Compiler/TMS320F28379D: Error: rts2800_fpu32_fast_supplement.lib<sqrt_f32.obj> redefined in rts2800_fpu32.lib<e_sqrtf.obj>

I will need to loop in someone from the C2000 team.  Note that there are examples fro Fast RTS in C2000Ware.  I just tried importing them and they are building ok for me.  One thing I noticed is that the "standard RTS CCS Project" in Resource Explorer imports the Fast RTS project and the "Fast RTS CCS Project" imports the standard one.  I will file a bug for that.

However if you import both of them you can compare the build options, specifically the link order.  The capture below is for the Standard RTS project and you can see that the regular RTS is ahead of the fast one.  In the Fast RTS project it is the reverse.

As far as 28379D support I will need to differ to the C2000 team.  The examples are for F28335.  

Regards,

John

John,

First of all thank you very much for your reply. I have already done what you said (also what is written in the pdfs) but didn't work for me.

Then I created a new empty project and managed to run fastRTS and SFRA together on F28379d without a problem. But I didn't understand why it doesn't work for my actual project. Then I tried to use SFRA FTMU lib instead of SFRA F lib in my project and it worked!!!

I am really busy right now, but when I have got time, I will move everything step by step from my actual project to this new project and find out what is the cause of this error. (just for engineering curiosity :)

Regards,
Hakan

Manish,

I have already put that in my cmd file and it didn't help. But I work on RAM, so I put FPUmathTables in RAM PAGE 1. I don't think putting tables into FLASH or RAM makes any difference, right?

By the way, let me give you some feedback for the new SFRA library. Maybe you can find some of them useful.

1) Instead of start frequency and step, you can define start and stop/end frequency, which is more logical to me. Because in frequency analysis, first I define frequency range that I am interested. Maybe its different in power electronics,  but for general, I think it is more natural way to define start and stop frequencies. I know I can calculate step value from stop frequency (and I do it that way), but it is better to put that calculation inside the library.

2)Instead of one amplitude value, you can define start and stop amplitude like frequencies. It is very useful, when your system response is very low in higher frequencies you want to test. Because the weak response of your system at high frequency, the output of the system can diminish in the noise during the test. So, you may want to increase the amplitude to have better SNR value at high frequencies. But increasing the amplitude can also damage your system in the low frequency during the test. As a workaround I do multiple test for lower and higher frequency ranges with different amplitude value, but it will be an improvement if I can do everything in a single test.

3) In SFRA, test frequencies are increasing in a logarithmic way using step value. I think it would be nice to have an option of increasing in a linear way.

4) Like DC offset, you may want to add phase offset to the generated sinusoidal ref. Maybe it is not important for power electronics, but it could be very useful if you have a mechanical control system with a limited motion capability.

5) You can eliminate jumps on sinusoidal reference during frequency switching and produce continuous sinusoidal signal with increasing frequency like MATLAB chirp signal.  This will decrease the settling time of the system output when the frequency is changed.

Best Regards,
Hakan

I have already put that in my cmd file and it didn't help. But I work on RAM, so I put FPUmathTables in RAM PAGE 1. I don't think putting tables into FLASH or RAM makes any difference, right?

If you are connected to the debugger it does not make a difference, if you are running standalone, you will need some routine in code that will copy the tables from FLASH to RAM, if you use the RAM option. Hence i recommend using FLASH to begin with.  

By the way, let me give you some feedback for the new SFRA library. Maybe you can find some of them useful.

1) Instead of start frequency and step, you can define start and stop/end frequency, which is more logical to me. Because in frequency analysis, first I define frequency range that I am interested. Maybe its different in power electronics,  but for general, I think it is more natural way to define start and stop frequencies. I know I can calculate step value from stop frequency (and I do it that way), but it is better to put that calculation inside the library.F32

Hakan, in the v1_30_00_00, the setup of SFRA will be independent of SFRA frequency points setup, i.e. the SFRA_init function which will be called SFRA_F32_config() .. now will not initialize the freq step array.

Instead, there will be a separate function that will set the frequency points. Note , even in the current SFRA library, even though we setup SFRA for log steps, we often over-ride the frequency step by a following routine to setup according to your needs. This is done at init time.

    SFRA_F32_reset(&sfra1);
    SFRA_F32_config(&sfra1,
                    SFRA_ISR_FREQ,
                    SFRA_AMPLITUDE,
                    SFRA_FREQ_LENGTH,
                    SFRA_FREQ_START,
                    SFRA_FREQ_STEP_MULTIPLY,
                    plantMagVect,
                    plantPhaseVect,
                    olMagVect,
                    olPhaseVect,
                    freqVect);

    SFRA_F32_resetFreqRespArray(&sfra1);

	
	i=0;

	#if SFRA_TYPE==SFRA_CURRENT// current loop
		sfra1.freqVect[i++]=SFRA_FREQ_START;
		for(;i<sfra1.vecLength;i++)
		{
			if(sfra1.freqVect[i-1]<10)
				sfra1.freqVect[i]=sfra1.freqVect[i-1]+2;
			else if(sfra1.freqVect[i-1]<50)
				sfra1.freqVect[i]=sfra1.freqVect[i-1]+10;
			else
				sfra1.freqVect[i]=sfra1.freqVect[i-1]*sfra1.freqStep;
		}
	#else
		sfra1.freqVect[0]=2;
		sfra1.freqVect[1]=4;
		sfra1.freqVect[2]=6;
		sfra1.freqVect[3]=8;
		sfra1.freqVect[4]=10;
		sfra1.freqVect[5]=12;
		sfra1.freqVect[6]=14;
		sfra1.freqVect[7]=16;
....

For example in the above, we set the frequency points to be more coarse at lower frequency, to make sweep go fatser and then finer at higher frequency etc.. 

2)Instead of one amplitude value, you can define start and stop amplitude like frequencies. It is very useful, when your system response is very low in higher frequencies you want to test. Because the weak response of your system at high frequency, the output of the system can diminish in the noise during the test. So, you may want to increase the amplitude to have better SNR value at high frequencies. But increasing the amplitude can also damage your system in the low frequency during the test. As a workaround I do multiple test for lower and higher frequency ranges with different amplitude value, but it will be an improvement if I can do everything in a single test.

Hakkan, We are adding a new function SFRA_F32_updateInjectionAmplitude , which you can call periodically to get the behavior you want. For example, the below can be used, you do not have to setup multiple runs of the SFRA. The reason why I do not want to add a linear scale in the library is because user may want different behavior depending on the system.

    SFRA_F32_runBackgroundTask(&sfra1);
    new_injection_amplitude = injectionAmplitudeStartFreq + slopeInjectionAmplitude * sfra1.freqIndex;
    SFRA_F32_updateInjectionAmplitude(&sfra1, new_injection_amplitude);
    SFRA_GUI_serialHostComms();

3) In SFRA, test frequencies are increasing in a logarithmic way using step value. I think it would be nice to have an option of increasing in a linear way.

As i mentioned earlier, SFRA does not care if the freqStep is log, even now, you can override the freq points by writing your own values to the freqVect array after the init routine. 

4) Like DC offset, you may want to add phase offset to the generated sinusoidal ref. Maybe it is not important for power electronics, but it could be very useful if you have a mechanical control system with a limited motion capability.

Can you elaborate a little more on this?

5) You can eliminate jumps on sinusoidal reference during frequency switching and produce continuous sinusoidal signal with increasing frequency like MATLAB chirp signal.  This will decrease the settling time of the system output when the frequency is changed.

I'll have to look at this more.