Anyone familiar with LEA-M8F?

[mojoe]

I just got one of these modules on a board. It seems to work with u-center and Lady Heather. I'm still reading through the documents that I downloaded, but my understanding is that this module is essentially a mini-GPSDO, with a reference output of 30.72 MHz that is frequency locked to GPS. The internal TCXO is also supposed to be locked to GPS, eliminating sawtooth jitter. There was a previous GPSDO that locked it's internal oscillator to GPS, but I forget which one it was.

BTW, that oddball frequency is used for certain cellular systems. It has something to do with audio modulation. In fact, you can divide it down nicely for 96 KHz, 32-bit, 2-channel audio. I guess if you're an audiofool, you could GPS lock your stereo 

In u-center, it seems to work very much like an M8T. One curious thing, when I set the TP2 to various frequencies, the ones that look stable, without lots of jitter are the same frequencies that are integer divisors of 24 MHz, not 30.72 MHz. The maximum TP2 frequency I can set is 24 MHz. So, is the M8F using the same TCXO as the M8T, and somehow still locking that oddball 30.72 MHz to GPS? I'm confused.

In LH, the M8F looks pretty much like an M8T. However, the device name is shown as "Ublox timing", and not a specific module. The sawtooth error shows as exactly 0.000000 nS. The accuracy is exactly 6.000000 ns (it started out as something over 1000, then dropped to 9 ns, then 6 ns). It just went up to 8 ns, then dropped to 7 ns (I'm indoors, so perhaps that's why it's jumping). Since LH reports sawtooth error from other u-blox receivers, can I assume that this M8F actually has no sawtooth error?

One anomaly regarding the reference output: it's supposed to be 30.72 MHz, but I'm reading 32-34 MHz, and it's not stable.

Does anyone have experience with the M8F, and can provide some insight? I'll continue to read the documentation.

The M8F is also supposed to allow disciplining an external oscillator, with the addition of an external I2C DAC. This is the reason I got this module. If it works, this should make a dandy GPSDO with an external 10 MHz OCXO.

[mojoe]

I solved the reference output anomaly. It doesn't like 50 Ohms. When I put a x1 scope probe on it, I got 30.72 MHz. The waveform isn't exactly a sine wave. It's sort of a cross between a sine and a triangular wave.

The TP2 output worked just fine, feeding 50 Ohms. Pretty decent looking square waves.

[SCSKITS]

Have a few of these parts and tried to use the DAC with success on one of the M8F modules and no results on the other.
I am concerned about counterfeits, but the EBAY part worked and the part I received from a 3rd party as new did not.
I did not find the DAC controlled OCXO performing as well as I expected, but I did not dig into it any further at this time.
The board that worked has an M8F with the DAC. The OCXO was on a second board.

The board that did not work had both the M8F, DAC, and the OCXO on the same board.

I have to find time to look into the two boards some more.
I attached a short plot of the M8F 30.72MHz output.
I will try to get a longer plot in the next few days..

ed

ed

[SCSKITS]

Latest plot of the LEA-M8F 30.72MHz output.

ed

[ch_scr]

I also got drawn in by the "Sirens call" of these modules. See the kind of "eval board" I've made to mess around with it (DAC still missing in that picture). Came so far to hook the extIN1 to a 10MHz OCXO and the Vtune DAC through it's LPF directly in parallel onto the OCXO's Vtune potentiometer wiper. Some poking around in the settings, but eventually it "pulled it". Some observation: it only started disciplining once I've manually set the "is calibrated" flag. I've also let it do the cal cycle and watched the vtune sweep. It then correctly set the "Hz pro Volt" sensitivity very low (as the pot in parallel keeps it close to center), it took about 15 minutes IIRC. Haven't yet messed with switching the module over properly to the OCXO as it's "disciplining source" or telling it to survey in and then setting it to stationary. Also have yet to used it to measure another OCXO on it's other input. Pretty good fun so far though!

[SCSKITS]

I got the bug as well and made a couple boards for testing. Used JLCPCB, only takes about a week to 10 days.
The smaller board is just the M8F with the DAC, no OCXO, the DAC has since been removed and placed on the larger board.

The small board worked with an off board OCXO but I did not have time to properly characterize it.
The large board does not control the DAC properly and seems to be different in configuration than the M8F on the small board.
Again ran out of time to investigate the issues. Wondering if the M8F is a counterfeit.

ed

[ch_scr]

After a long while, I came back to this, mainly to try out my measuring system.
I measured two OCXO's free running (one was still warming up, hence the drift) and then disciplined the lower noise one with the Ublox M8F.
Two measurements again, one in "normal" mode and one in "stationary" mode after a quick survey-in.
I'm using a "normal" gps puck antenna on the outside windowsill, likely performance would be better if half the sky wasn't blocked by the building.
But the survey-in seems to have helped a bit nonetheless. Now I sure hope (and think) my scaling factors are correct... but take this with a grain of salt 

[SCSKITS]

I have attached plots from a NEO-M9N with a PLL controlling a 5V Piezo Technik OCXO P/N PTOC32227.
The NEO-M9N is not a timing module, it is just one of the regular models and I did not do a survey in prior to taking data.

I suspect that the result could be a bit better using the LEA-M8F, but I am near the limit of my ability to properly test with the NEO-M9N.
I posted a frequency difference plot earlier of the 30.72MHz from the LEA-M8F which was about +/-7E-10.
That is much better than the NEO-M9N output (at 8MHz). I use 10KHz output on the PLL board.

I am not sure that the extra cost of the LEA-M8F ($105 at Mouser) vs the NEO-M9N ($27 at Mouser) is worth it if an external PLL is used.

ed

[ch_scr]

I have attached plots from a NEO-M9N with a PLL controlling a 5V Piezo Technik OCXO P/N PTOC32227.
The NEO-M9N is not a timing module, it is just one of the regular models and I did not do a survey in prior to taking data.

I suspect that the result could be a bit better using the LEA-M8F, but I am near the limit of my ability to properly test with the NEO-M9N.
I posted a frequency difference plot earlier of the 30.72MHz from the LEA-M8F which was about +/-7E-10.
That is much better than the NEO-M9N output (at 8MHz). I use 10KHz output on the PLL board.

I am not sure that the extra cost of the LEA-M8F ($105 at Mouser) vs the NEO-M9N ($27 at Mouser) is worth it if an external PLL is used.

ed

Hi ed,
not sure the latest plot got uploaded? I guess the appeal of the M8F is that it's almost a completely integrated solution, only add DAC and OCXO.
The external TCXO/OCXO doesn't have to be 10MHz, either. Regarding the price, there are reasonably priced desoldered modules on Aliexpress, for now.
I suspect when they run dry so will the hobbyist interest in these specific modules...

[SCSKITS]

Here are the plots

[Solder_Junkie]

I have some experience of using an M8T vs an M6N, the frequency stability when using these in a GPSDO referenced to a rubidium standard is that using the M8T results in half the short term frequency wandering, compared to using an M6N with the same loop time constants. You can see the kind of stability I am able to obtain at: https://www.qsl.net/g4aon/gpsdo/

Beware of "marking your homework" with Timelab, the results can look impressive but I am not sure it is really the best way to measure stability. The better method is to compare the frequency of your oscillator with a high stability reference, such as a rubidium. Using a phase frequency analyser, I can obtain repeatable results day after day.

Many seem obsessed with the long term accuracy measured over hours in some cases... the more relevant stability is short term, in the case of being used for frequency control jitter is a better measure, or perhaps measurements taken over a period of 10 seconds to 5 minutes if used as means of calibrating test equipment.

Incidentally, the Ublox part number suffix gives a clue as to the intended use, F = frequency, N = navigation and T = timing, although all will do basically a similar job.

There are fake modules on eBay, I have a couple of fakes that were cheap and are good enough for playing around with. A much better source of assembled modules is the GNSS Store: https://gnss.store/

SJ

[SCSKITS]

SJ:

The plots were made against an HP Z3801 GPSDO.
I also have a couple Rubidium oscillators, but have to permanently install them so they can run for a while before use.
For my purposes, the PLL type GPSDO is adequate.

What do you use for a Phase/frequency analyzer?

ed

[Solder_Junkie]

I use a TinyPFA phase frequency analyser, you can also load alternative firmware into a NanoVNA H4 and use it as a PFA (and swap back again to VNA firmware).

I have three GPSDOs, 2 home made (Lars Walenius and G3RUH type), and a Leo Bodnar GPSDO (uses a TCXO and navigation grade GNSS module). My rubidium is an Efratom FRS-C that is ex cellular use, it was a gift from an electronic surplus dealer.

Of my four "standards", the poorest stability one is the Leo Bodnar, but oddly enough it is the one that I use the most as it is almost instantly stable within a few seconds from switch-on and can output any frequency up to 800 MHz. It can also be powered from USB, so consumes very little power.

SJ

[SCSKITS]

SJ:

I have a TinyPFA coming in later this week. Looking forward to seeing how it performs.
My 3Ghz spectrum analyzer's bandwidth only goes to 10Hz so not low enough for phase noise measurements.
I have access to an MXA analyzer but have not had time to try it.
If the MXA has the phase noise measurement I will try to compare to the TinyPFA. I am sure this has already been done..

I had a Leo Bodnar two output GPSDO for a few weeks last year for a project, it is a nice unit. Unfortunately had to return it when done.
It uses a MAX-M8Q with a SKYWORKS SI5328C jitter correction chip.
The Si5328C has loop filters from 0.05Hz to 6Hz so it can clean up the UBLOX output.
I have been experimenting with the SI5328C to see if I can get similar results. but not quite there yet.

ed

[Solder_Junkie]

I had a Leo Bodnar two output GPSDO for a few weeks last year for a project, it is a nice unit. Unfortunately had to return it when done.
It uses a MAX-M8Q with a SKYWORKS SI5328C jitter correction chip.
The Si5328C has loop filters from 0.05Hz to 6Hz so it can clean up the UBLOX output.
I have been experimenting with the SI5328C to see if I can get similar results. but not quite there yet.

Ed, I have the dual BNC Leo Bodnar unit. The level of frequency jitter I measured is around +/- 4 parts in 10^10, which doesn't sound much, but is around ten times worse than my home made Lars Walenius unit. Both measured at 10 MHz. I tried powering the LB from both a wall wart phone charger (USB) and from a linear power supply to the DC input socket, both produced the same level of jitter.

When you run out of things to test with the PFA, you can have endless fun measuring inductors, capacitors and filters with it configured as a VNA 

SJ