SDG

[@IanScottJohnston]

"It's now 5AM"........love it!, but don't worry we've all been there.

[@sdgelectronics]

Yep, 0 sleep. Then found audio missing for half the video 😢 Haven't had a no sleep 'till its fixed for many years! Thanks Ian

[@NivagSwerdna]

Loving the PCB colour! I don't have a reflow oven yet but that hot plate looks perfect for my low volume tinkering.

[@sdgelectronics]

It works surprisingly well. I think the oven was probably a waste of money as it stands, but I'll cover it in a video

[@wizrom3046]

For f=10MHz and gain=1.57 an opamp was a poor choice. You're already using capacitive coupling so any decent transistor and a few resistors would do that amp job. Also, don't try to discipline to the 1Hz PPS signal. Reconfig the UBlox to send out a good midrange freq like 1kHz to 10kHz, then do phase locked loop to that in your micro or fpga. Better still if you can use a direct multiple (integer divisor) of the UBlox main freq which is also direct multiple of your PLL processor. Also be aware that the UBlox time accuracy will go out by approx 50 PPB every time a satellite drifts in/out of use, which happens every couple of minutes depending on weather so you need fairly long term digital filtering on that UBlox output. Interesting project! ⏰👍

[@Bllinker]

My first instinct was also to wonder if a discrete transistor solution would work here, but look at the available voltages, only a single 3V3 supply. I doubt you could get clean 3Vpp from a single stage amplifier with that, not to mention the 50Ω output impedance requirement. If I were in Steve's shoes I'd probably just try to work with the rails a little, maybe get a higher voltage supply (that 12V in looks interesting) or mod in a negative voltage converter just to get this board working and think about a permanent solution later, but that's just my two cents.

[@diggleboy]

This is a great project to build your own GPSDO. Voltage rails do matter with respect to opamps because of the high slew rates required to propagate signals fast enough for very high frequencies. It makes supplying power challenging because you also need a +- rail to rail voltage on an otherwise mostly 5vdc digital board. Trade offs in design. Looking forward to seeing how this project turns out.

[@UpcycleElectronics]

I like how this is coming together.

[@cncdavenz]

Looks like this is going to be an interesting project. It would be nice to have a LCD display showing exact time and date and other information relating to performance. you have a nice collection of multi meters, it must be hard to choose which one to use. Cheers Dave. Keep Safe !

[@sdgelectronics]

Thanks Dave. I'm thinking an LED readout of the time and maybe the date, but then another screen capable of all the stats etc. Probably expansion capabilities on the main board to add the possibility of using it as an NTP time server

[@WAEviltechie]

If you're going for broke, you could also stick some extra output types on there. I work in broadcast TV, so I'm thinking along the lines of timecode, blackburst/tri-level sync, or word clock.

[@jorditribo94]

You would not get an opamp that can drive the output to 0V in a single supply configuration. There are negative bias generator ICs that can generate a -0.2 V supply for the opamp. As an alternative you can get a negative voltage from an inverting charge pump.

[@George-gz5zm]

Great project, might build this one if I get chance. Op-amps are a real pain when they go wrong, especially the high speed ones. Avoid that input capacitance.

[@allthegearnoidea6752]

The reflow looked very good. What vacuum pick up tool did you use for placing the SMD components. Interesting i wondered who the other person in the world used proteus. Thanks regards Chris

[@sdgelectronics]

I'm tempted to move on from Proteus, but I know how it works and all the quirks. The pick up tool was this contraption: kzbin.info/www/bejne/Z3fReKR-lLyegpY

[@kissingfrogs]

It is 5am but I think you wanted 32768 not 32678. Curious if the stencil also had issues where the pads were malformed. Great series of vids BTW.

[@sdgelectronics]

Yes I saw that after I uploaded. Oops. Oddly, no the stencil was how it should have been so it's a bit puzzling.

[@craigs5212]

Just curious why are you locking the oxo to the GPS at 1pps vs 10 Mhz? I built a couple of these early last year to compare GPS timing with a Symmetricom SA22.C Rubidium oscillator. The design just phase locks the 10Hhz OXO to a u-blox 7 GPS receiver programmed for a 10Mhz output, using an HCT4046 phase detector & very low frequency cutoff loop filter to remove the GPS jitter. The interesting was to watch the Rb oscillator and GPSDSO on the scope. Once things stabilized and the Rb oscillator locked in. The two signals were rock solid for for 10 to 20 minute periods then you would see a period of several minutes where they would slowly phase drift apart and become stable again. I don't think I ever saw phase drift apart by more than one cycle period at 10 Mhz. They would drift one way for a while then drift back a while later then be stable for a while. Without a second Rb oscillator I was not able to tell which one was doing the drifting. Could be lots of things, the Rb oscillator, the loop filter, the oxo heater the power supply or the GPS signal propagation.

[@nicolas8414]

The output of the u-blox 7 is generated from an internal 48 MHz clock, therefore 10MHz will give you a lot of jitter. You want a frequency like 8 MHz, which can be created by an integer divider.

[@craigs5212]

@@nicolas8414 You are correct, there is significant phase jitter on the 10 Mhz due to the DSS and the 48Mhz txco. The sub hertz loop filter should make this not much of an issue as the phase modulation should be way outside the loop filter cutoff. Went back and looked my implementation I actually use 100kHz as my phase detector reference to minimize the issue while still maintaining quick lockup. Didn't have access to a spectrum analyzer at the time, will have to look at that again. Also I wasn't able to find the timing version of the u-blox 7 so there could be other things going on but most of these relate to timing and not the frequency.

[@nicolas8414]

@@craigs5212 I agree on this, with a long pll time constant the phase noise of the timepulse shouldn't be an issue. The time constant should be carefully selected with attention to the drift of the oscillator. With a very good OCXO this can be in the range of hours.

[@JxH]

Nice project.

[@YanickT]

Drift isn't very important in gpsdo. uC will correct the error over time and very negligible. I like the way you have include the ublox onboard without using an external gps board. It should be nice to see 2 waveform of different gpsdo at same speed (matching). And at the end, if you frequency counter have not a reference 10mhz. Your device (gpsdo) will be more accurate of your counter :) Thank for sharing

[@sdgelectronics]

If I can get my other unit working, I'll be able to compare 3 waveforms. Should be interesting.

[@rjordans]

Nice project! What kind of FPGA are you using? I have something similar with a LCMXO2-1200HC-4SG32C, nice to have something small to fit on the board with still quite some processing power. The ice40 series has some similar ones with small footprints as well.

[@sdgelectronics]

At the moment it's the LCMXO2-7000HC-4TG144C . Probably excessive on the pin count, but I don't know at this stage how many LUTs I need. I may swap out for a smaller device.

[@vincenzomastroianni9532]

dear, a great project, is possible have a file gerber for pcb ? thanks.

[@danw1million]

Hi just wondering if you continued with this project or if you have any more information on it ?

[@faidularcs]

Nice job 👏 👏 👏 👏 👏

[@largepimping]

I probably missed the answer from an earlier video, but... if I understood everything correctly, you're "only" using the 1 pps from the GPS module to discipline the oscillator. The ublox module can be reconfigured to output 10 mhz directly (as I'm sure you're WELL aware...). What's the advantage of not just using 10 mhz from the GPS module? I'm also curious about what you're doing differently vs. the (apparently) very similar modules widely available on ebay - was there any motivation beyond purely educational purposes? IOW, when you look at the circuit design of those products, do you see things that could/should be done differently/etc.? Lastly, thank you for showing that even someone that knows what the hell they're doing (you) can totally pick a component that is (in hindsight) so badly wrong!! It gives me hope because it can happen to the best of us!

[@oldblokeh]

The Ublox timepulse output is divided down from the internal reference clock which is 48MHz. 10MHz cannot be derived using an integer divisor, therefore would have very bad phase noise/jitter.

[@sdgelectronics]

Yes, as the others have mentioned, even the specific timing modules from ublox have "quite poor" jitter. They're actually fine for general hobby lab purposes but it's partly an acedemic project to tinker with and refine. Regarding the ebay units from China - I have one but no real specs so that should be interesting to test against.

[@YanickT]

Hi, this is the jitter of a 8m ublox at 1 mhz. Imagine the waveform at 10mhz. It's veru ugly. Look here step 2: www.instructables.com/id/GPSDO-YT-10-Mhz-Lcd-2x16-With-LED/

[@on8rth]

Why The OPAMP? there are better solution, like putting a few CMOS buffers in parallel. Also it is better for jitter to use square waves as sinewaves.

[@sdgelectronics]

I will likely have both sine and square outputs on the final design. Some equipment asks for sine output and it also reduces the high frequency content.

[@catalinalb1722]

Hi can you share at least the Gerber files and the BOM list? I would like to give it a try. How much did it cost in the end? Thanks

[@JxH]

@13m05s: "32678"

[@youcef_benchalel]

Hello , can you share your design schema please ? , i would like to make it for my radio 10mhz reference , thank you in advance

[@bsodmike]

Is there a part 1 for this?

[@MrHD71]

why did you make this PCB on pcbway?

[@gamerpaddy]

sponsored i guess he switches between jlc and pcbway from time to time depending on board size and additions either one is cheaper sometimes

[@sdgelectronics]

This project I'm doing on PCBWay so everyone can see the options and quality of suppliers

[@YanickT]

Always had good pcb from pcbway

[@jmfggbarros]

I have built www.eevblog.com/forum/projects/lars-diy-gpsdo-with-arduino-and-1ns-resolution-tic/ some time ago and the performance is great. It interfaces with lady heather which is a plus. The code is not that well written so it is not easy to follow but may give you some ideas.

[@sdgelectronics]

I'll take a look at that one, thanks.

[@johnshea135]

Pick up tool??

[@sdgelectronics]

kzbin.info/www/bejne/Z3fReKR-lLyegpY

[@andymouse]

What temp did you use please ?

[@sdgelectronics]

220C

[@andymouse]

@@sdgelectronics Ta, just got a plate myself and gonna try it out so that's a good place to start, I have played around and think this is a good way for a hobbyist to get into SMD, perhaps a video on this technique would be nice ?...cheers.

[@AlessioSangalli]

@@andymouse as a hobbyist that tries to avoid through hole as much as possible, but I don't really do big volumes, why wouldn't a regular solder iron be enough? Yeah it won't look as perfect but I've still done relatively complex boards. I tend to etch the first iterations of the boards at home and then order the "final" ones out of pcb manufacturers

[@andymouse]

@@AlessioSangalli Reguar iron is fine , I use one, this is just one of many options that are fun to try

[@thekaduu]

I know the board cost is not that bad, but you could have fit the same design onto a board that’s half the size of the one you have... Why make it large? Does this need to fit into a big enclosure? I don’t get it...

[@sdgelectronics]

This is a prototype board - it's purposely large in case I need to bodge parts on it. The next revision will be refined.

[@jaguarulgamer6802]

First! I like it

[@cmuller1441]

Sorry you're not... 😜

[@cmuller1441]

Nice I'm first!