The reason for that high looking ripple spec by Oscilloquartz might simply be on account of either excellent LPF filtering or very high quality LDO regulators being used to condition incoming raw smpsu power. Do you have a circuit diagram for your unit to check this out?
It's not unusual to use relatively high ripple supplies in audio amplifiers and similar kit where the high ripple content is less important in the final power output stages and is essentially ignored by the final stage (at least when it's not driven into clipping) and all the lower level pre-amp/pre-processing and driver stages will incorporate the additional PSU rail filtering using lower power handling rated high quality filtering circuits (including low noise analogue voltage regulators - not all analogue voltage regulators are inevitably noise free). IOW, it does rather depend on whether the design properly deals with the issue of high ripple content of "raw dc power" or not.
I'm one of the later contributors to that long running thread on the Feeltech FY6600 function generator where, amongst many other issues, this question of 'cheap and nasty' smpsu power has been raised (mostly concerning the issue of half live mains 'leakage' touch voltage (common to all such class II smpsus btw) and the potential ESD risk it can pose to any devices being tested.
In this case, the simple and effective solution was to replace the two pole C8 mains socket with a three pole C6 or C14 earthed connector and tie the ground rail via a 10K 'drain resistor' to the PE connection to attenuate the undesired 100vac 'leakage' down to half a volt without introducing the evils of ground loop induced interference (a precaution Feeltech overlooked in their own kack handed upgrade to an earthed C14 mains socket on the revamped FY6800 model).
Of course, this rather tarnished the perceived quality of the small smpsu board which was often replaced with bulkier analogue based DIY concoctions or higher quality 'open frame' smpsus with ripple figures barely better than the original, often needing to be overspecced on the +12 or +15 volt rail simply to get a high enough rating on the negative 12 or 15 volt rail. In most cases, this was a mod done before the issue of ineffective ventilation had been dealt with, thus adding an even larger thermal burden onto the already heat stressed components (although it has to be said that most, if not all, PSU upgraders did add a fan to the box simply because it was so conveniently opened up for the PSU work being done).
I spent quite a lot of my time 'improving' the existing PSU board by replacing the cheap and inappropriately chosen rectifier diodes used on the LV side along with a capacitor or two of larger value (but only so far since the HV switching IC sees too much of this 'good thing' as an overload and goes into a fugue when you overdo things in the smoothing department).
The reason for my not replacing the little PSU board simply being that even the higher quality open frame smpsu units are
shit. Staggeringly, I couldn't find a single two rail +/- 12 or 15 volt (let alone a three rail +5 with +/-12 or 15 volt) smpsu that didn't suffer this weird current rating imbalance between the two 12 or 15 volt rails.
In the end, after hitting upon the addition of a single buck turn to the 5v rail's winding to redress the voltage imbalance between the 5v and the 12 volt rails which neatly solved the issue of barely sufficient voltage to meet the requirement to produce 20Vpp into 50 ohm loads), I came to appreciate the elegance (although not the bean counteritus it had been subjected to) of its design in that, with no complication, the +/- 12v rails were inherently balanced in their output current ratings. Who knew?
Actually, better put, who knew that such a cheap 'n' nasty smpsu board could so resoundingly "outclass' all of these idiotically rated +/-12 or 15 volt open framed smpsus in being able to supply equal current outputs to satisfy the very common dual rail opamp loading requirement? Since the +5v is only used by the less fussy digital circuitry in the FY6600 and FY6800 function generators, all that is required to make it worthy of its task is another 1.5v lift on the +/-12v rails, with a couple of 12v LDO regulators and a nice screened 'open frame' box to house the existing PSU board into.
As it happens, it already does a pretty fair job with just the basic enhancements I've already applied. Even so, this is still something I'd like to sort out 'properly' in the future should I ever find a non-idiotically rated three rail smpsu to replace it, a task that appears to be on a par with discovering Unicorn droppings.
For the moment though, I'm preoccupied with a GPSDO project of my own and struggling with persuading a Motorola MC14046B PD to do what it's supposed to do having given up with a 7486 XOR based PD which resoundingly failed to lock my 13MHz 5v CQE OCXO (test setup - I'll pick one of my six remaining 10MHz 12v CQE OCXOs when I finally crack this vexing problem) to a cheap NEO-6M GPS module which is limited to a 1KHz max frequency output on its PPS line.
I suspect the NEO-6M's habit of choosing a duty cycle at random between satellite loss of lock events (antenna signal or PSU interruptions) which I only recently discovered didn't help in the XOR PD case but it shouldn't have effected the class 2 PD choice with the 4046 micro-power PLL IC I'd managed to find lurking in a 40 year old collection of cmos chips the night before.
When I discovered that 14046 chip lurking in my collection, I thought my non-locking PLL issues had been solved.
Never mind, it'll be a relief to reach the OP's stage of worrying about the GPSDO psu ripple noise conundrum.
JBG
[EDIT 2020-01-19]
The failure to achieve a phase lock eventually proved to be a peculiarity of the NEO-6M I had been using. I'd configured it to output a 1KHz square wave on its PPS line (the fastest pulse rate it could do) and I eventually discovered that it had been shrinking the positive pulses halfway through a 10 to 20 minute cycle down to nothing before immediately restoring the pulse width back to the 50% mark I'd configured it to.
No effing wonder the XOR gate PD couldn't get a lock and the 4046 only managed to sporadically lock the PLL! Even picking the narrowest possible pulse width setting to limit the steadily continuous narrowing of the pulse to just a brief one second period at the transition point from extinction back to configured pulse width failed to fix the problem when using the 4046 in frequency comparison mode. The 7486 stands no chance of phase locking the frequencies since both signals need to be 1:1 ratio square waves in this case.
I gave up using the cheap NEO-6M and ordered a relatively expensive fake NEO-8M (16 quid!) which, fakeness aside, worked in this project just fine, allowing me to ditch the 4046 chip. I ordered another M8 module for just 8 quid a few months later (also a fake but less so - it seems to include the SAW filter which the first fake appeared to lack as well as more typically the absence of the flash chip). Receiver performance of this fake is more on a par with the original genuine M8 module which I'd managed to burn out the PPS drive output line with a 12v jolt a few months earlier.
I'm now in the middle of rebuilding my breadboarded GPSDO onto veroboard sized to fit an extruded aluminium enclosure as my MK I version of a completed working unit and dealing with further improving on the 5.5mV P-P (700μV rms) switching noise and ripple after the LC filter I'd added to reduce the 30mV noise and ripple of a 5v buck converter module I'm using to power it from cheap 9 or 12 volt wallwarts.
I'm now looking to use a ferrite bead or two with a 1μF and another 10μF ceramic capacitor to get rid of the remaining 10MHz switching transient noise - a prerequisite if I wish to add LDO regulators to further stabilise the Vcc supplies later on if deemed necessary. However, this extra filtering might prove sufficient in itself without the complication of additional LDO regulators being thrown into the mix - time will tell.