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

[rhb]

I lost track of this thread long ago.  Is there FOSS  FW available that can be loaded on an STM32F103 that  will make borked units work when installed on the front panel?  I'll buy a socket and program some chips if it can be done.  Mostly because I despise F***Tech.

My last recollection was that fremen was using a blue pill, but it required a PC to control it.

[tsman]

Is there FOSS  FW available that can be loaded on an STM32F103 that  will make borked units work when installed on the front panel?

I don't think there is anything. fremen67 hasn't been around for months either so no source or newer builds for that avenue.

[Johnny B Good]

I lost track of this thread long ago.  Is there FOSS  FW available that can be loaded on an STM32F103 that  will make borked units work when installed on the front panel?  I'll buy a socket and program some chips if it can be done.  Mostly because I despise F***Tech.

My last recollection was that fremen was using a blue pill, but it required a PC to control it.

 I don't think any more progress was made on that front I'm afraid. I've read the whole of this thread twice now plus some reprising of some of the earlier hardware related sections. Once the problem 'went away' with the later production models from Feeltech, I lost interest in what was obviously a doomed firmware repair project due to the uncrackable protection that prevented retrieval of the binary code for reverse engineering a fix.

 I bought mine in the first week of November and it arrived on the 9th sporting the 3.3 version of firmware which seems to have addressed several issues aside from the major bug that caused it to brick itself as yours did. Fingers crossed that version 3.3 doesn't have a similar bug quietly waiting in the wings for just exactly the right moment to take centre stage to wreak similar havoc and mayhem.

 I can entirely understand your attitude to Feeltech and I sympathise but, as others have pointed out, when it's this cheap, cash strapped hobbyists are going to take a punt on it, especially if they've read through this EEVblog thread to discover that the fatally flawed firmware version is now ancient history.

 It seems, in view of the warranty voiding mods you'd already applied, that the best you can hope for is to try and sweet talk Feeltech into sending you a replacement front panel at cost price by way of a compromise as a gesture of good will since the fault seems to have resulted from a fatal flaw in the firmware they had supplied it with. Other than that, you're pretty well stuck with a mid priced paperweight if the PC based work around doesn't appeal.

 Regards, Johnny B Good

[rhb]

Thank you.  I couldn't face reading 60 pages.  I'd mostly dropped out by page 20 or 30 as I did not have a programmer or blue pill.  But I do now.

I got a replacement panel with V 3.1, but not by "sweet talk".  But at least 4 others did not.  I'd like to help them.  In a classic twist of irony I got a front panel  after I had purchased a Keysight 33622A from the Keysight eBay store.

I offered to do a review of the FY6600 against the 33622A and attached some 8560A spectrum analyzer plots.  But that did not result in the other 4 members with 3.0 units getting relief.  Nor did a 2nd less friendly approach.  So it's time for me to carry through with what I said I'd do if they did not supply new FPs to the other 4.  It will likely prove *far* more expensive than sending out 4 FPs.

If we have the bus protocol worked out I ought to be able to write the front end UI part.  It would be a good warm up exercise for writing FOSS DSO FW for Zynq based scopes (e.g. the Instek 2000E and the Siglent 1000X-E).  And  a pleasant diversion from reverse engineering the DSOs which are rather more complex.

I've sent a PM to fremen67.  My concern was that it might have already been done.

Subsequent to my V 3.0 adventure I bought a very nice set of bench gear, mostly early 90's HP  gear.  I'm 65 and finally gave myself permission to spend the money.  But I have many painful memories of when I couldn't afford test gear.  My first good scope was a Dumont 1060, a very nice dual trace analog scope which I bought from Tucker Electronics in 1991 for $325 with a 30 day warranty.  The horizontal sweep died about day 35.  Eventually I traded a couple of spare 144 MB ESDI drives for a wonky 465 which worked well enough to let me fix the Dumont which then let me fix the 465.

After that moves and the turmoil of life left me without a bench setup.  I was starting to rebuild a bench as all my other gear has died of old age and will require repairs.  But I was still in the penny pinching mode I was taught growing up so Rigol DSO, F***Tech AWG, Tenma 4.5 digit DMM etc.  But several friends dying and in particular watching my brother in-law deteriorate from Parkinson's dramatically over 9 months changed my attitude.  Though my goals are still the same,  good quality low cost T&M gear.

[Johnny B Good]

Thank you.  I couldn't face reading 60 pages.  I'd mostly dropped out by page 20 or 30 as I did not have a programmer or blue pill.  But I do now.

I got a replacement panel with V 3.1, but not by "sweet talk".  But at least 4 others did not.  I'd like to help them.  In a classic twist of irony I got a front panel  after I had purchased a Keysight 33622A from the Keysight eBay store.

I offered to do a review of the FY6600 against the 33622A and attached some 8560A spectrum analyzer plots.  But that did not result in the other 4 members with 3.0 units getting relief.  Nor did a 2nd less friendly approach.  So it's time for me to carry through with what I said I'd do if they did not supply new FPs to the other 4.  It will likely prove *far* more expensive than sending out 4 FPs.

If we have the bus protocol worked out I ought to be able to write the front end UI part.  It would be a good warm up exercise for writing FOSS DSO FW for Zynq based scopes (e.g. the Instek 2000E and the Siglent 1000X-E).  And  a pleasant diversion from reverse engineering the DSOs which are rather more complex.

I've sent a PM to fremen67.  My concern was that it might have already been done.

Subsequent to my V 3.0 adventure I bought a very nice set of bench gear, mostly early 90's HP  gear.  I'm 65 and finally gave myself permission to spend the money.  But I have many painful memories of when I couldn't afford test gear.  My first good scope was a Dumont 1060, a very nice dual trace analog scope which I bought from Tucker Electronics in 1991 for $325 with a 30 day warranty.  The horizontal sweep died about day 35.  Eventually I traded a couple of spare 144 MB ESDI drives for a wonky 465 which worked well enough to let me fix the Dumont which then let me fix the 465.

After that moves and the turmoil of life left me without a bench setup.  I was starting to rebuild a bench as all my other gear has died of old age and will require repairs.  But I was still in the penny pinching mode I was taught growing up so Rigol DSO, F***Tech AWG, Tenma 4.5 digit DMM etc.  But several friends dying and in particular watching my brother in-law deteriorate from Parkinson's dramatically over 9 months changed my attitude.  Though my goals are still the same,  good quality low cost T&M gear.

 Glad to oblige, I'm probably the only member who's read the whole 69 pages (now 70 thanks to my recent efforts) within the past fortnight or so who could answer your question off the top of their head. 
 
 So you did manage to get a replacement panel then. Pity about the other four members who've had to go without or use the PC control software work around. I'm interested in how you finally managed to get Feeltech to 'do the right thing' (or did they?). I can imagine the irony of fixing your cheap FY6600 after spending some $5K plus on an all singing all dancing Keysight AWG.

 Curiosity got the better of me and I found a review of the 33622A. I was rather surprised at just how small it was (not all that much bigger than the FY6600 as far as I could make out from the pictures). Obviously better made and a significantly more substantial piece of kit but quite bijou just the same.

 I can quite understand the appeal of Feeltech's "Poor Man's 33622A" with its order of magnitude lesser specification at around a two orders of magnitude cheaper price point aimed at cash strapped hobbyists who can work around the lack of the refinements of a 5 grand AWG. I think Feeltech know their market demographic far far better than it knows itself.

 If I had to describe Feeltech's marketing strategy using a simple to understand comparison, I'd say they're rather like a realtor (estate agent) specialising in fixer upper (run down) properties located amongst prime housing stock (up market areas), where the idea is not to waste money bringing the property up to the standard of its surroundings, an investment you'll have difficulty in recovering because you've excluded the wider market demographic, but to leave it open to a wider market of first time buyers trying to get on the property ladder who can see it as a low price ticket into an upmarket area.

 It's just a pity that they don't quite understand the value of good customer relations (in particular the notion of 'Good Faith' and the concept of 'Good Will'). Whatever you have in mind to effect an attitude adjustment at Feeltech, I wish you all the best success.

 It's not exactly clear but I assume your offer to review the FY6600 against the 33622A was to Feeltech rather than Keysight (the other way round wouldn't make much sense imo - Keysight, I imagine, would regard such a comparison as an insult, and one which would dent their sales figures). If Feeltech didn't jump at the chance of such a comparative review, they've either not got a customer relations/marketing department or else one that's totally oblivious to the value of such a priceless PR/marketing exercise.

 Interesting that you still want to write a UI from scratch for that FY6600 in spite of what's gone on and after all this time. I'm guessing, in part at least, that you're trying to spite Feeltech for their lack of support. I do hope you succeed in this FOSS project.
 
 However, you do realise that this will benefit Feeltech more than it will you since it will expand its appeal as a poor man's Keysight AWG even more. Still, a challenge is a challenge I suppose (and those four luckless members will no doubt offer profuse thanks for your efforts). Remember what I said about Feeltech knowing us better than we know ourselves? Well, I rest my case. 

 With regard to treating ourselves after a lifetime's frugality, I understand totally. Witness my blowing a whole £365 on a brand new SDS1202X-E DSO just the week before I then blew a whole £75.66 on that Feeltech AWG. Spending that sort of cash on test gear was something I'd have never dreamed of doing before. It's a whole new ball game now! 

 I'm now in my late 60s (I have to recalculate my age from the current year just to keep track - the novelty of birthday celebrations wore off decades ago) and it's finally dawned on me that I might as well "spend the children's inheritance" on myself rather than let them squander it on smart phones, smart TVs and other such trash - they're all well enough off to manage from just the sale of the house as 'our final parting gift' once the missus and I have popped our clogs.

 Like you, I want good/serviceable quality T&M gear at the less than obscene, eye wateringly high prices still surprisingly being commanded by the Likes of Tektronix and Keysight. I enjoyed your tale of using a wonky 465 to repair that Dumont scope in order to fix up the 465. It reminds me of the sort of things I used to do and it demonstrates that perfection in T&M gear isn't as important as knowledge and experience and a proper understanding of the limitations of the less than perfect test gear you already possess or can aspire to.

 As if to demonstrate that principle of "boot strapping" a repair using 'damaged goods' (and getting back on to the topic of improving my FY6600), I've been observing a rather curious variation in the wattage readings of the AWG's power consumption which I suspect may be due to one or more of the smoothing caps in its PSU board.
 
 The readings during the past few days since I fitted the cooling fan have varied from just over the 9 watt mark to a gnat's dick short of the 10 watt mark on the mirror backed 100W scale on my Metrawatt analogue wattmeter (around a 6 to 7 percent variation).

 Switching PSUs don't respond to mains voltage variations in the way that analogue PSUs do. Indeed, I wouldn't expect to be able to observe any variation whatsoever even if it varied between 200 and 265 volts, let alone the typical 239 to 246 volts we experience at this location, so I know that it has to be caused by changes taking place inside the box.
 
 Since I've got the AWG set up for 20MHz 20vp2p driving 50 ohm loads with my DSO hooked up in order to monitor the output and nothing has changed this configuration during all the time I've been monitoring the power consumption, my number one suspect is the PSU board with the main board and front panel as secondary suspects (I can't completely rule them out) and caps are the most likely culprit for such subtle strangeness in smpsu behaviour - I've got to make a start somewhere.

 I've built up a nice little collection of fine looking caps salvaged from scrapped MoBos over the years but, despite having no comebacks from repairs to PC MoBos using such recovered caps, there's still the possibility of the odd one or two bad ones hiding amongst all the good so I do need to test my most suitable candidate replacement caps to avoid my repair attempt making things worse.

 It's rather lucky then that I happen to have a 'scope and a sig gen on hand to perform some basic ESR tests to weed out any duds from my collection of salvaged capacitors before I use them to 'improve' the PSU board.    If I'd had another sig genny to hand, I could've tested the caps in situ and perhaps avoid swapping them out for testing later but, as it happens, I don't, so the job has to be done this way.

 You may well ask why I don't suspect the last item I'd added as extra loading on the PSU and the answer is that the fan has never faltered and when I do stall it to see whether it has any effect on the wattage reading, all I can observe is a barely perceptable increase of no more than a quarter at most of the slow variation I've witnessed so far (plus, I've had a notion that it had been doing this even before I added the fan). If a cap change makes no difference, then I'll have to test the PSU board with a resistor dummy load before I can cast my suspicions further afield.

 There's even an outside chance that this might be 'normal behaviour' for this AWG unlikely as that seems (it's Feeltech, so who knows what's possible?). However, I can't imagine too many of Feeltech's customer demographic will even have an analogue wattmeter or an expensive bench wattmeter capable of measuring in increments of 10mW or better in order to observe such subtle power consumption behaviour (analogue wattmeters are rarer than unicorn droppings as of over a decade ago), so I don't expect anyone here to pipe up with a similar observation even if they wanted to test for it with their "Kill-A-Watt meters" since a digital readout to one tenth of a watt accuracy will make such subtle variations quite hard to distinguish from the "dance of the digits" typically performed by such plug in energy consumption meters.

 Still 'n' all, if anyone is intrigued enough to try and repeat my observations, what I've observed has been a slow variation between, in my specific setup, just over the 9 watt mark increasing slowly and, as far as I could ascertain, smoothly to just short of the ten watt mark over a period of an hour or more with no sudden jumps before going the other way after perhaps an hour or so at the 9.9 watt level. I'd estimate a variation of around 750mW, roughly a 7% variation on a nominal 9.5W load. It's only a small variation but I can't see any obvious reason why such a variation should exist at all.

 Clutching at straws, assuming the FPGA comprises mostly of NMOS logic rather than CMOS, it's just possible that the average current consumption might be being modulated by the processing of a 20MHz sine wave at the 20vp2p level setting but I've not been able to find sufficient detail about the gate technology actually used in a modern day FPGA so that's just an educated guess with no hard evidence to support such a hypothesis. With any luck, a recapping of the PSU board may solve this riddle anyway. 

 Regards, Johnny B Good

[rhb]

I sent F***Tech 4 photos taken of the screen of an HP 8560A SA of the output of the 33622A and the FY6600 without identifying which was which.  I deliberately chose the output levels so that they looked almost identical.

The 33622A was used from Keysight's eBay store, so not as much, but still 30X the 6600.  But also 120 MHz rather than 60.

F***Tech told me I had lifetime support and similar BS for about 6 months without actually doing anything about the problem.

Not too sure that FOSS FW for the 6600 will help F***Tech.  Relative to a DSO the FPGA code is simple.  So writing a complete FOSS FW suite might well lead to a lot more competition.  To date I've spent over $1k setting up for the DSO FW project.  Right now I'm debating whether to also get a 1102X-E now or later. I already have an Instek GDS-2072E for testing. But long experience has convinced me that writing and testing SW on multiple platforms in 50-100 line increments avoids lots of bugs and other problems..

In a JDS6600 thread, a poster purporting to be with JDS said that they hired the guy who developed the F***Tech AWG who with another member of their staff designed the JDS6600.  Interestingly, the FY6600 uses a DAC and the JDS6600 uses an R2R ladder.

[Johnny B Good]

I sent F***Tech 4 photos taken of the screen of an HP 8560A SA of the output of the 33622A and the FY6600 without identifying which was which.  I deliberately chose the output levels so that they looked almost identical.

The 33622A was used from Keysight's eBay store, so not as much, but still 30X the 6600.  But also 120 MHz rather than 60.

F***Tech told me I had lifetime support and similar BS for about 6 months without actually doing anything about the problem.

Not too sure that FOSS FW for the 6600 will help F***Tech.  Relative to a DSO the FPGA code is simple.  So writing a complete FOSS FW suite might well lead to a lot more competition.  To date I've spent over $1k setting up for the DSO FW project.  Right now I'm debating whether to also get a 1102X-E now or later. I already have an Instek GDS-2072E for testing. But long experience has convinced me that writing and testing SW on multiple platforms in 50-100 line increments avoids lots of bugs and other problems..

In a JDS6600 thread, a poster purporting to be with JDS said that they hired the guy who developed the F***Tech AWG who with another member of their staff designed the JDS6600.  Interestingly, the FY6600 uses a DAC and the JDS6600 uses an R2R ladder.

 I'm quite surprised F***Tech didn't gift you a new AWG with the latest firmware for you to review alongside of the 33622A. Any company with a functioning marketing division wouldn't have let such an advertising opportunity slip through their fingers like that.

 BTW, the penny has only just dropped regarding your use of F***Tech. At first I just thought you were avoiding the use of their name to signify your reluctance to give it the "Oxygen of Publicity" but I guess you're simply expressing an expletive variation where the F*** is an anagram of FCUK (another company where their Branding division realised they could create a suggestive and memorable acronym from the initial letters of their company name, "French Connection UK").

 Sounds like you bought the 33622A for around half price. Here in the UK, RS and Farnell/Element14 are selling it for £5854.80 after the VAT is added on, some 77 times more than I paid for the FY6600-60M. JOOI, I checked Newark element14's one off price which was $6,856.00. I've no idea whether that was a tax paid price or not but I'd seen more than enough to make my eyes water (only metaphorically you understand - I'm not made of money so there's no way I'm going to spend 77 times more for a box that's by volume no more than 50% bigger).

 They say it's money that makes the world go round. If that's true, all I can say is that Keysight Technologies seem to be going above and beyond in keeping the world in a spin (mind you, Tektronix are no slackers in this endeavour either).

 I'll take your word on the simplicity of coding a FPGA. My coding expertise only extends to BASIC and Z80 assembler and I haven't done any coding since the mid 80s. You might want to put off getting the Siglent till later (and maybe pick the SDS1202X-E instead ).

 There's a damn good reason as to why the JDS6600 is cheaper (just!), the lack of a dedicated 14bit high speed DAC with laser trimmed resistors in its ladder network to maintain monotonicity throughout the whole range of sample values, is its downfall. I've seen the youtube review showing the artefacts on its sine wave output as a result of this nasty bit of cost cutting.
 
 There's cheap and then there's so cheap it not only can't be any good, it actually isn't any good. The FY6600 falls into the former category whilst the JDS6600 is most firmly stuck in the latter.

 Mention of 'interesting facts' raises the question as to why a seven thousand dollar, less a bit of pocket change, AWG comes with only a 2ppm XO as standard with a 0.1ppm OCXO as an optional upgrade. At that price point you'd expect to see a 0.1ppm TCXO as standard with an option to upgrade to a 3ppb OCXO at the very least. In fact, at a factory fitted option price of $715 why isn't the OCXO a 3ppb rated one? It seems, on that basis, that my 20 dollar 0.1ppm TCXO board is too good a bargain to be true. I sincerely hope that's not the case. BTW, do you happen to know which timebase option is installed in your 33622A? You may have got a better bargain than you'd thought.

 One final item I feel is worth mentioning, in view of my earlier ramblings about the anomalous power consumption readings I'd noticed in recent days, is that the problem has definitely progressed in that the average power consumption has increased ever so slightly with it spending more time now consuming exactly 10 watts. It now looks more and more like a mundane crappy PSU smoothing cap issue than the result of some rather exotic anomaly in the main and front panel boards.
 
 I guess it's time I stopped talking about testing the caps and actually get some testing done before it's too late and I'm left with no choice but to fit untested replacements.

 Regards, Johnny B Good

[rhb]

I'm surprised you didn't catch the F***Tech allusion sooner.

I was quite surprised that I offered to do several thousand dollars of consulting and marketing for under $400 and they ignored it.  That's when I started referring to them as F***Tech.

It's really easy to store corrected coefficients at final test.  But you have to care.  Why do you rate the JDS below the F***Tech?  I've not seen a good comparison, despite a number of people in this thread having both.

The 33622A is the only one of my HP instruments that does not have the OXCO option.  But as I'm about to hook everything to a GPSDO it shouldn't matter.

Subsequently I saw Keysight offering a 33622A for about $2800.

[Johnny B Good]

====snip====

 One final item I feel is worth mentioning, in view of my earlier ramblings about the anomalous power consumption readings I'd noticed in recent days, is that the problem has definitely progressed in that the average power consumption has increased ever so slightly with it spending more time now consuming exactly 10 watts. It now looks more and more like a mundane crappy PSU smoothing cap issue than the result of some rather exotic anomaly in the main and front panel boards.
 
 I guess it's time I stopped talking about testing the caps and actually get some testing done before it's too late and I'm left with no choice but to fit untested replacements.

 Following up on the above, I picked out a couple of 16v 470μF caps to replace the 220μF ones used by the 12v rails and a nice pair of 6.3v 1500μF (actually proved on testing as 1860 and 1920 μF) from my salvaged caps collection to replace the 10v 1000μF and the 16v 220μF used in the Pi network filtering on the 5v rail.
 
 Unfortunately, the PSU switching chip didn't take kindly to all this extra capacitance and refused to start up (similar overload current foldback symptoms I'd seen with my fullwave rectification upgrade experiment).

 After double checking I hadn't bridged any tracks during my resoldering of the replacement caps, I figured that maybe I'd placed too much smoothing capacitance into the secondary circuit so replaced the 1500μF cap on the output of the 5v Pi filter with one of the 220μF caps and tried again (the tall 1500μF cap had made accessing the output header a little awkward, making it my first choice for reverting to the original value).

 This proved a sufficient enough remedy to allow the PSU to fire back up and, since the wattmeter showed no change for the worse, I boxed the job off to carry out ESR tests on the other three caps that I'd pulled from the PSU board. They tested ok on ESR compared to the other caps I'd used as replacements so they're no longer suspect. The one I was obliged to revert to and wasn't able to test seems to also be ok since I haven't seen the wattmeter readings get any higher than about 9.8W, typically varying between 9 1/3 W and that high water mark.

 That happened last night and this morning, I managed to catch a glimpse of the power level suddenly, yet smoothly, transitioning upwards by a tenth of a watt from just over the 9 1/2 watt mark. Over the next 5 or 6 hours it remained clear of the 10 watt mark by a good 5th of a watt whilst I was running some more tests on the AWG. However, after shutting it down so as to restart it with my last default settings, it refused to power up in a repeat performance of the excess capacitance episode.

 It was only after allowing things to cool down for 20 minutes that it started up again so I knew I'd have to pull out the remaining "1500μF" cap and revert to the original 1000μF cap so yet again I'd have to delve into the AWG's innards.

 I realised several days ago now that I'd best not bother refitting those four long screws that hold the case halves secured lest I totally wear the screw threads out by the time I've finished all of my planned modifications, so it's now just a matter of unclipping the top of the case from the front panel. I've gotten quite adept at this now with all the practice I've had recently with half of a cut up credit card as a spudger.

 Anyway, this latest modification to my last PSU modification seems to have dialled the capacitive loading back to a more tolerable level for the controller chip to get on with its job. I've had it running flat out for the past 5 or 6 hours and my hot power up tests checked out ok so it looks like I've modded that PSU to within an inch of its life (hopefully... I don't want to do any more back pedalling or I'll be right back where I started. high performance rectifier diodes upgrade and transformer mod aside).

 That brings me to a rather curious feature of Feeltech's design wherein the relay switched attenuator network creates a horrible impedance mismatch when it cuts in at P2P levels of 500mV and below. I noticed this effect, which only produces a discontinuity when driving into a 50 ohm load, after I'd been perusing the reverse engineered circuit diagram of the main board trying to work out exactly just how the relay switching was meant to function.
 
 What happens is that for levels at or below 500mV, the THS3002 (or replacement 3091 or 3095 or 3491) opamp(s) are bypassed and the output from the OPA686N is routed to the output sockets via a kakmaimee attenuator network consisting of two 100 ohm shunt resistors linked with a 510 ohm series resistor element. When I first saw that suspiciously high value resistor, I had to assume it had been a transcribing error for a 51 ohm resistor even though that would still have resulted in a 10% mismatch error of 45 ohms.

 When the AWG is driving a 50 ohm load, there's a huge jump in output voltage at the transition from 500.0mV level to 500.1mV level and beyond. When I repeated this test without the 50 ohm dummy load, the transition was seamless. Intrigued by this, I tested the output impedance with a DMM to measure the resistance of the output channels either side of this transition level and obtained readings of 52.7 and 84.7 ohms on channel 1 and 54.7 and 85.9 ohms on channel 2 for settings above and below the critical 500mV mark respectively.

 No wonder there was such a discontinuity in level when driving 50 ohm loads! It turns out that the series element of this attenuator is 510 ohms after all, rather than 51 to the zeroeth power of ten as I'd thought would be the more likely scenario. Those measured resistance values were in close agreement with my calculated value of 84.45 ohms for the resistor values shown on the circuit diagram.

 Anyone using this AWG only for driving Hi-Z loads at audio frequencies would never notice this "Skoolboy Howler" in the design of this final attenuator resistor network which Feeltech chose to "fix in the firmware" and hope nobody would notice their sleight of hand rather than sort it out properly with a new BoM for the attenuator resistor network.

 Luckily, for those with an interest in radio frequency work requiring properly terminated loads, this mismatch effect doesn't effect signal level settings above the 500mV mark. The obvious work around when millivolt test signals are required is simply to employ an external 20dB/40dB attenuator instead of the rather shoddy built in one.

 I did, btw, figure out that the other relay switching was to bypass the high voltage output current feedback opamp stages for output levels of 5V p2p or less. These high output opamps are only used for outputs when p2p levels above 5v are selected otherwise the OPA686N chips drive the output terminals directly or via those kakmaimee attenuator resistor networks.

 I did notice a problem with sub 50mV level signals with common mode interference from the PSU board (if it had been switching ripple on the voltage rails it would have reappeared when the attenuator cut out at the the 500mV transition point). Most likely a problem exacerbated by my 'curing' the half live mains leakage issue when I removed that class Y1 EMC bodge capacitor the other week. 

 BTW, I still think the JDS6600 is a bag of shite. It's merely a tarted up poor man's FY6600 imho and not worth its £62.48 plus £6.99 economy delivery price tag.


Regards and a merry Xmas, Johnny B Good

[rhb]

....snip...

 BTW, I still think the JDS6600 is a bag of shite. It's merely a tarted up poor man's FY6600 imho and not worth its £62.48 plus £6.99 economy delivery price tag.

Regards and a merry Xmas, Johnny B Good

I'm going to have to read through all your ramblings on the F***Tech.  It's both entertaining and educational.  But you still didn't give a reason for your denigration of the JDS.

Still, it's a great pleasure to encounter someone who both knows what they are writing about and writes well.  My BA is in literature, so I have done a lot of reading.

If I didn't have an Instek GDS-2027E sitting open on my bench while I try to sort the JTAG and TTY port I'd have the F***Tech open and follow along.

[Johnny B Good]

....snip...

 BTW, I still think the JDS6600 is a bag of shite. It's merely a tarted up poor man's FY6600 imho and not worth its £62.48 plus £6.99 economy delivery price tag.

Regards and a merry Xmas, Johnny B Good

I'm going to have to read through all your ramblings on the F***Tech.  It's both entertaining and educational.  But you still didn't give a reason for your denigration of the JDS.

Still, it's a great pleasure to encounter someone who both knows what they are writing about and writes well.  My BA is in literature, so I have done a lot of reading.

If I didn't have an Instek GDS-2027E sitting open on my bench while I try to sort the JTAG and TTY port I'd have the F***Tech open and follow along.

 I had to google "GDS-2027E" to check what it was you were working on. No sensible results until I tagged "DSO" on the end. First hit was Farnell/E14's details page on this stock item (one in stock at £462 plus the 20% VAT). Now that's more within my spending range!  However, now that I have an SDS1202X-E (a snip at £365, VAT and free delivery included), I'm not in the market for another 70MHz DSO.

 The big downside to the JDS version is the use of R2R ladder networks to save the cost of a couple of high speed 14 bit DAC chips. I did see the expected artefacts demonstrated in a tear down video but can't find it on Youtube. Almost certainly it was a link in an EEVblog posting but a quick search for threads mentioning the JDS6600 failed to reveal where I'd seen that particular video and another search on youtube was just as fruitless as was a search through my browsing history.

 If you want your full amplitude sine waves to be free of glitches at the zero crossing and half amplitude points, it's best to avoid assembling an R2R ladder network out of even the highest precision resistors available unless you're only going for an 8 bit resolution in your DAC circuit (you'll need to specify resistor values to a precision better than 0.1% for the ninth bit then 0.05% for the tenth and ever tighter by a factor of two each time you add successive bits of resolution to your R2R ladder network).

 You wouldn't expect any sane EE to try and outcheap Feeltech's FY6600 design but the JDS6600 represents a reduced BoM costed version that sacrifices performance out of all proportion to the minuscule cost savings involved (it's essentially the core of the FY6600 with the DAC chips substituted for by four dozen or so (high precision?) resistors.

 Just how cheap do you want your poor man's version of a £360 Siglent dual channel 30MHz AWG to be? I'd happily pay another tenner for a version of the JDS6600 that used proper 14 bit DAC chips... oh wait, I did!  Not quite an extra tenner though, just an extra 6 quid as it happened. 

Regards, Johnny B Good

PS  I've not had any more problems with the PSU board since I dialled the capacitor upgrade back a couple of notches (the 470μF substitutions for the original 220μF caps on the 12v rails don't seem to causing any problems for the driver chip). That's not to say I've stopped seeing that small third of a watt variation in its consumption. It just means I'm happy to accept that it's a characteristic of the generator's operation when supplying steady output signals into 50 ohm loads rather than some sign of impending PSU failure.

PPS  That kakamaimee output attenuator resistor network seems to have been designed as a 20dB pad for an 85 ohm impedance line.   The firmware mediated compensation for the insertion of this pad for 500mV and below is only correct for the unloaded Hi Z case. What is required is to calculate the voltage attenuation ratio for this Hi Z case and design a 50 ohm attenuator pad that gives the same unterminated output voltage level and the extra 6dB of attenuation that's expected when properly terminated. I've only gotten as far as determining that it's an 84.755 ohm line attenuator.

 I'll try doing the rest of the calculations later on. I should be able to come up with a corrected BoM list for the attenuator resistor network compatible with  their 'quick fix' firmware compensation for the Hi Z error condition[1].

[1] They were probably aiming to introduce a 20dB drop in output level, ie a reduction to 10% of the (terminated) output voltage with an expected reduction of attenuation to 20% in the Hi Z case. What they actually got was a much greater reduction to 10.33% in the Hi Z case which they must have 'corrected' for in the firmware (but only good for the Hi Z output case). the 6dB drop expected with a 50 ohm load becomes an 8.6 dB drop in this case (I think).

 Trying to reconfigure for a 50 ohm attenuator pad that provides the same open circuit voltage attenuation and hence the expected 6dB drop when terminated with a 50 ohm load is doing my head in, even with the help of an on line dB calculator and a Matching Pi Attenuator Calculator. Never mind, I'm sure I'll find a solution... eventually! 

[EDIT 2020-04-14]

 I did find a solution! It turns out I'd been overthinking the problem and giving F***tech far too much credit for ingeniously disguising the cheap choice of resistors used in their attenuator networks (RS1 to 6) by a 'firmware bodge' when in fact all they'd done was simply pick a cheap resistor combo that would provide the 20dB voltage reduction in the Hi-Z out[put case only (and let the Devil take the hindmost - those users relying on the output remaining at a 50 ohm impedance - for settings at or below the 500mV p-p mark).

 Using preferred values from the more expensive E192 range, substituting the 510 ohm with a 249 ohm series element and each of the 100 ohm shunt elements with 61.2 ohm resistors produces a 20.03dB @50.11 ohm attenuator pad. Even relying on the +/-0.5% tolerance of the E192 range limits the variations to a worst case  low of 19.96dB @50.28 ohms and a worst case high of 20.1dB @49.94 ohms, still plenty good enough for use as 'drop in' replacements.

 Anyone looking for a 'quick fix' to this 85 ohm attenuator issue, only needs to know that the required replacement is exactly a 20dB @50 ohm attenuator pad. The existing resistors (W1 to 6) can be shunted with additional smd resistors (easily calculated) soldered over the top, saving having to deal with the ticklish business of extracting the originals without risking harm to the board. At the modest 60MHz upper frequency limit, the small additional stray capacitance introduced by doubling up resistors in this way can safely be ignored.

JBG

[bugi]

If you want your full amplitude sine waves to be free of glitches at the zero crossing and half amplitude points, it's best to avoid assembling an R2R ladder network out of even the highest precision resistors available unless you're only going for an 8 bit resolution in your DAC circuit (you'll need to specify resistor values to a precision better than 0.1% for the ninth bit then 0.05% for the tenth and ever tighter by a factor of two each time you add successive bits of resolution to your R2R ladder network).

I have some vague memories of reading mentions that the resistor network on JDS isn't all R2R, but something else at one end, apparently exactly for the reason to make it a bit more accurate.
Also, IIRC, the manufacturer itself mentioned something about them doing something with the R2R/resistors stuff so that the accuracy would be better than expected with standard way to putting one together. Which I read in between the lines that either it refers to that something-else-than-R2R, and/or perhaps they use binned components (i.e. higher tolerance relative to each other, but not necessarily closer to the nominal value). Maybe... maybe just marketing talk.
I bought such JDS unit (didn't even know about the FY-stuff at the time). The little I have used it so far, the biggest problem isn't tiny errors in the DAC, but larger issues at higher frequencies/amplitudes caused probably by the amplifiers, and power supply noises etc.  But I guess the biggest issue depends on what the device is used for.

[ledtester]

On the subject of replacing the power supply, here's a video of one guy experimenting with using DC-DC converters:

[Johnny B Good]

If you want your full amplitude sine waves to be free of glitches at the zero crossing and half amplitude points, it's best to avoid assembling an R2R ladder network out of even the highest precision resistors available unless you're only going for an 8 bit resolution in your DAC circuit (you'll need to specify resistor values to a precision better than 0.1% for the ninth bit then 0.05% for the tenth and ever tighter by a factor of two each time you add successive bits of resolution to your R2R ladder network).

I have some vague memories of reading mentions that the resistor network on JDS isn't all R2R, but something else at one end, apparently exactly for the reason to make it a bit more accurate.
Also, IIRC, the manufacturer itself mentioned something about them doing something with the R2R/resistors stuff so that the accuracy would be better than expected with standard way to putting one together. Which I read in between the lines that either it refers to that something-else-than-R2R, and/or perhaps they use binned components (i.e. higher tolerance relative to each other, but not necessarily closer to the nominal value). Maybe... maybe just marketing talk.
I bought such JDS unit (didn't even know about the FY-stuff at the time). The little I have used it so far, the biggest problem isn't tiny errors in the DAC, but larger issues at higher frequencies/amplitudes caused probably by the amplifiers, and power supply noises etc.  But I guess the biggest issue depends on what the device is used for.

 It's easy for me to point out the deficiencies of the poor man's version of the FY6600 (not that that didn't have its own defects including the very serious issue of the early firmware version 3.0 bricking the AWG). However, since people are still being tempted by its lower price point (about £6 cheaper than what I paid for my FY6600 just over a month ago) and improved 'haptics', ignoring the loss of the additional features of the FY6600, I think it's well worth reminding anyone, who has yet to sample one of these gloriously cheap AWGs for themselves, of this particular deficiency for which no practical DIY remedy exists (unlike the other issues of the PSU, the clock chip, the high voltage output opamp and the kakamaimee output attenuator that's automatically switched in for the sub 501mV p-p output amplitude range - not forgetting the need to add a small 40 or 50 mm cooling fan to stop it cooking itself into an early grave).

 I most definitely saw a video where the entirely predictable glitches at the zero crossing and half amplitude points of its sine wave output being unambiguously displayed on a 'scope trace and commented on by the reviewer. Unfortunately, I haven't yet been able to track down this review and tear-down video for all to see for themselves rather than simply just having to take my word for it. 

 Regards, Johnny B Good

[Johnny B Good]

On the subject of replacing the power supply, here's a video of one guy experimenting with using DC-DC converters:

====link removed====

 I've seen this guy's earlier videos on the subject. I can't recall watching this video before so thanks for that link. It was an interesting insight into how a project can get out of hand (BTDT&GTBTS!). This is yet another video on how not to eradicate the half mains leakage of a cheap class II smpsu free of the HF switching transients that get through via common mode conduction path on the output wiring when you remove that bloody 1 or 2nF class Y capacitor responsible for the problem as a "quick 'n' dirty fix".

 He's on the right lines but I do have to question the wisdom of incorporating the mains transformer and rectifier pack inside the box. It strikes me that a better solution would be to use an external 16 to 20 vac 15VA transformer supply, leaving the rectifier and smoothing inside from which to feed the +/-12 or 15 volt and +5v dc-dc converter modules.

 Such modules are quite tiny so could be mounted clear of the original PSU board to leave the way open for an improved smpsu board equipped with a ferrite transformer blessed with the shielding foil layer to screen the secondary windings from the high voltage switching transients on the primary winding, neatly obviating any need for that accursed class Y emc bodge capacitor to produce a supply with a similar to, if not better than, isolation of a supply using a small conventional mains transformer.

 If he'd taken that approach to keep his future PSU upgrade options open and placed the dc-dc converter modules in the space between the existing PSU board location and the front panel, he could have made a neater job of adding fan cooling as I've previously described using a slimline 50mm 12v fan powered off the 5 volt supply mounted onto the base between the PSU and the rear panel (and blocking off those useless rear vent slots for good measure).

 The side vents are actually quite effective once you remove reliance upon the thermo-siphon effect alone as a passive cooling solution and provide just a modicum of forced ventilation in this way. It also helps to "gasflow" the accidentally provided exhaust leakage path provided by the RHS front panel clip slot in the base of the case to help dissipate the heat which builds up under the main board. 

 Regards, Johnny B Good

[CDaniel]

I replaced that crappy cheap power supply with a transformer and linear stabilizators . Doesn't matter if the transformer is inside or not , as long it's a good one ( with some shielding ) .

[Miti]

There's nothing wrong with the power supply other than the stray voltage that can be easily corrected. But if you're having fun, sure, why not?

https://www.eevblog.com/forum/testgear/feeltech-fy6600-60mhz-2-ch-vco-function-arbitrary-waveform-signal-generator/1525/

[rhb]

I sent a pitch to @"Siglent America" for an SDS1204X-E loaner for 2-3 months to do a side by side comparison to the GW Instek MSO-2204EA a couple of days ago.  No response yet.  They are both Zynq based DSOs with comparable features.  The Siglent being about 1/2 the price of the Instek once you buy the LA and AWG options.

The FY6600  FP FW is looking unlikely as a PM to fremen67 did not produce a reply about getting the source code.  But people do travel on business or take vacations, etc.  So it might still happen.

The FP is  a pretty simple state machine.  My experience is almost entirely exotic DSP routines running on high end workstations and larger.  But I don't see anything that worries me.  It's just another thing I've not done before.  That was my stock in trade my entire career.  And the thing that made going to work so much fun. But at 65 with $50/bbl oil I can't ever go back to that.

Now visual persistence on a DSO, *that* scares me.  My LeCroy DDA-125 has up to 32 S of persistence and then infinite.  The latter is easy, but I can't figure out how you subtract what came in 32 S ago and remove it without eating a huge amount of resources.

All of this is tightly intertwined with the corruption of higher education and the cost of even a very modest test bench.  Most of my electronics efforts were when I had only a Heathkit IO-18 5 MHz recurrent sweep scope and Radio Shack VOM and DMM.  All my signal sources were built from datasheet circuits on perfboard.

Realistically, less than 5% of the population is university level science, engineering and math material.  And that's all we need as a society.  We need similar proportions of mechanics, plumbers, carpenters, machinists and such.  But the mania for trying to make silk purses out of sow's ears has destroyed education for all except the autodidact.

Unfortunately, lathes, milling machines, table saws, wrenches,  DSOs and such are so expensive as to be out of reach for most and  a serious barrier to the autodidact.

The Zynq and Gnu/Linux have been strong drivers in reducing the cost of DSOs.  A FOSS FW stack would reduce the NRE costs.  If I can write Verilog code which is portable across the Zynq and Cyclone V, the two FPGA lines with embedded ARM cores, then the barrier to market is lowered substantially and bug fixes become much easier.  During the workstation wars in the early 90's I ported 500K lines of VAX FORTRAN to six different Unix systems. After doing the first 4, the last two took 4 hours.  That was arguably the most important lesson of my career.  Develop and test simultaneously on as many systems as you can a hundred lines or so at a time.  I wrote two 15,000 line libraries on that contract that never had a bug reported in over 6 years. Though the test system did catch when Sun failed to set errno properly after a call to getcwd(3C).

Fundamentally I am doing what research scientist/programmers do.  Things I've never done before.  That's why it's so much fun.

And the pain of having limited financial resources, doing all the HW upgrades on a V 3.0 FY6600 and then it borking itself is quite palpable. 

But, Johnny, be good.

[Johnny B Good]

I replaced that crappy cheap power supply with a transformer and linear stabilizators . Doesn't matter if the transformer is inside or not , as long it's a good one ( with some shielding ) .

 That's true enough. The point I was making was that the bulky transformer(s) could be kept outside of the case, allowing the rectifier and smoothing cap to sit where the original smpsu had sat, leaving ample space for the three converter modules in the space between the front panel and the PSU location and still allow a 50mm cooling fan to be fitted in the bottom of the case in the remaining space to the rear of the PSU location, above which said fan, would be the ideal location to fit a 0.1ppm oscillator board (10MHz with the original clock on the MB replaced with a 5x pll multiplier chip or else as I've done, a 50MHz unit feeding the now vacated clock chip location) mounted at a rather jaunty 45 degree angle to keep it well ventilated with the incoming cooling air and free of the heat pollution it would otherwise suffer if mounted in the original clock chip location. In my case, this also improves the internal airflow by virtue of it acting as a deflector to direct the airflow more forcefully towards the front so as to increase the amount of air flowing past the heatsinked opamps.

 This leaves the option to fit a better grade of smpsu with a shielded transformer that obviates the nonsense of connecting the output ground return rail to the half live mains voltage via a 1 or 2 nF Y class capacitor as is, annoyingly, an all too common practice with such small Class II smpsu wallwarts and commodity PSU boards used in low wattage mains powered devices.

 The quest to find a suitable low leakage 12 to 15 watt rated wallwart is pretty hopeless but amongst my collection, I did discover three Nokia 5.7v 800mA phone charger wallwarts which read only 6.5 to 10.5vac on my 10Mohm DVM with respect to ground. I actually broke open the one of the three ever so slightly different implementations which had been not only screwed together but also glued (the other two were held together by only three screws in one case and two screws and a clip release in the other). Stupidly, I had assumed all three were identical in the finer detail of their construction, choosing on the basis of the one with the lowest leakage voltage reading only to discover that this was the one that not only needed the two screws to be remove but also the use of a junior hacksaw and some leverage with a screwdriver to crack it open.

 Never mind, I can use it to replace a similar wallwart cct board that I'd fitted into an ITX 140W PSU box to beef up its weedy 100mA 5VSB rail. The extra 0.7v will nicely compensate for the anti backfeed diode I used to stop the existing 5VSB from causing this supplementary 5VSB cct board to go into an overvolt shutdown state. It can go into my FreeNAS spares box for now. 

 What I haven't yet done is check for the presence of common mode induced switching hash on the output lead. Looking at the one I broke open, it's not obvious as to whether the transformer is one with a screening foil layer between the primary and secondary windings to properly account for the lack of that dreaded Y class 1nF emc bodge capacitor.

 It'll be interesting to see whether it's a manufacturer's decision to simply omit the capacitor to emulate the capacitorectomy mod I've already done with the original PSU or whether they've actually justified such omission legitimately by using a screened transformer to meet the EMC directive. From what I've seen, I suspect not.

 I think I might land up attempting to rewind the low voltage secondary of an existing 12v 1.5 or 2 amp wallwart transformer so as to insert a screening foil of my own. I've certainly got plenty of such wallwart smpsus to experiment with. 

 Regards, Johnny B Good

[beanflying]

Interesting reading back on this flawed but really good value for money Box as it gets to EOL. There are some very cheap 50MHz units on evilbay at present if anyone is in the market with not to many $$ to spend as of today.

Mine after 15 months or so is about to get mothballed (Replacement is a Siglent 2000X) rather than sold for the few $ I might get for it so I chased up the last versions of the Windoze software and any updated Labview drivers and such to pack away with it.

Overall unlike some here my experience has overall been a positive one after sorting out the floating power supply and with a little trial and the correct tongue angle improving some levels on the board. Frequency has always been out by a bit but generally this is a non issue and if it was I checked and tweaked a few counts if needed.

Sub 10Meg for the complete software package that allows Arb Signal creation and driving the front panel via USB kicks the butt of the Siglent by comparison and as per the couple of comparison waveforms @30MHz Sine and 10MHz Square it holds up fairly well against a Generator at current pricing 10 times more $$. Coin toss for the Sine and suffering a little on the square.

Remembered fondly as it goes to the 'cupboard' to maybe never return 

[GregDunn]

It sure does what I need with plenty of room to spare; mostly audio functionality testing, swept frequency, IM, etc. etc. and it can be quickly set up from the front panel to do rather complex stuff.  If I need low distortion, I use my Heath IG-5218, and if I just need a signal at an approximate frequency to spot-check something, it's the Wavetek 182 (the two of which together cost even less than the FY6800).  No, it's not a pro generator, or a real RF tester, but that's not what I expected when I bought it.

[Johnny B Good]

Interesting reading back on this flawed but really good value for money Box as it gets to EOL. There are some very cheap 50MHz units on evilbay at present if anyone is in the market with not to many $$ to spend as of today.

Mine after 15 months or so is about to get mothballed (Replacement is a Siglent 2000X) rather than sold for the few $ I might get for it so I chased up the last versions of the Windoze software and any updated Labview drivers and such to pack away with it.

Overall unlike some here my experience has overall been a positive one after sorting out the floating power supply and with a little trial and the correct tongue angle improving some levels on the board. Frequency has always been out by a bit but generally this is a non issue and if it was I checked and tweaked a few counts if needed.

Sub 10Meg for the complete software package that allows Arb Signal creation and driving the front panel via USB kicks the butt of the Siglent by comparison and as per the couple of comparison waveforms @30MHz Sine and 10MHz Square it holds up fairly well against a Generator at current pricing 10 times more $$. Coin toss for the Sine and suffering a little on the square.

Remembered fondly as it goes to the 'cupboard' to maybe never return 

 Hi Beanflying,

 I was beginning to think the next post would be another follow up from me. It seems a shame to retire your Feeltech AWG to the toy cupboard after all the fun times you've had turning it from a sow's ear into a silken purse. I guess you didn't put in quite as much effort as I did to be able to throw it aside like that (I can't remember just how much work you did put into it this past year, I just know you were a regular contributor to this ongoing thread). Mind you, new toys and all that, perhaps I'd do exactly the same if (when) I bought a more upmarket replacement. :-)

 Anyway, it's been a remarkable 'Bean counteritus mitigation upgrade project' that makes all previous such projects I've had over the past 40 odd years pale into insignificance. I never realised there could be so many ways to waste a perfectly good AWG main circuit board by rampant beancounteritus. It's almost as if Feeltech could read the minds of their target market and went all out to give it the upgrade project of the century. 

 After spending a few weeks obsessing over the calibration accuracy of the 0.1ppm 50MHz TCXO module upgrade and lamenting the lack of any WWV transmissions (notably the 10MHz broadcast), I finally decided to tackle the PSU issue once and for all.

 I've spent the past two or three weeks searching for low leakage smpsus I could use with or without add on DC-DC converters to create an ultra low leakage three rail PSU. I even considered medical grade 12W (5 or 15 volt) wall warts from which to extract the smpsu board but gave up when I realised the price premium on something that only cost a penny or two more for its shielded transformer would make a mockery of my penny pinching signal generator purchase.

 The penultimate nail in the coffin of that "Holy Grail" search came when I discovered a complete lack of suitable DC-DC converter modules I could drive from a conventional mains transformer and rectifier smoothing pack to obviate the accursed Y cap. The final nail was when I discovered just how inefficient those specialised mains isolating transformers were as a consequence of their split bobbin construction.

 This was when I decided to 'Think Outside The Box' and came up with the clever idea of linking the common zero volt rails of a pair of smpsus wired to the mains in 'anti-phase' so their leakage currents would cancel (I was thinking of using the original for the +/-12v and an additional one for the 5v rails). I was so convinced that this 'neat solution' would work, I jumped out of my bed in the middle of the night to put my clever hypothesis to the test. Sadly, it didn't work because it was based on faulty reasoning so was doomed from the beginning.

 Never mind, I had another neat idea concerning a method to 'null out' (or buck) the unwanted half live mains leakage voltage. This idea was based on using a 1VA pcb mount 240 to 120 volt isolating transformer to provide an oppositely phased source of half live mains I could use to cancel out the unwanted touch voltage. Unfortunately, this was also doomed to failure (and for essentially the same reason that my "two smpsus in anti-parallel connection to the mains supply" hypothesis had been doomed).

 In both cases, bucking (or nulling out) the unwanted leakage voltage relies upon tying the anti-phase generator to the neutral line wrt to the supply rather than to an arbitrary mains input wire on the smpsu(s) concerned as I'd originally considered. The key to such a scheme is that you need to identify the neutral by which to reference your anti-phase source in order to successfully cancel out the leakage voltage. Although it's not too difficult to automatically detect which way round the supply is connected and automatically correct the polarity so as to ensure such nulling out exercises will succeed, such complexity is needless.

 The whole nulling out process becomes somewhat redundant since you can simply bridge the 0v rail to the neutral via a 100K high voltage safety resistor which will bog down the half live mains voltage on a 240v supply right down to a mere 7.5v - low enough not to harm any ESD sensitive devices under test. With that in mind, it then begs the question as to why invoke all the extra complications just to link the 0v rail via a safety resistor to a ground referenced connection (the neutral) to shunt the leakage voltage to a safe level when one might just as well use the safety earth to achieve the same effect without such complexity and less risk of a non-fatal but unpleasant electric shock when Sod's law inevitably messes up the auto polarity detection. Better yet, a real earth lets you use a 10K resistor, still large enough to avoid hum loops but even lower so as to reduce the leakage voltage to just half a volt on a 240v supply.

 It was this which led me into finally using my variant of the "Upgrade from class II to a class I earthed" configuration by adding a 3 wire IEC connector (C6 or C12/13) solution which most everyone else had chosen as one of their very first modifications to the FY6600, typically electing to use a 10 to 100nF capacitor in place of my 10K leakage voltage suppression resistor or else directly bond the safety earth connection to the common ground rail, sometimes using a "Grounded/Floating" option switch to obtain the best of both worlds (risky if the switch isn't bridged by a leakage mitigation resistor in the range of 10K to 100K and also prone to operator forgetfulness).

 I chose the slightly more tricky C6 socket option to avoid the "Tail wagging the dog" syndrome of hanging a 700g box on the end of a 10A 3 core mains lead, spending most of yesterday's free time fabricating the aluminium reinforcement and support plate required to accept the slot in C6 connector I'd recovered from a scrapped laptop charging brick. Unlike others' socket upgrades, this plate is simply bolted onto the flimsy rear panel which measure is more than adequate to prevent it bending and popping out of the case half retaining slots.

 It's taken me this long to get around to doing this job because I wanted to avoid this messy business of a brute force and ignorance pragmatic fix since I'd once held out hope of a more refined and elegant solution to eliminating the half live mains leakage hazard. It's only now, after days of fruitless searching for affordable low leakage smpsus and trying to think up cunning ways around the problem that I've come to realise that the pragmatic brute force and ignorance approach turns out, in this case, to be the one and only sensible option after all. Who (honestly) knew? :-)

 Following on from that, this modification seems to have cured a rather annoying feature whereby it would stop generating signals on both channels whenever connecting anything earthed produced an ESD event depending on the charge state of the Y cap at the time of connecting. TBH, I'm not sure whether this new symptom is a consequence of the TCXO modification or an existing feature that I only happened to notice after doing the TCXO mod.

 Anyhow, as I'd been hoping, the PSU leakage fix does seem to have cured this annoying habit of disabling both channel outputs whenever connecting up to an earthed device under test (and even when disconnecting) requiring a power down restart to restore the signal. I've just assumed that this is an unmentioned foible of the FY6600 series in general as a consequence of the psu leakage current issue, perhaps largely limited to those of us running it from a UK 240 or a European 220v mains outlet. I wonder whether anyone else has noticed this particular quirk?

 Incidentally, whilst on the question of funny quirks, has anyone else experienced the random loss of digit selection when using the left and right arrow digit selection buttons? This one is less annoying only because a simple disable/enable sequence on the channel select buttons restores the digit selection operation rather than require a power down/up reset. It's just a possibility that it might also be related to the psu leakage issue but I won't know whether this is the case until I've done some more testing.

 I think the only outstanding issue now is that of the attenuator used to handle the 1 to 500mV output range using the wrong resistor values, causing a terrible mismatch to 50 ohm loads. Feeltech have fudged this in the firmware by compensating for this error when driving Hi-Z loads. I did look into fixing this one with a change of the resistor values after working out what would be needed to create a 50 ohm pad with an attenuation that matches the Hi-Z loading condition they've actually compensated for (ie one that results in the 6dB drop when loaded with a 50 ohm load). Since this involves reworking smd components, I only want to do this job the once and get it right first time round so I've put it on the back burner for now until someone else can offer a sanity check on my recalculated resistor values. Has anyone else tackled this particular quirk?

 The resistor values they've used are a 510 ohm series element with a couple of 100 ohm shunt elements which turns out to be a 22.26dB attenuator for an 84.456 ohm impedance line. I suspect they were aiming for a 50 ohm 20dB attenuator using 56 ohm shunt resistors with the 510 ohm series element and somehow landed up  placing 100 ohm resistors instead of the specified 56 ohm ones through some sort of cockup in the manufacturing process and, rather than rework the boards with the correct resistors, they've chosen to compensate for the Hi-Z condition in the firmware and hope nobody important notices this blatant manufacturing error.

 Luckily, there is a hardware fix for this cockup (just as well seeing as we're not likely to see a firmware plus hardware fix out of Feeltech support) which involves calculating a new 50 ohm attenuator network that produces an attenuation figure 6.02dB lower than the Hi-Z attenuation factor of the current attenuator network for which they applied their firmware compensation fix. I think, going from my rough notes on the back of an envelope (literally!), this calls for a 25.71dB 50 ohm attenuator network to correct the problem with the current firmware left unchanged.

 Plugging the appropriate values into an on line matching Pi attenuator calculator, the ideal resistor values are 55.465 for the shunt elements with a 481.141 series element. trying alternate 'preferred values' in the hope of minimising the reworking, leaving the existing 510 series element as is and replacing the 100 ohm shunts with 56 ohm resistors will produce a reasonably close match giving a slightly higher attenuation value of 26.09dB with a 50.706 ohm matching impedance. Although this will produce a slight departure from monotonicity (not obviously apparent under Hi-Z conditions as things currently stand) as the level is stepped through the 500mV threshold, the very obvious discontinuity under 50 ohm loaded condition should all but vanish. I haven't delved any deeper into a more precise analysis to check just how big a glitch in voltage amplitude this might represent using those preferred 'quick fix' values. I'll leave that as an exercise for the interested reader to tackle. :-)

 Regards, Johnny B Good

[beanflying]

Putting in the cupboard and moving on is really an extension of gaining some more accurate equipment that you 'know' will output or input closer to the truth and provide a better series of ranges.

I made do for a very long while without one at all. Having one made life easier and saved me time using bread boarded or lashed up worse performing oscillators. So for someone without they are still a good thing with a simple mod to the power supply to improve the floating outputs. Much beyond that time and money spent to still get a box with still questionable accuracy and output limits doesn't make logical sense to me.

On a budget or after a first Sig Gen you can do way worse for more money. 

[Johnny B Good]

Putting in the cupboard and moving on is really an extension of gaining some more accurate equipment that you 'know' will output or input closer to the truth and provide a better series of ranges.

I made do for a very long while without one at all. Having one made life easier and saved me time using bread boarded or lashed up worse performing oscillators. So for someone without they are still a good thing with a simple mod to the power supply to improve the floating outputs. Much beyond that time and money spent to still get a box with still questionable accuracy and output limits doesn't make logical sense to me.

On a budget or after a first Sig Gen you can do way worse for more money. 

 Hi Beanflying,

 I do take your point but, right now, I'm not at that stage where I'd care to invest double what I paid for my Siglent SDS1202X-E dso just to get a slightly more refined version of my modded FY6600-60M signal generator (actually, as far as the front panel buttons go, hugely refined! - it's only the 600 quid saving which makes this lack of refinement acceptable... for now).

 If I was relying upon such T&M gear for my living, I'd certainly be prepared to make a greater financial investment but I'm now three years retired from the business of repairing desktop PCs, a trade which more or less died off some 4 or 5 years before I finally called it quits to collect my pension. Whilst I do have a reasonable lump sum languishing in low interest rate accounts to be burnt through, I can't really justify blowing some 10 to 15 grand of it on 'low end' decent gear, at least not right now.

 As you say, it's best to just shelve the FY6600 as a "get you by spare" rather than just sell it off for a paltry sum, especially when you've invested some time and effort improving its performance into the low end quality gear territory with price tags an order or two of magnitude greater. The factory upgrade option to a 0.1ppm TCXO on the Keysight 120MHz AWG is 700 quid alone which rather puts my 15 quid investment to do the same upgrade somewhat into perspective. 

 I've applied several mods in the past two months since I purchased the unit last November. The list (in order of priority with the benefit of hindsight rather than the actual order I did them in) is:-

 Add a cooling fan, a highly desirable upgrade on the woefully inadequate passive ventilation, destined to cook it into an early grave.

 Add a C6 mains socket (in preference to the larger C12/13 socket if wish to avoid the "Tail wagging the dog" effect) and use a 10K resistor (not a 10 to 100nF cap) to link the 0v line to the protective earth pin to kill the unwanted half live 'Touch Voltage' without creating an earth loop problem,

 Modify the PSU to raise the 5v rail (4.95v measured) to 5.49v in order to raise the +/- 12 rails from the 11.4v mark into 13.7v territory upgrading the 12v rectifiers to heavy duty 20A rated dual shotcky rectifiers, leaving the 5v diode as is to maximise the voltage boost obtained by winding an extra two turns per "12volt" winding on the transformer (don't bother fitting 25v caps) and definitely don't try converting the 12v rectifier into a fullwave bridge by adding the 'missing' diode elements, DAMHIK, IJK. 

[Edit 20190722]
 I found a much better alternative (which I described in a later posting) to adding two turns onto each end of the 12v windings - leave the 12v secondary untouched (I'd had to undo this modification) and wind a single turn to buck the 5v winding output voltage (easily accomplished without the need to remove the transformer).

 I'd used thicker wire than was needed, making a rod for my own back when Sod's Law ensured I'd connected it the wrong way round - put half a dozen single turn windings wired in parallel (plenty of room on the bobbin) to get a well coupled low resistance/leakage single turn winding to connect in series with the 5v rectifier diode (lift the anode out, wire one end in the vacated hole and the other end soldered to the lifted out anode lead).

 In spite of the voltage increase on the +/-12v rails (now at +/-13.7v mark), the power consumption actually dropped a fraction of a watt. Mind you, I'd replaced the 47K resistor in the voltage feedback network with a 200K to reduce the 5.49v on the 5v rail down to 5.09v which would have made some modest contribution to the achieved energy savings but I suspect most would have come from the efficiency improvement I'd achieved over the original adding of two turns at each end of the 12v winding modification.
[EndEdit]


 Upgrade the single THS3002 dual opamp to a pair of THS3001 (at a minimum) or, better yet,  THS3095 or ( a little OTT)  THS3491 opamps to eliminate sine wave distortion at 20V p-p when close to the 20MHz limit and driving 50 ohm loads.

 Upgrade the really shitty little commodity 20ppm rated 50MHz clock chip to a 0.1ppm TCXO module that can be located away from the vicinity of the three analogue voltage regulators that raise the board temperature up to 70 deg C, causing the original shitty oscillator chip to run at 50 deg C - the TCXO module can be located in the incoming airflow provided by the cooling fan.

[EDIT 2020-04-14]

 Finally, correct the output attenuator network that's used to reduce DAC and amplifier noise in the voltage output range of 1 to 500mV by...

paralleling each of the four 100 ohm shunt resistors (marked RS 1,2,5,6) with 124 ohm resistors and placing an 8K2 resistor in parallel with each of the two 510 ohm series elements (RS 3,4).

 You can use 120 ohms instead of the 124 ohm values since you'll likely have to trim the 8K2 resistors anyway to get an exact match to the unloaded attenuation corrected for by the firmware for the original 84.46 ohm impedance 22.26 dB attenuator that had been erroneously created (it looks like the 100 ohm shunt elements were placed in error for 56 ohm resistors which would have created a 26dB 50.706 ohm attenuator with that 510 ohm series element value). You can compensate such an error in the firmware but only for either the terminated condition or the unterminated Hi-Z condition, not both, so it looks like Feeltech opted to compensate for the Hi-Z condition with their firmware fix and hope no one notices the glaring error in the terminated case.

 I'd originally envisioned this modification as an smd reworking task but, after looking at pictures of the mainboard, realised it could be done using the smallest of wire ended resistors in parallel with the existing smd resistors which otherwise would have been replaced with lower value smd resistors (480 ohm series element resistor with a pair of 55.36 ohm shunt resistor elements per attenuator pad assuming 0.1% tolerance resistors in the original circuit). Since it's highly likely Feeltech only used 1% tolerance resistors at best, a better way to make sure of a good match to the original Hi-Z attenuation, they seemed to have precisely compensated for in the firmware, is to wire up a variable 10K resistor across the series element after soldering in the 124 or 120 ohm shunt resistors into the circuit and adjust for no discontinuity in level as you transit the relay switching point at the 499/500/501 mV mark other than the selected mV setting. This test should be done at a low frequency (say 1KHz) so that the stray L and C in the foot or so of wire between the board and the 10K pot have no detrimental effect - we're just finding the exact resistive value required in order to pick out a fixed value resistor by selection testing suitable candidates from our parts collection.

 The same 'select on test' procedure could be used to gather the desired 124 ohm 'rogues' from the 120 ohm parts bin - desirable but not essential although avoiding samples that are on the minus side of the tolerance range are best weeded out just the same.

 This might seem to some as rather a lot of trouble to go to (selecting on test and a variable resistor to tweak to a precise Hi-Z attenuation value) but if you're going to have a go at fixing this manufacturing defect at all, that extra faffing about is a trivial thing compared to the process of soldering in the additional wire ended resistors. After all, if a job's worth doing, then it's worth doing well or not at all. The motivation in this case being a more tightly specced output impedance than you typically see in gear costing one or two orders of magnitude more.    Any EE (amateur or pro) with any pride at all wouldn't need reminding about the worth of doing a job well.

 Forget all that tosh. The design and its firmware had been predicated on the multiply by 10 add 20dB rule. That cockamamie 85 ohm attenuator just happened to be the cheapest way to get 20 dB in the Hi-Z load case. It had replaced the proper 20dB 50 ohm attenuator pad originally specified and made up from the most expensive E192 (0.5% tolerance) preferred values range using 61.2 ohm shunt resistors with a 249 ohm series pass element. I guess the chief bean counter must have threatened the chief designer with the sack if he didn't find a seriously cheaper alternative (meaning anything that could substitute for a real 20dB attenuator in the Hi-Z loading case and forget all about maintaining any semblance of 'Technical Standards').

[END_EDIT 2020-04-14]

 The final improvement I may end up attempting, would be to improve the action of the rather horrible front panel buttons, in particular the left and right arrow ones beneath the rotary encoder with the emphasis on the right-hand button which still causes the selected digit cursor to vanish from time to time, necessitating a cycling of the channel select button to get it back (at least it doesn't require a power cycling reset like the esd earthing contact transient used to do prior to my mains socket and 11k earthing resistor mod).

 The problem seems to be the result of the excessive force required in just the right place on the arrow button to get it to register the key-press possibly causing a cracked solder joint (hopefully not a cracked trace) nearby to open up on the front panel PCB causing a glitch in the cursor positioning subroutine. Whilst the left hand button produces a satisfyingly tactile toggling effect, the RHS one lacks such tactility, suggesting a faulty switch behind the membrane (hopefully an actual solderable into/onto the board micro-miniature momentary contact pcb mounted switch rather than a rubber membrane switch with a carbonised tip to bridge contact traces on the underlying PCB after the fashion of a cheap TV remote controller).

 The front panel is the one component I've not removed for closer inspection since, until recently, I've not had cause to apply an 'upgrade' to it. It's only now that I'm considering taking a closer look, not so much as an upgrade exercise (although a repair could be deemed as an upgrade from 'not working' to 'working') so much as a repair job, so I'm in blissful ignorance as to what lies beneath the actual switch locations underneath the membrane bumps. Has anyone taken a closer look at the front panel switches?

 At some time in the not too distant future, I'll be taking a very close look at the 'arrow' buttons with a view to 'upgrading them (or at least the RHS one) into a properly working condition. I don't know what awaits such a close examination so there's a chance that nothing short of a replacement front panel will resolve this issue. Since the FY6800 model appears to be an FY6600 with an upgraded front panel, I might, in that eventuality, land up buying a 60MHz version from which I may elect to cannibalise its front panel as a replacement for the FY6600 in view of all the work that's already been invested thus far. However, that does rather depend on just how similar the main boards actually are between the two models. If the front panel transplant proves to be impractical, that just means I'll be treading a well worn upgrade path to bring the FY6800 up to scratch.

 I'm not quite ready to 'throw in the towel' and splurge 700 quid on a Siglent SDG2122X  just yet! 

Regards, Johnny B Good

[Noy]

Ist it worth to get a FY6800 if you already have a Toellner 7401 and the build in AWG from Rigol MSO5000?

Or better a old wavetek Model 275? I think the wavetek is to old or? Waveform has to be tipped in over the buttons because i don't have a GPIB..