TMS320F28035: query regarding dividing _iq variable by integer

Richard McAleer

Part Number: TMS320F28035

I am using the IQmath library within our DSP code and am finding it tricky to get the correct method for integer division of an _iq variable

I have:

the numerator: a measured _iq value of around 60.0
the denominator: a Uint16 counter running in interrupt with value between 100 and 10000

Global_Q is 24 for our application

I would like to divide the _iq value by the Uint16 but I am unsure of the correct way to do this.

It might be that I am inspecting the variables incorrectly in CCS but for two options in the test code below both give negative numbers once den equals 1000
- this makes me suspect that I am hitting the limits of the global iq24?

#define kNoTests 4
Uint16 den = 10;
_iq num;
_iq testIQOut;
_iq testIQOutTwo;
static inline void _testRoutine(void){
  num = _IQ(61.2);

  Uint16 i;
  // testing 10, 100, 1000, 10000
  for (i = 0; i < kNoTests; i++ ) {
	  testIQOut =  _IQdiv(num, _IQ16toIQ((_iq)((Uint32)den << 16)));
	  testIQOutTwo = _IQdiv(num, _IQmpyI32(_IQ(1.0), (Uint32)den));

	  den *= 10;
  }
}

Any suggestions for how to cleanly do this would be gratefully received

Best regards,
- Richard

over 6 years ago

0 Richard McAleer over 6 years ago

Expert 1485 points

Suggesting my own answer it looks like it works if I promote to IQ16 for the complete calculation:

static inline void _testRoutine(void){
  num = _IQ(61.2);

  Uint16 i;

  for (i = 0; i < kNoTests; i++ ) {

	  _iq16 numTemp = _IQtoIQ16(num);
	  _iq16 outTemp = _IQ16div(numT,(_iq16)((Uint32)den << 16));

	  testIQOut =  _IQ16toIQ(outTemp);

	  den *= 10;
  }
}

I would still be grateful for any suggestions or recommendations for how this is best done

Best regards,

- Richard

0 Sira Rao80 over 6 years ago in reply to Richard McAleer

TI__Mastermind 26875 points

Richard,

It looks to me like you are handling the variables consistently in your second "suggested" case. The only thing that looks suspicious to me in your original code snippet is _IQ16toIQ((_iq)((Uint32)den << 16)) for the denominator, where in your second case, you have (_iq16)((Uint32)den << 16)). In other words, I believe the first case might be performing the same operation (IQ conversion) twice, thus corrupting the den?

Thanks,

Sira

0 Richard McAleer over 6 years ago in reply to Sira Rao80

Expert 1485 points

Thanks Sira

I think the issue is more that the interim value I need does not fit in the IQ24 (but my ultimate result does)

So if I shift back using _IQ16toIQ while the value is greater than (or equal to) 128 I get a saturated IQ24 value

In essence I need to preserve the dynamic range during the whole of the operation