Ha I just scored three of those LEA-6T modules off ebay ($30 a pop - bargain!), because I'm interested in a very accurate GPS clock for (ironically) playing with radio multilateration as a fun pet project..
I also have some of those V.KEL modules in the mail ($8 on Aliexpress!) which apparently contain a ublox 8 - UBX-M8030-KT chip, ~30ns timing, but only up to 1khz time output. We shall see how they do.
I was interested in these LEA-6T modules b/c it has dual time outputs allegedly configurable up to 10Mhz (but in practice the u-center software will let you go up to 24mhz, which works fine!)
The appnote on LEA-6T timing is a good read
https://www.u-blox.com/sites/default/files/products/documents/Timing_AppNote_%28GPS.G6-X-11007%29.pdfThey mention the internal osc runs at 48mhz so you pick an integer factor of this to get a relatively jitter-free output (e.g.6mhz, 8mhz, 12 etc)
Because I have several of these units (and no proper time source other than that) I figured I'd play one off against the other - I hooked two of them up, set them to 3Mhz output (48mhz/16) and put them on two scope inputs. As I write this now they've both oddly decided they can't get a satellite lock but earlier today they did lock fine, and there was roughly 30ns constant jitter between the two clocks, and no drift at all. The phase jumped around a bit as the units improved their satellite lock, the best I saw was in the region of 20-40ns offset between the edges, the worst was a few hundred ns. (this was measured with cables of exactly the same length from each unit to the scope).
The ~30ns constant jitter figures because their internal 48Mhz clock has a granularity of 20.8ns, so one naturally expects random jitter of up to 41.6ns when measuring between two of them. Eyeballing it, it looked slightly little better than that, but close enough.
I thought that was pretty good - these two GPSes were right next to each other - if there were e.g. 20miles apart they might have more variation in their outputs, but for $30 a pop it doesn't suck at all. Next stop is to use it as a ref clock for a PLL, I see there's a cheap board on ebay with a ADF4351 that can run off a 10mhz reference and generates 35Mhz - 4Ghz (but only a sine output >=2Ghz else it's a square, meh). Obviously I could get a real GPSDO or whatever but I'm just playing around here.
I'll shoot a bit of video of the two GPS outputs next to each other on a scope for you.. once I figure out why now they suddenly won't get a sat fix (they're in same position near a window that they were earlier.. weird).. when they don't have a fix, the two outputs drift from each other quite rapidly as you'd expect; right now I'm seeing about 200ns/sec of continuous drift between the two time pulse edges.
EDIT: I'm pretty sure the reason they couldn't get a fix just now is because I had them both generating a 6Mhz clock temporarily driving long, unshielded wires... As soon as I turned the timepulse down to 10hz they suddenly both started seeing a ton of satellites and got a nice fix. The rise/fall time on the clock is ~2v in 10ns so at 6Mhz that'd be quite enough mush radiated to screw them up (esp as they only have a moderate view out of a side window). Note - use shielded cable with fast clocks to avoid screwing up the incredibly weak GPS signals!
Anyway, pretty sweet modules for $30. Planning to use a 2mhz output from the LEA-6T into the OSC_IN input of an STM32F103, which can PLL it up to 26mhz and output it again to run a CC1101 radio, the second internal PLL can clock the CPU core at 78Mhz, and if I'm not mistaken the whole thing should then be nicely synchronous together (gps+radio+cpu) meaning I can hopefully transmit radio packets with the only jitter being the ~20ns from the GPS module itself.
Coming up next, some $8 VK2828U7G5LF GPS modules (which according to the spec at
https://github.com/CainZ/V.KEL-GPS/blob/master/VK2828U7G5LF%20Data%20Sheet%2020150902.pdf look pretty well suited as a regular GPS unit for a different project I'm pondering)...