hdl coder work flow adviser block compability error

Dear all,

I have a model so as to control a pmsm. when I run the hdl workflow adviser to generate VHDL code, an error which is not clear for me has been occured on the step "check block compability". The error message and related blocks are given as follows;

Simulink BlockWarnings/Errors

fpga_Id_Iq_control/pmsm_id_iq/speed_controller/PID Controller w/Integrator/Discrete/Integrator

fpga_Id_Iq_control/pmsm_id_iq/speed_controller/PID Controller w/Integrator/Discrete

fpga_Id_Iq_control/pmsm_id_iq/speed_controller/PID Controller w/Sum/Sum_PI/I

fpga_Id_Iq_control/pmsm_id_iq/speed_controller/PID Controller w/Sum/I

fpga_Id_Iq_control/pmsm_id_iq/speed_controller/PID Controller w/Integrator

Error: Slope and bias scaling 'sfix32_S1p953125en7' is not supported for HDL code generation. Slope must be equal to a power of 2, and bias should always be 0.

I need help about solution of the error.

Thannks in advance.

Regards

Fahri

Answer 1

[Kiran Kintali]

Running hdlsetup command on the model targeted for FPGA always helps in terms of data type selection.

web(fullfile(docroot, 'hdlcoder/ref/hdlsetup.html'))

web(fullfile(docroot, 'hdlcoder/ug/check-for-safe-model-parameters.html'))

One of the things this command does is setting ProdHWDeviceType Device vendor to ASIC/FPGA and this greatly helps in coming up with optimal full-precision types for HDL Code generation.

Answer 2

[Kiran Kintali]

HDLCoder only supports binary-point scaling. The product currently does not support slope/bias scaling.

You need to choose fixed-point types with a sign bit, fixed wordlength and fraction length.

For ex: A signal with type fixdt(1, 32, 16) as shown on the signal like sfix32_En16 has the following range.

>> a = fi(pi, 1, 32, 16)

a = 

    3.1416

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Signed
            WordLength: 32
        FractionLength: 16
        
>> range(a)

ans = 

   1.0e+04 *
   -3.2768    3.2768

          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Signed
            WordLength: 32
        FractionLength: 16

Answer 3

[Fahri Gürbüz]

Dear Kiran Kintali,

First of all, thanks for your fast answer. I have used all data either single or fixed-point and as you know hdl coder supports both. Thus, I think there is another problem that I could not see. Input signal data type of speed controller is ufix16_En15. I could not see, what is the problem. Maybe, data length exceed the maximum limit. However, input is 16 bits. If you have any idea, I can follow it.

Regards

Fahri

[Walter Roberson]

The sfix32_S1p953125en7 has to be in there somewhere... Possibly a value near 1.04918032786885e-06 that was not properly converted?

[Fahri Gürbüz]

Hi Walter,

sfix32_S1p953125en7 is in PI control block. Integrator subblock output data type is sfix32_S1p953125en7. Nevertheless, Integrator is generated that datatype and I think my data is pushing integrator so as to generate it. Do you have any idea why integrator is generating this data? How can I fix it?

Regards,

Fahri

[Kiran Kintali]

Can you share a sample model that reproduces the issue?

[Fahri Gürbüz]

Hi Kiran,

The error was occured integrator output of PI controller and I could not make any change on it. I have omitted the PI controller and set it with discrete element and set the data type. The problem is solved.

Thanks for your attention and help @Kiran Kintali @Walter Raberson

[Fahri Gürbüz]

Hi,

Sorry for late answer. I have passed the problem by setting data types again.

Thanks for advices.

Regards,

Fahri

Answer 4

[Angel de Castro Martín]

Hi. Similar problem here. This thread gave me the hint for solving it. I don't know why, the I (integrator) and D (derivative) try to use "strange" data types if let to choose by default (Inherit: inherit via internal rule). In my case, in which I was trying just a PI, I got to compile correctly setting only the "Additional data type" of "Integrator output". The rest can use inherit, but not this one. I also set the type of the input, but not of the global output. I attach a screenshot. I hope this helps future users.