Delfino Floating Point Motor Control

Question for those of you using floating point.

There are two options for how we write our floating point code

1. Use IQMath (target IQ based DMCLibs with the float switch)

PROS: Gives single source set to allow easy targeting of float or fixed point CPU

CON: Does not look like native floating point for someone unfamiliar with IQMath; won't really be used by tools like The Mathworks

2. Use native floating point (create special floating point DMCLibs)

PROS: Universal floating point

CON: No modularity with fixed point, have to create and maintain new floating point blocks

For a new user only interested in floating point #2 would seem to be the clear favorite, but for "portfolio" users who may want to target a range of our devices #1 will make the most long-term sense.

Please let us know how you feel, and if you consider yourself a "FPU only" or "Portfolio" user.

Thanks!

I would have thought both would be provided, as having the native floating point libraries saves a lot of hassle creating one. But if I had to choose one, I would vote for the native floating point.

Hi Chris

Do you know if there are any floating point ports for the PowerLib library used in the AC/DC software kit. Iam using a F28335 control card together with the powerlibs which are all written (as far as i know) in none floating point instructions. Though i would like to use floating point instead. So are there any ports to floating point.

Thank you

Flash microcontrollers with floating point are going to be available in low cost and low pin-count DSCs (less than U$10 for 1K units), like Renasas RX62T and the Cortex-M4 and maybe Texas in the future. So I also vote for native floating point.

2. Use native floating point (create special floating point DMCLibs)

Chris,

I agree with most of the posts, create native FP with the FP libs. In general, we would look at the application and design for the FP micro if needed. Using a less featured micro across all apps is not usually our game. It might be different if we were a lo-$/hi-volume shop.

Steve

Bjorn,

We hadn't planned to support this as the vast majority of our power customers use the cheapest MCU possible to do the job. Floating Point is a luxury. Perhaps we can look at this and see if it makes sense to use IQ for some of this, so that they could be easily targeted to float. I know on many of these we are optimizing assembly, and for some we will have them optimized for the floating point CLA.

Ari,

Yes, I think you can expect low cost floating point MCUs from TI as well..... :)

Chris,

Any more to report on getting to floating point since May?

Delfino (floating point) F28335 controlCARD has been included in the HV Motor Control + PFC Kit since this summer. The floating point version of this code, and the floating/fixed point Motor Lib are in controlSUITE.

We have also recently added updated floating point DSP and math libs.

For the new, low cost floating point......look for an announcement on January 11th :)