Need good and inexpensive GPS component

Take a look at

Meindert

Oops, that should be

Meindert

1. AFAIK, no one has a single-chip GPS solution. Perhaps there is just too much process difference between the RF section and the processor section.
2. The smallest modules are less than one sqare inch, so you're not going to have to waste a lot of board space.
3. The UBlox modules are about 0 in small quantities and you can also get information at
   formatting link

Mark Borgerson

As you say there are no GPS single chip products. We have just completed a survey of the available GPS modules and there are currently two that are neck and neck. u-Blox and Fastrax.

u-Blox has LEA-4A and LEA-4P (programmable) as well as brand new high sensitivity units LEA-4H and LEA-4S. They run between about $100 in small quantities (the high sensitivity units are about $20 more) and under $30 in large quantities (2500 or more).

Fastrax has the iTrax-03S which is not quite as sensitive, but is the same price range.

I don't know how supportive either of these vendors will be to a small volume user. But I may be working on an Open Source GPS project later this year and if so I will arrange a supply of units to anyone who wants them.

That's the difficult part. Time data, etc. are just feebies.

We are considering the Atmel ATR0600 RF front end chip. It extract 4MHz IF signals for additional processings. A good ARM/PPC/X86/FPGA can process the 4MHz signal for GPS data.

That's still too expensive for us. We need it less than $10

In case anyone is interested, I just came across this GPS module from a UK supplier:

35.19 UKP including VAT, antenna extra. No experience of them, I just came across them today while searching for a low cost SBC.

Simon.

--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
If Google's motto is "do no wrong", then how did we get Google Groups 2 ?

Yes, but these devices will cost much more than $10. Your design time will be significant as well. Buying the module gives you a full solution that spits out position data a few times a second and requires no work... all for a low, low $25 in quantity. If you need a price of $10, how can you use an FPGA and CPU to finish the task???

If you need the FPGA and ARM for other things, then you still have to pay the incremental cost of larger/faster parts required to do the GPS processing in addition to the huge NRE expense. That is why they make the baseband chip and roll it together with an ARM. I have been told that the GPS processing is using about 85% of the CPU on an ARM7 which you can buy for under $4. To gain any headroom for other tasks you would need to switch to a part that uses an ARM9 for another $10. So pay me now or pay me later...

If your quantities are small, why do you need such a low price?

Good luck meeting that price point. :-)

not,

The SiGe SE4100 or Atmel ATR0600 would cost less than $10.

But we don't need a few times per second, perhaps once every minuties.

That's for 10K or more, or $100 in hundreds.

Only the incremental cost of GPS is important.

But we can't use that ARM for anything else, so it would cost us another ARM and LEG.

Again, we don't need sub-second response time.

only

not,

My understanding of GPS processing is that you have to keep doing it to keep the correlations and time differences up to date, or else you have to start all over again. In that case you have to do all the processing at a fairly intensive rate for 15 to 30 seconds to reacquire the ephemeris data and or update the time differences.

The only way to meet the cost and processor goals you are seeking may be an assisted GPS solution similar to that in cell phones. In that case, the GPS incremental cost may meet your $10 budget, by you need the cellular infrastructure to do the communications and calculations. If you're not building a cellular enabled gadget, you are probably out of luck at the $10 price point.

No one yet has mentioned the additional cost of the antenna and cables. Just those components alone will eat up your budget in small quantities.

I think it's going to take more than luck!

Mark Borgerson

only

If not,

That's fine. We are not flying, so we don't need the GPS to tell us that we are in the same spot (+/- meters) every seconds.

Yes, there will be another processor doing the calculations.

Yes, but they are priced as options. It would be difficult to make a 100+ QFP chip optional. For systems not using the GPS option, they don't have to pay the extras. $100 would be more than most of the other components.

I am sure the ready-make module would be attractive for some. It's just not for us.

Personally, I don't think the math adds up. Your incremental costs of addiing CPU horse power to do the calcs and adding FPGA gates to do the correlator is at least as much if not more than the extra $15. If you are only buying 100 units, then how can you possibly justify the NRE to reduce your recurring cost by less than $10,000??? I can assure you it will take N*$10,000 to do the design work you are describing. There are also restrictions on such designs. I have no idea how they are enforced, but I have been told that if the unit is going both faster and higher than some threshold values the unit has to shut down and stop providing data. This seems to be to prevent using commercial GPS units for missiles or something.

If you think you can do it this way, great. Please come back and let us know how it goes.

The CPU and FPGA are for other processings as well, so they do not have additional cost.

We are not ruling out IP licenses for the whole or in part. There are existing IP libraries for digital PLL, correlations and matrix equations. We are not trying to re-invent all that stuffs. We are just ruling out using a dedicated processor for GPS only.

I can assure you that we are not building missiles, or any fast moving objects. We only need some rough geolocation data once in a while. It does not matter if we restart the acquistion process every time.

Motorola had a single package GPS, the FS-Oncore. I spent 3 years working on it. I am not sure what became of it after Mot sold their GPS operations to Sirf. It needed a pwb, a few capacitors, a patch antenna and level shifters to talk to standard RS232. Sensitivity was in the

-146 -150dB range. The package was a 60 something ball BGA package I think. I don't know what the selling prices was in less than rail car quantities.

Good Luck, Bob

Sorry to follow up my own post. I forgot to mention the TCXO in the required parts list. Also, these people are selling eval kits of it:

Of course they have additional cost. FPGA incremental pricing is far from cheap and even the IO pins cost you extra as you need more. If you are using the smallest, slowest, cheapest FPGA and CPU you can get, then I guess you can't save by going with a smaller ones, but are you really using a 4 bit processor and a 22V10 CPLD???

Don't kid yourself. Doing the GPS calcs yourself is likely the most expensive way you could possibly do it.

What does IP have to do with it? The trade off is between saving $15 on each of 99 units since you are saying you can't even get the 100 piece discount. Actually, I did MY math wrong, that is only $1500 you are saving on recurring costs. The expense of discussing this with your management will easily top that.

The application is not the point. I belive there are regulations on building GPS receivers that you may need to meet. I think the issue is called the COCOM limits. I don't know a lot of detail, but a GPS receive that does not have these limits might not be exportable or have other restrictions. This seems to be so that this units can not be adapted to weapons.

I don't want to sound negative on your project. It is just that it doesn't sound completely realistic or you are leaving out some important details.

Don't worry, we are going to have a Xilinx Xc3s200 (200K) FPGA anyway. It will have a 500MHz CPU as well. They are needed whether we are getting GPS or not.

True for any other designs (GPS or not) as well. The easiest way is always to buy everything.

The difference is adding an additional QFP for $90 more, so $9,000 for first 100 (@90 ea) and approx. $50,000 for first 1,000 (@$50 ea). We would not hit $250,000 for 10,000 (@25 ea) any time soon.

Don't worry. We are not building weapons and not exporting them either. By not having sub-second response times, we are well within any security related limits.

Nothing other than the cost.

An interesting product. It is the smallest unit I have seen yet at 12 x 16.6 mm. But the power consumption is nearly double the leaders in the market at 180 mW. I will have to contact the makers and add it to our eval list.