
Using a low power on/off wakeup switch like what they used could really extend battery life, if the sleep interval is long enough. It may not be worth it though if the intervals are varied or short, or if sensors need to be continually powered for low power monitoring.
I’ve got a quote for producing a few Noras from a local fab house. I pulled the Rev5 files from this site. I think I’ll go ahead with that production just so I can do some testing with Nora.
On the Nora page there was the wakeup cycle test, but it didn’t say the interval between wakeup cycles or what the system’s sleep current was, so I wasn’t seeing what the potential battery life could be. Did you do any of those tests?
Really the low power switch does not extend battery life much, because batteries have internal leakage - if you don’t burn that power, they will burn it themselves. Also, you lose things you may want like an RTC or low power non-volatile storage, and then those components (and their idle currents) will have to be added, directly connected to the battery.
The Nora design turns the power supply totally off when the imp is asleep, because the imp can keep time, preserve SRAM, and be able to wake even down at 1.8v (2x0.9v cells).
Nora with 2xAA lithiums on 15 minute wakes is 2.5-3 years (depending on network conditions). If you look in the battery powered product design guide it walks through battery life calculations.
Once again, thank you. I had a great conversation on Friday with Eric and Brandon and will keep doing more investigation. I recommend you update the Nora page with more details about battery life and experiments, since what is posted now is a bit tentative.
Thank you for taking the effort to make those sample designs and to document them so well. I find them to be really useful.