RFID, what DSP Controller?

Hello,

My goal is to make a 16 channel fixed RFID reader with DSP. 
But I don't really know what controller to use. 
I will be using the ISO 11785 standard. The HDX transponder uses FSK
modulation, 124.2 kHz is logical "1" and 134.2 kHz is logical "0".
Questions are how fast do I want the cpu to be, do need a fpu? etc.
I need a 16 channel ADC, so preferably a controller instead of a
processor.
If you need more info, please ask me.

Is there anyone out there who can help me choose the right one?

Thank you,

M

dsp_questions wrote:

> Hello,
> 
> My goal is to make a 16 channel fixed RFID reader with DSP. 
> But I don't really know what controller to use. 
> I will be using the ISO 11785 standard. The HDX transponder uses FSK
> modulation, 124.2 kHz is logical "1" and 134.2 kHz is logical "0".
> Questions are how fast do I want the cpu to be, do need a fpu? etc.
> I need a 16 channel ADC, so preferably a controller instead of a
> processor.
> If you need more info, please ask me.
> 
> Is there anyone out there who can help me choose the right one?

This is going to be rather fat DSP with expensive ADC. You would be 
better off if you change the approach. Use the simple hardware to 
downconvert/demodulate, then sample the signal into a microcontroller.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

On Fri, 04 Sep 2009 08:43:57 -0500, Vladimir Vassilevsky wrote:

> dsp_questions wrote:
> 
>> Hello,
>> 
>> My goal is to make a 16 channel fixed RFID reader with DSP. But I don't
>> really know what controller to use. I will be using the ISO 11785
>> standard. The HDX transponder uses FSK modulation, 124.2 kHz is logical
>> "1" and 134.2 kHz is logical "0". Questions are how fast do I want the
>> cpu to be, do need a fpu? etc. I need a 16 channel ADC, so preferably a
>> controller instead of a processor.
>> If you need more info, please ask me.
>> 
>> Is there anyone out there who can help me choose the right one?
> 
> This is going to be rather fat DSP with expensive ADC. You would be
> better off if you change the approach. Use the simple hardware to
> downconvert/demodulate, then sample the signal into a microcontroller.
> 
For a lot of applications this would definitely be the way to go.  For 
some you may want to downconvert externally then demodulate in the DSP, 
to take advantage of the better signal processing opportunities in the 
DSP.

But you're still left with needing quite a lot of DSP horsepower.  My 
suggestion would be to pick an economical processor and build a module 
that can do 1, 2 or 4 channels, then combine as many as necessary to get 
up to your required 16 channels.

-- 
www.wescottdesign.com

On Fri, 04 Sep 2009 07:32:44 -0500, dsp_questions wrote:

> Hello,
> 
> My goal is to make a 16 channel fixed RFID reader with DSP. But I don't
> really know what controller to use. I will be using the ISO 11785
> standard. The HDX transponder uses FSK modulation, 124.2 kHz is logical
> "1" and 134.2 kHz is logical "0". Questions are how fast do I want the
> cpu to be, do need a fpu? etc. I need a 16 channel ADC, so preferably a
> controller instead of a processor.
> If you need more info, please ask me.
> 
> Is there anyone out there who can help me choose the right one?
> 
> Thank you,
> 
> M

In some ways, if you have to ask that question then the answer won't help.

To size the processor you have to know what algorithm you're going to 
use, and what analog preprocessing (see Vladimir's and my comments).  
Were I to do this I would do a block diagram of the signal processing, 
both in and out of the DSP, I would prototype the algorithm in Scilab and 
get it working, _then_ I would estimate the DSP horsepower required.

The only definitive comment I can make here is that if you need floating 
point for this, then DSP probably isn't the field for you.  Floating 
point is handy, but it makes the processor expensive -- and you're 
already implying an expensive processor.

(Note, too, that if you're bound and determined to do this with one 
monolithic board and no analog preprocessing you're probably damning 
yourself to a signal processing chain with separate ADCs, an FPGA, and a 
processor, or a processor on the FPGA).

-- 
www.wescottdesign.com

"dsp_questions" <mariusfrielink@gmail.com> wrote in 
news:z5Sdnc_sIK9xlTzXnZ2dnUVZ_gCdnZ2d@giganews.com:

> Hello,
> 
> My goal is to make a 16 channel fixed RFID reader with DSP. 
> But I don't really know what controller to use. 
> I will be using the ISO 11785 standard. The HDX transponder uses FSK
> modulation, 124.2 kHz is logical "1" and 134.2 kHz is logical "0".
> Questions are how fast do I want the cpu to be, do need a fpu? etc.
> I need a 16 channel ADC, so preferably a controller instead of a
> processor.
> If you need more info, please ask me.
> 
> Is there anyone out there who can help me choose the right one?
> 
> Thank you,
> 
> M
> 
> 

If you mux the front end, you could use a modern SHARC with its PDAP (fifo) 
input. Our dspblok 21369zx would be one choice.

We also have DSP-FPGA solutions but I don't think you will need the FPGA for 
downsampling. 

Perhaps you could just undersample the inputs (as long as the input S/H is 
fast enough).

Al Clark
Danville Signal

On Sat, 05 Sep 2009 14:59:49 +0000, Al Clark wrote:

> "dsp_questions" <mariusfrielink@gmail.com> wrote in
> news:z5Sdnc_sIK9xlTzXnZ2dnUVZ_gCdnZ2d@giganews.com:
> 
>> Hello,
>> 
>> My goal is to make a 16 channel fixed RFID reader with DSP. But I don't
>> really know what controller to use. I will be using the ISO 11785
>> standard. The HDX transponder uses FSK modulation, 124.2 kHz is logical
>> "1" and 134.2 kHz is logical "0". Questions are how fast do I want the
>> cpu to be, do need a fpu? etc. I need a 16 channel ADC, so preferably a
>> controller instead of a processor.
>> If you need more info, please ask me.
>> 
>> Is there anyone out there who can help me choose the right one?
>> 
>> Thank you,
>> 
>> M
>> 
>> 
>> 
> If you mux the front end, you could use a modern SHARC with its PDAP
> (fifo) input. Our dspblok 21369zx would be one choice.
> 
> We also have DSP-FPGA solutions but I don't think you will need the FPGA
> for downsampling.
> 
> Perhaps you could just undersample the inputs (as long as the input S/H
> is fast enough).

And the sampling jitter is low enough.  I hadn't thought of one of your 
boards -- if the OP is just doing one then it's probably cost effective 
to throw a bunch of processing power at it, and buy it back with reduced 
engineering time.

But it could just be an idle question, because the OP hasn't responded to 
any of the responses.

That particular RFID protocol is odd because the bit lengths are 
different.  I suspect it was done to minimize the logic in that itty 
bitty RF-powered tag, but it means that PLL bit synchronization is out of 
the question (or at least prohibitively weird); that makes it hard to 
treat it as anything but a strong-signal mode.

So you either treat it as strong-signal and hope the tag's return is 
always good and strong, or you treat it as a weak-signal mode and you go 
crazy trying to synchronize to the bits, and wondering if you should be 
using bit synchronization in your bit detection...

-- 
www.wescottdesign.com

Thank to everyone for the replies, I haven't responded yet because I'm not
at my workplace. My colleague is working on the analogue part and has made
an analogue active bandpass filter for each input channel. The exact
details of the filter I don't know right now, but will post it as soon as I
do. 
So more information will follow. 

Thanks so far.

M

Some time has passed and I've been looking especially at TI's TMS320F2812.

I like it because it has a fast 16 channel adc (80ns) and it is powerful
enough for the dsp algorithms I will be using. 

But what I would like to know is, are there similar alternatives and does
someone has (good) experience with one. I do want it to be a dsp
<u>controller</u> with 16 channel adc. 

Thank you.

M

dsp_questions wrote:

> Some time has passed and I've been looking especially at TI's TMS320F2812.
> 
> I like it because it has a fast 16 channel adc (80ns) and 

The ADC of TMS28xx is junk. It is wildly inaccurate and noisy. Moreover, 
you can't relate the sampling instants of the sequencer to the real time 
events; this includes the other activities of the TMS28xx.

The TMS28xx is a motor control, not a DSP. Its performance as the DSP is 
fair. You can relate the performance of 28xx at 80MHz to a real DSP chip 
at ~25 MHz. The 28xx is a power hog also.

> it is powerful
> enough for the dsp algorithms I will be using. 

How did you decide that?
Direct sampling of 16 channels at ~ 300KHz = 4.8M samples/s. This is the 
job for a powerful DSP or FPGA. You can use the bandpass sampling, but 
that will take the appropriate analog part, and the performance is 
likely to take a hit.


> But what I would like to know is, are there similar alternatives and does
> someone has (good) experience with one. 

You would be better of if you change the your approach to the problem in 
the subject of the thread.

> I do want it to be a dsp
> <u>controller</u> with 16 channel adc. 

If you want a DSP with an ADC, take the DSP and the ADC separately.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

On Wed, 09 Sep 2009 09:51:50 -0500, Vladimir Vassilevsky wrote:

> dsp_questions wrote:
> 
>> Some time has passed and I've been looking especially at TI's
>> TMS320F2812.
>> 
>> I like it because it has a fast 16 channel adc (80ns) and
> 
> The ADC of TMS28xx is junk. It is wildly inaccurate and noisy. Moreover,
> you can't relate the sampling instants of the sequencer to the real time
> events; this includes the other activities of the TMS28xx.
> 
> The TMS28xx is a motor control, not a DSP. Its performance as the DSP is
> fair. You can relate the performance of 28xx at 80MHz to a real DSP chip
> at ~25 MHz. The 28xx is a power hog also.
> 
>> it is powerful
>> enough for the dsp algorithms I will be using.
> 
> How did you decide that?
> Direct sampling of 16 channels at ~ 300KHz = 4.8M samples/s. This is the
> job for a powerful DSP or FPGA. You can use the bandpass sampling, but
> that will take the appropriate analog part, and the performance is
> likely to take a hit.
> 
> 
>> But what I would like to know is, are there similar alternatives and
>> does someone has (good) experience with one.
> 
> You would be better of if you change the your approach to the problem in
> the subject of the thread.
> 
>> I do want it to be a dsp
>> <u>controller</u> with 16 channel adc.
> 
> If you want a DSP with an ADC, take the DSP and the ADC separately.

I've designed with the '2812 -- it was a very nice fit for the problems 
that we were solving, and is a continuing success for my customer, but we 
didn't use it's on-board ADC for anything critical.  One of our analog 
guys referred to it as "12 bits going on 8", which I liked.

You just won't find a processor with a good on-board ADC, because the 
semiconductor processes that make fast, inexpensive processors make 
really crappy ADCs.  You _will_ find 12-bit ADCs advertised, but when you 
start analyzing the specifications for the effective number of bits 
you'll find that by most measures it's less than 10 and often barely 8.

So do like Vladimir says, and use an external ADC (and probably follow it 
with an FPGA or CPLD for the first-line signal processing).

-- 
www.wescottdesign.com