I’m trying to figure out the maximum amount of data that we might need to store in one day.

My worst case usage scenario ignores battery life, and collects data at the maximum sustainable rate.

TS05 collected accelerometer data at 10Hz.  I was able to work perfectly well with that, but let’s assume we may need to log at 100Hz, i.e. 100 samples from all the sensors each second.  So I’m going to estimate the maximum number of bits in each sensor sample.


At 100Hz, the time field needs to measure a new record every 10 ms.   If we assumed we required an accuracy of 1uS.  A 64 bit time field can represent 500,000 years.  8 bytes!


Most accelerometers produce about 12 bits for each of the three axes.  Let’s assume 16 bits/axis = 6 bytes.


A magnetometer is similar, but it responds very slowly so polling at 100Hz is probably pointless.  Anyway, let’s say another 6 bytes


A barometer is  slow to respond – let’s assume 10Hz is the best it can do, and you get 2 bytes of P, and 2 of Temp.  There’s 4 bytes

BLE RSSI to one anchor

If you were taking RSSI measurements from the BLE beacons, each measurement is 16 bits, and let’s assume the beacon address is a full iPv6 address (16 bytes), and we don’t do anything smart to compress it. That’s 18 bytes.  Each polling cycle would hit a different beacon.  I’d expect to rotate through some subset of the beacons.

DW Range to one anchor

We use the relatively inexpensive BLE measurement, and any previous knowledge of the layout to guess the best three anchors to range.  This is only done when the user settles because it’s too power intensive.  But for the sake of argument let’s say it is gathered at the same  rate as the RSSI and in the same format.  That’s another 18 bytes.

Battery Coulombs?

Battery status is 2 bytes.

Adding it up

8+6+6+4+18+18+2 = 62 bytes.
At 100Hz we generate 6200 bytes /sec.  and 535,680,000 / day = 536MB.

When the subject comes to a complete rest then the device figures out it’s position and goes to sleep.
If we assume that people are stationary for 80% of their day. (pulled that number out my arse), and no samples will be gathered, then we will only generate data 20% of the time, and only produce 107MB / day.

At 10Hz it would be 10MB / day.

How much power would this consume?

We talked about how to record data. A MicroSD card is too large for a ring, obviously. However, a uSD card is good because it contains a controller that takes care of things like mapping bad memory sectors and doing load leveling. It’s not clear to me whether or not these functions need to be performed with a discrete flash memory IC.

Here is an SPI-connectable 1G-bit flash chip from Micron that is only 6mm x 8mm:
This device has 4 stacked 256 M-bit die. The memory is therefore arranged as 256Mbx4, or 128 M-Bytes. 

I am hoping that this device could be used both for data storage and program storage. For example, if the Dialog  DA14580 were used, the program could be loaded from the memory device to the 42kB system RAM.

If there were swapping involved that would most likely kill the power budget. After development is completed the OTP memory of the DA14580 could be used. Maybe an OTP memory saves some power.

Writing data to a flash memory takes a lot of power. Maybe it’s not practical to consider it.

Adesto 32Mbit memory, 6mm x 6mm BGA:  http://www.adestotech.com/sites/default/files/datasheets/doc3686.pdf

David’s email about DecaWave.

April 16

30 dev kits ready for shipment: http://www.digikey.com/product-search/en/rf-if-and-rfid/rf-evaluation-and-development-kits-boards/3539644?k=decawave. Only $606.67.

April 18

Two EVK1000 have sold sometime in the last two days. Now only 28 available

April 28

Today there are 18 EVK1000 available and the price has dropped slightly to $590.91. The sales rate is about one per day.

The DWM1000 is now listed at Digikey for $31.84 each. There’s a product brief. There’s no stock yet.
There are 3242 DW1000 in stock today. 3482 were in stock on 4/10/14. The sales rate has increased from 6.1 per day from March 12 – April 10 to 20 per day since .April 10.
I feel the urge to buy a Dev kit.

I tripped over this brochure in my travels.  My eye was drawn to the bit that said that the update voltage was 0.5 micro-amps per cm2, and that it runs off 5V.  Maybe this is outside the realm of possibility, but I love the idea that it draws no current when it is not updating.  Now this is a 33-segment display, and not a bitmap, but I assume you can have whatever you want if you pay enough.


I’d quite like to futz around with a Dev Kit to see how much power it draws in practice.
DigiKey has one for $70

I have poked around a bit and made a list of the BLE SOCs I could find easily. Then I tried to figure out what chip they are using.


Bluegiga’s BLE112 and BLE113 – is an 8051-based TI module!


Panasonic’s PAN1323 is an 8051-based TI module!


Nordic’s nRF51822

Cambridge Silicon Radio

CSR µEnergy® Product Family -Processor is TI CC2540 – 8051-based processor

Dialog Semiconductor

Just a few snippets drawn from David’s email of April 28, 2014 at 9:58 PM
DA14580 – http://support.dialog-semiconductor.com/downloads/DA14580_DS_v1.63.pdf


  • 32 kB One-Time-Programmable (OTP) memory
  • 42 kB System SRAM
  • 84 kB ROM
  • 8 kB Retention SRAM

Other than the lack of flash it is an attractive solution. The smallest package is only 2.5mm and only a few board components are needed.


EM9601 – http://www.emmicroelectronic.com/webfiles/product/other/EM9601_FS.pdf
Well spotted David!  EM – http://www.emmicroelectronic.com/Sitemap.asp They seem to be based in Sao Paulo. The web-site is a bit amateurish. They have a wide variety of products, and don’t seem particularly focused on BLE. However their BLE system is based on the STM32F051x8 32-bit CortexTM M0 which is a good feature. It’s pretty similar to the Nordic system. 200 Ohm differential antenna – does this matter? It’s a BGA.
Typical current consumption:
12mA Tx current at 0dBm output power 13mA Rx current 14µA in connected BLE Sleep Mode (including Host MCU current) 10µA in Deep-Sleep Mode (with memory retention)
64K of Flash, and 8K of SRAM (woo hoo) and a Calandar RTC!


BLUENRG: Announced September 2013.  And somehow I managed not to hear about it … or did I just forget.
64K Flash and 12K
8.2 mA maximum TX current (@0 dBm, 3.0 V)
Down to 1.7 µA current consumption with active BLE stack
Accurate RSSI to allow power control. Uh huh 🙂
Demo boards  the one with the rubber antenna is already available from DigiKey, but the dongle version is not.  Same board is on backorder from Mouser.
EWARM Compiler 6.60 version is required for building the BlueNRG_DK_x.x.x demonstration applications.  Oh joy.  IAR again.  $3500
User docs for kits.

Which actual ST ARM M0 system is the Bluetooth chip?  32Lx…?  I still have not seen any evidence that it can be programmed.

The STEVAL-IDB002V1 is a product evaluation board based on the BlueNRG device. The BlueNRG is a Bluetooth low energy 4.0 compliant low power network coprocessor. The STEVALIDB002V1 is composed of an RF daughterboard and a microcontroller motherboard. The RF daughterboard features the BlueNRG device, an SMA connector for an antenna or measuring instruments and an SPI connector for external microcontroller. The motherboard is based on the STM32L, acting as external microcontroller driving the BlueNRG device. A JTAG connector allows development of firmware on the microcontroller.

I went to TI’s booth at EE Live earlier this month. I asked about the ARM-based BLE SOC and he said he could not tell me anything about it unless I signed an NDA. Under NDA, he said he could tell me something interesting.

I called TI and opened a Service Request. Today I received this email:

RE: Service Request # 1-1399505934

3:58 PM April 30, 2014

Hello David,

Unfortunately, in order to obtain an NDA, you will need to contact an authorized distributor.  The authorized distributor will obtain the necessary information and start the process with the local field sales rep if need be. I have provided a list of our authorized distributors in your area below. I apologize for any inconvenience this may cause you.

Arrow- 800-777-2776  www.arrow.com
Avnet- 906-632-4500   www.avnet.com
Digikey- 800-344-4539  www.digikey.com
Mouser- 800-346-6837   www.mouser.com

Best Regards,
Matthew Fenley

I’ll call Digikey.

extracts from emails exchanged with Bob Mayo, starting April 18th, 2014

My current stack is:

Jade (template language)
Express (web framework)
Nodejs (server-side runtime)
Mongoosejs (database object modeling)
MongoDB (database)
There should be tutorials with this exact stack, perhaps advertised as node or express tutorials.
The example on the front page of http://nodejs.org/ is a place to start.

Next step could be http://expressjs.com/guide.html


Energy & Environmental Science Facebook page

This is the one from Korea Advanced Institute of Science and Technology (KAIST):
Wearable thermoelectric generator could extend your smartwatch’s battery life
Thermoelectric generator on glass fabric for wearable electronic devices

Tin selenide
Ultralow thermal conductivity and high thermoelectric figure of merit in SnSe crystals
A Great Improvement in Thermoelectric Material Found
Surprising Material Could Play Role in Saving Energy
However, this material appears to be most useful at hightemperature.

4mm square Peltier

Energy scavenging ring project

I wrote this email on 6/14/13:

I think the best way to measure power consumption is to implement the Linear Technology fuel gauge / coulomb counter IC that we have on TS05b. Once we know the number of coulombs that have been consumed then we have to look at the current draw to de-rate the battery because of the power lost through the battery’s internal resistance. Hopefully battery life will be long enough that it’s not practical just to wait for the battery to run down to see how long it lasted (hopefully it will be many months).

I want to look into using a regulator to draw 0.19mA from the coin cell for longer periods to charge a big capacitor so that the capacitor is what powers devices when they come on. There would have to be a second regulator to buffer the supercapacitor output.  It’s hard for me to believe that somebody isn’t already making an all-in-one IC to do this but I can’t find it. There are several very interesting energy harvesting IC’s from Linear.

This device might work really well: http://cds.linear.com/docs/en/datasheet/3105fa.pdf. The drawback is 80% efficiency at 0.2mA which might approach the loss from the battery’s internal resistance, but having regulators will make the design a lot more stable than just running off a battery, and the battery can be used even when its voltage gets very low. This device has a power adjust pin (MPPC) that can be used to prevent too much current from being drawn from the battery. The boost voltage would be set to maximum (5.25V) to charge a supercapacitor and then that voltage would drive a buck-boost regulator (maybe http://cds.linear.com/docs/en/datasheet/3534f.pdfto output the system voltage (maybe 1.8V or 2.5V). Since the LTC3105 will work with an input as low as 225mV we can completely drain the coin cell. The absolute maximum input voltage of the LTC3105 is 6V so putting two CR2032 in series (to increase capacity) is questionable. At 20 degrees C the output of a CR2032 is 2.9V but at 60 degrees it is 3.2V.
Or maybe it’s best just to put a big ceramic cap across the battery to minimize pulsed current and have no regulator. That’s probably what everybody else is doing.
Now there are more energy harvesting products from LT. Here’s the portfolio: http://www.linear.com/parametric/Energy_Harvesting

Article: How much energy can you really get from a coin cell?

The article references this white paper from TI entitled Coin Cells and Peak Current Draw: http://www.ti.com/lit/wp/swra349/swra349.pdf
I assume for now that the way to achieving the longest battery life possible is to extract the most energy possible from the battery, and therefore the highest priority is to limit the amount of current drawn from the battery. The regulator chosen must limit the current drawn from the battery to 200uA or less, but must be able to deliver 300mA in short bursts.
The acronym to look for when selecting a device is MPPC – Maximum Power Point Control. To use the MPPC feature in a regulator, I think the coin cell must be used as the the energy harvesting source. The “battery” powering the regulator will be a large capacitor (“supercap”).
To use an energy harvesting source, such as a thermo-electric generator (TEG), multiple regulators would have to be used.
TI bq25504 (datasheet http://www.ti.com/lit/ds/symlink/bq25504.pdf) looks to be a flexible device. The startup voltage is 330mV but it will run down to 80mV. However, it can only supply about 80mA. A capacitor on the output might be able to absorb the 300mA spikes.

TI paper: Indoor Light Energy Harvesting Reference Design for Bluetooth® Low Energy (BLE) Beacon Subsystem. It uses a bq25505 and CC2541.

Papers that might contain relevant information:
Development of a Wearable HRV Telemetry System to be Operated by Non-Experts in Daily Life
Using a Supercapacitor to Power Wireless Nodes from a Low Power Source such as a 3V Button Battery
Parallel Battery Configuration for Coin Cell Operated Wireless Sensor Networks
A 700-uW Wireless Sensor Node SoC for Continuous Real-Time Health Monitoring
Battery Capacity Measurement and Analysis using Lithium Coin Cell Battery
Ultra Low-Power Smart Medical Sensor Node for In-Body Biomonitoring
Power Management Subsystem with Bi-directional DC to DC Converter for μ-Power Biomedical Applications

To be continued…


http://tclbattery.en.ec21.com/Polymer_Lithium_Ion_Battery_Cell–604293_604294.html https://www.sparkfun.com/products/11316 http://www.powerstream.com/thin-lithium-ion.htm http://www.aliexpress.com/item/Wholesale-Small-Rechargeable-Cylindrical-Lipo-Battery-3-7v-45mAh-60220-100pcs-lot/707397136.html http://www.aliexpress.com/item/Wholesale-3-7V-40mAh-PL031220-Small-Rechargeable-LIPO-Battery-100pcs-lot/707964273.html

This is an interesting battery: http://www.fdk.com/battery/lithium_e/pdf/data_sheet/coin_p_type/CR1.3N_spec-sheet.pdf.

But it’s too high.

With this battery: http://www.digikey.com/product-detail/en/BR-1225/P183-ND/31915the enclosure could be 1/2″ in diameter and all the components could sit on top.


This buzzer claims almost 90dB at 4kHz.http://www.aliexpress.com/item/RUIDA-ELECTRONC-BUZZER-SMD-3V-5x5x2-5mm-PIEZO-TRANSDUCER/754733076.html


The TI BLE SOC is 6mm square. The new Invensense IMU is 3mm square. We need a 1206 capacitor and space for the antenna. Maybe the total size could be about 1/2″ by 1/4″. How would connection be made for programming? It would need pads and a special fixture to fit into.


This beeper only draws 3mA but the rated voltage is 9V (78db) so at 3V I’ll bet it barely squeaks: http://www.aliexpress.com/item/SMD-120025F-09041TJ-SMD-piezo-disc/623852162.html. And it’s really big. I think it’s better to go with a low voltage, electromagnetic beeper (shown in the drawing, 5mm x 5mm) and just give it really short bursts of pulses. With a short pulse duration, short burst, and long duty cycle (perhaps once every three seconds) the power consumption may be acceptable.

Current Consumption

CC2541 CC2540



Operating voltage 2V – 3.6V 2V – 3.6V
Shutdown current
Idle current
Max current
Task 1 current
Task 2 current
Task 3 current
RX mode, standard mode, no peripherals active, low MCU activity 17.9mA 19.6mA
RX mode, high-gain mode, no peripherals active, low MCU activity 20.2mA 22.1mA
TX mode, –20 dBm output power, no peripherals active, low MCU activity 16.8mA
TX mode, –23-dBm output power, no peripherals active, low MCU activity, MCU at 250 kHz 21.1mA
TX mode, 0 dBm output power, no peripherals active, low MCU activity 18.2mA 27mA
TX mode, –6-dBm output power, no peripherals active, low MCU activity, MCU at 250 kHz 23.8mA
TX mode, 4-dBm output power, no peripherals active, low MCU activity, MCU at 250 kHz 31.6mA
Power mode 1. Digital regulator on; 16-MHz RCOSC and 32MHz crystal oscillator off; 32.768-kHz XOSC, POR, BOD and sleep timer active; RAM and register retention 270uA 235uA
Power mode 2. Digital regulator off; 16-MHz RCOSC and 32MHz crystal oscillator off; 32.768-kHz XOSC, POR, and sleep timer active; RAM and register retention 1uA 0.9uA
Power mode 3. Digital regulator off; no clocks; POR active; RAM and register retention 0.5uA 0.4uA
Low MCU activity: 32-MHz XOSC running. No radio or peripherals. Limited flash access, no RAM access. 6.7mA 6.7mA
Timer 1. Timer running, 32-MHz XOSC used 90uA 90uA
Timer 2. Timer running, 32-MHz XOSC used 90uA 90uA
Timer 3. Timer running, 32-MHz XOSC used 60uA 60uA
Timer 4. Timer running, 32-MHz XOSC used 70uA 70uA
Sleep timer, including 32.753-kHz RCOSC 0.6uA 0.6uA
ADC, when converting 1.2mA 1.2mA

Battery Power

As I look at the capacity of these coin cells and the power requirements of the components, I don’t see how it could work for more than a few months. The IMU draws 6.4uA with the accel running at 1Hz and the CPU needs 0.5uA with all clocks off just for RAM retention, and the sleep timer needs 0.6uA with the internal RC oscillator. If the crystal oscillator is running it needs 60uA-90uA. Whatever “BOD” is it takes more than 200uA.

Hopefully the IMU has a deep sleep function to get its power consumption way down.

At least things look good for making AA-battery-operated anchors to put around the house. They could potentially last years with a lithium AA battery (more than 3000 mAh).

This paper has a table that lists the power consumption for various medical devices and how long they typically last using a CR1025, which is the battery I have been considering.

I also attached the paper.

 If the battery is 10mm diameter the enclosure has two be at least another 1mm thick so the diameter will be at least 12mm. The height will have to be at least 7mm.

Inline image 1