Re: FeelElec New Arrival FY-6900 Signal Generator

[Johnny B Good]

@Roman oh

 Thank you (and @ms963) for your reports. If I didn't already have a very nicely modified FY6600 to perform the more demanding tasks that my SDG2042X is currently incapable of through the lack of a decent UI (FFS! Wake up, Siglent!), I'd be tempted to buy an FY6900 myself.

 The only concern being the change of FPGA and its deficiencies reported by Labrat101 which may or may not be limited to his particular example or the particular revision he got lumbered with. Everything else, such as that 85 ohm attenuator pad and lack of effective cooling and so on are trivially easy to fix. Addressing the cooling issue is even easier than it had been with the FY6600 and 6800 models - just fit a 40mm sq 12v fan powered off the 5v rail, no cutting out of a 38mm hole in the base of the case or wherever. Even better, the hard work of upgrading the C8 mains socket on an FY6600 has already been done with the 68 and 69 models, leaving just the task of undoing the wilful earth loop vandalism perpetrated by Feeltech before adding a 1 to 10K bleed resistor to suppress the half live mains "Touch Voltage".

 That 85 ohm attenuator pad simply needs to be replaced with a standard 20dB 50 ohm pad since it had been designed to use one in the first place. All Feeltech had done was save a few pennies on the BOM costs by substituting the more expensive 1% E96 resistors (60.4 - 249 - 60.4 Pi network), normally used, with a cheap set of 5% E24 resistors (100 - 510 -100 Pi network) which could fake 20dB only in the Hi-Z case (>10K loadings).

 Upgrading from a single 5v rail smpsu to an analogue transformer based psu makes elimination of the radiated and conducted switching noise a relatively simple task in this case. You can use an analogue regulator but I think you'd be making a rod for your own back in regard of the extra heat dissipation. However, testing with a small 12v SLA and a 5v regulator is a lot easier to set up than a 3 rail battery powered supply when determining whether such an upgrade would be worth all the effort and expense, You can even compare analog versus switching regulators.

 Since the design relies upon on-board boost converters to generate the +/- 13.7v rails, the modest level of noise and ripple on the 5v rail using the mini 360 buck converter mentioned below, will likely be drowned out by that of those on-board boost converters anyway, so certainly worthy of serious consideration.

 My own personal choice would be to use one of those 3A (max) rated mini 360 buck converters sold by Banggood for use in flying drones (same size as a TO-220 7805, up to 97% efficient - synchronous rectifier- and ultra low noise compared to the larger general purpose buck regulators also sold by BG). If you need a separate 12v to power an OCXO, you could use another one of these buck regulators and solder blob the 12v jumper after cutting the trace that presets these to 5v (but see below).

 Their maximum input voltage rating is only 24v so you need to be careful in your choice of secondary voltage - a 12v secondary (for a 5v only psu) seems to be a wise choice given all the unknowns (is it an unloaded at maximum or minimum mains voltage input rating, or is it the rms value when fully loaded at exactly the nominal mains voltage after the 3 to 5% volt drop is accounted for and so on?). A 15v secondary would allow a second buck converter to supply 12v but this seems too risky an option regarding the 24vdc max limit of these buck regulators.

 If you need a separate +12v for an OCXO, it might be better to stick with a 12v secondary and add another 5v 3A buck converter to feed a 7W rated 5 to 12v boost converter. This was the arrangement I'd used in my MK II GPSDO to retain the flexibility of the MK I's 6.8 to 24vdc input voltage range.

 Eliminating the extra TTL chips used in the MK I to obtain 10MHz locked to the original 13MHz OCXO I'd used had reduced power consumption from 1.8W down to a surprisingly low 1.4W in spite of the additional boost converter to drive a 10MHz OCXO with exactly the same at-temperature 900mW power consumption as the 13MHz unit I'd powered from the 5.17v rail in the MK I (it had tested ok on 5v but I'd had no datasheet to verify this, my very first and only, OCXO's actual voltage requirement so I'd played it safe).

 I've been meaning to test the effect of powering my FY6600 from battery power with a 7805 supplying the 5v from the SLA but, with having to deal with three voltage rails plus the 12v feed to the OCXO, it's an experiment that's simply been left on the backburner all of this time. I have to admit that I do envy those of you looking to upgrade to a quieter 5v only psu in these later FY6900 models - it's so much easier to test and concoct a replacement single rail psu.

[Roman oh]

Addressing the cooling issue is even easier than it had been with the FY6600 and 6800 models - just fit a 40mm sq 12v fan powered off the 5v rail, no cutting out of a 38mm hole in the base of the case or wherever.

Not quite so simple - I found that the new 5V connector on the V2.1 pc just fouled the 40mm fan by enough to stop the rear cover locating cleanly, so I removed the connector and soldered in a flying lead (using a heavier gauge wire than the original) to the power supply (kept the connector at the ps end, of course). Also, one has to cut out the little plastic moldings to open up the slots in the bottom of the unit, to enable the airflow. 2 minutes work with a sharp knife. Otherwise the (totally inadequate) 30mm2 or so of opening comprised of very fine (about 0.6mm slots) will very quickly block up with fluff and dust and become totally ineffective.

[Labrat101]

Just sticking a fan on the back to to cool it down .
Just does not always work as you may be cooling some parts of the PCB @ different rates its called turbulence so it will cause different components to
drift at different rates .
IE  1k 1%  can be 0.9987 at say 24.0 degree and at 25.5C be 1.002 k . (Example Only) 
so if the PCB is not held at an even steady temperature including the thermal heat transfer from IC, and any SMP the drift and errors will be made worse.
 And with the already bad design can only get worse .
The Power supply should be cooled separate to the main PCB . By using a thermal baffle to isolate the SMPS & the main board .
and let the main board stabilize to the box temp with no air movement from the fan cooling the SMPS. .
After about an hour main PCB will stabilize.   

Example .. Jumping into a swimming pool the water is nice and warm at the top . BUT the bottom can be freezing 

Have Fun

[Trader]

[Johnny B Good]

This is the other Robert's Smorgasbord in depth review video mentioned. Sadly it too suffered from a lack of understanding of the limits inherent in even DDS ARB generators costing tens of thousands of dollars (nor the basics of measuring the performance of such kit - the DSO used was not exactly a shining example of its class). Also, I'm assuming he'd thought to use the 10x setting with his kludged probe to bnc adapter setup since a 1x setting would have seen the bandwidth limited to 8 or 10MHz at most. Presumably he didn't have the 50ohm terminator and bnc cable kit to hand to make up a tidier setup.

 I forgot to mention that I've never seen such bad jitter on a 60MHz sine wave with my own FY6600... until just now Since I knew the waveform had never been as bad as the one in Robert's video, I tried out the autosetup button on my trusty SDS1202X-E and even the "default" button for good measure despite the autosetup turning the weedy jittery trace into a nice steady full fat sinewave - it must be a good three years since I last had to resort to those measures 

 Anyhow, after that bit of twiddling, I was able to restore "normal" service, literally! and retain the full fat steady sine wave trace I've come to expect from my much modded FY6600 with just the merest hint of the 200 to 300ps of jitter noise, which obviously becomes more apparent at 60MHz compared to the usual 10MHz setting I've been using during my RFS project over the past year or so.

 Since I'd seen exactly the same unsteady thin and dim trace as Robert's video had shown, my guess is that his Rigol had also been in need of similar persuasion tricks to make it fly straight again. Unless he sorted it out later in the part of the video I haven't yet seen, it would seem an injustice was being perpetrated against that poor defenceless FY6800. I'll watch the rest of his YT video - who knows, perhaps he figures it out and maybe even that the real culprit for the DC offset is the mains safety earth connection.



 I gave up when he tried to adjust the DC offset and hit the max output voltage swing limits, a DC offset that was probably due to the hard mains earth connection of the BNC grounds (as one of our members had fallen foul of with his newly purchased FY6800 until I'd pointed out this possible source of ground loop interference that simply hadn't been an issue with the earlier FY6600 models due to the lack of such a mains earth connection which had created the half live ESD risk to sensitive DUTs instead - cured by simply replacing the C8 socket with a 3 pole one such as a C14 or C6 and only linking to the socket's earth tag via a 1 to 10k resistor to kill the ESD risk stone dead without providing a low Z path for DC and noise interference from the mains earth wiring to pollute the link between the generator and the DUT).

 However, having said all that, I'm in total agreement over tomtektest's high regard for these cheap AWGs. Admittedly they start off as Pig's Ears for the sake of penny pinching but it really doesn't take an awful lot of effort to convert them into something to be proud of, even to the extent of outclassing the Siglent SDG2000X series in some aspects of their operation, mainly, it has to be said for lack of a decent UI that doesn't throttle the frequency adjustment range capabilities of its DDS and an odd way of handling a single external 10MHz reference BNC socket that can act bidirectionally, resulting in pollution of a real external reference when you disable it as an input, allowing it to assume it can send its internal reference out of said BNC port -  was that all about!

 All of these Feeltech FY series are in need of a damn good modding to cure them of their penny pinching induced shortcomings. The 6600 model needed a lot more work than the later 6800 (it already had a C14 mains socket which just needed Feeltech's earthing vandalism to be removed and a 1 to 10K resistor fitted in place of the stolen zero volt return wire in the PSU to mainboard ribbon cable connection - ditto for the 6900 btw).

 The 6900 saved any need to cut out a hole to mount a fan since it already possessed one (aperture in the case, not a fan) making it an easy essential cooling mod. The latest version using only the 5 volt rail from the exact same psu board as used in the FY6600 makes the option of replacing it with quieter analogue mains transformer type an almost trivial mod compared to similar exercise attempts with the earlier models.

 That just leaves all the other mods used on the FY6600 to be applied to cover the exact same penny pinching induced issues to turn your Pig's Ear into, if not a silk purse, then at least a nice looking rayon purse.

[Labrat101]

Hi Johnny B G .
I thought this FY had died . All said and done.  Great time waster
No one will get rid of the jitter until the c¥clone chip is replaced with a real Cyclone . Instead of the installed C¥clone . And also the bad design of the pcb between c¥clone & dac's.

[Johnny B Good]

Hi Johnny B G .
I thought this FY had died . All said and done.  Great time waster
No one will get rid of the jitter until the c¥clone chip is replaced with a real Cyclone . Instead of the installed C¥clone . And also the bad design of the pcb between c¥clone & dac's.

 I had quite a few tabs open on almost zero activity forum threads which I was in the process of setting them to give me an email notify alert before closing them when, after refreshing each page to check for any new replies, I spotted Trader's contribution to this rather moribund thread. Naturally, I viewed the youtube video which lead me to check out the Robert's Smorgasboard video both of which were irritating enough for me to pass comment on.

 The irritation stems from the fact that they both lack experience of the product they're attempting to review, one of the reasons why I don't attempt to post any youtube vids in response to set the record straight - by the time I've gained a more thorough (not necessarilly complete) understanding of a T&M's quirks, the urge to contribute has been tempered by the realisation that doing so would be giving google yet more data than they've already gleaned from my internet activities thus far.

 Trying to stop google from data mining the hell out of my existence is almost certainly a lost cause by now but I'm a man of principle, largely about the one relating to that truism "Knowledge is Power" of which google is a shining example - the less detail about my life that I'd be obliged to give to them in order just to 'put the record straight', the better.

 Even signing up to contribute to the comments is a big no no, quite apart from the fact that my contribution would be drowned out by all the AOL responses and the downright misleading contributions - it would be a lost cause not worth the sacrifice of my privacy to google. Privacy is not so much a matter of 'having something to hide so much as having everything to hide, hence my contribution here to a more targeted audience where, for the most part, the misleading and AOL type responses are mercifully rare aberrations.

 I'm glad I gave the 1202 a break from monitoring the 64Hz PWM gate drive pulses controlling the cooling fan that regulates the LPRO-101's base plate temperature to within +/- 0.2 deg C of the 36 deg setting in order to compare my FY6600's 60MHz sine wave output against Robert's Rigol traces otherwise I'd have been dissing his Rigol more than it was actually deserving of.

 The only reason I knew why my initially poor traces were down to the DSO and not the FY6600 was because I knew from past experience that they should have been a lot better and also that I had seen the 1202 get itself, in the dim and distant past, into a state of confusion (something that had been exclusively the prerogative of the operator in the days of CRO technology) leading to this result.

 In this case, I knew it was merely a matter of giving the 1202 a chance to sort itself out by pressing the "Auto" button in the trigger panel. Using the "Default" button was an unnecessary 'icing on the cake' measure (but if in doubt, worth resorting to for 'good measure' in such cases since it's not an entirely 'Bad Thing' ).

 It wasn't just the business of "DSO confusion" that irked me, it was also, for instance, the inept way he kept pressing the waveform button to cycle through the options when he should have been using the rotary encoder knob and pressing the OK button twice to jump straight back to the sine wave option along with missing the trick of long pressing the ok button prior to a power down in order to preserve the current settings (which is why I put up with the beeps since the resulting long beep confirms that I'd held it down long enough to store the settings).

 If he'd taken more time to acquaint himself with the control feature set, he could not only have shortened the video, he could also have passed this useful knowledge on as as a 'Value Add' service to his viewers.

 Moving on, I think the reason for my initial poor looking 60MHz trace is down to a minor timebase configuration bug (seemingly one common to the Rigol 'scope) which seems related to changing from seconds or milliseconds per division to the microseconds per division settings. Most likely unaddressed as a minor bug easily worked around by judicious use of the 'Auto' button and/or the 'default' button if required.

 As I mentioned above, I'd been monitoring the 64Hz PWM signal in the thermal fan controller breadboard lashup I've been experimenting with to stabilise the LPRO's base plate temperature so I guess the adjustment from a 2ms to a 2ns per division setting must have invoked the timebase instability issue in this instance.

 My earlier attempts to stabilise the rubidium oscillator (RFS) by tight base plate temperature control had been less than stellar due to undesired changes of temperature gradient from ambient temperature variations through insufficient insulation of the other five surfaces of the RFS for lack of space in my initial choice of enclosure. As a consequence, my modified FY6600 was used as a proxy for the RFS, allowing me to trim the frequency in calibrated 10μHz steps avoiding the need to try tweaking the RFS's trimpot, thus saving needless wear and tear.

 The SDG2000X just wasn't up to this task in spite of having its own (unfortunately bi-directional) external 10MHz reference socket since its UI prevented such fine tuning control. It was therefore relegated to a supporting role as a plinth for the FY6600 as you can observe in the attached photo.

 About a fortnight ago, I decided to correct this by enclosing all but the fan cooled heatsink in a 20mm layer of polystyrene foam and ditch the enclosure entirely until I could acquire or build a more suitable one. I figured that with enough insulation, an enclosure would become almost but not entirely, redundant (at least for the purposes of further thermal stability testing). As a result of this improvement (and a well padded out heli-pot connected to the C field adjustment interface pin), I've been able to remove the FY6600 out of the circuit, allowing the RFS to 'fly solo' as it were.

 The results over the past week or so of testing now seem to have borne out my hypothesis regarding the lack of insulation. I think I've almost reached the limits of what tight thermal regulation can achieve (obviously a suitable enclosure will provide a final refinement) and it looks like I'm now seeing the effect of barometric pressure variations starting to make its presence known - the weather station is currently reporting a figure of 1016mBars (a slight drop from the 1017mBar peak of an hour or three ago). Over the past fortnight it had ranged from a low of 984 to a high of 1007 mBars, generally hovering within +/- 5mBar of a 1000mBar average for a large part of that time.

 This is the hottest day so far this year in my part of the UK (early afternoon temperature reaching a high of 28.5 deg C (currently now a mere 25 deg at 17:30 BST (16:30 UTC) and I've closed the windows to my office come workshop to raise the room temperature to 30 deg or higher in order to test the RFS's performance against a wider temperature swing.  The room is already at 27.6 deg from the 25 deg it had been just half an hour back when I shut the windows and I'm already starting to feel the heat.

 I've been taking scope screenshots of the GPSDO and the RFS 10MHz sine wave traces with infinite persistence enabled to track the phase variations between them. I've been seeing phase shifts of less than 100ns per 24 hour period - the last, current run, allowing for the ever present 7ns pk-pk phase shifts in the GPS signal due to ionospheric disturbance and other system errors produced a shift of just 31ns over an observation interval of 19 hours and 40 minutes before starting the high room temperature test to check the limits of my temperature management of the RFS.

 With the aid of a 2.5KW heater, I managed to zoom the room temperature right up to 33 deg C before switching it off - it had done its job, stressing my thermal control setup beyond its current limit which was now running the fan full bore without let up. It only managed to regain control when the temperature had dropped to 32 C - the heatsink had gotten up 36.6 C before settling back to the 36 deg set temperature. Even so, this temperature excursion had taken its toll on the frequency stability of the RFS as can be seen in the final two screenshots

 I've attached a photograph of the RFS test setup, along with 9 screenshots depicting that aforementioned observation period. Whilst some may consider this an off-topic posting, it does have relevance in demonstrating how useful these cheap and easily modifiable AWGs can be when you upgrade the XO to an OCXO with an external 10MHz reference input socket by which to phase lock (or, in my case, injection lock) the OCXO to an atomic derived frequency standard. Besides, you can see my FY6600 sat on top of its Siglent made plinth in that picture.

 In the foreground you see a breadboarded thermal regulator for the heatsink fan seen attached to the baseplate of the LPRO-101 encased in a 20mm thick polystyrene foam overcoat. Just to the left of the breadboard lash up can be seen a 5K Bournes heli-pot padded out with a pair of 22k resistors connecting to the 0v and 5v rails on the breadboard which is powered from a separate 12v wallwart which also supplies power to the fan, feeding the  5 and 3,3v regulators on the power adapter module plugged into the right hand end.

 The RFS and breadboard grounds are tied together since the RFS is powered directly with a separate 19.5v laptop charging brick. The 10MHz half volt sine wave output from the RFS feeds CH2 of the SDS2000X Plus behind the RFS with the GPSDO (perched on top of the FY6600) feeding CH1. I'm triggering from the RFS on CH2. The SDS1202X-E that can be seen on the shelf above the 2000X+ is displaying the PWM pulses driving the gate of the FET which feeds the fan.

 The two DMMs to the left are, from left to right, displaying the thermistor and 33k potential divider voltage going to a 5v cmos rr opamp (located beneath the blue electrolytic cap in the middle of the board) configured as a comparitor comparing the setpoint voltage from the black multiturn trimpot at the left hand end of the breadboard. The DIP to the right of the hidden opamp is a 32KHz clock oscillator and divider chain IC I'm using to get a stable 64Hz triangle wave for the PWM (using a schmitt trigger oscillator just didn't cut the mustard in this application).

 The second Mestek DM91A is displaying the voltage difference in millivolts between the VFC monitor output from the GPSDO and a temperature stabilised 2v reference derived from its OCXO's reference voltage pin. This allows me to get the same 100μV resolution as a rather more expensive 5 digit bench voltmeter would provide. It might not match the absolute accuracy of a 5 digit bench DMM but it does provide the 100μV resolution needed to detect the smaller voltage variations that would otherwise remain hidden by using a 10000 counts DMM to directly measure the VFC. The absolute accuracy of the 2v reference is of little to no consequence in this application just as long as any offset error remains unchanged for reasonably long periods of time against ambient temperature changes. My experience over the past 12 months or so indicates that both this two volt reference and the DMM's internal reference have been very stable indeed.

 The 'scope screenshots helpfully provide a real time date/time stamp in the bottom RH corner, a feature sadly lacking in the SDS1202X-E which would otherwise have done the job. It lacks any form of RTC or a built in webmin interface, limiting its usefulness in documenting the results  from this method of data acquisition by screenshots for periods longer than an hour or two. The 1202 only uses 22W versus the 54W of the 2000X+ which has become a bit of an irritation after running it 24/7 for near on the past two weeks

 The screenshots should be self explanatory so I'll quit whilst I'm ahead. This been a very long post already but, hopefully one of some interest to whoever may still be watching this topic thread.

[Labrat101]

Hi Johnny B G .
It looks like you have been busies . I must say very nice setup on the RFS  .  I would have offered you some BadAss coolers but they weigh
over a Kilo as they are solid copper and very large . and the shipping to the UK is ugly . I use them here and the ambient temperature is 38C +    .
I also saw Robert's Smorgasboard videos quite some time back . They are really good if you suffer from insomnia  .
As you had already covered almost every thing . Yes, this guy really did not know how to use the FY6800 to its full potential and he never
mentioned the frequency Doubling from the user section . which I found, well must be almost year ago , I did manage to get a good sine
upto 90Mhz from my fy6800 60Mhz . . But the square was something else .
Due to C¥clone . ("Say no More... Know what I mean!!, A Peek is as Good as a Poke to a Blind Chip ")
But as you know that I converted FY to an  Aβ 40hz brain sink . which has done and served a very productive Use .For my wife with Demencia .
 The Best thing That I Ever Made & the frequency is locked to 40hz (10^-9 ) .
.
I have added some more cool test equipment to my Home Lab . A fluke Calibrator & HP3457A which I got from a local Lab and had already been Modded & Cal .  . Just love seeing 7 zeroes  after the 10 volts . But it was cheaper than what I Paid for my Fy6800 from BangUgood.

PS loved reading your post . Where Have you been Hiding ??

[Johnny B Good]

@Labrat101

 Thanks for considering those copper heatsinks. I'd been considering using scrap copper busbars as a wrapping for the whole LPRO underneath the insulation layer to keep the electronics board a few degrees cooler but gave it up in view of the problems involded in maintaining a good thermal contact with the top cover and sides and the baseplate. In any case, I came across an old NIST paper where they did exactly what I did in order to use a 1970s commercial predecessor to the Efratom LPRO as a transportable frequency standard, assuring me that I wasn't going to be thermally over-stressing the electronics with such a thick layer of polystyrene foam - if it's good enough for NIST, then it's good enough for me.

 I DDGd that HP3457A you mentioned and landed up checking out Ebay prices which seem outrageously high - circa 700 quid which is about the price of a modern Siglent SDM3065X version with discount. As good as the vintage HP kit was (and in most cases, still is) the asking price in the UK is far higher than it has any right to be. The availability of decent vintage T&M kit in the UK is virtually zero.

 Luckily, I haven't reached the stage where I'm thinking 5 1/2 digit resolution just doesn't cut it any more so all these Ebay Bastard dealers can go shove it as far as I'm concerned. If I ever needed a 6 1/2 digit bench DMM, I'd rather pay that fraction extra and get a brand new Siglent with a modern feature set and a three year warranty. That time may come but for now, that cheap (13 quid at the time) 9999 counts Mestek and my temperature stabilised 2 volt reference work around the lack of a 5 1/2 bench meter nicely serves my immediate needs.

 As to the question of where I've been hiding, I've been occupied with that RFS project, one way or another, ever since I acquired the LPRO-101 last August and more recently with upgrading a 22 by 10 foot corrugated iron garage that was already past its prime when we moved into our current address just on 40 years ago. I curse the 'cowboys' who'd erected it, I'm guessing, way back in the late 40s or early 50s but of course they've long since gone to meet their maker.

 With much assistance from my youngest son, we've cleared it out, pulled it down and disposed of the scrap some 6 or 7 weeks back. I'm still waiting for builders' estimates for making good the existing (and rather badly laid) concrete base before ordering a concrete sectional garage of almost identical size (it's just an extra 3 inches wider).

 When I first decided to turn the LPRO into a lab grade RFS, I'd located a still unused steel enclosure in my basement collection of 'handy stuff' which seemed just the ticket but I'd not allowed for all the extra thermal insulation required to make the fan cooled base plate the sole arbiter of its internal temperature.

 As a result, I got rather bogged down trying to source a more suitable enclosure from Ebay but couldn't find anything 'cheap and cheerful' enough to save myself the effort to fabricate a custom sized case from out of a larger pre-owned aluminium case that had been fabricated using aluminium sheet and lengths of aluminium angle drilled and tapped to take cheese headed Allen screws to house a Transam Tuscan S100 bus kit computer motherboard way back in the early eighties.

 I've disassembled that case into its constituent parts to await my final decision on how to reconstruct them into a 300mm wide by 180mm high by 220mm deep enclosure (exact dimensions subject to further refinement). However, that is going to be a lot of metalworking which I'd prefer to do in a well equipped workshop, hence the sudden urge to replace the rotted out corrugated iron garage with a modern concrete sectional garage that I can use primarily as a workshop (I don't ever plan on parking any cars inside of it).

 The RFS project got stalled for the want of a suitable enclosure but about three weeks ago, I came across a rather handy polystyrene frozen foodstuffs box in the attic, complete with lid out of which I'd cut the pieces used to insulate the LPRO. It was this find that had triggered my resumption of the RFS project since I realised that with this much insulation I could continue without an actual enclosure until the final stage of the project.

 Yesterday's extreme thermal stress test seemed to have a more enduring effect on the RFS which wasn't helped by my attempts to recalibrate it hours after things had cooled to more typical summer temperatures hours later. Until the day before, I'd been triggering from the GPSDO on the basis that this had the long term stability of the World Atomic Time reference in spite of its shorter term 7ns pk-pk subsonic phase modulation wobbles. I had switched the trigger source to the RFS on the basis that this didn't have any such subsonic frequency wobble to shift the effect on the wave traces to where it rightfully belonged to the GPSDO trace.

 Unfortunately, this led to a bit of confusion on my part as to which way to retune the RFS to syntonise it to the GPSDO. It wasn't until this afternoon (Sunday) that I realised why I hadn't had any success with this task. Once I had started carefully working out which way I had to rotate the heli-pot for each and every attempt to trim the RFS onto frequency, I finally started to see success in getting the RFS syntonised to the GPSDO. I made the last adjustment just before 9pm BST (19:53 UTC) Sunday evening and it's been running without further adjustment ever since. It's now 3:14am BST Monday and I've just captured my final screenshot for 'tonight' which you can peruse, along with the 8 preceding screenshots during that final period which I've attached to this post.

 Allowing for at least a 6ns pk=pk GPS wobble, the region swept by the infinite persistance suggests a worst case drift by the RFS of 10ns in just over 6 hours. Of course, the GPS may be contributing more than that 6ns of wobble , so a much longer run is needed to get a better assessment of the RFS's stability. I'll have a better idea in another 6 to 18 hours time.

[Labrat101]

@ Johnny B G ,
That last photo is starting to look very nice . As with all your projects you put a lot of brain power into getting it
to Perfection . I would be careful NASA may sneak in a steal your ideas .
As you mentioned earlier what with Google Mining everything . Even from EEvBlog I have seen pictures up on
Google within a few days if not less.
Maybe Dave will give an option to set photo's and text to Private registered users Only .
 To keep the spiders  & other Bugs --ers  out .
I am sure there are many great ideas that were thought up on this site and stolen by the Chinese etc .
and a month later up on BangUgood etc with no credit to the original Brains  .
Ok may save you building it boxing it all for a few $$  . Interests & profit in there pockets 

 Good luck on the construction outside

[Johnny B Good]

@Labrat101

 Whilst I like to get a project as close to perfection as practical, there's little to no original design effort involved unless you count my independently arriving at solutions already tried and tested by others as 'original'. Of course, it's all rather gratifying to discover that ones own modest, independently arrived at solutions have been endorsed by previous developers' efforts.

 My main problem in such endeavours is a tendency to overthink a solution. For example, whilst I bemoan the fact that the cpu cooler fan's high cogging effect prevents a completely smooth control of fan speed from zero to max rpm to minimse the +/- 0.2 K 30mHz variation in the middle of the quarter inch thick aluminium heat spreader that links the fan cooler to the base plate, in practice, considering the thermal inertia involved, it's unlikely to be of any consequence provided there's no significant shift in the average temperature between the stop/start cooling mode below the 28 deg threshold and the continuous run mode above that point.

 Even a simple on/off control of the fan without using PWM to 'soften the blow' is probably more than sufficient for this task, especially when you take the effect of ageing drift and barometric perturbations into account which will mock your best efforts at creating the most perfect of thermal management solutions you can possibly concoct.

 The ageing drift and barometric effects I can deal with after I've located or built a suitable enclosure to assemble it in where I can add yet more foam insulation, keep it protected from random air currents and create a balanced flue effect in the ventilation pathways to minimise passive convective cooling to extend the lower ambient temperature limit when the fan is completely stopped.

 The key to the improvement over my previous attempts to stabilise the temperature has undoubtedly been the use of a 20mm thick covering of polystyrene foam versus the two or three layers of micro-bubble wrap I'd previously used for lack of space in the original enclosure. Having the room to add another 20mm layer of polystyrene foam and create a fan intake plenum out of the same material to draw in fresh air via a hole in the base of the case and exhaust the heat out via ventilation holes low down on the rear panel should allow me to achieve my design goal. The case will be standing on 10mm tall rubber feet and I'l glue a 10mm square length of soft rubber foam to the underside as a further barrier to unwanted pollution of the intake by warmed up exhaust air.

 All those final touches to the thermal management should leave me just the ageing and barometric issues to contend with. Since I'll need to use a Nano 3 to interface to the barometric sensor in order to generate an analogue control voltage, I'll likely have a crack at the Lars GPSDO design (probably using a TIC in place of the original resistor/capacitor arrangement he'd used) and convert it into a GPSDRO capable of days long holdover performance. However, that's a project in itself which I'll deal with much later.

 Creating a temperature stabilised Rubidium frequency lab standard reference is more than enough of a project for now. I'd like to be able to take time out to bask in the glory of this achievement before gathering enough 'nits' to compel me to make further refinements.

 I've attached another ten screenshots continuing on from the last set I posted in order to provide 'closure' on this last test run.

[Johnny B Good]

 Here's a final 36 hour run's worth of screenshots.  I think I've posted more than enough such screenshots but I thought a final single screenshot would be of some interest here. Since I've now got to the state where I can syntonise the RFS to within less than a cycle's worth of daily drift, I can give the power hungry SDS2000X+ a well deserved rest and let the 1202 do its share for a change since I don't need to record any more screenshots. The lack of a date/time stamp and any means to capture such images direct to the PC via the lan connection is no longer an issue in this case.

 Although the FY6600 is no longer part of the setup, that wasn't the case some six months back when it had played an important role in the development of my RFS project.

 As is always the way with such postings, I forgot to mention the need to knock off a conservative allowance of 6 or 7ns from the total of 70ns swept by the infinite persistence to account for the GPS 'wobble' (which possibly may in reality be significantly greater than this at phase modulation frequencies down into the hundreds of μHz due to limitations in the GPS system itself).

 I also noted the slight 100Hz discrepancy in the 'scope's frequency counter reading which until recently would have displayed the 10MHz GPSDO output as 'exactly' 10.0000 MHz.

 When I first got the 'scope about three years ago, it had been showing 10.0003 MHz which over an initial period of 5 or 6 months, settled onto a 10.0000MHz reading. The fact that the reading has risen rather than continue downwards suggests this is just a temporary effect due to the unusually high temperatures I'm now experiencing rather an ageing trend.

[EDIT 2021-07-26] My hypothesis about the 1202's frequency counter  being slightly off due to the unusually high temperature appears to be correct. I was reading through what I'd written above and decided to take another look now that we're back to more sane British summertime temperatures (currently enjoying a less stressful room temperature of 23.1 deg C) and it's now back to displaying the GPSDO and Rubidium output frequency (and, of course that of the FY6600 when it's set to output an indicated 10,000,000.000000Hz) as 10.0000MHz.

[robot.golem]

Hello:

I got a FeelElec FY6800 60 MHz and generally liked it.  However, it stopped working after a few months: Channel 1 is completely out, and channel 2 has positive DC bias and unregulated amplitude.  I contacted FeelElec about getting the warranty service (they promise 1-year warranty), but in return they send me a series of emails amounting to what I would call a run-around.  Still waiting for them for instructions on how to obtain the warranty service.

Anyone has repair manual/schematics for the instrument?  -- This could be a very long shot for me, but this is the only option before trashing a fairly expensive device.

Also thinking of writing to the Communist Party leaders in these guys' province and city, how their company gives China a bad name in commerce.  Maybe that will catch their attention.

[Labrat101]

Hello:

I got a FeelElec FY6800 60 MHz and generally liked it.  However, it stopped working after a few months: Channel 1 is completely out, and channel 2 has positive DC bias and unregulated amplitude.  I contacted FeelElec about getting the warranty service (they promise 1-year warranty), but in return they send me a series of emails amounting to what I would call a run-around.  Still waiting for them for instructions on how to obtain the warranty service.

Anyone has repair manual/schematics for the instrument?  -- This could be a very long shot for me, but this is the only option before trashing a fairly expensive device.

Also thinking of writing to the Communist Party leaders in these guys' province and city, how their company gives China a bad name in commerce.  Maybe that will catch their attention.

This is FY6600 . I have the same FY6800 the diagram is very close and will cover 99% of what you require.
This subject hits many sore spots and there are 100s of pages in this forum by Me , Johnny B Good , Bad driver and the list goes on . Getting anything out of Feeltech  makes mission impossible look easy .
This will start you off . The FY6800 was just a bodged update of the FY66 . along with Non original cloned chips
There is a chance your output ops have oopsed .
Johnny B Good May remember were all our old links are covering this subject . and I think it goes back quire a few years and a lot of reading .

[garryk]

I've learned a lot about the characteristics of these units from this forum (great thread, thanks!), in my quest to find a good Sig Gen for my developing home lab and am quite attracted by the promise of their features for the price point. So I have been looking at which model/option "FeelTech" AWG (FY6600/6800/6900 -10/20/50/60 MHz etc) would be best for my needs, and was tending towards the FY6900-60MHz option, when I noticed that there appears to be a new(?) '100MHz' offering appearing for sale on the shopping channels!? Namely a FY6900H-100.

Does anyone know anything about this seemingly new model? Is it likely just FY6900-60MHz hardware with a few parameters adjusted??

Is it for real capable of 100MHz? Or (without wanting to be unkind about it) am i better to assume it's likely not worth the extra cost due to all the 'considerations' attached to real world performance of these 'hobby grade' units?? On the other hand, if it's just a stretched 60MHz unit, what changes are likely to have been made to push it to 100MHz?

[Trader]

Is it for real capable of 100MHz? Or (without wanting to be unkind about it) am i better to assume it's likely not worth the extra cost due to all the 'considerations' attached to real world performance of these 'hobby grade' units?? On the other hand, if it's just a stretched 60MHz unit, what changes are likely to have been made to push it to 100MHz?

These devices are hobbyist grade (about noise, THD, etc).

Thinking about cost-benefit, the 100 MHz costs roughly 40% more than 60 MHz, and is rare to use more than 30 MHz.

[Labrat101]

Does anyone know anything about this seemingly new model? Is it likely just FY6900-60MHz hardware with a few parameters adjusted??

Have a look at the  UNI-T  UTG962  It is 60Mhz and has some really good functions . I bought one about a year ago and
I am very happy with its performance . Check out the spec sheet and see see if it covers your personal requirements .
It also does not suffer from Jitter on the square wave like the FY series .
It also has a smaller footprint .

[Johnny B Good]

@garryk

 The FY6900-100M differences to its predecessors is obviously going to be a modified firmware and, most likely, a revamp of the anti-aliasing filter components to increase the turnover frequency from around the 75MHz mark to 100MHz (assuming the filter hadn't been designed with a maximum 100MHz option in mind to begin with).

 The DAC sampling rate has always been an overclocked 250MSa/s since the first of the FY6600 models which sets the Nyquist limit at 125MHz so such an upgrade to 100MHz has always been a possibility, even though it may have required an increase in the AA filter's turnover frequency.

 Setting aside doubts about the clone FPGA chip, the current FY6900 models are better featured than its predecessors, reducing the remedial work to installing a cooling fan, undoing FeelTech's earthing hack job and replacing the 85 ohm impedance 20dB attenuator resistor network with a proper 50 ohm one.

 The change from a 50MHz xo chip to a 10MHz one has somewhat simplified the business of upgrading to a better 10MHz tcxo chip or the even better option of installing a 10MHz ocxo. Also, the recent main board design change that has allowed a single 5v rail smpsu board to be used (in reality, exactly the same psu board as before with its +/- 12v outputs simply left unused), makes upgrading to an analogue based psu, free of the switching noise typical of cheap smpsus, a significantly easier to achieve DIY project for those concerned over this source of (mostly radiated) noise.

 Unfortunately, as has been reported many times in these topic threads, FeelTech or FeelElec offer little to no customer support, even to the extent of neglecting to update the user manual to reflect the actual pk-pk voltage limits against frequency, claiming in all models, a limit of 10MHz for the 20 (or 24) v pk-pk limit before it drops to 5v pk-pk when in fact, this limit has always been at 20MHz

 FeelTech's penny pinching goes far beyond mere cost cutting on the BoM since their target customer base is quite clearly cash starved home hobbyists only too prepared to remedy its shortcomings with warranty voiding modifications ranging from the easiest yet most urgent fan installation mod (made almost irresistible in the case (literally!) of this latest version of the FY6600, the FY6900 with its unoccupied fan mounting point just crying out to be filled with a suitable and cheap 5 volt cooling fan) to a more ambitious OCXO upgrade.

 The bastard in control of the FeelTech operation is quite clearly determined to have his cake and eat it.  If you are still of a mind to buy an FY6900 after all is said and done, I suggest you refrain from heeding Dave's advice of "Don't turn it on. Take it apart!" for at least a month or two of stress testing the hell out of it before voiding the warranty.

[Labrat101]

My FY6800 failed the stress test after only 3 hours when the power supply sent out smoke signals  S O S .
(smoked out Smoke) . I asked for a replacement . They gave me $20 refund and said get it fixed   
 Why would you trust their Warranty     Its Void ed

[Trader]

l read all the topics complaining about jittering in Square waves at 9.00024MHz, 9.00027MHz, 7,777,777 Hz, 7.7778MHz, etc, and I tested all of them and other combinations.

For me (IMHO) the FY-6900 behavior is as expected for a $90 AWG.  I only can see some jitter in the 20MHz and seems to be 1.5ns.

I'm not sure if the UNI-T UTG962E has low jitter in all frequencies.

[Trader]

My FY6800 failed the stress test after only 3 hours when the power supply sent out smoke signals  S O S .
(smoked out Smoke) . I asked for a replacement . They gave me $20 refund and said get it fixed   
 Why would you trust their Warranty     Its Void ed

I let my FY6900 running both channels: 1) max frequency and vpp, 2) 24vpp, 3) Sweep for 24+ hours without problem, and I never read about people having the magic smoke.

[Trader]

Even with a small jitter, the wave seems to be "stable", after 1 hour the signal is still inside a fixed boundary.

[Johnny B Good]

@Labrat101

 That issue of the 4ns jitter on square wave output in the case of the FY6600 and 6800 models has been addressed (or at least mitigated) in the FY6900 series. The UTG962 has a few more shortcomings compared to the FY6900-60M as I'm sure you're only too well aware.

https://www.eevblog.com/forum/testgear/uni-t-utg932utg962-200msas-function-arbitrary-waveform-generator-220394/msg3614478/#msg3614478

 Anyone looking for a cheap AWG will need to take this into account when deciding which will best meet their needs. All of these cheap AWG have their shortcomings - even Siglent's otherwise fine range of AWGs have some rather unexpected ones such as a bi-directional external reference socket that can pollute a 10MHz lab reference distribution with its own mediocre internal 10MHz clock due to ill thought out firmware and a horribly afflicted frequency setting UI that makes the Feeltech's keypadless frequency setting interface look like a work of genius to name the two most egregious annoyances

 Out of curiosity, I checked out their pricing on BangGood to see if it would make any sense to choose the UNI-T over the FeelTech. At just under 3 quid more for the less lightweight FY6900-60M I don't see much of an incentive to go for the UTG962 option other than perhaps to avoid having to deal with a company that's had a bad customer relations record. Since I'm not in the market for another cheap AWG, I've not looked for reports on UNI-T's quality of after sales service which, for all I know, could be nearly as bad as FeelTech's. I just might find out if I care to read through all 11 pages of that UTG932 thread I linked to.

[Johnny B Good]

My FY6800 failed the stress test after only 3 hours when the power supply sent out smoke signals  S O S .
(smoked out Smoke) . I asked for a replacement . They gave me $20 refund and said get it fixed   
 Why would you trust their Warranty     Its Void ed

 JOOI (I can't remember if you've already mentioned this) Did you manage to repair it? I might have accepted a 50% refund after verifying that it was only the smpsu board that had emitted the magic smoke and after verifying that the the main board hadn't suffered by testing with a temporary psu setup, otherwise I'd have insisted on a full refund, escalating the dispute with Ebay if needed.

 I wonder if this had been a consequence of the earthing vandalism inflicted by FeelTech to address the rather vociferous complaints against the half live mains voltage ESD hazard made in the long running FY6600 topic thread? Did you examine "The Remains"?

[Labrat101]

Yes that's what started the whole upgrade thing about a year ago .
There are photos way back of the upgraded power supply & and the oscillator upgrade. Etc also had problems with the relay feed etc.
And then I made the big discovery that my cyclone chip was a FAKE.
so I recycled the unit into an Alfa Beta pulse gen . For my wife with demcia.  As this works at 40hz .
And this unit was very stable after GPSDO calibration.  With a ocxo and PLL  mod. 
But as a signal generator there was too much jitter over 1Mhz.
Due to the cyclone chip being a Fake . The internal DAC is missing a feed point.  .
When I have more time I will look for the links . Or search there 100s of pages on this .I upload a lot of photos of all the upgrades .

Update

The Link to photo mods
https://www.eevblog.com/forum/testgear/re-feelelec-new-arrival-fy-6900-signal-generator/msg3093974/#msg3093974

also note the Cyclone chip all the label washed off with alcohol   

.