Hello,I want to develop an system which is an voltage quality recorder, i mean it measures the various voltage disturbances like transients, swells, sags, THD, etc,I was comfortable in analyzing the concept and with the coding using matlab. but for the hardware implementation i am confused of which kind of TI microcontroller or microprocessor is suitable for my project. i will work on the real time calculations and will use FFT in analyzing the voltage parameters. I am searching on dsp processors of C2000 series of TI, and its datasheets. But not able to wisely choose among them.
On what major parameters should i look upon while choosing them, my system is supposed to work at around 200Mhz, and would need good on chip memory, and interfacing, and precision calculations. how can i really meet the high speed calculations, and 32bit signal controller is it something i should look for or anything more is available!? i have looked at the TMS series to meet my considerations.
Or should i use arm processor??
I would be helped if anyone can recommend me which one would precisely be used for this cause.Any other suggestions related are welcomed.Thank you.
Tejas,
Without detailed information regarding the algorithms you're going to run its hard to say exactly which device best suites your application. That being said i would recommend you look at the Delfino series devices. There is a 150MHz flash version and a 300MHz ram version. Both devices devices are floating point and have the capability of performing the type of calculations you'll need to monitor power quality. We also have math libraries for things like the FFT you will be calculating. Finally, have you looked at Matlab's code generation capabilities? You may be able to modify your matlab models slightly so they can be compiled and run on the Delfino device.
Regards,Trey
Trey German
C2000 Applications
Dear sir,
i checked out the Delfino series i chose TMS320F28335, floating point series with 150Mhz and 256k flash, will that be something i can use or anything better would be required? and it has no external memory interface.
And i would like to know about the Matlab code generation capability which you just mentioned, like how will i actually optimize the code, can you give some link for knowing more about it, so that i would get a clear picture about it.
Thanks & Regards,
Tejas
The F28335 is certainly an easier chip to design into a system than its RAM only equivalent. Depending on your algorithmic requirements this device may have enough performance or may not. Keep in mind you'll probably want to run the speed critical portions of your application from 0 wait state RAM.
I know the mathworks tools are capable of doing code generation, but I'm not familiar with them myself. I found this page about it on their site here:
http://www.mathworks.com/discovery/ti-piccolo.html
Regards,
Trey
Trey,
I am really confused with deciding. I totally agree
Keeping speed in consideration even Omap series can make it. In general what exactly should I look upon while choosing a processor for my requirements.
It really just depends on the requirements of your algorithms. What kind of FFT are you running exactly? How many points does the FFT have? What other algorithms are you doing? I would outline all the processing you need to to and how often you need to do it and then estimate or benchmark the MIPs needed to perform each algorithm. Then you should have an idea of the horsepower your application needs and can select an appropriate processor.
Thank you very much Trey,
Analyzing in this perspective what i was lagging. I will definitely work on the way you have just suggested. I will do it soon and finalize based on my requirements, I will approach you once I have done with it.
Thanks lot
Hi Tejas,
If you use the C2000 family then you can and want to use onboard ADC. This makes whole system much simpler especially with 28335 which also has onboard FLASH. But if I would be in your place I wuld check the specs of onboard ADC if they meet your criteria. If they don't then you are gonna need external ADC. This will make your design (and Life) a bit more complicated and it might even push you to choose different MCU/CPU/DSP. Because when you have and use external memory interface for ADC it is not require much more effort to also use external FLASH...
This is just another thing to think about.
Regards, Mitja
Hello Mitja,
Apart from the on board ADC, i dont think it would meet my conditions,for this purpose i am planning to use ADS1610 adc which has 16 bit res ,10msps, or ADS1278 with 24bit and 144ksps, and i think this sampling rate would be comfortable. since I am using an external ADC how would i actually, but when i checked out for TMS320F2812 and TMS320F2, it has good built in ADC with 12MSPS, is it necessary to use another adc for my high speed calculations??
Basically my design needs an high precision ADC with sampling frequency of 10Mhz, i would use an voltage ref to determine the kind of parameter i want to study whether an transient, sags, swell, and other interruptions. for processing this data i thought an dsp would make good for the real time calculations.
As you have mentioned about the external adc, this will improve the system speed, but have to meet certain power considerations too.
How should i proceed with it, can give you give me suggestions regarding my dsp and adc boards. There are lot many C2000 32bit realtime mcu's, kindly help.
Selecting a MCU can (and sholud) take some time (and effort). Final selction depends on a lot of factors besides technical requirements. Obviously you should write down the technical requirements as it will make your life much easier. Judging from your selection of ADCs I get the feeling that you haven't nailed down the technical requirements yet. Is this a one man band project or are you part of larger group? Have you done this kind of things before? Is this a hobby project, research project or do you have a client? Can you share a bit more about the application?
This is part of the university project conducted by TI. I have a 3members in my team. I can request for development kits upto 200USD, and I am able to choose IC's in TI free sample lists. Keeping these constraints in mind, i have kept F28335 development board as my option, it costs around 99usd. It should do 16bit fft operations, the sampling rate is also desirable.
I was too confused with the selections so i merely read whatever i could find on the site, and yes I have written my technical requirements and judging based on it, and i think the on board adc would be sufficient for me.
can i have an interface for the external memory device, to store the processed data??
To be honest this is my first hands on experience with dsp's, but I know what i can do from it.
To tell you about the application it is basically an Voltage quality meter, i intend to calculate and record the number of occurrences voltage disturbances, and prevent the electronic system which are prone to it. I can say its a surveys the AC main voltage supply. For proper working of the devices, in home or industrial environments, electric vehicles, or smart grid applications, this would be useful.
I have made a list of IC's for the system power management, and other necessities.
I just had confusion with choosing the processor. Have i chosen the device which can meet my requirements?? Or should i look upon a few more processors.
Your suggestions on design are welcomed.
Well with budged available, and as it is an university project (I assume its main purpose is pedagogical) it makes no sense to complicate stuff. Therfore I would recommend 28335 with internal ADC. If you are still tempted to use external ADC make sure that your voltage sensors are accurate to 0.1% or better and you will need to use 32 bit FFT due to numerical loss of precision. Also you should use temperature compensated clock for DSP and all your voltage references, opamps, ... should hav minimal temperature drift. Otherwise you might think that you are accurate to 16 bits but in reality you will not be.
If I understand correctly you will be measuring grid voltage parameters? With fundamental frequency of 50 or 60 Hz? Ask yourself how fast do you really need to sample? Make sure that you use anti-alias analog filters.
Do you have a deadline?
Exactly!! yes I will be measuring the grid voltage parameters, with frequency of 50Hz. With sampling speed 10samples/us. I guess built in adc provides me with anti aliasing filters if not wrong!! will i need an external too? the processor also provides with clock, isnt it sufficient enough!!? and yes i have chosen the kind of opamps and voltage references which you have mentioned.
The deadline is Jan26,2013.
To add upon to this i have planned to have an power control system too, using mov's and other filters!! kind of feedback loop, with both monitor and control.
And i need to work on the algorithms more. Can i get any help in programming the processor, like how to proceed with the matlab code generation, if you can provide any links or tutorial documents related, i have CCStudio 3.3, what other software tools do i need?
Thank you very much for your kind support, it is great to get this help from you.
Your guidelines and suggestions regarding are very welcomed as it makes me analyze things better and easy from all your experiences.
Thanks lot Mitja
HI Tejas,
Regarding sampling, you will need external analog anti alias filter (maybe the voltage sensor bandwidth acts like one). I recommend that you read a text regarding digital signal processing, I'd recommend "Understanding Digital Signal Processing" by R. Lyons. Also why do you need to have such high sampling frequency?
By my experience you will be lucky if you'll have microcontroller sampling and doing FFT by the deadline.
When it comes to matlab code generation I am afraid that I can not offer any help.
Best regards, Mitja
I bought that book yesterday, nice coincidence...
For low frequency transients sampling 40kilosamples/ sec would be sufficient, but if I consider for a power distribution system, the sampling rate for recording the impulse transient should be higher than 2 megasamples per second(Msps) and IEC standards recommend 150 mega-samples per second.
Though I am not well confident that i can work at these high sampling frequency its just a mere hope, but i can measure low frequency transients.
As for matlab is concerned i have expertised professors here, i believe they can help me out in someway.
I will work on it for the best results, by all these conversations and experiences i learned many things.
Anytime if you get any kind of information related to this subject please share it, it would be a great help for me and others working on similar topics.
Thanks lot Mitja for your kind support, I wish I get the same till i complete my project.
Thank you,
Best Regards,