Sunday, May 12, 2013

Current Measurement

DSC_9218

Minor coding changes can have a major effect on current consumption. The above was taken running a while loop to wait for the DRDY line to de-assert itself. Notice the nRF51 is using 4.53ma.

DSC_9219

However, if it’s coded to just check the status and keep running in the main loop this drops to 1.86ma. When I press Button0 to simulate the reed switch (or likely hall effect sensor) this peaks for an instant to 1.99 – 2.05ma.

Below is a table outlining the potential circuits. Right now there would likely be an additional 0.5 to 1ma for the ADS1248.

image

This is still a little high for coin cell operation. The big problem is the strain gauges being 350 ohm and there being two Wheatstone bridges made up of 4 gauges (essentially 4 x 700 ohm resistors in parallel). The CR2540 is rated at 3.0ma continuous. I have a bit of a ways to go on this.

One thing that could halve the consumption is to use two LM4140, one for the left measurement and one for the right. They both have enable lines and therefore only one needs to be working at a time. I could alternate between the left and right measurements then. This drops consumption 3.4ma bringing things down to 5.26ma on the LM4140 configuration or about 100 hours (max) on a CR2450. Still shy of what I’d like and I don’t have a nice method of using two LM4140’s together at the same time in a continuous read operation.

The holy grail is a DC-DC down to 1.2v at 90%+ efficiency then, LM4140 to bring it down to 1.024v. In theory this would get the 3V consumption down to 2.2ma using “Top Gear Maths”.  However, I might have to settle for a little more than this because the LM4140 requires a 1.8v input to generate 1.024v. This pushes me close to 3ma then which is the max continuous discharge for the CR2477 like Quarq Saturn Cinqo uses.

It’s been suggested to put a resistor in series with the Wheatstone bridges (which are in parallel). After thinking about this, this would retain the ratiometric nature needed for the ADC. So I might look at LM4140 + high quality resistor, though in theory it doesn’t need to be high quality. The potential downside is that as the voltage across the bridge goes down, the susceptibility to electrical noise increases. I’ve had issues with this back on V1 which was little more than a couple of gauges on an old FSA Vero square taper crank. It was very unpleasant to deal with.

IMG110

BRC Electronics in the Netherlands are working on the CMSIS-DAB located here. This little guy is similar to the SEGGER J-Link. I’m going to trial one in the coming months  for inclusion with the beta units. This is especially important for when I mail prototypes to people who aren’t local. I’ll need a way to fix problems for those far away. Looking forward to trying this out, and hopefully including it for people to update firmware. This works with Keil uVision IDE currently which is all I need, however they are trying to make it work with other things.

2 comments:

  1. Just so I understand correctly.
    Currently the nRF51x is drawing 1.8mA constantly?

    Because from page 32 of the nRF51 I can't particular see how it can be drawing this much. Everything quoted is 'typical', but it only adds up to about 0.9mA which is 50% less than your measured current.

    ReplyDelete
    Replies
    1. Actually I completely missed pages 34 etc.

      I'll try to do a theoretical calc and see what I come up with.

      Delete