FeelTech FY6600 60MHz 2-Ch VCO Function Arbitrary Waveform Signal Generator

[DaveR]

Hi josh - BIAS is just the screen shorthand for the modulation frequency of the output, which can be set from 1uHz up to 1MHz.  As for the default output voltage, there is no way to change it from the front panel, but it could be changed by tweaking presets W5 (for Ch1) and W3 (for Ch2) on the main board.  I couldn't say offhand over what range you'd be able to adjust it, though, and you'd obviously have to live with incorrect screen vs actual figures until you recalibrate it.

Regards,
Dave

[yo0]

Hi Pio - those screenshots are probably just taken from an early model as the 6800 went into production, and still used as examples for advertising on ebay etc..  I doubt very much whether you'll get anything other than v1.6 or 1.7 if you buy one now.

Regards,
Dave

Thank you Dave!!

Best regards.

Pio

[Johnny B Good]

Hi, cybermouse,
Just to be sure, I tried powering the modified board with new op amps from the original switching power supply. Sure enough, the waveform is clipped and heavily distorted in high amplitude settings. So, the voltage from that supply is not enough. +-15V is a must (no wonder, this is to be expected).

 That's a surprise considering the 5v p2p limit above 20MHz. I wouldn't have thought the +/- 11.5 ish volts of an unmodded PSU board would have been a problem at a mere 5v Pk to Pk setting. The voltage swing limits pretty well just about save the generator from gross distortion of the sine wave output even for the THS3002 dual current feedback opamp chip that Feeltech have stuck with 'through thick and thin' despite the need to swap over to the later single CFB opamps they'd made provision for in the unpopulated U21 and U22 locations on the main board once they went beyond the original 15MHz limit in the first version of the FY6600.

 Apologies for the *very* late response (and first posting) but I didn't discover this EEVblog forum until about a fortnight ago as a result of a posting in the sci.electronics. design (sed) news group (Usenet) under the thread titled: "amazing ARB pricing" way back in October 2018 (the 25th) where my initial contribution had been to seek advice on sourcing a cheap oscilloscope.

 This set me off on a quest to buy myself a cheap Chinese 2ch DSO which I resolved by purchasing a nice SDS1202X-E from Siglent's UK agent (Labtronix) just 5 days later which was delivered just two days after that. It was only then that it occurred to me that a signal generator might be worth obtaining as a useful accessory to my DSO purchase, hence my taking note of the link to this forum that had been provided by the OP in the sed "amazing ARB pricing" thread.

 At that point, this forum thread had grown to a massive 69 pages and I read every last page (three days!) before deciding, in spite of the few serious negatives with the earlier firmware, to purchase an FY6600-60M model for the princely sum of £75.66 delivered.

 FYI, the model I received reports firmware version 3.3 which seems to have addressed a lot of the issues mentioned such as the long press of the "OK" button (the rotary control knob) to set the current settings as subsequent defaults on power up failing to provide a feedback 'beep' - it now does so (after about one second's worth of keeping the knob pressed).

 The first mod I applied was to add a 47K resistor to the PSU to boost the 4.95v (loaded) to 5.49v so as to raise the 12v rails from circa 11.5 up to circa 12.5 volts. A few days after that I decided it might be a good idea to replace the 12v rectifier diodes with something better than those 1N5819s (not the best  choice imo) with some proper 30A 45v rated dual shotky rectifiers (TO220 package - it was a bit of a tight squeeze!) recovered from scrapped smpsus which gave me a quite surprising boost to 13.7v on the +12v rail and 14.05v on the slightly less loaded negative rail.

 I didn't bother improving the 5v diode since this will reduce the +/-12v rail voltages. The just below 9 watts consumption 'flat out' has now dropped by about a 100mW in spite of the voltage boost which indicates a reassuring improvement of PSU efficiency. 

 I haven't bothered trying to tackle the half mains voltage 'tingle current' issue since I know this is the inevitable consequence of the mandatory EMC 'grounding capacitor' (typically a 1nF Y class) in a class II double insulated PSU. The best solution imo, is to track down a better class of smpsu where the need for such a capacitor has been obviated by the use of an HF ferrite switching transformer with the necessary screening foil built in to shield the secondary winding(s) from the HF voltage switching transients on the HV primary.

 In a class II double insulated PSU, this shielding foil will still be connected to the 'half mains' voltage grounding reference but will only present a few pF  of leakage coupling of the 50/60Hz half mains voltage rather than the 470pF or higher of a class Y capacitor.

 Since I did these PSU mods prior to recording my DSO traces as a 'before the opamp upgrade' record, I can only show attached jpg images for the 20Mz at 20v into 50 ohm load after having improved the PSU, so I have no proof that my mods are of any benefit with the existing opamp chips. Nevertheless, I've attached them for reference.

 I ordered a couple of THS3491 opamp chips from Farnell/E14 last Thursday and I'm expecting them to finally get here, as promised by UPS, this Monday ("By the end of the day"). It'll be interesting to see whether these will offer any further improvement over that seen with the lower spec THS3095s that had been used by one of the posters to this thread earlier this year.

 The problem in this case being the performance limitations of the OPA 686N chips driving them and the relay switching which bypassess the THS3002 (or 3091/3095/3491 upgrade) for frequencies above the 20MHz limit. It could well be that the crafty relay switching employed in Feeltech's design to minimise the shortcomings of the THS3002 chip will probably turn the THS3491 upgrade into an 'overkill solution'. Still "Better safe than sorry.", after all, there wasn't that big a price premium in opting for the 3491s over the slightly cheaper 3091 or 3095 chips and, who knows, I or someone else might come up with a mod to take full advantage of these opamp upgrades. 

 Regards, Johnny B Good

[Johnny B Good]

Hi, cybermaus,
The problem is mostly due to the "30021" op amp. As you say, it is only just above 7Vpp into 50 ohms. And I had tried at first powering the generator from my lab supply with +-15V. The improvement was miserable if any. So, it must be the op amp.

With the THS3095 the generator easily outputs 20Vpp (10Vpp into 50R), no problem. Still, if to be picky, the behavior of the THS3095 is in line with what the manufacturer shows in their datasheet for the component. The pictures bellow are taken from the datasheet. At high frequencies, as those graphs also show, I see some distortion (not clipping though but exactly the same as manufacturer graphs show). The manufacturer suggests properly paralleling two op-amps for each channel (second picture) to remove most of the distortion. That is what major manufacturers, like Siglent, and others do in their designs. For the sake of interest, I might have a look at that later (accidentally, I do have two more THS3095s in my drawer). By the way, Texas Instruments have recently introduced a new power opamp - THS3491. If what they say is true, that one is a real beast with unmatched performance. But they do not have them yet.

Sure, the generator steps down to 5Vpp at 20MHz. But that is to be expected. Normal behavior. Difficult to achieve frequent swings at high voltages.

 Although the subject of replacing the THS3002 dual opamp seems to be just over a year old (I've not been able to track down later opamp references), rather than directly follow up on my previous, first time post,  I thought a 'reply' to this post would be quite apposite in view of the fact that I've just replaced the stock 3002 opamp in my FY6600-60M (purchased almost three weeks ago) with a pair of THS3491s and so can verify that 'what they say' is indeed true! :-)

 I've taken the liberty of reposting Vytautas's  oscillographs showing the improved 20MHz, 20V p2p waveforms along with my own 'before and after pics' which do show a further quality improvement over the earlier 3001 and 3095 chips. Unfortunately, this modification doesn't improve the quality of square waves beyond what was available with the original dual opamp. I can only suppose this is a limitation imposed by the anti-aliasing filter.

 Regards, Johnny B Good

[DaveR]

Welcome to the 6600 Club, Johnny!  You've just reminded me that I still haven't got round to fitting the 3491s I bought for my FY6800 several weeks ago.  I'd always planned to do a side by side comparison with the 3095s I put in the FY6600, but I got sidetracked onto other projects and forgot about it.  It's now back on the list of jobs to be done!  I'd recommend doing the D75J TCXO upgrade to your 6600 as well - it gives a huge improvement to frequency accuracy and stability.

Regards,
Dave

[Johnny B Good]

Welcome to the 6600 Club, Johnny!  You've just reminded me that I still haven't got round to fitting the 3491s I bought for my FY6800 several weeks ago.  I'd always planned to do a side by side comparison with the 3095s I put in the FY6600, but I got sidetracked onto other projects and forgot about it.  It's now back on the list of jobs to be done!  I'd recommend doing the D75J TCXO upgrade to your 6600 as well - it gives a huge improvement to frequency accuracy and stability.

Regards,
Dave

 Hi Dave and thanks for that welcome.

 I've had the benefit of a lot of hindsight (over 12months of it!) by the time my interest had turned to the FY6600. :-)

 Reading all 69 pages took about 3 days on my initial full read through in order to glean as much guidance in my decision as to which cheap Chinese ARWG to blow my hard earned upon (it was quite a compelling read!). I really owe everybody who has contributed to this discussion over the past 12 months or so a debt of gratitude so, thanks to everybody for your invaluable insights and suggestions. As a matter of interest, after having gotten hold of an FY6600-60M, I reread the whole of this discussion thread again! I just wanted to clarify a few of the points made in the various upgrades I was interested in.

 I might be an old hand on Usenet but I'm a bit of a newbie when it comes to web based forums (fora). My one experience of participating in web forums was about a decade ago when I tried to stay in contact with the exiled zetnet clientèle after Breathe Networks Ltd (BNL) hoovered up the last of the decent small ISPs when the original owners were forced to sell up/their customers down the river (take your pick).

 Zetnet had set up several private news groups on their news server (this was back in the day when many such small ISPs ran their own news servers complete with a few private news groups to serve their customers needs). Zetnet had its own USP to which BNL were entirely oblivious, so much so that I got myself a whole year's worth of free service through their total and utter technical incompetence.

 Zetnet had a very loyal following in their customer base of which a large part had developed a strong sense of community through the private/closed zetnet news groups. When the end finally came, a web forum had been set up for the "Zetnutters", as we referred to ourselves, to maintain contact. The problem was, as it generally is, the rather klunky interface and the need to log on every time. Web fora are, in general, a piss poor substitute for Usenet imo, and I soon gave up on this rather klunky substitute for the old zetnet news groups.

 Since then, my main experience of web forums has been in the form of anonymous guest read only sessions typically as the result of a google search for answers to various technical issues, where, as a silent witness, I''d often find myself screaming inside my mind at the abysmal lack of expertise and ignorance being demonstrated before my very eyes. Sometimes I'd find inspiration for solving a problem but most times, I'd just find the experience an exercise in frustration and move on.

 Even where I feel I could join in simply to straighten out a few misconceptions, there was very little point since the thread I'd tracked down through a search had run its course, often several years earlier, leaving me with a sense of frustration that I've since come to terms with as the unfortunate side effect of such guest visits.

 Here, in this case, the situation is very different. For one thing it wasn't the result of a search engine's attempt to direct me to a particular topic - it was a side note in a usenet posting in SED about the remarkably cheap price point of modern Chinese AWG signal generators where one of the contributors (possibly the OP himself) had posted a link to this discussion thread. The other point being that despite this thread's remarkable age, it was quite obviously still ongoing and a quick skim had revealed a very high signal to noise ratio with a high quality of technical competence which, for me at least, is an extremely unusual experience when it comes to web forums.  Anyhow, here I am, actually participating (in my own gauche way) in a web forum! :-)

 This leads me to explain that I'm currently seeing an error message (which I suspect is no longer actually relevant - I hope!) in this reply form, stating:-

 "The following error or errors occurred while posting this message:
The message body was left empty."

 as a result of my wanting to cancel my reply so that I could go back to click on the "Insert Quote" button by way of starting over again. Having opened the reply page, I could see no way of adding 'quotage' (hence this attempt to start over) other than to ignore the 'error message' and go back to your post and click the "Insert Quote" button in the hope that it would clear the error message - which it hasn't. I figure this is a a basic bug or 'feature' of the data entry input system that's handling the 'reply to function'. It's probably waiting on my hitting the 'send button' to recheck for content before clearing the error message. [Actually, all it needed was for me to press the preview button after having written something for it to display. :-)]

 The point I'm making is that you might be (actually, you are!) reading a copy of what I've been typing into the message reply box (I'm going to copy and paste my efforts into a backup text file just in case :-) ). Also, of course, in my haste to make my first posting, I'd been anxious to get something written up to test my understanding of the reply mechanism, hence the terse nature of my introduction to the group.

 Anyhow,  now that I've explained my distracted first posts and sorted out the distracting error message in this attempt to reply, I can provide a little more info about my opamp upgrade. Having managed to cobble a special de-soldering bit (short length of 2.5mm solid copper wrapped around the heating element of my 25W Antex) to retrieve the original 3002 chip non-destructively (I did manage to slightly dislodge a tiny smd resistor in the removal process but I was able to  to resolder it back in place with the aid of a tooth pick and my 18W Antex). I was able to solder the 3491s onto the U21/U22 locations with some difficulty (and trepidation) using my smaller 18W Antex.

 I was aware of the issue with the extra Vref and /PD pins on the 3095 and the 3491 chips occupying isolated pads on the board. Not a problem for the 3091 but a problem certainly for the 3491 (can't recall if it was the same for the earlier 3095 chip) so I had a very careful read of the data sheet to confirm that whilst the /PD could be left floating, the Vref pin couldn't - it must be connected to ground or -Vss - I chose to strap it to the -Vss pin.

 The 3491 has the added complication of the thermal pad on its underside. I was considering cutting out a tiny piece of thermally conductive silicone rubber to act as a 'gasket' but decided this was an unnecessary complication since it's electrically isolated anyway (although there is a circuit track or two in the landing zone that would only have the solder mask to electrically insulate them from the thermal pad - not a problem once the chip is firmly soldered into place) so I decided the best compromise was simply to place a very thin coat of heatsink compound in the landing zone between the solder pads. I also dabbed tiny blobs of heatsink paste on the opamps and the third chip covered by the heatsink prior to reassembling it back to the board. It's worth noting that the heatsink just about fully covers the U21 and U22 chip locations rather than just 60% of the original dual opamp chip - just as well since at full chat (20Vp2p 50 ohm load at 20MHz) the total mains input wattage has gone up from 8.8W to 9.9W measured with my trusty Metrawatt analogue watt meter.

 I've more or less had the FY6600 running flat out since I completed the mod around 3:30 Monday afternoon which, as I type this, represents a 12 hour 'soak test'. It runs a little warmer but I've seen no sign of the 3491's fabled thermal protection kicking in.

 Using a cheap IR thermometer, I got readings of 50 or so deg C off the base of the heatsink as well as the top of the Cyclone FPGA. Aluminium has a lower surface IR factor than most other materials hence my pointing the IR thermometer at the base where I'm probably picking up the PCB temperature (closed up, the base of the case feels just as warm as the two warm spots on the lid). I can touch the top of the FPGA with my finger without any discomfort but the aluminium heatsink is almost too hot to maintain finger contact with - but this is to be expected anyway, even when both items happen to be the same temperature.

 With regard to that D75J TCXO upgrade, that's also on my to do list. I've been checking out availability of 0.1ppm 50MHz TCXOs and the cheapest units on Ebay claiming 0.1ppm, are all from the same Chinese based trader, yuyan-industrial with 3 different models ranging in price from £17.90 to £22.0). The question in my mind is, "Are there such TCXOs with 0.1ppm stability and accuracy to be had from Chinese suppliers via the auspices of Ebay, or is it case of the impossible  AH claims made for Surefire 18650 Lithium cells and I simply land up with a 5 or 10 ppm "TCXO" (if I'm lucky)?".

 TBH, I don't want to rush into spending any more money just yet until that FY6600 has survived the "Infant Mortally" phase of its service lifetime. Those THS3491IDDAT chips from Farnell/Element14 have already added an extra £27.36 onto the original £75.66 I paid for the sig generator in the first place, turning it into a £103.02 sig generator (still cheap for the improved performance, mind you). A two or three orders of magnitude improvement in frequency accuracy and stability would be a nice touch, especially if it can be done using a simple drop in replacement for the existing clock generator module. Shoehorning an OCXO module into the box is something I'd prefer to avoid if possible.

 I think I should be looking for a better quality smpsu which uses the slightly more expensive type of ferrite isolating transformer with a screening foil layer to shield the secondary windings from the HV switching transients that get capacitively coupled onto the output terminals rather than rely on the EMC capacitor bodge responsible for the unconscionable levels of half mains voltage leakage current.

 The real hazard of using the EMC cap bodge lies with its potential (pun intended) to discharge a high voltage into the DUT at the moment of contact - the 'half live tingle current' is merely an annoyance otherwise. If you want to retain the option to eliminate ground loop currents by letting the generator float free of any local safety earth connection, a properly designed smpsu is the best way to go imo (even if you do intend to make this a switchable option). I suspect the existing smpsu is now running close to its maximum rated output so worth an upgrade anyway.

 Right now, I'm looking to source cost effective TCXO modules (preferably 'drop in replacement if possible) and a decent 15 to 20 watt rated three rail smpsu so I'd welcome any advice in regard of achieving those goals. :-)

Regards, Johnny B Good

[EDIT for future reference 2019-10-22]

 A lot of water has since flowed under the bridge and many discoveries from hard won experience have led to a change of mind over wasting time in a holy grail search for any type of class II smpsu that exchanges the cost of a class Y2 cap for that of the penny expense of a shielding foil in the HF ferrite stepdown/isolating transformer to eliminate this common mode interference on the outputs free of any such 'touch voltage leakage' from the mandated by EMC directive Y cap to short the high voltage switching transients back to their origin to attenuate such unwanted switching noise by some 10 to 20dB (if you're very very lucky!).

 In the end, after trying out some rather fanciful 'nulling out' schemes, done in the name of avoiding an earth loop issue, it became all too obvious that the only practical way to "Have your cake AND eat it!" was to fit a 3 pole mains socket (C6 or C14 - I chose a C6 to minimise the "Tail Wags Dog Effect"(tm) with this extremely lightweight signal generator) purely to give access to a connection to ground to permit a 3 to 10KR 'static drain' resistor connection to the common return rail (0v and BNC shield ground returns).

 On a 240v mains supply, this knocks the 80 to 90vac leakage voltage seen with a typical DMM down to just half a volt rms (and less with a 3K resistor). There is absolutely no need for a hard earth connection in this class II mains powered device and the 10 to 3KR 'drain resistor nicely drops the Y cap leakage voltage to well below the safe limits for even the most delicate of electronic components you might otherwise damage with a 170 to 340vdc transient discharge from that accursed Y cap at the moment of contact at the worst possible point in a cycle of 50 or 60Hz mains voltage.

 Assuming a BNC cable shield resistance as high as one ohm between the signal generator and the DUT's ground reference points, a 10KR will give you some 60dB and a 3KR some 50dB reduction in unwanted earth loop interference from noise sources in the mains earth wiring (neutral to ground potential difference voltages due to volt drops from the total house load current on the incoming supply, including possible thermocouple effects from hot junctions and the more likely galvanic dc sources of interference through electrolytic corrosion from damp ingress which could easily compete with any dialled in DC offsets in the tens of millivolts range).

 Also, it's worth pointing out that the 'static drain' resistor will also attenuate any common mode switching transients being conducting back into the mains safety earth wiring, reducing radiated interference from the HF switching transients leaking onto the output ground return via the smpsu's transformer's inter-winding capacitance.

 If you want to avoid replacing the existing smpsu board with an analogue bodge psu (for the time being at any rate whilst you choose the most optimum analogue based design), I suggest you upgrade the output rectifier diodes to proper Schottky types, double up the 220μF 16v caps on the +/-12v rails to 470μF 16v 105 deg C types (voltage rating _isn't_ the issue here) and wind a single turn buck winding on the transformer to buck the 5v winding voltage (half a dozen single turns of thin insulated wire in parallel to give better coupling and minimal loss) which neatly forces the +/-12 rails to something like 13.5v each on load without resorting to increasing the 5 volt rail to 5.5v as per the earlier trick to get a decent voltage level on the 12v rails. You may have to reverse the connection between the lifted out cathode of the 5v diode and the pcb hole it was disconnected from.

 The details were posted here, with pictures on page 74 of an 86 page (at the time of this edit) topic thread:

https://www.eevblog.com/forum/testgear/feeltech-fy6600-60mhz-2-ch-vco-function-arbitrary-waveform-signal-generator/msg2310768/#msg2310768

 I'm still using the original (if many times modified smpsu board) at this time but I do have plans to replace it with an analogue hybrid (classic 3 terminal voltage regulators for the +/-13.5v rails powered from 15 or 16vac secondary winding(s) for lowest ripple noise on the analogue voltage rails with a 5v 1.5A rated output dc-dc converter connected to any ac winding voltage in the range 7 to 18 volts (to give me more transformer secondary options - ideally a separate 8vac winding to maximise isolation of the digital and analogue rails and optimal voltage choice for use of a 7805 regulator if deemed beneficial in spite of the modest additional heat load this would introduce).

 However, I do want to test the benefit of completely eliminating the switching noise of the existing psu by putting a recently acquired pair of 6v carbon zinc lantern batteries in series for the -12v rail and a 12AH SLA to power the +12v rail and a 7805 regulator for the 5v rail to provide noiseless sources of dc power to allow me to determine the potential benefit of an analogue psu upgrade. It's now just a matter of finding a Round Tuit to run such a test setup.

[DaveR]

Hi Johnny,

That's quite a tome you've just sent: forgive me for not replying in detail, but it sounds like you're an old-time tinkerer like myself.  I started off on various bulletin boards in the very early days of the internet, before the likes of Compuserve and AOL appeared, and well before the WWWeb took off.  Usenet served me well for years, but with everything being web-based now, you just have to accept the quirks and foibles that come with it.  You soon grow into it!

Back to business, no-one has come up with anything better than the D75J TCXO so far.  It's cheap, but the only place you can get one is Digikey (order in the UK, get it in a couple of days from the US, but postage doubles the cost unless your order is over £40), and it is a drop in replacement - but you'll need a hot air gun to remove the old one unless you can rig up a piece of wire on your soldering iron to melt the four contacts at once.  I remember looking at those Chinese TCXOs a few months ago, but they are rather expensive to take a chance on, and they'll probably need mounting on stilts, or a daughterboard, as there'll be no room on the PCB for them.

On the subject of soldering irons, my 18W Antex bit the dust about 5 years ago and I had to continue with the 30W Antex I bought in the 1980s, using a bit that was too large for most modern needs.  A couple of years ago I switched to Hakko T12 stuff - it was like driving a new Jag after getting out of a Ford Anglia.  Even the Chinese clones are good, and the right tools make modding a breeze!

I need to have a read of the 3491 datasheet again, going by what you say, as I was sure they were a simple direct replacement for the 3091/3095 even with the floating pins.

Regards,
Dave

[DC1MC]

BTW, I will do a Digikey Germany order by the end of the week, if someone needs a TCXO or anything else, he can PM me.

 Cheers,
 DC1MC

[bugi]

...but you'll need a hot air gun to remove the old one unless you can rig up a piece of wire on your soldering iron to melt the four contacts at once.

Commenting on the basis of only looking at tear-down photos of it, I have to guess which components is the osc. But if I guessed right, it seems to have 4 exposed pins/pads. In such case, the Chip Quik removal alloy might also work. I've used it couple times, and it worked really nicely, though I have only used it up to an SO-8 package so far. It worked so nicely with the small "kit" I bought first, that I bought a bigger bunch of it soon after.

[DaveR]

Yes, bugi, you're right - it would be a good alternative in this case.  After watching a video of it in use a couple of weeks ago I almost bought some myself, until I saw the price for a couple of strands of the stuff .  I suppose the hot air gun has the advantage that you don't have all the cleaning up and re-tinning to do after the chip removal, and you can use it to attach the new chip as well in most cases - the D75J possibly being the exception as the contacts on its underside don't quite match the pads on the PCB.

Regards,
Dave

[bugi]

The cleaning afterwards is quite easy, at least if the area is not tight, but it still needs to be done. Price is indeed high, so one needs to consider the price of the device being "fixed" (or modded) and prices of other solutions. However, I think in this case the device is valuable enough that using a bit of that alloy would be justifiable.

If one already has a nice hot air station, then that would likely be simpler solution.

[Johnny B Good]

Hi Johnny,

That's quite a tome you've just sent: forgive me for not replying in detail, but it sounds like you're an old-time tinkerer like myself.  I started off on various bulletin boards in the very early days of the internet, before the likes of Compuserve and AOL appeared, and well before the WWWeb took off.  Usenet served me well for years, but with everything being web-based now, you just have to accept the quirks and foibles that come with it.  You soon grow into it!

Back to business, no-one has come up with anything better than the D75J TCXO so far.  It's cheap, but the only place you can get one is Digikey (order in the UK, get it in a couple of days from the US, but postage doubles the cost unless your order is over £40), and it is a drop in replacement - but you'll need a hot air gun to remove the old one unless you can rig up a piece of wire on your soldering iron to melt the four contacts at once.  I remember looking at those Chinese TCXOs a few months ago, but they are rather expensive to take a chance on, and they'll probably need mounting on stilts, or a daughterboard, as there'll be no room on the PCB for them.

On the subject of soldering irons, my 18W Antex bit the dust about 5 years ago and I had to continue with the 30W Antex I bought in the 1980s, using a bit that was too large for most modern needs.  A couple of years ago I switched to Hakko T12 stuff - it was like driving a new Jag after getting out of a Ford Anglia.  Even the Chinese clones are good, and the right tools make modding a breeze!

I need to have a read of the 3491 datasheet again, going by what you say, as I was sure they were a simple direct replacement for the 3091/3095 even with the floating pins.

Regards,
Dave

 Hello Dave,

 I didn't get started with the internet (and usenet) until around 1996 when I joined up via a dial up service to a small, highly recommended ISP (based in, of all places, the Shetland Islands!) known as zetnet. For the first 4 or 5 years, I was totally reliant on their dial up PoP until my teenage daughter paid NTL to install their cable TV service with 512Kbps broadband at the behest of her boyfriend / fiancée (now our SiL) who was/is a web developer with some expertise in Linux. I assembled a lowest power consumption PC and had him install a Linux based router/firewall/gateway OS so the BB could be shared to the Cheapernet(tm) cabled house LAN for all to access with a tenner a month towards their rental costs as the inducement.

 This was just about the time when zetnet's resident wizard (Tim Cole, RIP) had added additional connectivity to the custom email/usenet client, ZIMACS to allow authenticated connections via other dial up access points provided by the likes of Freeserve et al, allowing me to connect via the NTL BB cable connection, neatly eliminating the cost of a dial up connection (zetnet had started offering ADSL packages by then to their customer base). I learned soon enough of the crapiness of ADSL compared to the real deal of a cabled broadband connection from all of the ADSL connection woes being endlesly reported in the zetnet groups for the next 5 or 6 years).

 When my daughter left home, I took over the NTL account just at the time when they first offered broadband internet connectivity as a stand alone service sans what had been a mandatory pre-requisite of TV and broadband. Having discovered I was paying an extra 20 quid a month for an unnecessary TV and BB service, I ordered a cessation of the TV package and transfer to BB only (at the minimum speed of 128Kbps 15 quid a month service).

 Unfortunately, their 'engineers' (two of them) turned up whilst I was in hospital as a result of an emergency admission and simply recovered the Pace STB without replacing it with the Ambit modem to effect the change-over to BB only. So through no fault of my own, when I finally got home after a fortnight or so's stay in hospital, I found myself without a connection and had to wait another month for NTL to schedule another visit to correct their cockup. However, I did receive compensation for not only the month or so I'd been without service but also another few months' worth for free. If I had been on hand at the time of the first 'engineers' visit, I would have stopped them taking the Pace box away until they'd returned with the required Ambit modem to complete the task as per my request.

 Other than that major cockup of a simple change of service request, NTL (now Virgen Media) have provided a very reliable service that puts to shame the service typically experience by many relying on ADSL over their phone lines. Over the years, what had started out as the slowest basic broadband service (initially 128Kbps for 15 quid a month until they decided to bring it in line with the 600Kbps upgrade of their next higher level of 512Kbps service by increasing it to 150Kbps as a ruse to charge an inordinate 3 quid increase) has enjoyed "free upgrades" which have, so far, culminated in a speed boost to 80Mbps down, 5Mbps up. Mind you, inflation has just about doubled the monthly rental to £37.74. Still, I suppose a connection speed increase of 640 times for a mere doubling of cost is hardly a basis of complaint. 

 Mind you, I wasn't entirely new to the concept of Usenet since I'd been making use of the radioham BBS packet radio service since the mid 80s prior to gaining direct access to the internet. The transition to Usenet was a natural upgrade on the radioham BBS packet radio based setup. It's these web based forums that seem so alien and klunky by comparison. 

 Anyway, back to the topic. I took note of the D75J TCXO reference and googled it (I wasn't sure whether it was an actual part number or named after its proponent (like the classic G5RV antenna)). I see that it's merely a 1ppm part rather than the 0.1ppm ROJON TCXOs being offered by Chinese only sources (the cheapest option yet is a small PCB module containing a TCXO for approximately £15.59 now (yesterday, it had been approximately £15.42) available here: <https://www.ebay.co.uk/itm/1pcs-50MHZ-hifi-TCXO-0-1ppm-Ultra-precision-Golden-Oscillator-CLOCK-power-supply/132652757688?hash=item1ee2b872b8:g:zFoAAOSwlgNbGBqS:rk:5:pf:0> or https://tinyurl.com/y8he5bqa).

 If it wasn't for the fact that I'd like to have a decent very low leakage high efficiency smpsu based PSU with a little more margin in hand, I'd have ordered it there and then. However, one step at a time - I don't want to find that by the time I'd received the TXCO, mid January 2019 (knowing my luck), my FY6600 had suffered a horrible infant mortality failure no longer insured by its warranty due to my modifications to date. An ultra low leakage three rail class II smpsu can always be found a use elsewhere, the 1 or 0.1 ppm 50MHz TCXO, otoh, is a rather more niche item to repurpose.

 This leads me to the problem of tracking down an affordable 3 rail ultra low leakage class II smpsu PCB module. Try as I did last night - the reason I'm only replying just now - I couldn't find anything remotely suitable. Such low leakage class II smpsus are possible - I've actually got one in the form of a Nokia 5.7v 800mA "For LTE use only" phone charger 110 to 240vac wallwart which only shows as 12.5vac leakage voltage to ground with my DMM (presumably 10 or 11Mohm impedance) versus the more typical 75-79vac seen with the Feeltech and almost every other wallwart smpsu I have to hand. Better yet, its no load consumption is circa 50mW (barely perceptible when viewing the mirror backed 100 watt scale of my Metrawatt analogue wattmeter with a magnifying glass) which bodes well for its efficiency.

 It looks like my best bet is to track down a low leakage class II 90 to 265vac 5v 3 or 4 amp smpsu and hang a couple of dc-dc 5v to 15v (one plus, the other minus) boost converter modules onto its output. Low leakage 3 rail PSUs seem to be about as rare as unicorn droppings. 

  With regard to my soldering kit, I'll just carry on with my existing 18 and 24 watt Antex irons and copper wire adaptor bodges for now. I'll consider soldering tool upgrades when (and if ever) I get a decent workshop sorted out (we're considering moving from this large 6 bedroom, with basement, Victorian semi-detached house, to a decent sized 3 bed bungalow with an outhouse or two (or space to build a decent workshop)) so I'm in a bit of a Limbo situation right now.

 Regarding the need to tie the Vref to ground or -Vs on the THS3095, I think the advice (emphasised so clearly in the THS3491 data sheet but overlooked in the THS3091/3095 data sheet) to NOT let the Vref float still applies. If you read the description on page 26 of the 3095's data sheet, you'll see the need to tie the Vref to ground or -Vs rail since this would otherwise float to the V+ rail, some 4 volts above its recommended voltage level. The /PD can be left floating since this also swings to the Vs+ rail. The /PD pin has to be 2v or more positive than the Vref pin in order to guarantee enabling of the amp. This condition is clearly impossible when both pins float to Vs+, hence (in the 3491 data sheet) the exhortation to NOT let the Vref float free.

 The residual distortion shown in the DSO trace of the THS3095's output waveform at 20v P-P 20MHz may simply be a symptom of allowing the Vref pins to float. I chose to tie the Vref to Vs- rather than ground since it saved scraping through the solder resist to solder onto a groundplane and a handy VS- connection lay only three pins away from the pin 1 Vref on pin4 of the chip (also if I must rely on just letting the /PD pin float in a vulnerable  hiZ state to the VS+ rail, the extra voltage margin wouldn't hurt).

 If I were you, I'd be inclined to strap pins 1 and 4 on each of those 3095s and repeat the waveform tests at 20Mz 20v P-P both in HiZ and 50 ohm terminated conditions to compare against the earlier tests. You might not need to upgrade to the 3491s (unless you're planning on doing a mod that over-rides the crafty relay switching designed to mask the 3002's shortcomings).

For anyone who might be contemplating some of the modifications that have successfully been applied by members of this discussion thread, It seems to me that the best order to follow is firstly to modify the existing PSU to improve its efficiency on the +/- 12v rails by replacing the shoddy RF1007 diodes with proper 20 or 30 amp 40v PIVr rated dual shottky TO220 rectifiers and  fit a 100K or 47K resistor across the lower arm of the 10K+10K divider network to raise the 5v line from 4.95 to 5.25 or 5.49 volts to take the 12v rails another half to one volt higher again.

 It's a simple and effective mod but it doesn't address the HiZ half mains voltage leakage issue. The solution to that problem needs a better class of class II PSU, analogue if you must but preferably a good quality smpsu which uses an HF ferrite transformer with the required shielding foil layer between the high voltage primary and the low voltage secondary windings to obviate the need to bodge an EMC countermeasure in the form of that 1nF class Y cap.

 Secondly, upgrade the existing dual opamp setup to a pair of single opamps using either THS 3091, 3095 or 3491 SOT8 chips, remembering to strap pins 1 and 4 together in the case of the 3095 and 3491 chips.

 Thirdly, if you feel the need to match the existing 50MHz 100ppm frequency reference to the counter's display resolution, then install a 50MHz 0.1ppm TCXO (if you're going to bother at all, why only take half measures with a mere 1ppm rated TCXO when you can get hold of a 0.1ppm ROJON part from a Chinese supplier on Ebay for as little as 15 or 16 quid?  ).

 Fourthly, if the issue of mains leakage still troubles you, replace the PSU board with a class II unit (analogue or smpsu) that only has a few pF's worth of half mains voltage leakage (mains transformer 50/60Hz leakage of analogue PSUs or ditto for the foil layer screen connection via a 10nF capacitor to the half mains voltage virtual ground in a high quality ferrite transformer used in properly designed smpsus free of the need to include that abomination of the classic 1nF EMC bodge capacitor).

 The classic analogue mains PSU has the charm that such low leakage is inherent without having to specify it explicitly whereas the smpsu needs to have this requirement clearly spelled out in its specifications - good luck in finding such smpsus by design. 

 [EDIT 2021-11-25] Please disregard my rather pompous advice above and take note of the edit below. 

Regards, Johnny B Good

[EDIT 20190425]  I eventually changed my mind about replacing the PS board with a medical grade low leakage smpsu just to avoid having to upgrade to a 3 pin socket and thicker, "Tail wagging dog" mains lead some two or three months later.

 I eventually realised that even with such low leakage, there'd still be enough residual leakage voltage with a potential to destroy the more electrically fragile DUTs I might land up testing. In the end, the pragmatic solution also proved to be the most optimal. However, I used an 11KR resistor (three 33KR resistor in parallel) as a low impedance static drain which knocks the 90vac down to a mere 500mV without introducing earth loop effects which the 100nF capacitor option used by others would introduce at high frequencies.

 All the PE connection was doing was to provide a convenient way to automatically connect an 11KR static drain between the zero volt reference and ground. The PS board is still a class II double insulated smpsu so a 'hard' protective ground connection is not needed (nor desired). Leakage voltage risk suppressed without the introduction of a troublesome grounding loop as per Feeltech's bodgery with the FY6800.

 The details of this are in my later posts - I'm mentioning it here for anyone who may be plodding their way through this thread for the very first time. :-)

[soundtec]

Hi Johny ,welcome.

Wow a three day read through , and then a re read .

I probably read each post at least twice by now , but in small chunks ,

I have both 66 and 68 hunderd , the 68 was a won in a competition held by Feeltech itself , there was several other people very worthy of a free machine for their contributions ,but if your not in you cant win .

Ive recently got myself a pair of Hewlett Packard HP200cd tube sine oscillators , in the 50's these were state of the art and in fact the product line lasted well into the mid 80's tubes and all , they did drop the tube rectifier in the later models though ,about .5 % thd was as good as they do normally , now with the Feeltech I can easily get very very low distortion audio frequency sine waves ,the key to this is to use an external attenuator and run the unit full throttle ,thus minimising any digital artifacts and using up the full bit rate of the unit . Theres no doubt the original power supply would be marginal with higher powered op amps and under volted too .Any offset dialed in at 20v pk-pk instantly clips the outputs , theres no margin .
I went linear ,with a transformer and regulators ,just like a conventional op amp audio set up .
I did fool around with a few other switchers as well , I tried to run the original psu with the 5 volt unloaded but the  +/- rails dropped way low .
Deriving the three voltages from one switch  psu I dont like , you'll find that under load the original psu droops under high voltage swing on the +/- rails , thats no good at all for fidelity. There have been a handfull hoping to make further improvements , including the custom software from Fremens, he ironed out some of the waveform compatibility issues for us in the Feeltech package , and of course allowed further tweaks if you used a programed STM32 microcontroller . I guess everyone's just settled into using their units now and are happy with the improvements they made , I would seriously think about a larger heatsink for the newer TI op amps though maybe even extra wings of aluminium extrusion screwed into the original .

[Johnny B Good]

Hi Johny ,welcome.

Wow a three day read through , and then a re read .

I probably read each post at least twice by now , but in small chunks ,

I have both 66 and 68 hunderd , the 68 was a won in a competition held by Feeltech itself , there was several other people very worthy of a free machine for their contributions ,but if your not in you cant win .

Ive recently got myself a pair of Hewlett Packard HP200cd tube sine oscillators , in the 50's these were state of the art and in fact the product line lasted well into the mid 80's tubes and all , they did drop the tube rectifier in the later models though ,about .5 % thd was as good as they do normally , now with the Feeltech I can easily get very very low distortion audio frequency sine waves ,the key to this is to use an external attenuator and run the unit full throttle ,thus minimising any digital artifacts and using up the full bit rate of the unit . Theres no doubt the original power supply would be marginal with higher powered op amps and under volted too .Any offset dialed in at 20v pk-pk instantly clips the outputs , theres no margin .
I went linear ,with a transformer and regulators ,just like a conventional op amp audio set up .
I did fool around with a few other switchers as well , I tried to run the original psu with the 5 volt unloaded but the  +/- rails dropped way low .
Deriving the three voltages from one switch  psu I dont like , you'll find that under load the original psu droops under high voltage swing on the +/- rails , thats no good at all for fidelity. There have been a handfull hoping to make further improvements , including the custom software from Fremens, he ironed out some of the waveform compatibility issues for us in the Feeltech package , and of course allowed further tweaks if you used a programed STM32 microcontroller . I guess everyone's just settled into using their units now and are happy with the improvements they made , I would seriously think about a larger heatsink for the newer TI op amps though maybe even extra wings of aluminium extrusion screwed into the original .

 Thanks for persevering. 

 I had a lot of material to work with (69 pages!) and a lot of points I wanted to address. Sorry for unloading it all in two large lumps, so to speak. Hopefully, I can dial my post lengths back to more manageable portions now that I've got most of the comments and observations I've had bottled up over the past couple of weeks since I read through the whole thread off my chest.

 Congratulations on winning the 6800 btw. I guess that makes you our "Go to Guy" for comparing these two Feeltech products.    It's an interesting point you've made about the HP200CD only managing a half percent THD. Considering the revived interest in all things tubed (valved) by the HiFi fraternity, that must come as a bit of a disappointment for the more extreme proponents of tubed kit. TBH, it's a bit of surprise to me that valved test gear couldn't do much better.

 Understand about the point of setting the output level to the 'sweet spot' and applying attenuation to get the required level (and minimise amplifier noise into the bargain as well). As you (and everyone else) have observed, the PSU in the FY6600 is barely up to the job even with the modifications suggested here (does the FY6800 suffer the same deficiency I wonder?). Reducing the loading on the +5v doesn't do the +/-12 volt rails any favours.

 It's an old problem with multi-rail power supplies. The original PC and XT then the AT and now the ATX psus suffer exactly the same effect but in the case of desktop PC supplies, the variation on the 12 rails was never an issue, they just had to remain within the wider +/-10% tolerance that was allowed on those rails. In this case, it's the +/-12 rails that are the more critical rather than the 5v which seems to be used to power the 3.3v and 2.5v (and 1.5v?) regulators on the main board. Hopefully, the 3.3v regulator is a low dropout version (I haven't checked but it doesn't matter so much to me since I raised the 5v to 5.49 volts purely to increase the 12v rail voltages).

 Putting my FY6600 into standby mode only saves just under 4W - it drops from a high of a gnat's dick shy of 10W to similarly just over the 6W mark. I'd say at least half the power supplied by its tiny PSU must be via the 12v rails when it's driving 50 ohm loads at the 20v p-p limit. Like yourself, I did observe quite bit of sag on the 12v rails between the zero and the maximum output levels. It could certainly do with a better quality PSU, preferably a higher efficiency smpsu to minimise the heat load. Small smpsus like the one used here by Feeltech, typically have an efficiency of 80% at full output so I figure the PSU is contributing a bad two watts to the heat loading inside that not very well vented case.

 With only half a watt total being accounted for in the 50 ohm dummy loads, that leaves a balance of a further 7.5 watts of waste heat being generated by the front panel and main board. Going by the noticeably hotter feel of the underside of the case compared to the two 'Hot Spots' (fsvo 'Hot') on the topside, I'd guess a good (or bad if you prefer) 6 watts or so must be coming from the main board. I think drilling holes in the underside of the case below the main board would be the best place to make a start if you want to improve the ventilation.

 From my experience of running a 24 hour heat soak test, drilling vent holes in the underside of the case will likely be enough to save the opamps from cooking themselves to death. Allowing some airflow round the underside of the main board will likely reduce the board and the heatsink temperatures by a noticeable amount. You might not need to add any extra metal to the existing heatsink.

 BTW, from what further research I've managed to do into this question of mains leakage currents in class II smpsus, it seems the problem isn't an easy one to fix without increasing the level of common mode conducted switching noise on the output wiring. One treatise (see attached pdf) did note that the use of a screening foil layer between the primary and the secondary windings was the most cost effective way to deal with this 'conflict of interest' between mains leakage and EMC requirements.

 The Y cap is simply a means to press the secondary into service as a shield against the high frequency switching transient voltage capacitively coupling into the output as a common mode conducted interference signal. Unfortunately, since the 'ground reference' is tied to a source of half mains voltage, the half to 2 nF class Y cap also provides a Hi Z leakage path for this voltage to manifest itself on the output terminals. Using a separate shielding foil between the primary and secondary windings as a dedicated screen where the capacitive coupling to the secondary windings is reduced to a mere 40 or 50 pF instead of the 500 to 2200 pF of a Y cap reduces the half mains voltage coupling considerably without compromising the EMC requirements.

 As I see it now, I think my best approach may be to get hold of a low leakage class II 3 or 4 amp rated 5v smpsu and add a couple of DC-DC converters to generate the 14 or 15 volt rails. However, I'm still mulling over alternative solutions to this conflicting requirement of EMC and low mains voltage leakage.

 Regards, Johnny B Good

[DaveR]

Hi Johnny,

Have you seen these SMPSUs?
https://www.ebay.co.uk/itm/SMPSU-Switched-Mode-Chassis-Power-Supply-12v-0-3A-12v-0-2A-5v-2A-SMPS1-432-/141117111577?var=&hash=item667e84f884
https://www.ebay.co.uk/itm/SMPSU-Switched-Mode-Chassis-Power-Supply-5v-2A-15v-300mA-15v-200mA-Toko-783-/361575419603?hash=item542f930ed3

They look the part, if nothing else, but the seller is an anus - so caveat emptor!!  See the negative feedback on him (which I can corroborate from personal experience.)

That TCXO module you found must be a new one, and it's cheaper than the oscillator by itself from other suppliers.  There's also a Rojon mini TCXO for £22 if you want to be really extravagant:
https://www.ebay.co.uk/itm/Mini-TCXO-50-000MHz-50MHZ-0-1ppm-Ultra-precision-Oscillator-audio-DIY-GQ1-XH/382556866867?hash=item59122a8933:g:nIwAAOSwA3dYL7Cj:rk:3:pf:0

I can't see a problem with leaving pins 1 and 4 floating in the case of the 3095: both pins are unconnected on the PCB, so both will float to Vs+ and the device will be permanently enabled (no differential just means no power down control, but the default state is enabled, as would be expected).  In the case of the 3491, although the datasheet doesn't recommend leaving PD floating, the device will still be permanently enabled if it is (subject to a probably unlikely stray signal event ).  I wasn't intending to upgrade the 3095s in my FY6600 to 3491s - I was planning to install them in my FY6800 (when I get round to it!) to get a direct comparison between them and the 3095s, more for curiosity and experimentation than anything else.  If there happens to be a problem with them I've got spare 3095s I can drop back to as the upgrade from the single 3022.

Regards,
Dave

[Johnny B Good]

Hi Johnny,

Have you seen these SMPSUs?
https://www.ebay.co.uk/itm/SMPSU-Switched-Mode-Chassis-Power-Supply-12v-0-3A-12v-0-2A-5v-2A-SMPS1-432-/141117111577?var=&hash=item667e84f884
https://www.ebay.co.uk/itm/SMPSU-Switched-Mode-Chassis-Power-Supply-5v-2A-15v-300mA-15v-200mA-Toko-783-/361575419603?hash=item542f930ed3

They look the part, if nothing else, but the seller is an anus - so caveat emptor!!  See the negative feedback on him (which I can corroborate from personal experience.)

That TCXO module you found must be a new one, and it's cheaper than the oscillator by itself from other suppliers.  There's also a Rojon mini TCXO for £22 if you want to be really extravagant:
https://www.ebay.co.uk/itm/Mini-TCXO-50-000MHz-50MHZ-0-1ppm-Ultra-precision-Oscillator-audio-DIY-GQ1-XH/382556866867?hash=item59122a8933:g:nIwAAOSwA3dYL7Cj:rk:3:pf:0

I can't see a problem with leaving pins 1 and 4 floating in the case of the 3095: both pins are unconnected on the PCB, so both will float to Vs+ and the device will be permanently enabled (no differential just means no power down control, but the default state is enabled, as would be expected).  In the case of the 3491, although the datasheet doesn't recommend leaving PD floating, the device will still be permanently enabled if it is (subject to a probably unlikely stray signal event ).  I wasn't intending to upgrade the 3095s in my FY6600 to 3491s - I was planning to install them in my FY6800 (when I get round to it!) to get a direct comparison between them and the 3095s, more for curiosity and experimentation than anything else.  If there happens to be a problem with them I've got spare 3095s I can drop back to as the upgrade from the single 3022.

Regards,
Dave

 Thanks for the links to those PSUs,

 I've just looked them over. Unfortunately, the first thing that strikes me is the marginal current rating of the 12 and 15 volt rails (0.3 and 0.2 amps) which I suspect won't be enough to support the demand from the opamps.

 I noticed the use of a couple of HT smoothing caps (presumably 200 or 250 volt rated ones wired in series) with a flylead jumper on the 12v version for the 120/240 volt mains voltage selector which is a bit of a spoiler (meaning user intervention is required when moving between mains voltage standards - no automatic handling of mains voltage adjustment).

 The 15 volt version is using a single 400v rated 68μF smoothing cap suggesting that it's probably a "Universal" smpsu type. That doesn't preclude the possibility that it may be 200-240 volt only of course, but that seems rather improbable given the extra cost over the optimum choice of 22μF since a 68μF would provide ample smoothing even for a 60W 100v 60Hz supply working off the apparent design rule of a microFarad per watt's worth of loading on a 100v 60Hz supply.

 TBH, neither of them are, imo, suited for the task. I think the best option in this case (looking for a psu system with the lowest possible level of half mains voltage leakage current with good voltage regulation on all three rails) is to find a 15 or 20 W rated 5v low leakage smpsu and use dc-dc converter modules to derive the two independently regulated 15 volt rails rated for 500mA each. This way, you only need to track down one low leakage psu and not have to put up with the indifferent regulation typical of secondary 12/15 volt rails loosely tied to a precisely regulated 5 volt primary output rail courtesy of secondary turns ratio alone.

 I started writing this reply about 12 hours ago but "Real Life"(tm) got in the way and I was still distracted by the, now seemingly impossible task of tracking down a suitable 15 to 25 watt rated Class II 5v smpsu containing a shielded transformer and no frelling Y1 class capacitor (as per that Nokia 5.7v 800mA phone charger that proved to be totally devoid of such vandalism yet was virtually free of the half mains voltage leakage).

 I'm still mulling this problem over but for now, I've changed my mind and decided to upgrade the shite rectifier diode in the FY6600's psu. I've just discovered that at 100Khz 20v p2p, I can offset the voltage by +2.9 and -3.6 volts before clipping becomes evident suggesting  that I can afford the resulting drop on the 12v rails as a result of such an upgrade to the 5v rail after all!

 Regarding TI's advice in the THS3491 data sheet about not letting the Vref pin float, perhaps the absence of such explicit advice in the 3095 data sheet is because it doesn't matter after all... perhaps. I just felt it was a point worth raising considering the distortion at a mere 20MHz that was evident in your(?) 'scope traces.

 Bear in mind that the similar distortion shown in the data sheet was for 32MHz 20v p2p with 100 ohm loading. The adjacent 32MHz trace was with two of them paralleled up to demonstrate the effect of halving the load on each amp to the equivalent of a 200 ohm load per amp, the benefit of which was even more starkly demonstrated in the 64Mhz plots. I'm just sayin' is all. 

[DaveR]

 Thirdly, if you feel the need to match the existing 50MHz 100ppm frequency reference to the counter's display resolution, then install a 50MHz 0.1ppm TCXO (if you're going to bother at all, why only take half measures with a mere 1ppm rated TCXO when you can get hold of a 0.1ppm ROJON part from a Chinese supplier on Ebay for as little as 15 or 16 quid?  ).

I bought one of these not long ago:
https://www.ebay.co.uk/itm/External-TCXO-clock-PPM-0-1-for-HackRF-one-GPS-Applications-GSM-WCDMA-LTE/173432677234?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649

and have just looked up the spec in the seller's listing: I thought it was 1ppm, but it's shown as 0.1ppm in the headline, and referred to as 0.1 - 0.5ppm in the description.  It's Rojon, but only half the price of those referred to earlier.  I wonder how much confidence you can have that 0.1ppm means anything like that accuracy, or that the £15 and £22 TCXO cans contain anything different to this £8 one?  The answer, of course, is "none".  Caveat emptor, again!

[Johnny B Good]

 Thirdly, if you feel the need to match the existing 50MHz 100ppm frequency reference to the counter's display resolution, then install a 50MHz 0.1ppm TCXO (if you're going to bother at all, why only take half measures with a mere 1ppm rated TCXO when you can get hold of a 0.1ppm ROJON part from a Chinese supplier on Ebay for as little as 15 or 16 quid?  ).

I bought one of these not long ago:
https://www.ebay.co.uk/itm/External-TCXO-clock-PPM-0-1-for-HackRF-one-GPS-Applications-GSM-WCDMA-LTE/173432677234?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649

and have just looked up the spec in the seller's listing: I thought it was 1ppm, but it's shown as 0.1ppm in the headline, and referred to as 0.1 - 0.5ppm in the description.  It's Rojon, but only half the price of those referred to earlier.  I wonder how much confidence you can have that 0.1ppm means anything like that accuracy, or that the £15 and £22 TCXO cans contain anything different to this £8 one?  The answer, of course, is "none".  Caveat emptor, again!

 Hi Dave,

 I had I a reply nicely typed up and waiting to go until I decided to hit the refresh whilst previewing my missive. Apparently, hitting the refresh button is a shortcut to sending replies to the bit bucket.  Whoops! I did it again (but not without copying  my text beforehand). There's no doubt about it, hitting refresh is a sure fire way to lose a nicely typed out reply... 

 Apologies but you're gonna have to wait a little longer for a proper reply. I'm done with this klunky bug ridden system for today. I'll leave it till tomorrow (er, later today - it's already "tomorrow").

 Regards, Johnny

[soundtec]

Hi Johnny ,
Its does happen on many Blogs ,Ive found ,long and carefully typed responses can simply end up gone .
A very simple way to help prevent loss is to type into a text file on the desktop ,then once your done ,copy and paste to the Blog page . Alternatively before hitting send , copy and paste what you've written to a text file and save .

[Johnny B Good]

 

Hi Johnny ,
Its does happen on many Blogs ,Ive found ,long and carefully typed responses can simply end up gone .
A very simple way to help prevent loss is to type into a text file on the desktop ,then once your done ,copy and paste to the Blog page . Alternatively before hitting send , copy and paste what you've written to a text file and save .

 Sensible advice. In this case, it's best to save the text before even taking a chance with the preview button (in my case, just in case I forget the lesson about refreshing the view being so terminal). I'm typing this into a text file rather than into the rather cramped reply box. I suspect this is probably what many of the longer established members do.

 Anyway, it was probably for the best. I was going to suggest converting the seemingly bi-phase half wave rectified 12v circuit on the PSU to full wave, forgetting that with the simpler (and cheaper if you discount the need for a doubling up of the transformer rating compared to that needed in the more efficient but complex push pull circuit) single ended drive smpsu topology, only half wave rectification can be used as I discovered to my chagrin, luckily without damage when I tested my hypothesis that this would give a small boost in voltage without the need to add a couple of turns to the 12v bi-phase secondary winding, along with a halving of the ripple voltage.

 I tested my adaption with the PSU fully connected up, voltmeter poised, eye on the wattmeter scale ready to hit the off switch should it go north of the 10W mark. Switching on showed a mere 2W reading as if the PSU had gone into current foldback due to an overload. The 5v was showing just 2.5 volts and the 12v lines, afaicr, less than 6 volts (possibly just 2 or 3 volts).

 Thinking I'd miswired the extra diodes, I removed the PSU to my 'bench' for a closer look but as far as I could tell from visual and electrical inspection, I had managed to create a full wave centre tapped bridge circuit exactly as planned so back it went into the sig generator for another test where it did the same thing.

 More curious now, I left the output disconnected to try again and observed some 21 volts on the 12v rails (the 5v was still its usual 5.48 and the wattage was still the 2 watts of earlier. Needless to say, I didn't stop and stare at this marvel of excess voltage and hit the off switch quick smart.

 What I'd seen as a rather curious abuse of a bi-phase secondary winding where the opportunity to use either a single winding to generate a centre tapped voltage doubled supply or else a missed opportunity to use the bi-phase winding to create a centre tapped full wave bridge circuit, turned out to be simply a means to allow all three rectifiers (the +5 and 12 and -12 volt rectifiers) to conduct on the same phase as the active single ended drive pulses. Lucky for me that my "Bull in a China Shop" approach hadn't resulted in a fatal mistake. :-)

 I realised my 'skoolboy howler' during my musings of the problem whilst waiting for the iron to warm up for the diode-ectomy. This led me to wonder whether replacing the extremely efficient 30A per diode element 45Vrrm rated diode on the 5v winding with a slightly less efficient SB340 (3A average 40Vrrm DO-201AD) I'd recovered from an old external HDD power supply would provide the slight lift on the 12v rails to give a little more than just a tenth of a volt's worth of postive offset range at the 20Vp2p into 50 ohm loads I'd been getting with the +12.35v/-12.7v rail voltages after the transformer mod. Since it looked like the component symbol printed on the PCB, I swapped out the giant 30&30A diode and mounted my prize diode in its stead.

 Retesting revealed a rather surprising boost of the 12v rails to +13.55 and -13.70 volts which proved sufficient to let me go +2.0v and -2.3 volts on the offset adjust without clipping. A rather ample margin for such a modest boost on the 12v rails. The maximum power consumption after the diodes have warmed to a more efficient operating temperature (takes about a minute or two from the 10.25 watt initial cold switch on state) is now just a whisker over the 9W mark.

 I reckon this final modification to the PSU is just about the optimum between its improved efficiency and the overall energy consumption, the vast majority of which results in raising the internal temperature of the generator. It can be reduced by another watt but this leaves you with no offset margin.
 
 For anyone else interested in modding the PSU for better efficiency and voltage stability on the 12v rails, it's sufficient to replace all three rectifier diodes on the outputs with SB340s all round. For those of a less nervous disposition, a nice finishing touch is to lift the transformer out to add a couple of extra turns to each end of the bi-phase secondary winding which serves the 12v rectifiers.
 
 This, of course assumes the obvious addition of the 47K resistor across the lower 10K resistor of the 5v sense network to raise the 5v (and the 12 volt rails tied to it via the transformer winding ratio) to 5.49v (still just within the +/- 10% Vcc spec for 5 volt logic chips).

[EDIT 20190330]

 I've since undone that transformer mod (removed the two turns from the 12v windings) and added a single turn over-wind on the bobbin, made up with 4 separate single turn windings in parallel for better coupling and minimum I sq R loss and leakage inductance to wire in series with the 5 volt diode to buck the transformer output voltage (effectively removing one turn from the 5v secondary winding).

 This forces the switching module to raise the 12v rails as it compensates for the effect of the lower transformer output to the 5v rectifier and smoothing circuit which drives the voltage feedback for all three rails. It's a far more elegant method to rebalance the voltage distribution between the regulated 5v rail and the "we'll take pot luck" turns ratio mediated voltage output level on the "Twelve Volt" plus and minus rails.

 Adding this single turn to buck the voltage on the existing 5v winding which lies beneath the HT windings, neatly avoids interference to any of the existing windings and their connections (very unlike the case with my first transformer mod) since it simply terminates directly onto a vacated PCB connection and the lifted out anode of the 5v rectifier diode.

 The only gotcha with this is that you may have to reverse the connections to achieve the buck winding state. When I tried this earlier today, that was exactly how it went -Sod's holds its sway as per usual. What made it a little more awkward for me was my choice of needlessly heavy gauge wire which obliged me to unsolder the transformer again in order to preform the ends after reversing the polarity in order to refit it back to the board.

 Just because there IS ample room on the bobbin, that's not a valid reason to try and fill the gap with heavy gauge wire when thinner and more flexible wire will be more than adequate to keep the I sq R losses vanishingly small anyway. It would have been trivial, given a more sensible choice of wire gauge, to reverse the buck winding connections without having to lift the transformer from off the PCB. Indeed, the whole mod can be done without removing the transformer at all if sufficiently flexible wire is used. If it hadn't been for the need to undo my earlier mod, I'd most likely have made a better choice of wire to begin with.

 Anyway, this alternative solution DOES work, and rather well at that! Before starting this modification work, I measured the voltages provided by the previous modifications (diode upgrades, the two turns add ons to the + and - "12v windings" and the 47k shunt of the voltage feedback potentiometer lower arm to raise the original 4.95v to 5.49v just to get a an extra boost on the 12v rails). The voltages were +5.46, -13.78 and +13.60 volts. After reverting the transformer to its original state and adding a single turn to buck the 5v secondary and removal of the 47K resistor, I then got the following voltages, +4.94, -12.98 and +12.83. I then fitted a 200K resistor to bump the +4.94v rail to a less marginal (for the 12v cooling fan) 5.07 volts which boosted the "twelve volt" rails to -13.29 and +13.13 volts which still provides a useful DC offset margin at the 20v p-p level.

 One final thing I can add before ending this edit, is that you can forget what I said about upgrading to 'medical grade' class II smpsu modules using transformers blessed with a proper foil shield layer as the ONLY effective solution to the problem of the half live mains 'touch voltage' and the risk of ESD damage to any fragile devices being tested.

 In the end, I realised that such ESD hazards still existed even with such medical grade' psus and after trying several schemes to cancel out this touch voltage whilst avoiding the use of an earth connection which would, as it does in the case of the FY6800, introduce troublesome mains earth loop interference, I finally admitted defeat in achieving such an impossible ideal and rather pragmatically, replace the IEC C8 connector with an IEC C6 (trefoil) connector just so I could link the zero volt rail via an 11K drain resistor to flatten down the 90vac half live mains touch voltage to a mere 500mV ac which nicely solves the ESD risk issue yet avoids the earth loop issue that stupid, stupid Feeltech have introduced into their FY6800 - Fools that they are!

[END_EDIT]


 Anyway, back to the point that Dave made regarding the TCXO upgrading options, I had been about to mention (as a result of my original transformer mod prior to my doomed bridge diode mod) that I'd changed my mind about hanging fire on ordering that £15.53 "50MHZ hifi TCXO 0.1ppm Ultra precision Golden Oscillator CLOCK power supply" with its JYEC TCXO (the object of my desires).
 
 It's a simple enough job to transplant the TXCO from the board onto the FY6600's main board to directly replace the original cheap 100ppm XO module and the little circuit board it came with won't be going to wast since I have several XO modules to hand that I can test on this board (I'll fit a set of gold turned pin sockets to the board to turn it into a plug in XO test rig).

 The order went in on Friday and I have an estimated 3 to 8 weeks delivery (12 Dec to 25 Jan). Just as well I'm not in a blazing hurry to upgrade this aspect of the FY6600-60M. :-)

 Regards, Johnny B Good
 

[Johnny B Good]

 Thirdly, if you feel the need to match the existing 50MHz 100ppm frequency reference to the counter's display resolution, then install a 50MHz 0.1ppm TCXO (if you're going to bother at all, why only take half measures with a mere 1ppm rated TCXO when you can get hold of a 0.1ppm ROJON part from a Chinese supplier on Ebay for as little as 15 or 16 quid?  ).

I bought one of these not long ago:
https://www.ebay.co.uk/itm/External-TCXO-clock-PPM-0-1-for-HackRF-one-GPS-Applications-GSM-WCDMA-LTE/173432677234?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649

and have just looked up the spec in the seller's listing: I thought it was 1ppm, but it's shown as 0.1ppm in the headline, and referred to as 0.1 - 0.5ppm in the description.  It's Rojon, but only half the price of those referred to earlier.  I wonder how much confidence you can have that 0.1ppm means anything like that accuracy, or that the £15 and £22 TCXO cans contain anything different to this £8 one?  The answer, of course, is "none".  Caveat emptor, again!

Hi Dave,

 That would require the addition of a 5 times pll multiplier since it's the all too ubiquitous 10MHz 0.1ppm TCXO that I kept bumping into in my search for a 0.1ppm 50MHz unit. That £15.53 JYEC TCXO on a handy test board is worth shelling out the extra 7 quid for imo in order to avoid the extra cost and work of adapting a 10MHz reference to the requirements of the FY6600 so, as I mentioned in my reply to Soundtec just now, I've got that module winging its way to me as I type.

 A point I was making in the original posting attempt that fell foul of the refresh button was my observing that every picture of a 0.1ppm TCXO I'd looked at showed a trimmer adjustment access hole in the can. Since I didn't observe any such hole in the cans of the 1ppm TCXOs I'd seen pictures of, I think one can safely conclude that if there is such a hole, it's got to be a 0.1ppm part.

 That discrepancy between the headline +/-0.1ppm and the wider +/-0.5ppm accuracy/stability figures is just a matter of the wider -10 to +60 deg C range compared to the less arduous room temperature range of 0 to 25 deg C.

 In the case of the part I've just ordered, I notice a degradation of temperature stability over the -10 to +60 deg C range to just +/-0.2ppm versus the more typically quoted 0.5ppm value. Hopefully, the TCXO module is actually as good as this. The internal temperature of the FY6600 is more likely to be 40 to 50 deg C rather than the 25 degrees C upper limit on its 0.1ppm rating and I can't see any easy way[1] of ensuring it stays at or just above typical room temperature.

 Since the degradation of temperature coefficient in this case would appear to be only an extra 0.1ppm rather than the more usually quoted 0.4ppm worsening, it shouldn't be too difficult to trim it to better than 0.1ppm at the elevated temperature in any case.

Regards, Johnny B Good

[NOTES]
 
[1]  I can see a hard way to achieve 'room temperature' operation for the TCXO which involves cutting a 47mm diameter hole in the base, fitting a slimline fan[2] and adding a deflector to divert the incoming fresh air straight towards the TCXO end of the main board. This would probably be a good thing to do regardless if only for the sake of everything else in the box but I suspect a tweak of the trimmer to compensate for the higher temperature will produce a result almost indistinguishable from that of the 0 to 25 deg C operational temperature range.

[2] Yep! A footnote to a footnote. 

 I just found a neat little 50mm square 10mm deep 12v 120mA cooling fan in an icecream tub marked "Fan Cooler Bits" that's been sat on a cupboard shelf in my 'Office' for probably the past two decades. A quick test with a 5v wallwart proves it will run reliably (and, of course, almost silently) at this low voltage and draws just under 43mA so won't be any strain for the under-utilised 5v rail. If anything, the extra 43mA on the 5v rail should provide a small but welcome boost to the 12v rail voltages.

 There is room between the back panel and the PSU board to mount this fan onto the base of the case but, just to present myself with a bit of a quandary, I also found a 60mm square 13mm deep cooling fan which only draws 35mA from the 5.15v wallwart despite its higher 170mA at 12v rating. Obviously, its larger size provides more airflow per mW but the only place it could be installed would be the top cover just over the area between the heatsink and the TCXO location. This would have the bonus of eliminating the complication of ducting but would spoil the top cover with a wire finger guard covering a 57mm diameter hole.

 This leaves me with the choice of style over function versus function over style - decisions, decisions. This is something I'll have to give some consideration to. Oh, and BTW, I found yet another 50mm square 10mm deep fan that draws just 28mA from the wallwart ( a mere  144mW) and seems to offer about the same CFM. I don't seem to be short of cooling fan options at any rate.

[DaveR]

Hi Johnny,

I wasn't suggesting that you could use one in the FY6600, just alluding to the same item being on sale for wildly different prices from different suppliers.  As for the 10MHz one I bought for £8, I've tested it against RWM on 9996kHz and it comes in at 0.2ppm, which I'm not going to complain about, and I've had a look in the trimmer hole and the trimmer is there, so it's identical to the £18 version here:
https://www.ebay.co.uk/itm/TCXO-10MHz-10-000000MHz-0-1ppm-Ultra-precision-Crystal-Oscillator-DIY-AUD-GC1-XH/381746677180

I've also put a piece of kapton tape over the can, so now I've got the golden version as well! 

Regards,
Dave

[Johnny B Good]

Hi Johnny,

I wasn't suggesting that you could use one in the FY6600, just alluding to the same item being on sale for wildly different prices from different suppliers.  As for the 10MHz one I bought for £8, I've tested it against RWM on 9996kHz and it comes in at 0.2ppm, which I'm not going to complain about, and I've had a look in the trimmer hole and the trimmer is there, so it's identical to the £18 version here:
https://www.ebay.co.uk/itm/TCXO-10MHz-10-000000MHz-0-1ppm-Ultra-precision-Crystal-Oscillator-DIY-AUD-GC1-XH/381746677180

I've also put a piece of kapton tape over the can, so now I've got the golden version as well! 

Regards,
Dave

 Actually, Dave, that would have been a better idea. I could have used Arthur Dent's technique of using an NB3N502 14 MHz to 190 MHz PLL Clock Multiplier to generate the 50MHz clock from the 10MHz OCXO that he'd installed into his Fy6600[1]. In my case, substituting the OCXO with the less power hungry 0.1ppm 10MHz TCXO instead.
 
 I've taken another IR thermometer survey of the main board, in particular the area adjacent to the 50MHz clock chip packed out with three linear voltage regulator chips raising the board temperature to circa 70 deg C, leaving the clock chip, as best as I can determine, at a rather toasty 50 deg C. My airflow deflector idea just ain't gonna fly in this case.

 Closer inspection of the picture of what I've ordered  suggests a circuit board size of 50 by 35 mm which will sit nicely over that 28mA @5v 50 x 50 x 10 mm fan that I plan on fitting into the bottom of the case in the space adjacent to the main board between the PSU and the rear panel. It looks like I won't be recycling that little PCB as a test mount for other XO clock generators after all since I've decided that, in this case, it'll be easier "To bring Mohamed to the Mountain".

 JOOI, I'm looking at a 3N502 chip right now on Ebay for £3.49 delivered from a London based warehouse <https://www.ebay.co.uk/itm/NB3N502DG-Frequency-Multiplier-8-Pin-SOIC/173385949750?epid=648417593&hash=item285e9ba236:g:psUAAOSwxDtbNlVW:rk:3:pf:0>

(https://preview.tinyurl.com/y8mosmkv).

 Of course, now that I've ordered the 50MHz part, the die is set and there's little point in my ordering a 3N502 frequency multiplier chip. However, since a 10MHz 0.1ppm TCXO is a damn sight cheaper to buy than a 50MHz TCXO, plus the fact that it's better to relocate the replacement TCXO away from the heat of those three voltage regulators anyway, for anyone else contemplating a similar TCXO upgrade, the 10MHz TCXO plus freq multiplier chip is not only cheaper, it's also the better option imo even if you aren't planning on fitting a small cooling fan.

 Nice touch BTW, using kapton tape to 'upgrade' your TCXO to a 'Golden' version. :-)

 I noticed that the 10MHz TCXO you linked to is from the same trader that sells those three 50MHz TCXOs I mentioned earlier, ranging from £17.90 (9p cheaper than that 10MHz example you mentioned) to a top end of 22 quid for one in a square can. When it comes to 50MHz 0.1ppm TCXOs on Ebay, it's a very small world indeed.

 I had to google "RWM". I'm more familiar with WWV which operates 4KHz above RWM's frequency allocations at precisely 5, 10 and 15 MHz with the additional frequencies of 2.5 and 20 MHz. I've used the 10MHz WWV signal in the past (some 20 odd years ago now) to calibrate my Kenwood TS140S HF transciever which I recently rescued from a damp and neglected basement radio shack come workshop.

As far as I can tell, its frequency calibration appears not to have suffered in spite of the decades of neglect but I no longer have an external antenna by which to tune into the 10MHz WWV signal to confirm this. Even with a good antenna, it wasn't always possible to get a good enough signal from WWV to perform such calibration checks.

As far as I remember, I only had to recalibrate it the once shortly after I'd purchased it. Aside from the initial warm up from cold, it remained within one or two hertz of the 10MHz signal, demonstrating remarkable frequency stability for a late 80s example of radio technology.

 Regards, Johnny B Good
 

[NOTES]

[1] Reply #151 Posted just over a year ago (page 7).

[2] I've just had another look at the description of this "oscillator clock power supply". The dimensions given are "size: 20.8 x 13.2 x 15mm" which is clearly a nonsense unless they're that of the bare TCXO itself which, now I think about it, seems more likely. Double checking against the 74HC04D chip on the board to get a sense of scale confirms I was in the ball park with my original size estimate.

 My new estimate of the board's dimensions are now ever so slightly revised downwards to 45 by 30 mm which gives me a touch more leeway on exactly how I'll be mounting it over the fan to catch the incoming airflow without creating unwanted turbulence near the fan blades (which is a sure fire way to drastically reduce the efficacy of these high volume/low pressure axial fans).

 I'll want to provide ducting to divert the incoming air past the oscillator board and towards the front in order to minimise losing too much of that hard won cooling air straight out of the rear panel vents which I don't want to block off. I need all the exhaust venting cross sectional area I can recruit to exhaust the warmed over air, hence the ducting so that what does escape via the rear panel vents has already managed to pick up and carry away some heat from the innards of the box.

 I've got a few more weeks to install the fan before that TCXO board arrives anyway so I might as well do some testing to make absolutely sure there's not going to be a problem in stealing power from the 5v rail to drive the fan before I start cutting holes in the bottom of the case. I rather doubt there will be a problem but you can't take this stuff for granted. The most likely side effect I'm expecting is a slight boost in the 12v rail voltages which, in moderation, will be no bad thing in this case.

[Johnny B Good]

=====bulk of previous reply snipped====

[NOTES]

[2] I've just had another look at the description of this "oscillator clock power supply". The dimensions given are "size: 20.8 x 13.2 x 15mm" which is clearly a nonsense unless they're that of the bare TCXO itself which, now I think about it, seems more likely. Double checking against the 74HC04D chip on the board to get a sense of scale confirms I was in the ball park with my original size estimate.

 My new estimate of the board's dimensions are now ever so slightly revised downwards to 45 by 30 mm which gives me a touch more leeway on exactly how I'll be mounting it over the fan to catch the incoming airflow without creating unwanted turbulence near the fan blades (which is a sure fire way to drastically reduce the efficacy of these high volume/low pressure axial fans).

 I'll want to provide ducting to divert the incoming air past the oscillator board and towards the front in order to minimise losing too much of that hard won cooling air straight out of the rear panel vents which I don't want to block off. I need all the exhaust venting cross sectional area I can recruit to exhaust the warmed over air, hence the ducting so that what does escape via the rear panel vents has already managed to pick up and carry away some heat from the innards of the box.

 I've got a few more weeks to install the fan before that TCXO board arrives anyway so I might as well do some testing to make absolutely sure there's not going to be a problem in stealing power from the 5v rail to drive the fan before I start cutting holes in the bottom of the case. I rather doubt there will be a problem but you can't take this stuff for granted. The most likely side effect I'm expecting is a slight boost in the 12v rail voltages which, in moderation, will be no bad thing in this case.

 Apropos of installing the whole "oscillator clock power supply" PCB into the case as a sort of airflow deflector rather than just transplant the actual TCXO onto the main board in place of the original XO chip, I completed the electrical tests with both the 28 and 43 mA at 5v fans to check the effect on voltages and power consumption.
 
 As expected, the 12v rails enjoyed a modest 30 to 40mV boost with the 28mA fan whilst the 5v rail dropped about 10mV with the overall consumption rising by just under half a watt to the 9.5W mark when driving 50 ohm loads at the 20MHz 20v P2P amplitude setting (maximum consumption conditions). Thus encouraged, I've cut out the hole and fitted the fan exactly as I described in NOTE 2 above (except, of course, for the fan ducting).
 
 With the lid perched loosely to allow convenient test access, the temperatures do appear to have dropped a good 10 deg C all round despite a lot of the incoming air getting blown straight out of the rear panel vents. I've now clipped the lid into place to seal up the gaps and, for the time being, I've taped over the rear vent slots to increase the cross flow air circulation over the main board.
 
 The "Hot Spots" on the top of the case are now merely perceptibly warm spots and even the underside hot area is now significantly cooler than before and I'm now inclined to permanently block the rear vents and open up the RHS vent slots to increase the airflow across the main board rather than faff around with additional fan ducting. Those rear vent slots are now more trouble than they're worth (and they were pretty worthless to begin with imho).
 
 I'm still planning on mounting the oscillator board over the fan to both keep it at little more than ambient and to act is a deflector to direct the incoming air towards the front panel. I think this is likely to prove just as effective a solution as the more fiddly ducting idea I'd had to prevent the rear vents from exhausting incoming cooling air straight out the back.
 
 It can be all too easy to overthink things when trying to find an optimum solution. Initially, it seemed a shame not to put the rear vent to good use but they were a horrible half hearted exercise in passive ventilation to start with, being as they are in a far from ideal location, especially considering the built in tilt to the case before it's made even worse by being propped up by the pull out front support stand.
 
 The kindest thing I could do after addressing the woefully inadequate passive ventilation with a cooling fan was to block this vent and look to improving the existing right hand side vent (the LHS one is perfectly fine as it is). However, I'll deal with that when I finally get hold of the oscillator board.
 
 As things stand right now, simply adding a  five volted 50mm fan and blocking the rear vent slots has considerably reduced the internal temperatures so yet another upgrade that's well worth adding to the list. There's no need to be concerned about 'fan noise' since it's barely audible against the louder fan noise of the Siglent SDS1202X-E DSO which is whisper quiet just as had been remarked upon by most, if not all, of the reviewers who'd checked it out.
 
 IOW, I had to shut the Siglent down before I could ascertain the level of fan noise coming from my modified FY6600. Even then I had to pick it up to place it close to my ear before I could hear any noise. Unless every other bit of kit in your workshop is passively cooled, you'll be hard pushed to detect fan noise from the FY6600 (assuming, of course, that you haven't used the crappiest 50mm fan from your Junk Box collection).
 
 Now that I've had it running for a good half hour or so since I clipped the lid back into place, I can no longer detect any warm spots other than on the underside of the case. That, I have to say, is a remarkable improvement for a relatively simple modification. Opening up the RHS vent slots as a final touch should noticeably improve the effectiveness of this forced ventilation modification to the benefit of significantly improved reliability and an extended service life.
 
 It looks like it's going to be several weeks before I'll have anything new to report so I'll take this opportunity to wish everyone still monitoring this thread a Merry Christmas and a Happy New Year. :-)
 
Johnny B Good.
 

[Johnny B Good]

====snip====
 
 Now that I've had it running for a good half hour or so since I clipped the lid back into place, I can no longer detect any warm spots other than on the underside of the case. That, I have to say, is a remarkable improvement for a relatively simple modification. Opening up the RHS vent slots as a final touch should noticeably improve the effectiveness of this forced ventilation modification to the benefit of significantly improved reliability and an extended service life.

====snip====
 

 One final final post for this year (unless that oscillator board arrives early). I couldn't resist taking another look at improving the RHS vent slots this afternoon. Although there is the potential to open them up without altering their external appearance, the job looked just too fiddly and time consuming to be worth all that effort, not to mention the risk of spoiling the appearance of the case, so I decided against it.

 However, I did spot the gaps where the front panel clips onto the bottom half of the case so, with careful application of a small file, I bevelled the inside of the RHS engagement slot to increase the effective CSA without obviously altering its external appearance and used BluTak(tm) to seal the gap in the LHS engagement slot.

 This represents a compromise between riddling the base under the opamps with a dozen or so 4mm holes to encourage additional airflow under the main board as I'd suggested in an earlier post and improving the existing airflow accidentally provided by the gap in the RHS securing slot.

 Having applied BluTak(tm) to seal up the LHS securing slot in order to favour the improved RHS with more airflow, you might well ask why I didn't block or partially block the LHS vent slots to favour the RHS slots. Well, the answer is that you can overdo this balancing act when restricting airflow in one part without reducing the restriction elsewhere since the increased back pressure can significantly reduce the efficacy of a high volume low pressure axial fan to such an extent as to become counter-productive not just to the total flow of air through the whole box but also to the targeted areas of concern.

The LHS vents might be letting out cool air, seemingly in an unproductive fashion but it's still allowing a higher speed of airflow within the confines of the box to encourage better heat transfer to the circulating air some of which heat energy, whilst diluted, will still escape via the LHS at a reduced temperature whilst the air exhausted via the RHS vent will be carrying more energy as a result of its higher temperature, courtesy of the improved energy transfer from the faster air speed within the box around the heatsink and the main board and its components.

 One might also wonder why I agonised over drilling a dozen 4mm holes in the base after cutting out a whacking great 47mm diameter hole for a cooling fan. Well, TBH, the 47mm hole is an inescapable consequence of blessing the box with a much needed active cooling system, aka, a fan and, as such, is as much a statement as a necessity whereas a bunch of 4mm holes, let's face it, just looks a little naff. 

 I reckon I've done about as much as I can to reduce component temperatures short of lending the case a 'Swiss Cheese' look. Installing the oscillator board (once it finally arrives) over the fan location at a jaunty angle to better deflect the airflow towards the front panel should offer a further, if modest, improvement in the efficacy of my active cooling modification.

 Before I forget, I think it's worth noting that despite the optical illusion, that little heatsink is actually square (19mm by 19mm in my case). Whilst I'd had the main board out to modify the retaining slot, I'd thought to check the dimensions with a view to reorientating it to favour the anticipated cross flow of air.

 When I'd first refitted it after the opamp upgrade, I'd replaced it in its original orientation with the fins aligned front to back, thinking it might not completely cover the opamps if I did otherwise (yet another ASS U ME moment). It now sits on the board with the fins parallel with the front and rear panels which should improve heat transfer under the new regimen of my active cooling system.

 Indeed, I think it can even be fitted at +/-45 deg without fouling anything and still cover the opamps if that gives a better alignment with the airflow direction wherever you might choose to locate a cooling fan. Obviously, you need to determine or guestimate which direction the airflow will be taking in order to make an effective choice but at least you have room for manoeuvre to maximse the heatsink's efficacy. In my case, I've simply guestimated that, on balance, an east west orientation rather than the original north south one is the better choice for my setup.

 Anyhow, that's the latest development for the time being. As modifications go, it's been one of the most satisfying ones I've done so far (it's right up there with the opamp upgrade with the PSU mods not too far behind - all vital and worthwhile improvements imo).

 Not only does this mod improve reliability and extend component life, it's also an essential part of the 50MHz clock oscillator upgrade since it should hold the temperature of the TCXO to within a degree or two above room ambient rather than have it subjected to 50 or 60 degree temperatures. There'll be virtually no warm up delay from switch on as a result which is a nice touch to my mind. 

 Well folks, that's all for now. Have a Merry Christmas!

 Johnny B Good