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

[bitseeker]

Those results sound great, DrTune. More than good enough for my needs. Looking forward to your vids.

[DrTune]

Vid!  Two LEA-6T's head to head. Looks a bit rough but IMO it's a pretty decent result. The jitter is pretty much down to the 48Mhz clock granularity.

 


10Minute persistance view (after GPS units had been on overnight)

[Gyro]

At last, someone with the wealth to buy 2.

Good comparison data, thanks. 

BTW I'm in the process of developing a very basic GPSDO based on these modules, in case you haven't seen it...

https://www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-%28very-scruffy-initial-breadboard-stage%29/

[Robaroni]

Gyro,
Love to see what you come up with. I've been watching this on the bench for the last few days and it is rock solid to my rubidium.
Rob

[splin]

You said that the unshielded cables were screwing up GPS reception (thus  explaining the > 30ns rms and 99% <60ns specifications for timing error between the two) but the Ublox software seemed to show at least one was seeing quite a number of satellites and thus working ok. Did you see that the second wasn't locked? Does the LEA-6T have an output that shows when the timing output is locked and valid?

At the end of the video when you have the timebase set to 10ns/div the 2nd trace shifts by approx 10ns at a time. How can it do that when its clocked at 48MHz?

[jpb]

Does the LEA-6T have an output that shows when the timing output is locked and valid?

According to the Ublox white paper on timing you can set the time signal amplitude to be zero when it is not locked so the signal is only present when it is locked. (It is shown in one of the figures but I don't have it to hand to be more precise.)

I have a couple of these Ublox 6T modules (the cheap e-bay ones) and also a couple of older GPS timing modules with 10kHz output and at some point I'd like to compare them. Unfortunately at the moment I keep moving between rented places whilst I try to sell my house after a job move so I no longer have a GPS antenna setup.

The timing module has a correction mode to correct for the 48MHz jitter (quantization). In the white paper they show results with rms on the timing pulse of only 6.6 nsecs uncorrected and only 3 nanosecs corrected.

A quick back of an envelope calculation showed this was consistent. The spec in the datasheets is much wider (30 nanosecs and 15 nanosecs) to allow for low to the horizon satellites and atmospheric effects (according to the white paper). You can send a command to tell it to ignore low angled satellites.

James Miller (of the simple GPSDO fame) told me he has his GPS antenna in a chimney pot (mainly to avoid spoiling the looks of his house I guess) but such an arrangement where it is forced to look up more might work very well for timing.

[DrTune]

You said that the unshielded cables were screwing up GPS reception (thus  explaining the > 30ns rms and 99% <60ns specifications for timing error between the two) but the Ublox software seemed to show at least one was seeing quite a number of satellites and thus working ok. Did you see that the second wasn't locked? Does the LEA-6T have an output that shows when the timing output is locked and valid?

Yeah I wasn't being super scientific with the unshielded cable comment - the ublox s/w was showing quite a lot of sats but it was also saying "no fix" in red (instead of 2d or 3d fix in green) and kinda alternating between no fix/fix. There's a sat signal strength waterfall and that wasn't looking very good when I started shooting the video.   I could run the test again OR I could simply not be a dumbo and use shielded cable. When the unit is trying to receive -160dbm or whatever GPS is, having two long bits of wire radiating squares wave edges 12 million times a second probably doesn't help.

Yes I'm pretty sure it does have a 'locked' output you'd have to check the datasheet but I think so; obviously it does tell you over the serial/usb output - one really nice feature is that the clock period/duty cycle can be set to two different values for "locked" and "not locked" so for example you could have a 20% duty for locked and 50% for unlocked and it'd still be a useful signal, I thought that was a neat idea.   It's not at all hard to solder extra wires onto the Ublox module (see previous pics in this thread by other folks).

At the end of the video when you have the timebase set to 10ns/div the 2nd trace shifts by approx 10ns at a time. How can it do that when its clocked at 48MHz?

I think that's just an illusion; obviously at 1hz the update is pretty slow so you don't have many samples to look at; if you look at the other faster clock stuff earlier in the vid you can see the jitter is evenly distributed (as it should be, there's no physical connection between the two clocks, the position of the jitter just depends on the phase of the two TCXO's in the GPS modules).  If I put the clock on say 10 or 100hz it looks as random as you'd expect it to.  It'd actually be quite a technical feat if somehow there was granularity in the jitter!

[Gyro]

Yes, you can configure the outputs to generate a constant low, constant high, or free run clock of any frequency before lock. After lock you can, if you want, output a constant high or low too instead of a clock as a permanent lock signal... or any duty cycle or pulse interval between these states. It's pretty flexible.

I'm running mine with the timing app note recommended settings: SBAS turned off and reject low angle satellites (it's 5 degrees by default). At 45 degrees I could drop down to two or even 1 satellite briefly at my location and indoor antenna. 30 degrees seems like the best compromise for me at the moment (at least 3 satellites used at all times). Of course all satellites are still tracked, it just selects the ones being used to calculate the time fix.

[DrTune]

The timing module has a correction mode to correct for the 48MHz jitter (quantization). In the white paper they show results with rms on the timing pulse of only 6.6 nsecs uncorrected and only 3 nanosecs corrected.

Mmm I thought that was interesting, it looks like you basically switch it to a low frequency output (e.g 1hz) and then "after the fact" of each 1hz tick you read out over UART/whatever an error value in nanoseconds.   Personally I'd rather they just had e.g. a 200mhz clock inside the GPS but hey, I'm nitpicking :-)

[Vgkid]

I have my small taoglas gps antenna in my window, i have no problems there. With the 6T , it stays locked to atleast 8 sats at all times.

[jpb]

I have my small taoglas gps antenna in my window, i have no problems there. With the 6T , it stays locked to atleast 8 sats at all times.

For timing using just one satellite is better according to Ublox. I guess though the problem is when the unit needs to switch between satellites.
Multiple satellites is better for navigation of course.

[Gyro]

I guess though the problem is when the unit needs to switch between satellites.

Yes, that's the one that's puzzling me. I'm working on the basis that at least 3 high angle satellites might be the best compromise for switching effects but I don't know if that's the case.

P.S. I haven't found a setting in ucenter to specify the number of satellites rather than the angle, that would probably be more useful.

[DrTune]

Setting # of satellites to use:

It appears from looking at the protocol spec that you can limit the number of satellites (a.k.a. "SV" = Space Vehicles) using the NAVX5 setting;
From page 196 of the u-blox6_ReceiverDescrProtSpec document:

"The minimun number of SV default setting is set to 1 in a firmware with the timing premium feature enabled (LEA-6T)"
which totally figures, the default minimum for non-timing GPS modules is 3.
There's also "Maximum number of SV" = 16, the GUI will let you set this as low as 2.

Furthermore on the "NAV5" setting there is "Min SV elevation" (default 5 deg) described as "Minimum elevation of a satellite above the horizon in order to be used in the navigation solution. Low elevation satellites may provide degraded accuracy, due to the long signal path through the atmosphere."

Not tested either but that looks promising if you want to limit angles/and #SV's.

I like "SV", it sounds kinda techy and also suitably open-ended for future enhancements, i.e. if in future instead of satellites we launch, say, Donald Trump into a 12,550 mile high orbit.

[Gyro]

Well there were some of the old Routemasters that had clocked up close to a million miles. 

I think I tried the NAVX5 settings page without much success. It's a bit confusing because the min #SVs setting also looks like a min C/N. The Min SV elevation setting certainly works.

[Robaroni]

I'm setting mine for min angle.
Rob

[edpalmer42]

Normally, you should set your minimum elevation according to your local conditions i.e. what angle do you need to clear nearby obstructions like trees and buildings.  This lets the receiver reject signals from satellites that will be distorted by travelling through trees or bounced off buildings.

Artificially restricting the number of satellites you use isn't usually a good idea.  Yes, you may see a glitch when a satellite joins or leaves the constellation, but you minimize that by having a good location fix and you will probably get your best fix with lots of satellites.  The timing receivers have extra smarts to select the best satellites at all times.  If you restrict the number of satellites used, it could impact the performance of those features.  I've never tried something like that and I suspect that each receiver would react differently.  It could cause 'churn' where the satellites used change frequently to always choose the best of the limited number you allow.  If not, you could find yourself using a few bad satellites as long as they are in sight.  Tread carefully!

Ed

[jpb]

The Ublox white paper says

If the position is known, the receiver can provide an accurate time solution by tracking only one satellite.

This is very ambiguous. I took it to mean that you got better results from one satellite, but it equally could be a statement of the fact that you can still get time from one just satellite but that it is still better to look at more satellites (excluding low angle ones).

But the next paragraph (in the white paper) refers to the UBX-CFG-NAVX5 message for reducing the number of satellites. If the hardware is clever enough to select best satellites why allow the user to reduce the number? All very confusing.

I guess the best way to resolve it is to experiment - I wish I had more time!

[uncle_bob]

The Ublox white paper says

If the position is known, the receiver can provide an accurate time solution by tracking only one satellite.

This is very ambiguous. I took it to mean that you got better results from one satellite, but it equally could be a statement of the fact that you can still get time from one just satellite but that it is still better to look at more satellites (excluding low angle ones).

But the next paragraph (in the white paper) refers to the UBX-CFG-NAVX5 message for reducing the number of satellites. If the hardware is clever enough to select best satellites why allow the user to reduce the number? All very confusing.

I guess the best way to resolve it is to experiment - I wish I had more time!

Hi

A *normal* GPS will stop giving you time information when it drops below 4 locked on sats. Some units have a feature that helps in this situation. That is what uBlox is talking about when they refer to 1 sat timing. As long as you are doing autonomous timing, the more sats you can see the better. The errors will sort of go down by the square root of the number of sats in view (it's a lot more complex than that...).

The feature to mask out satellites is there so you can take out a "known bad" SV. At various points over the last few decades this might have been useful.

The stuff about elevation mask is there to reduce multi path. For timing (as opposed to survey) you don't need the widely dispersed constellation. You can also reduce ionosphere issues this way.

Most of the above *assumes* you have a very well surveyed location. The fast survey done by the module is pretty crude. If you are really going to use this as a reference, you should do some longer term data collection.

Bob

[DrTune]

A touch off topic but I got my $8 Aliexpress GPS modules today - V.KEL VK2828U7G5LF
Datasheet:
https://github.com/CainZ/V.KEL-GPS/raw/master/VK2828U7G5LF%20Data%20Sheet%2020150902.pdf

Aliexpress e.g.
http://www.aliexpress.com/item/Smart-Electronics-VK2828U7G5LF-GPS-Module-with-Antenna-TTL-1-10Hz-with-FLASH-Flight-Control-Model-Aircraft/32634474566.htm
also on Banggood.com

Rather than turning it on, I took it apart (well, peeled off the can)...
So it's legit - but only kinda - the datasheet claims "Chip UBX-M8030-KT"
But my microscope says otherwise... I see "G7020-KT" on the board.

Other than that I'm a happy camper for $8.50 or so; has onboard flash, battery backup, remote power-off, etc.

The SPI flash chip on there (for autonomous data logging! sweet!) is this 512kbyte part...
http://www.elm-tech.com/en/products/spi-flash-memory/gd25q41/gd25q41.pdf

Also rather handily this chip (like the LEA-6T) has PPS output that goes up to 10mhz...  (if it's like the 6T it actually will go up to 24Mhz)
[edit] - ooh! actually it has TWO timepulse outputs (like the 6-T) which is super-handy. Getting to the second one (4th pin from the right on the top edge in my pic below) will require a steady hand with the iron :-)

When I have a sec I'll head-to-head it against the LEA-6T (which is $30 vs this being $8)...  it's a GPS BATTLE! :-)

[Koen]

The product is correct, the datasheet is wrong. U7 are G7020, U8 are M8030.

Quick note : the M8030 offering is limited but both chips are drop-in-compatible so you can order an U7 module with an M8030 chip.

Anyway VKel modules are a great alternative to uBlox modules distributors and sample shop idiotic prices.

[bitseeker]

That's very interesting indeed. Looks like it uses the same 48MHz internal clock. So, same caveat on choice of frequency output as with the LEA-6T.

http://innovictor.com/pdf/UBX-G7020-Kx_DataSheet_%28GPS%20G7-HW-12001%29_Confidential.pdf

I'm looking forward to the shootout!

[uncle_bob]

A touch off topic but I got my $8 Aliexpress GPS modules today - V.KEL VK2828U7G5LF
Datasheet:
https://github.com/CainZ/V.KEL-GPS/raw/master/VK2828U7G5LF%20Data%20Sheet%2020150902.pdf

Aliexpress e.g.
http://www.aliexpress.com/item/Smart-Electronics-VK2828U7G5LF-GPS-Module-with-Antenna-TTL-1-10Hz-with-FLASH-Flight-Control-Model-Aircraft/32634474566.htm
also on Banggood.com

Rather than turning it on, I took it apart (well, peeled off the can)...
So it's legit - but only kinda - the datasheet claims "Chip UBX-M8030-KT"
But my microscope says otherwise... I see "G7020-KT" on the board.

Hi

One of the things you get with a "real" LEA-6T is very good workmanship on the board. They also use very good components on the board. The boards made by "some guy named Bob" may not have as good parts (TCXO and SAW matter a lot). The one you shared the pictures of has some obvious workmanship issues.

How much is that all worth? Certainly not the delta between the eBay "guy named Bob" prices and the single piece price direct from uBlox ...

Bob

[Koen]

I have both and more, it's completely worth it. And ease of order is another big plus.

[uncle_bob]

I have both and more, it's completely worth it. And ease of order is another big plus.

Hi

I have both as well. I just would not pay the $280 each that they want when you buy and ship one.

Bob

[bitseeker]

So, Bob, you're in favor of the cheap modules? At first it sounded like you were against them. Re-reading your prior post again, I can see how it could read the other way. I agree that $8-30 vs. $280 is quite a premium for buying directly from uBlox.