Discrete GPRS

Of course. How do you think the modules are made?

If you are asking this question, you do not have what it takes (purchasing power, access to "secret" data, base station simulators, &c) to carry it out.

This type of answer is what I really do love about newsgroups. Somebody wants to satisfy his / her curiosity an then this response...

Not "it could be very difficult for you, read this book to get some more information". Not even: maybe this is too difficult.

NO: you do not have what is takes

Now: English is not my first language, but do I a detect a weeny bit of arrogance here? You do not have what is takes?

Yes,

You will need to

1.choose GSM/GPRS Silicon Manufacturers

1. Spend 0K in development equipment and qualifed RF engineers
2. Convince the Silicon Manufacturer that you will purchase a very very high volume of IC's (1M pcs)
3. Spend a few months in development
4. Spend lot of money in GSM approvals

So proababaly overall $1M

So yes it can be done.... but..... me.... I just will get the premade modules.

Joe

This is nothing to do with English and nothing to do with arrogance.

We routinely get questions in here of the form: "I need to build a cruise missile. I understand I will need diodes, what else and please give plans, schematics, sourcecode". This question falls into that category.

If the OP was capable of carrying out this project he would not be asking the question because anyone who can execute such a project already has sufficient background knowledge to understand the complexity of the domain knowledge required.

In c.a.e it is often possible to make this determination with utter assurance of being correct, and this is such a case.

I tend to agree with the you do not have what it takes (money) answers.

But if someone wants to do it in the lab for curiosity, they should probably read up on GSM and look into software defined radio experimentation platforms (GNUradio, HPSDR, etc). This would not be a portable project of course - I can recall years ago interviewing at (soon to be famous chipset maker) and being shown a "phone" that was about a shoebox's worth of FPGAs, proving the algorithms for what soon would be asics for a handheld.

If you had a suitable shielded enclosure, trying to build a "base station" that could interact with an existing handset could be a very interesting project!

un

entrevistas

" que se

bar lo que pronto

er una muy

Ok, thanks !

And if you could do this on your kitchen table, you would certainly command a huge salary from any one of the GSM module manufacturers!

I'm curious, what part of building a lab-demonstration stand-alone GSM femtocell do you think is so challengingly incompatible with a kitchen table?

The one question in my mind is if you'd need to be able to custom program a sim for the phone.

The lack of access to an ASIC to do it all? Developing and debugging a "discrete" (no ASIC) design is a many-man-year activity.

Initial GSM deployments were something like 15 years ago... which suggests to me that a lot of the functionality that one required a tabletop worth of FPGAs can probably be done in software on a fast processor today (much easier debugging) with only key parts adjoining the ADC/DAC kept in FPGAs.

What was once a multi-hundred-million dollar activity is now a mere multi-hundred-thousand-dollar activity, yes. But still well off the kitchen table budget, unless we're talking about Bill Gates' kitchen.

Debugging these designs at the software level isn't sufficient, you need considerable RF equipment and man-hours. Just a spectrum analyzer and signal generator covering the right frequency range and with the features you need to verify the design's RF performance will be in the region of $75k (at least - the sig gen alone will probably be about $40k).

If you purchase a suitable RFdigital front end you should not need any physical test equipment at all.

In many cases even if you did need it, a second instance of the RFdigital board will do the job - this is not a high performance application where things like the noise figure, instantaneous dynamic range, and phase noise of the analyzer matter. Should you for some reason need an analyzer with high performance, do what is common in industry: rent it.

This task is quite beyond the capabilities of a $99 "USB spectrum analyzer WOW! L@@K! 100% POSITIVE FEEDBACK!" as found on eBay!

?!! Are we sure we're talking about the same task? Admittedly it's not quite as complicated as say IS-95, but just making sure the txceiver is settling quickly enough without noise spurs appearing all over the place is DECIDEDLY nontrivial.

If size or weight is not an issue, I would simply use a linear transverter to generate/decode the signal at something like 70-140 MHz and convert it up to 900/1800 MHz.

Since the GSM TDMA signal is only 200 kHz wide, you could even generate it with a DSP+DAC/ADC at 2 MHz, use one transverter to go to

100 MHz and a second to 900/1800 MHz. The first transverter could even be implemented using traditional SBL-1 diode ring mixers :-).

However, these days I/Q modulator/demodulator chips good for at least

2450 MHz are available, so doing the GSM signal with zero-IF approach at the final frequency would be much easier to do.

If you are not constrained with cost or power constraints, you could use quite different approaches.

You might even implement it with tubes, of course getting hold of lighthouse tubes or klystrons for the transmitter, might be a bit problematic these days.

GPRS/SMS implementation might even be realistic, since you would not have to implement the audio codec :-).

Paul

Just put a crystal controlled receiver converter in front of your cheap spectrum analyser. Of course you will have to calibrate your 0 dB settings with a known signal source.

Do the signal handling at some low frequency and use some crystal controlled transverter to switch the frequency to the actual RF frequency.

Paul

I was talking about something akin to the gnu radio USRP, or if that is not sufficient a more modern design along those lines. If USB causes too much latency it would need a closer connection to the processor.

Simply solved by using fixed tuning in the RF and implementing frequency agility on the digital side.

Remember we are talking about a lab demo, not a product. Lots of things that could be problematic in productizing something can be avoided in this type of project.

Hetting it down to a lower frequency will hide the unwanted spurs if they're outside the bandwidth you picked, though.

I do understand what you're saying in principle, but do not think that it is achievable in practice (and certainly would not be good enough for type approval).

Depends on your receiver converter front end selectivity. On a wide front end, the spurs will show up as image responses in the narrow band spectrum analyser, so at least you know if they are there.

To identify which spur you are seeing, just slightly move the transmitted frequency and observe in which direction and how fast (2x,

4 x etc.) the alias is moving and you can calculate the actual frequency of the spur.

Of course, such mental exercise would be too much for most younger people working in an RF lab these days :-).

For far away spurs, simply put sufficient filtering before the transmitting antenna.

Commercial vendors can not afford this due to cost constraints, so they have to design a clean enough transmitter that will suffice without any extra filters. Paul