If you think you can easily solve the noise and ripple issue with cheap AM1117 LDO ICs, think again! For this trick to work, you'll need special LDOs designed to reject noise and ripple into the low MHz region - your standard off the shelf AM1117s just ain't going to cut the mustard for this job.
There is a very simple solution for that: use an RC filter in front of the LDO. The resistor will also help to dissipate some of the energy lost in the LDO.
Hi nctnico,
Apologies for the delay in responding but that recently completed basic DIY GPSDO I mentioned had been holding up a frequency injection locking add on project for my already very much modified FY6600 since May of last year and I had resumed this project to the exclusion of almost all else once I finally, and at long last, had a working GPSDO to play with.
For anyone interested in locking their already upgraded to an OCXO FY6600 or 6800 generator to an external 10MHz reference, you can check out my Topic thread: Injection locking the 10Mhz OCXO to external reference (upgrading a FY6600) posting here:
https://www.eevblog.com/forum/chat/injection-locking-the-10mhz-ocxo-to-external-reference-(upgrading-a-fy6600)/msg2418711/#msg2418711Returning to the matter in hand, using an RC filter is not a bad idea, especially when dealing with critical low current demand areas of a system that need to be kept isolated from the switching noise of a bulk power supply. A major plus point for the use of a resistor in place of an inductor is the avoidance of self resonance in a filter that's expected to deal with switching noise components that could extend to 100MHz or more.
When I was deciding how to power my little GPSDO, I looked into the pros and cons of using LDOs with a buck converter powered from cheap commodity smpsu wallwarts in the 9 to 19 volts dc range and ultimately decided to forego the (in my case dubious[1]) benefit such an arrangement might bring.
Since everything runs off a 5vdc supply (actually 5.16v), including the 13MHz OCXO, I'd initially considered using a 7805 (or modern day equivalent) to handle the voltage conversion from a 9 or 12 volt wallwart, but swiftly decided against this on the grounds that it would complicate the mechanical construction with the need to bolt its heatsink tab to the aluminium case as well as nearly double up the energy consumed with a 9v wallwart and more than double it in the case of a 12v supply.
All of this energy input needs to be dissipated within the confines of a small 50 by 100 by 112mm extruded aluminium box that I'd prefer not to compromise its RF screening properties any more than was absolutely necessary by riddling it with ventilation slots.
Also, I was concerned about the indifferent voltage regulation of the AMS1117-5 used on the solderless breadboard psu board (YwRobot MB102) compared to the boost converter used in the Poundland 1200mA powerbanks with which I'd been testing the CQE 13MHz OCXO to determine whether it was a 12 or 5 volt unit I'd acquired for cheap (I couldn't track down a datasheet for this particular model but the search had lead me to a cheap supply of 12v 10MHz CQE OCXOs
)
I decided in the end to use the buck converter on its own to directly create the 5 volt supply and do away with any further complication. A bit of LC filtering reduced the 25mv p-p noise and ripple to less than 10mv with additional filtering for the OCXO and the GPS module making the need for any LDOs redundant in this case. Since the modestly insulated OCXO only consumes around 850mW[2] or so once warmed up, the total input power in the 7 to 20 volt range remains within 20mW or so of 1.8W (just under 3W mains voltage input to the 9 or 12 volt wallwarts I've used to power it up). Power supply noise and ripple doesn't appear to be an issue in this case so I'm quite pleased with the result.
[NOTES]
[1] I figured the extra heat penalty of allowing another 1.5 volts margin for a pair of LDOs (one for the OCXO and GPS module - 180mA warmed up, the other to serve the logic gates involved in converting the 13MHz to 10MHz and phase locking at 100KHz, also another 170 to 180mA) just wasn't worth the trouble, especially when there was every chance of poorer voltage regulation with load and temperature compared to that of the buck converter I'd chosen.
I might have been able to reduce the load on the logic supply by some 30 to 50mA but the 74HC193 ICs just just don't seem to have the cojones of the old skool 74193 I was obliged to use to divide the 26MHz (clock doubled 13MHz OCXO output) to get the minimum frequency needed by the 2nd 3N502 clock multiplier I used to multiply the resulting 2MHz back up to the 10MHz sine wave reference I needed. Relying on additional LC filtering just seemed the better solution in this case.
[2] This 13MHz square wave output OCXO draws a warm up current, limited to 280mA which ultimately settles down to 190mA @4.82v after some seven minutes of warm up time which is reduced to a less agonising 5 minutes delay on a 5.16v supply. Wrapping it in a sponge rubber 'overcoat' knocks some 20 to 30mA off the at temperature operating demand.
What had made me decide that it was a 5 volt version was the fact that for each volt increase on the Vcc pin up to a maximum test limit of 7.5v, results in an extra volt on the p-p output voltage, a trend that if continued to a 12v supply would result in (the never specified for any OCXO I'd ever seen datasheets for) of 11v p-p output.
All the signs were that the oscillator was being powered directly off the incoming Vcc rather than via an LDO of any sort and although cmos can allow logic levels of 10 volts or more, this seemed a rather unlikely scenario so I decided not to push my luck any further than I already had and assume it to be a 5 volt unit (at 5 volts, it had proved stable enough without my taking any further risk in burning it out on a full 12 volt supply).
However, the 12 volt 10MHz examples I have are, rather curiously, also limited to a maximum heater current of 280mA which of course drops to a lower (uninsulated, at temperature) current of circa 90mA which leads me to reconsider my decision to treat the 13MHz example as a 5 volt unit (especially since I'd been able to get the 12v units to output from a 5v supply, albeit with an even longer delay to start actually generating a recognisable sine wave output than the 13MHz unit had taken on a 4.82v supply). I may experiment further at a later date if I ever find any evidence that any OCXOs with 10v p-p cmos logic level outputs were even manufactured but, for now, I think I'll leave this sleeping dog to lay undisturbed lest it bite me in the backside for my trouble.
[EDIT 2020-08-12]
Just to update the record, today, after finally commissioning my MK II GPSDO using a 12v 10MHz CQE OCXO powered from a 97% efficient 5v to 12v boost converter fed from the 5.34v output of a 3A rated mini 360 buck converter (allowing me to retain the option of 6.8 to 24 volt DC power sources of its MK I predecessor), I opened the MK I up to isolate my "Five Volt" CQE 13MHz OCXO to run some voltage tests which I hadn't dared to run for fear of smoking it with a 12 volt 'over-volting' event.
It turns out that this is also a "Twelve Volt" part that just happened to be sufficiently obliging to produce a stable output (after a protracted 7 or 8 minute warm up delay) from as little as 4.82 volts (I was actually using a 5.17v supply rail in the MK I).
I'd been harbouring a sneaking suspicion for the past few months that my "5V" OCXO was actually a "12V" one in disguise - the same 280mA peak heater current limit during warm up and the 'unregulated two thirds of the input voltage' on its Vref pin (which finally stabilised at 5.114v once the input supply rail started going north of the 10.8v mark). Just for good measure, I ran the supply rail up to 14 volts using a 0 to 32v 10A bench supply I hadn't possessed at the time I'd purchased my 'precious' one and only 13MHz OCXO for which I hadn't been able to find any voltage data for.
The only good thing about strip board construction with a 13MHz OCXO generating a 10MHz reference is that the noise from even a noisy switching buck converter pales into insignificance, drowned out as it is under a storm of TTL rail supply noise generated by the times two - divide by 13 - times five stages required to arrive at 10MHz using 3N503 clock multiplier chips with a 2MHz minimum input frequency requirement.
The MK II is essentially the same design but sans any 13 to 10 MHz TTL jiggery pokery since it conventionally uses a 10MHz OCXO and is built onto a single sided copper clad board to keep ground noise to a minimum. In spite of the use of an extra dc-dc converter to generate the 12v rail from the 5.34v rail, the MK II actually consumes some 400mW less power (1.4W versus the 1.8W of the MK I) as a result of eliminating all of the 13 to 10MHz TTL conversion circuitry.
I was going to restore the MK I but having to undo most of the connections to the 13MHz OCXO to keep the rest of the circuit isolated from the 12v test voltages, I'm now going to strip it down for the not so cheap and cheerful ICs I'd had to use in this design (3N503 clock multipliers ain't cheap!
) and the 13MHz OCXO will be kept as a "Keepsake" memento of my very first experiences with OCXOs and diy GPSDOs.
After the experience gained in proving that a 13MHz OCXO could be used in a 10MHz GPSDO, I don't see any further use for it. By the time I'd come to the actual construction of my MK I, I had already gotten hold of seven of the 10MHz 12v CQE OCXOs but having developed a working circuit to use that "Five Volt 13MHz" OCXO, I'd decided to go ahead as planned and see it through to the bitter end, especially as I'd had it in the back of my mind that it would save the (in hindsight over-rated) "complication" of safely mixing 5 and 12 volt supply rails on the one PCB.
I'm now awaiting delivery of an EFRATOM LPRO-101 and any future DIY GPSDO projects will be microcontroller based designs such as the Lars one or something along those lines, hence my stripping the MK I down for spares. It had served its purpose well as a test bed to reveal the limitations of a single frequency navigation class of GPS receiver module driving a hardware PLL controlled OCXO - the MK II is using a u-blox M8T which is better but still far from perfect, hence the LPRO-101 purchase to see just how far from perfect my MK II actually is.
And, to think that all of this was sparked off by my desire to improve the frequency stability and accuracy of my FY6600-60M AWG.
JBG