@AIMuz,
Thanks for the speedy response considering the mass of screenshots you'd probably had to peruse in some detail. In hindsight, I realise it would have been better for all concerned if I had actually been a little more selective over which screenshots to attach to my reply. Usually, the limiting factor is in the total file size but, in this case, the whole collection was less than a MB's worth which wasn't the best criteria in choosing what to send. Apologies for the screenshot overload.
The output scaling is easy to overlook, especially if you're unfamiliar with the basic circuit layout as I had been prior to downloading DerKammi's reverse engineered circuits, some 12 months ago now, for the FY6600 which he'd posted to github for the benefit of the members contributing to the FY6600 topic thread.
Sadly for me, the 1202X-E doesn't have that web server facility built in. A later firmware update might add this feature (I think I've applied two updates so far and I don't think there has been another one since).
=============================================================================================
[EDIT 2021-12-24] Actually, it does, but only a 'refresh' and a 'screenshot' facility rather than a 'live view' option. I think I'd tried the lan connection but quickly dismissed its usefulness out of hand for the lack of a 'Live view' option, totally ignoring its usefulness in obtaining screenshot files blessed with a valid date/time stamp.
=============================================================================================
Your comments regarding trying to assess noise characteristics are ever so true.
The built in FFT SA feature is limited by the use of an 8 bit ADC but using the ENOB option might improve the limited dynamic range (I haven't tested whether this is actually compatible with the FFT function yet).
I started using it just after I posted all those 'scope traces and with some experimentation with the settings and winding the timebase down to 20μs per division, along with an averaging factor of 16, I eventually got a decent looking spectrum covering the 0 to 100MHz range. I'll attach a couple of screenshots (the first with the generator switched off, leaving only the OCXO powered up, with the second showing the noise spectrum when turned on at a 0.3v DC output setting). You can be the judge of the usefulness of my FFT settings, which to me, seem to be the best I've ever managed to to set it up to so far.
Even after getting the FFT settings optimised, I was still seeing inexplicable spectra which led me, belatedly, to look for a baseline.
In the process, I discovered that one of my BNC dummy loads (the first one I happened to pick out of the tray) was acting like an antenna - it actually raised the baseline noise spectrum above that from an empty BNC socket contrary to my expectation of seeing a slight reduction from the 50 ohm input loading.
The rest of my dummy loads however, were free of this 'antenna effect' so it seems to have been just a weirdly faulty dummy load (testing the resistance with a DMM shows the expected 50 ohm resistance value (51 actually but all the others showed readings in the range of 50.4 to 50.8 ohms so, allowing for the test lead resistance, within a 1% tolerance range). I've put that weirdly broken dummy load to one side for a later more detailed examination.
Apropos of the quality of BNC patch cables, I've also discovered some serious problems. The half metre BNC to BNC cable supplied with the FY6600 is no shining example and, what's worse, the BNC plugs on my collection of 3 metre 'Cheapernet(tm)' patch cables are out of spec on the outer barrel dimensions, making them a little on the iffy side in regard of getting a good contact with the ground return on most sockets I plug them into. However, with a bit of adjustment on 'The Angle of Dangle', they're still an improvement over that supplied patch cable.
The actual cable quality might be ok (I'll have to chop the plugs off one to check - no great loss in this case). I've got a small collection of good quality BNC plugs I can use as replacements if the quality of the cable proves good. I guess even just over 20 years ago, the quality of Chinese manufactured patch cables was even worse than it is today.
Not being sure about how much noise the little 12v half amp smpsu board that powers the OCXO module might be contributing, I landed up pulling the plug to completely isolate it from the mains supply. When I was still seeing interference artefacts in the FFT display, I landed up disconnecting the 9999 counts DMM I use to monitor the OCXO's tuning voltage from my homebrewed GPSDO along with the BNC patch cable plugged into the 10MHz output socket.
This merely reduced the level of 10MHz spaced spurs rather than totally eliminate them so I pulled the plug on that as well which only gave a slight further reduction. In the end, I unplugged the mains cable from the C6 socket on the signal generator, leaving the connection to the 'scope as the generator's one and only connection to the outside world. Even in this state, I was still seeing noise that only reduced by unplugging it off the end of the scope connecting cable which then only completely disappeared once I'd unplugged that from the 'scope's CH1 input socket.
It seems that the plastic case of the FY6600 makes a very lousy shield, leaving the circuitry inside exposed to external interference. I plan on taking it and the 'scope (plugged into a long mains extension cable) outside, weather permitting, to repeat this test outside of what I suspect is a very polluted electromagnetic environment. It could turn out that a "Noisy PSU" may prove the least of this signal generator's woes although I rather doubt that eliminating this noise source won't at least offer a significant improvement.
If this proposed "Outdoors Test" demonstrates a significant (unpowered state) noise reduction, I'll have to run my battery power tests outdoors too if I want to collect any meaningful test data. It might seem to involve some effort just to get a baseline but I'd rather know the score before upgrading the PSU in the blind hope that I'll get a tangible improvement rather than discover that I've actually made things worse (or at least no better) for all my trouble.
Regarding what you said about setting too fast a timebase to see the slower yet higher amplitude mains frequency related ripple component is ever so true. It's a problem I became all too painfully aware of about a year ago when I first started trying to diagnose PSU noise and ripple from waveform traces alone. This sort of investigation is where the services of a half decent spectrum analyser comes into its own. If the use of ENOB is compatible with the FFT SA function on these DSOs, it might just provide a poor man's version of a half decent SA. I'll do some testing later today (it's now 04:27 UTC this Sunday morning right now).
Picking up on the points you raised over the BNC patch cables, your hypothesis in regard of the RG316 having a higher capacitance per metre due to it being thinner than RG58 doesn't apply. Being 50 ohm cables they will both show close to 100pF per metre (if you have an LC meter, you can test this quite easily). Although the conductor diameter does have some effect on its per unit length inductance, at this scale of physical dimensions seen with typical co-axial cables, it's an order or three less in magnitude (I wasn't able to find any figures in the literature on this aspect of co-axial cable properties - it's just a gut feeling) compared to the changes in capacitance per unit length that result from altering the diameters ratio and dielectric permitivity.
Since the impedance of co-axial cable depends on the ratio of L (which only changes slightly) and C which can be radically altered by the conductor diameter ratio, the measured capacitance for any diameter of 50 ohm cable remains very close to the 100pF per metre I've measured with various sizes of cable. Obviously, the capacitance value is lower for 75 and lower still for 93 ohm cables. I don't know the figures off the top of my head but comparing the capacitance (using an LC meter or DMM with a capacitance measuring feature) against length offers a convenient way to quickly ascertain an unknown (dirt cheap) cable's nominal impedance.
In this case when using a dummy load matched to the cable's impedance, you're using it as a transmission line and the LPF effect simply doesn't exist (other than for the unfortunate effect of the 15 to 20pF loading on the scope's input in the 100MHz and beyond range when it has no built in 50 ohm termination option and you're obliged to use an external in-line terminator). The lower noise level using the RG316 cable will simply be on account of its much better quality in regard of screening over that of the cheap cable that was used in the manufacture of that half metre BNC patch cable supplied with the signal generator.
All this discussion of patch cable quality (lack of in my case) has made me realise that I urgently need to upgrade my collection of test cables one way or another. I'll probably land up assembling my own set of patch cables unless I can track down a source of reasonably priced ready made patch cables of acceptable build quality which, these days, is easier said than done.
The following attached screenshots include the requested traces (but on the 500μV per division setting needed to raise a usable trace in my case) as well as the FFT plots I mentioned above.
The first two traces after the FFT images were captured with nothing plugged in and then with only a dummy load (I can't recall which way round they were captured - since there's no perceptible difference, the distinction seems rather academic). The fifth shows the Cheapernet(tm) 3 metre patch lead unterminated with the far end not plugged into anything whilst the sixth is terminated at the scope end. In both cases, the cable is simply coiled up on the test bench. The triggering source was CH1.
For the last four captures displaying the results of using the supplied half metre patch cable, I decided to use the AC line triggering option (hence the hardware frequency counter readings of 49Hz in the top right corner - obviously limited to truncating to the nearest whole Hz). In this case, this short cable was simply laid out on the bench in a straight line.
The first of this final sequence depicts an unterminated cable and I was obliged to reduce the Y axis sensitivity to 1mV per division for this capture. Grabbing the middle of the cable increased the amplitude by some 2 to 3 dB, confirming the abysmal quality of the cable itself. Holding onto the BNC plug at the far end reduced the level. The source of the external interference - the induced 50Hz voltage that my body was picking up - was being shunted to ground via this contact so no great surprise here.
The second image shows the result of connecting a 50 ohm terminator at the scope end of the cable, the third image shows the result of transferring the terminating load to the far end (scope end unterminated) and the fourth image shows the result when both ends are terminated. As before, I was once more obliged to switch back to the 500μV per division sensitivity setting for these last three screen grabs.
There was no discernible difference between the final three screen captures, confirming that the ingress of the 50Hz interference was due to incomplete coverage by the screening braid, allowing the E field component to leak through and induce a voltage onto the Hi Z input (1M ohm with circa 18pF shunt capacitance) when unterminated. Connecting a single 50 ohm terminator at either end was more than sufficient to attenuate it well below the 'scope's input noise level. Terminating both ends would be just "Gilding the Lilly" as far as this source of interference is concerned.
It's clear from these results that my Cheapernet patch cables could prove to be worthy of a connector upgrade. I've got enough BNC plugs to upgrade three patch cables at a pinch, two of which are of the screw onto the cable type (one with spring type centre pin socket making it solderless, the other needing the centre pin contact to be soldered for best reliability of this connection), the other four being the more conventional braid clamp and soldered centre pin "Burndept" style.
I've just replaced the bad plug on the cheapernet(tm) patch cable I'd been using with the solderless screw on plug and it seems to be quite acceptable (the outer screen proved to be mylar foil overlaid with a tinned copper braid with the centre wire being stranded tinned copper). I'll check out the other five patch cables. They might not all be as bad as the one I've just repaired and some, if not all, of the bad ones might only require replacement of a single plug to effect a repair. My patch lead situation may not be quite so dire as it first seemed. I guess you could include verifying your cables as a vital part of establishing your "baseline". The refurbished patch cable hasn't altered the original spectra I'd previously collected so I guess I must have got the "Angle of Dangle" just right on the earlier runs.
JBG