ebay u-blox LEA-6T GPS module teardown and initial test

[Gyro]

At the time there was no way that I could find to buy a bare module off the roll.
You could only buy a prebuilt module with optional ceramic antenna.
I recall Ublox ignoring me when requesting a sample of the bare module,
and some old fashioned integrity prevented me from stealing from my employer.

My goal to be consistent with the larger project was to construct everything down to the lowest reasonable component.

Some DC short circuits coupled with a capacitor make the best antennas!

Wow @RT, that's an awesome construction, I feel quite embarrassed now about having chickened out and bought an ebay module (I always use Manhattan style breadboarding if I can possibly get away with it), though as you say bare modules are almost impossible to get.

That helical antenna is a thing of beauty!   

[@rt]

The Garmin 60csx handheld had QFH antenna. Comparative drift tests with later models that have ceramic antennas,
clearly favoured the earlier 60csx so far as actual GPS hardware goes.
I’d always go with it for GPS where possible now... obviously not for aircraft!

Don’t be embarrassed, these modules weren’t designed for protoboard. I'm much more proud of the antenna.
I even made the optional EEPROM board to fit that box, but opted out of using it. Just so surprised how fast it was

[Gyro]

An Interesting development (well maybe for just a few anyway!).

I've been tracing out the RF section of the board and I think I've found an error. The 6 pin package near the RF pins of the module is a MMIC LNA, the board has amplification. I'm not sure of the exact part number, the package is marked 'LY' (but it's not a TOSHIBA dual AF transistor!). The pinout matches the AVAGO MGA-61562, and I think there's a Chinese version with the same part number, but no hits for 'LY'.

The photo shows what I found on the board (top diagram) versus what seems to be the standar application circuit (bottom). On the board the RF supply choke to the output pin is taken from the Bias Pin rather than decoupled VCC. That doesn't look right to me, I'm not sure how  much performance impact it might have though. It might just possibly account for performance being better with the circuit removed and an external passive patch antenna fitted (as Lars has done).



The other thing is that I'm not sure what value the bias resistor should actually be. The one fitted is 10R, but that coincides with the resistor recommended by u-blox for antenna overcurrent sensing, the Avago datasheet shows 680R for the bias resistor.

Section 2.2.2.7 of the LEA/NEO 6 Hardware integration manual Fig 22 shows the internal Ant power switching and how it should be connected, The RF input pin is tolerant of Antenna supply, but u-box don't recommend applying the supply directly to that pin (because it bypasses the internal switching and current sensing). I'm not sure if it's worth messing around with that though as the LNA is always powered.

I'm thinking of diving in and modifying / correcting the LNA circuit, any thoughts on whether this is deliberate (maybe required by whatever part it actually is) or just a design error ... and possible performance impact if so?

Chris

[uncle_bob]

An Interesting development (well maybe for just a few anyway!).

I've been tracing out the RF section of the board and I think I've found an error. The 6 pin package near the RF pins of the module is a MMIC LNA, the board has amplification. I'm not sure of the exact part number, the package is marked 'LY' (but it's not a TOSHIBA dual AF transistor!). The pinout matches the AVAGO MGA-61562, and I think there's a Chinese version with the same part number, but no hits for 'LY'.

........

Chris

Hi

If it really is the Avago part and if the resistor is really 10 ohms, the part should be pulling about 80 ma. That should be pretty easy to figure out. Is it blazing hot when you put your finger on it? At 80 ma, its running at max gain, reasonably low noise figure, and pretty good IP3. A quarter watt is not a good thing to dump into a normal battery powered design, the data sheet focuses on more rational power vs performance regions of use.

One of the most basic points about GPS is fairly simple:

As received, the signal is buried deep in the noise. Noise figure is more important than gain with these modules. The only real reason to have a gain block is to overcome losses between the antenna and the module. If you have no gain and your coax has 3 db of loss, your noise figure is 1 + 3 = 4 db. With a (hopefully) low loss pc board trace as your only connection to the antenna, gain may not be needed at all. To the extend that gain can overwhelm the AGC on the device, it can actually hurt performance.

Bob

[Gyro]

Hi Bob,

The same thoughts had occurred to me, no it isn't burning hot, in fact the whole unit is pulling well within the u-blox LEA current spec. That's why I'm wondering what sort of 'look-alike' part it might be - surely they wouldn't have fitted an Avago one (actually the device marking doesn't match). The absolute max current for the bias pin on the Avago part is 10mA anyway so it would be long departed by now - maybe it is dead and just not getting in the way of the signal too much.

Sure, the only time I've seen LNAs is at the head end (LNBs etc) to compensate cable loss, the u-blox manual warns about the dire consequences (edit: in performance that is) of overload, so why go to the cost and hassle of putting one on the board? 

Now I'm wondering whether I can bridge out the parts well enough to look anything like a controlled impedance stripline still.

Chris

[uncle_bob]

Hi

I'd bet it's some sort of LNA block that just happens to pinout like the one you found. It's not real clear what happens with their part from their data sheet at 10 ohms. Smoke and crater in the board is a good guess. It might work ... who knows.

The part is small enough that you probably can take it off the board and put in a tiny wire jumper. It should be small enough to not matter much.

Bob

[Gyro]

Yes probably. 

I'll try at bypassing it tomorrow after a bit more searching, the trouble is that it will be a distinctly one-way journey - there's no way I'll get it back on!

Chris

[uncle_bob]

Yes probably. 

I'll try at bypassing it tomorrow after a bit more searching, the trouble is that it will be a distinctly one-way journey - there's no way I'll get it back on!

Chris

Hi

Jumpering around it is going to be futile. You will destroy it trying to do that. It's either "pull it off" or "don't do anything".

The rated noise figure on the LEA-6T is only 3.2 db. That's not really great. Something in the 1 db range is pretty typical for a front end spec on GPS these days. With a good sat view that would be the difference between a 52 db c/n and a 50 db c/n. The real issue is how it impacts acquisition.  Unless you run two devices side by side, you will never see the 2 db delta after things lock up. There are simply to many other variables.

One thing you *will* see is better jamming (crud) immunity. If you live in the real world (where everything seems to emit RFI) that is more likely to be a limiting factor than noise figure.

Bob

[Gyro]

Well that was an interesting day...

In order to try and get some sort of qualitative assessment of the modifications I ran multiple runs in each configuration using cold-restarts each time. Each time it was positioned on the same spot on the bench.

1) I started with the original (slightly strange looking circuit as I documented in reply #27). Average time to 3D lock from cold start approx 1':30", HDOP 2.5. Number of satellites (green) 6.

2) I then moved the LNA output supply inductor sideways so that it was feeding directly off the supply decoupling cap rather than via the 10R. Result 3d lock a little quicker at around 1':20", HDOP around 1.8, No of satellites 6-7. OK, improving.

3) Removed the 10R resistor feeding the LNA (subsequently found to be feeding pins 4 and 5). Big improvement. 3D lock in 45-55 secs, HDOP around 1.3 (lowest I saw was 0.9), No of sats 8+ (10 max).

4) Fingers crossed. Time to remove the LNA, bridged pins 3-6 on the footprint with fine wire. Also removed the output supply inductor (obviously). 3D lock in ~2 mins (sometimes need to hold it above my head to get it to lock). HDOP 2.0, No of sats around 6. Oops 

In original configuration (1) It would occasionally lose lock, especially when I was leaning over it on the bench. In configuration 3 it was pretty much impossible to make it lose lock, even cupping my hands over the antenna. In configuration 4 it loses lock fairly easily when lean over it or cover it.

Conclusion... One modification too far! The LNA was obviously doing something useful, I tried to reflow it back on, but even on my Leister Hotjet S's lowest flow rate it kept blowing away. On the second attempt it pinged out of my tweezers and into oblivion. 

The LNA's identity is still unknown, The Maxim MAX2659 has a /shutdown on pin 5, but it also has VCC on pin 4 and doesn't need the supply fed to the output via an inductor, so that doesn't tie up. There are others that have a shutdown/ bypass mode. I suspect that it's probably a Chinese part, badly integrated onto the board.

My best bet at this point seems to be to implement u-blox's LNA supply voltage circuit properly and try it with my active external antenna.

I hope this will be of some help to others tinkering with the RF performance.

[PA0PBZ]

Did you measure the voltage at pin 4 in the original configuration? I don't think there will be a significant drop over 10R and it would not make any difference to put the coil directly on the rail.
Also, satellites move  so you can't really compare the performance like you did.

Anyway, I feel your (lna) loss 

The LNA's identity is still unknown

This looks close:

[jpb]

I've not been following this thread very closely, but I thought this module based on open-source hardware may be of interest since you can get the schematics and layout of a working timing system:

http://shop.sysmocom.de/products/osmo-lea6t-gps

(There is a link on the page.)

[Gyro]

Thanks a lot guys, both for the links and the sympathy!

Yes, it was that google image that first put me onto the 61563, I couldn't get access to pin 5 under the package so I assumed that it was tied low. Being high rules that one out and opens up the ones with /shutdown-bypass on that pin. No there wasn't a significant drop across the 10R resistor, however in RF terms anything hanging off it is a long way from the supply decoupling cap. I still can't make sense of Pin 4 possibly being a supply when power is being supplied to the output pin via the inductor - actually a step I didn't document, removing the inductor caused no-lock, so power must have been passing to the LNA that way.

Yes, good point on the satellite positions, and I guess density with time, I was relying mainly on cold-reset lock time which was significantly faster in config 3 every time.

Thanks again,

Chris

[PA0PBZ]

I wonder if the choke from pin 6 to rail is a red herring, if you google " gps lna marking ly" you find a few links to aliexpress showing a MAX2659 with the text "Marking LY". I don't think the choke would hurt a MAX2659.

[Gyro]

Hey PA0PBZ, well spotted, my first confirmed link between the two!  Actually I'd never have found it from here (I did a lot of searching for those terms yesterday), a search through google UK doesn't bring up the relevant hits until you add the magic word aliexpress - curious.

The good thing is that RS components sell them at £5 for a pack of five. Surely I must be able to get one of them down successfully, even if dead bug style.

I think you're right about the choke (although it did seem to make a difference), the Maxim schematic shows an internal series capacitor as part of the output matching circuit - maybe it's different on Chinese clones.

Thanks very much. 

[uncle_bob]

Hi

The interesting thing about HDOP / PDOP / TDOP and all the other  DOP's:

On most receivers, this is purely a geometric calculation. If the sat's are at the following angles, plug it in and out comes a number. It does not mater what the signal to noise is as long as the sat is being tracked. Please do *not* ask how many hours of staring at data and how many receivers it took to confirm this.

Time to first fix (cold start) is also a bit wonky. The GPS makes guesses about where it will find a sat. The more modern modules have a lot more horsepower (correlators). They make way more guesses so the likely first fix is at the first integrator dump. Signal to noise *does* degrade the time to first fix. Unfortunately, having a set of sats in a "bad geometry" also does the same thing.

Yes, it's painful. The only real answers are a common antenna and a group of devices *or* a fairly expensive signal generator. (which breaks and costs $10K or so to fix ...).

Bob

[Vgkid]

This is like my module.

[Gyro]

Another update for those who can still stand it.

Firstly an apology for the rather downbeat post yesterday. It was rather depressing (not) to see the then unidentified 1.5x1mm LNA disappear over my shoulder into the gloom together with apparent performance.

@Bob: Many thanks for the additional background, I am not a GPS expert I'm afraid (in fact my RF experience is limited to HF band military transceivers, consumer satellite/terrestrial STB design and far too many EMC tests over the years!). Agreed my measurements were not at all definitive, although over multiple restarts in each configuration there was definitely a 'feel' of better/worse. Not being in the position to do side by side comparison or  any relevant GPS test gear it will have to stand as the best I can do.

@Vgkid: Thanks, it hadn't even occurred to me to check out the u-blox eval kit schematics. They seem to have had several 'goes' at each one, from basic direct connector to LNA + SAW. I note that u-blox say that performance improves with an external LNA, but their eval design always pairs it with a SAW filter, which makes sense because additional gain would always be needed to overcome the filter loss.

Today's progress then... As I would need to order some MAX2659s, this was the best opportunity to try the external active antenna (the LNA would need to be removed or isolated by track cut anyway to allow clean access to the LEA-6T's RF input).

The first step was to properly implement the LEA-6T's LNA supply by adding the 10R overcurrent sense resistor between the VCC_RF and V_ANT. I fitted the resistor and then found a short between the two pins, after removing the resistor I found that the pins were shorted together by an inaccessible pcb track (these guys really can't follow a datasheet!). With some solder braid I was able to clean and then remove one of the PCB pads (sigh) and refit the resistor directly to the LEA module. On trying to enable RF power using u-center (actually the default setting)  I found no voltage at the RF input pin, 3V3 was present at the resistor and no drop across it. Ho hum, I'm not sure it this was due to thermal damage from the above or because the original configuration had 3V3 applied directly to the RF input via the choke on the LNA output. Not the end of the world, I could do the same again.

I've fitted the RF coax from my cheap ebay active magnetic patch antenna to the board (I'd previously inserted an in-line RF bulkhead connector in this for mounting in a box (with the Trimble Resolution-T) so it can be unplugged as needed.  To supply power in place of the LEA's faulty internal supply I've bridged a low vaue RF inductor from the adjacent VCC decoupling cap to the RF input (see photo).



The result is WAY better than the on-board antenna in any of yesterday's configurations! Using the same unscientific (sorry Bob) metrics on a series of 10 Cold restarts over several hours I found: 3D lock in <50 secs (often under 40), HDOP 0.8-1.1, Number of active satellites 8-11 (after 15 mins). Unscientific but clearly much better! The testing was done with the Antenna mounted on a small steel plate, on top of the shelves, just below the ceiling but beneath a rather a lot of storage in the loft. Its also been wet today so the roof tiles were wet. Cable length is rather long at 3-4m.

This is the ideal configuration for me (and I guess what I would have hoped for all along). It allows me to encase the module and have the clock signals accessible (rather than having to buffer them from a 'head unit'), better for USB integrity too. For the additional cost of the active antenna it seems a good solution.

I've attached a photo of what is probably the final configuration. I've removed the patch antenna from the 'back' of the board so that I could drill holes to add strain relief for the antenna coax.

Chris.

[uncle_bob]

Hi

The uBlox modules do not have an internal antenna supply. They simply route the voltage on the "antenna DC in" pin to the antenna pin via a choke. If you put +5 on the pin, you get +5 on the antenna. Same goes for +3.3V on the pin.

Bob

[Gyro]

Thanks Bob, but according to the Hardware Integration Manual the module puts out an LC filtered VCC on VCC_RF (pin18). Linking this to V_ANT (pin 19) via a 10R resistor is supposed to provide overcurrent protection/detection. Yes, for a +5V Antenna you would feed the 5V directly into V_ANT. I was definitely seeing 3V3 on V_ANT, it just wasn't making it through the internal switch to the RF_IN pin when commanded.

I've attached the block diagram.

Chris

[uncle_bob]

Thanks Bob, but according to the Hardware Integration Manual the module puts out an LC filtered VCC on VCC_RF (pin18). Linking this to V_ANT (pin 19) via a 10R resistor is supposed to provide overcurrent protection/detection. Yes, for a +5V Antenna you would feed the 5V directly into V_ANT. I was definitely seeing 3V3 on V_ANT, it just wasn't making it through the internal switch to the RF_IN pin when commanded.

I've attached the block diagram.

Chris

Hi

In about 99% of the places you see the module used, Vcc RF goes to nothing and there is a seperate supply going to Vcc Ant. The 10 ohm resistor is pretty much useless as over current protection.

Some math:

P = E^2/ R

If E = 5 and R = 10 you need a resistor that will handle 2.5W

I = E/R

Your short circuit current will be 1/2 A. That pretty much takes out the choke they use to put the DC on the antenna. It *may* take out the ferrite bead that does the Vcc RF.

The next layer to the problem is that there is no ESD protection (or lightning protection) on the antenna input. With a 10 ohm resistor, you may / may not get a "spike" back into your main system.

It is ...errr ... possible ... that ... somebody... might ship a few of these into the field before working that all out and fixing the others.

Bob

[Gyro]

Ah, I didn't say it was a GOOD system, just what is in there  It should have at least given me volts on the Antenna feed if there wasn't an internal fault.

The VCC_RF looks like a bit of a waste of a pin to be honest, but I suppose it might save a couple of external passives in very space constrained designs (mind you, those would probably be using the NEO anyway).

Agreed on the resistor dissipation, a little better at 3V3 but not by nearly enough. I think the intention is probably that something notices (don't know if the internal  f/w has this capability) and turns it off before cooking becomes a problem, not exactly failsafe though. I note that the datasheet max V_ANT current is 100mA.

My current solution seems fine anyway, I have no need to switch the antenna supply and the external inductor is just taking the place of the internal one. The 3V3 regulator current limit and thermal shutdown should provide a worst case blunt short protection.

Chris

[uncle_bob]

Ah, I didn't say it was a GOOD system, just what is in there  It should have at least given me volts on the Antenna feed if there wasn't an internal fault.

The VCC_RF looks like a bit of a waste of a pin to be honest, but I suppose it might save a couple of external passives in very space constrained designs (mind you, those would probably be using the NEO anyway).

Agreed on the resistor dissipation, a little better at 3V3 but not by nearly enough. I think the intention is probably that something notices (don't know if the internal  f/w has this capability) and turns it off before cooking becomes a problem, not exactly failsafe though. I note that the datasheet max V_ANT current is 100mA.

My current solution seems fine anyway, I have no need to switch the antenna supply and the external inductor is just taking the place of the internal one. The 3V3 regulator current limit and thermal shutdown should provide a worst case blunt short protection.

Chris

Hi

Coming up with an inductor that is really good at 1.5 GHz is not as easy as one might think .... Then you have to get it onto a trace that does not mess up the RF matching.

Bob

[Vgkid]

Has anyone delved into the manual to determine how to interface with the chip(using usb) while the chip is already powered up. I have just plugged a usb cord into the module, and I reported usb power errors.

[DimitriP]

It seems to me the device is either powered from the USB port, or not.
There is some hint of an explantion in section 1.6.2.1 in the Hardware Integration Manual  and it sounds like an either/or but not both kind of situation.
As for having it powered on and then connecting it to a USB port..I guess it depends on how the board you have is wired up.
What's the  error ?

[Gyro]

The Hardware Integration Manual section 2.1.2 gives a further hint:

2.1.2 Design considerations
For a minimal design with a u-blox 6 GPS module the following functions and pins need to be considered:
•  Connect the Power supply to VCC.
•  VDDUSB: Connect the USB power supply to a LDO before feeding it to VDDUSB and VCC. Or connect to
GND if USB is not used.

It looks to me as if in Self-powered applications, VCC is powered and VCCUSB becomes a 'USB present' indication (though still needs to be driven by an LDO).

There probably needs to be a diode-OR from the USB supply to the input of the main VCC LDO so that both VCCs still get powered if the USB is plugged in while the unit is turned off - I assume it is illegal to Power up VCCUSB while VCC is powered off.

Chris