Nice
You mentioned you have a GPSDO
Can you share some information / pics on the unit you have?
I just ordered (Snail Mail) this device from Amazon ...
Any comments / feedback welcomed (don't be too harsh, as its already been dispatched)
I have two GPSDOs, both homebrewed units based on a basic hardware PLL configuration similar to Gyro's design shown here:
https://www.eevblog.com/forum/projects/my-u-blox-lea-6t-based-gpsdo-(very-scruffy-initial-breadboard-stage)/msg1493431/#msg1493431 In fact my initial solderless breadboard builds were almost exactly that circuit but using an M8N module in place of the LEA 6T and without the cmos RRO buffer amp - the filtered PLL output simply went direct to the VFC pin on my CQE 13 and 10MHz OCXOs.
When it came to my first attempt to build it onto veroboard (stripboard), I'd decided to make full use of the 5v cmos rro opamp to allow me the freedom to use much higher resistor values than the 12 and 100 k resistors Gyro had used to get longer time constants than the 38s TC he'd been running with. I tried 100K and 1M (same cap values) but the loop wouldn't converge onto the target tuning voltage so landed up dropping down to 47 and 220 K to tame the instability in my MK I build. I did eventually get it to work using 100K with 1M then upping to 470K and 5.6M with a 10M bias resistor to allow it to settle within about 25mHz of 10MHz in the unlocked state.
I'd based this MK I version on the "five volt" 13MHz square wave output OCXO which needed an old skool 74193 divide by N counter to divide the doubled up 13MHz (26MHz) down to 2MHz (74HC193s just don't have the jojones to handle such a high input frequency) to feed a 3N501 clock multiplier with this minimum specified input clock frequency to generate the required 10Mz frequency reference.
All of this extra old skool TTL (74193, 7486 and a, probably unnecessary 7427 triple 3 input nor gate thrown into the mix to even up the 2MHz output pulses from the 74193) along with a couple of 3N501 clock multipliers must have added an extra 100mA of very noisy loading on the Mini 360 5v output buck converter I was using to power the M8N gps rx module, logic and the OCXO making the use of veroboard a very poor choice indeed for this application.
Checking the noise and ripple on my well filtered 5v buck converter supply revealed only the TTL power rail noise components, only a very close look at the FFT plots allowed me to spot the almost lost in the noise 1.23MHz switching component on the buck converter's output (and here was I thinking the dc-dc buck converter noise would be an issue
). I'd included a 5 pole butterworth LPF (Fc of 11.5MHz with an extra inductor across the middle shunt cap to give it a boost at 10MHz) but had to resort to inserting a 10MHz xtal series tuned to reduce the noise and jitter I was seeing on the 10MHz output. The xtal did a good job of sanitising the output of all this TTL noise pollution.
With all of these issues in mind, I decided it was time to build a MK II version without all of the bad design and layout choices I'd made with that MK I. I'd decided to build essentially the same design but using one of my 10MHz 12v OCXOs to eliminate all the TTL tomfoolery of the MK I, removing four ICs at a stroke, halving the IC count to just three 74HC ICs and one dual 5v cmos rro opamp.
The veroboard was replaced with a single sided copper clad board with the intention of mounting the ICs 'dead bug' or 'squashed bug' style. However, with such a low IC count, I decided to drill the board to mount the ICs the right way up, chamfering the holes where I didn't need a ground contact and using point to point wiring on the underside of the copper clad board to join everything up.
It was extra work to build it this way but I do like to be able to see the IC markings and it soon proved useful when it came to replacing the blown 74HC14 that had mysteriously failed a few hours after starting my initial testing in allowing me to fit a dip socket to make any future replacements less of a pain.
I'd simply had no idea why it had failed until two burnt out rro opamps and one overcooked buck converter later finally revealed my design error in upgrading the 10uF cap between ground and the non-inverting input on the rro opamp to a 47uF without considering the possibility of this cap dumping excessive energy into the opamp's input clamp diodes on switch off (it had been just dumb luck that the MK I hadn't suffered the same fate).
Simply including a 1k resistor in series with the non-inverting input pin was all it needed to prevent rro opamp number three going the way of its predecessors
I've had no more mystery failures since then (but I did take the precaution of replacing the 1.3A buck converter with its much tougher 3A rated cousin just to avoid any such further embarrassing failures from a less than perfect overload and over temperature 'protection' so amply demonstrated by the original 1.3A module).
At present, I'm running tests on the MK II sans the LPF filter that will ultimately be installed in the final build (useful at the moment in identifying which 'scope trace is which when comparing the two against my stand in for a Rubidium reference (the FY6600 with its 10MHz OCXO).
I've attached images of my DIY GPSDOs The first shows the initial stage of the mark one's build from last August. The other four show various angles of view of the mark two as set up on my bench. The layouts are almost identical although the final layout of most of the components of the mark one had yet to be determined at this early stage.
I don't have any other photos and I don't want to pull it apart just for another photo shoot - sorry! I may follow up in a few weeks time when I reach the stage where the mark two has been completed allowing me to finally retire the mark one and test that "Five Volt" 13MHz OCXO, to destruction if needed, and prove my suspicions as to it being in reality just another 12 volt OCXO.
The mark one could be powered from any DC voltage within the range of +6.8 to +24 volts and I wanted to retain this feature in the mark two despite the use of a 12v OCXO, hence the additional 5 to 12 boost converter module that can be most clearly seen in the last image.
Working from back to front, that yellow blob upstream of the DC jack socket is a 480uH common mode choke to block any common mode noise conducted from a wallwart. Next is the mini 360 3A buck converter configured to output 5.3v which goes towards the front edge via a 6 element LPF. The 5 to 12v boost converter is mounted at right angles, taking its feed directly from the buck converter's output and passing its 12 output to the OCXO via that 3 turn ferrite inductor (20uH) in the corner formed by the two converter boards.
The use of copper clad board should maximise the efficacy of the LPF components and in spite of the extra dc-dc converter to power the OCXO from the 5v buck converter, the whole unit takes about half a watt less than its predecessor, thanks largely due to the elimination of all the TTL jiggery pokery of converting the 13MHz output from the "Five Volt" OCXO into a 10MHz signal (warmed up both OCXOs consume the same power - just at different voltage levels). The mark one consumed 1.8W off its DC power source (7 to 24 volt range) whilst this mark two version only consumes around 1.3W (yet to be more precisely measured) a drop from 2.7W down to just 2W of mains voltage power consumed by the 12v wallwart supply.
The major shortcoming of such a basic design of GPSDO is in my choice of cheap navigation M8N gps receiver modules. I'm awaiting delivery of a NEO M8T timing module from a Chinese seller on Amazon with the expected delivery being the beginning of August. Hopefully, this will reduce the ionospheric effects on timing accuracy to a more acceptable level if the performance achieved with the old Jupiter T gps modules using a similar hardware only PLL is anything to go by. As they say, "Only time will tell."
As for the BG7TBL GPSDO you've ordered, the last I heard about them, afaicr, was that they'd either been programmed as an FLL with a slight offset error of 15mHz or else there had been a bug in the mathematical algorithm used to PLL discipline the OCXO. That was over a year ago when my resolve to build my own gpsdo was beginning to crumble and I'd started searching for a reasonably priced "Readymade". The issue may now be sorted. Others may have more recent information to offer on this subject.
Whenever I'm looking for anything like this, I always do as much internet research as I can before spending any non-trivial amounts of cash. What renewed my resolve to DIY was largely due to the high price compared to what could be achieved by DIY with well chosen components. The problem in this case was the stupidly high pricing of gps timing receiver modules. Even the Pixhawk boys have cottoned onto the value of the ancient LEA T6 modules that had inexplicably been used in the older Pixhawk drone navigation modules, adding a 100 quid premium onto a 20 quid M8N based Pixhawk which can perform better than the LEA 6T in this navigation centric role.
This NEO M8T I'd tracked down on Amazon was a snip at only 41 quid delivered. If it does prove to be exactly as described, I'll order a second one at that price, assuming the seller has any left to sell by then. GPS timing module bargains seem to be as rare as Unicorn droppings these days, hence my "jumping" at this chance of a reasonably priced M8T. Fingers crossed that it's the real deal.
JBG