Tiva performance

Hello Hector,

What is the computation requirement in terms of DMIPS required for this operation? That is not clear from the post.

Regards

Amit

hector hector said:

I would like to migrate  the code of ads1298 kit to TIVA 120mhz cortex M4

 No reference was for code.. please can you provide detail?

hector hector said:

Does TIVA speed enough to do the 8 channels ? in 125usec ( 8kps)

 Hi this I think  was done in MSP430 transferring data to do post processing of data, or in the DSP C55x?

 TIVA has much power and floating point capability on board than MSP but not more than DSP and if full 24bit with 400taps filter and a notch also if just 50 Tap on all channel may be at least it require a DSP as suggested on eva kit.

 Which processor where you using collecting data and where you where processing all data?

Hi Amit

How can I calculate this ? ( DMIPS needed)

as I told in my previous post we need to migrate the code of ads1298 to tiva ; 8 channels 8kps  with 2 FIR filters (400 tabs ) and 50 tabs  for each channel .

Thank you

hector hector said:

How can I calculate this ? ( DMIPS needed)

 From filter resolution and bandwidth.

hector hector said:

as I told in my previous post we need to migrate the code of ads1298 to tiva ; 8 channels 8kps  with 2 FIR filters (400 tabs ) and 50 tabs  for each channel .

 And we told you migrate from what?

 We aren't in your mind so please provide detail otherwise we can't do any little help than try ask again same detail is again missing from.

Hello Hector

On what processor platform were you running it earlier? That would give comparative data.

Regards

Amit

Right now we use stm32f4  168mhz ; but we decide to move for Tiva ;

tiva is maximum 120 mhz .

The kit of ads1298 is proccessing the 8 channels using dsp I think c55xx.

The main issue is we need to go with cortex m4 solution .

we will use TIVA for collecting data from ads1298 and proccessing it .

 Just to report, I checked all your question to various forum to see if some information on previous code.. other than first question none was verified also if valuable information was gift. All where vague just like this on unrelated arguments....

 So what do you think we can answer when you insist on your pointless view?

Robertro I dont know what is not clear to you .

I am telling 400 tab fir ; it means 400 multiplication + 400 adders

for 8 channels = 3200multiplication + 3200 adders  operations 

50 tabs nortch filter = 50 multiplication + 50 adders opertaions

for 8 channel = 400 multiplication + 400 adders

the total operation:3600 multiplication + 3600 adders 

to be in the safe side lets say 4000 multiplications + 4000 adder operations

the sample is 8kps =125us period

so can TIVA do a 4000 multiplication + 4000 adds in 125us ?

I hope it is clear now .

hector hector said:

The main issue is we need to go with cortex m4 solution .

 Main issue??? your language is not so close to what f who know theory behind the lines....

 start here filter theory...

 http://en.wikipedia.org/wiki/Finite_impulse_response

 Tapped filter theory.. what are tabs???

hector hector said:

Right now we use stm32f4  168mhz ; but we decide to move for Tiva ;

 And why ??

hector hector said:

The kit of ads1298 is proccessing the 8 channels using dsp I think c55xx.

 which C55xx member are you using?
 What about filter parameter??

hector hector said:

we will use TIVA for collecting data from ads1298 and proccessing it .

 Too many information missing .. processing to do what and to transfer where?

hector hector said:

I hope it is clear now .

 Yes it is clear you don't know in deep what you are speaking about...

 ECG filtering is generally bandwidth centered near 400Hz, filter can get from AD about 20 samples, not enough to do a so fine filtering... Collecting 400 samples get sampling windows freq down to 20Hz that is appropriate to ECG.. so why do filtering in as low as 125uS ?? FS 8KHz when sampling rate of ADC is this? So Shannon and Nyquist theory get lost.

 I don't know if TIVA support MAC operation in hardware like MSP430 or DSP do..

 Filter and operand resolution to scale filter to result and .... Are you missing nothing like tap coefficients?

Robert be cool I am not here to fight :D . I agree my language is the main problem .

let me start from begining .

We have 3 products based on ads1291 and ads1292R and we are ok with them , the main MCU is stm32 .

now we want to move for Tiva because cheaper  and better support , we have keil for it , so we are not interesting in getting new compiler for dsp .

we have a n ew product will based on ads1298 ; we need a mcu ( TIVA ) that could be able to collect data from ads1298 @ 8kps and proccess the data and then send the data to PC via usb  or may be serial with high buadrate..

the proccess is just filtering a low pass filter and a nortch filter .

As we can predict that wee need total of about 10000 operaions in 125us

we get Tiva running at 120MHZ (150DMIPS) .

so the question is TIVA is ok for this ?

Sorry Robert for misunderstanding and thank you for keeping following.

sorry Rober for miss understanding .

Thank you Robert;

let me explian more for you .

We still designing the product and  how can we do it faster , cheaper and easier  .

The product will be 12 leads  ECG monitor .

we will have two mcu (tiva) ; one will collect the data from ads1298 and filter it ( baseline - 50/60 nortch- low pass filter ) , detect R , detect pacemaker , then send the data to the main mcu vai usart or spi ...

the main mcu will do the tft tasks and other simple communication to pc or printer and so storage  .

The main problem that we must go with cortex M4 , we cant change to go with dsp and so (company limitations ) .

You can suggest   a way to do that ; for example what do you mean post proccessing ?

I believe that what I need  is already done with C5515 ; here in the attachment the original code .

You can still suggest us other solutions ;for filters we still can get less taps I go down to 50 taps .

3348.ADS1298_ECGSystem_C5515EVM.zip