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

[rhb]

It appears to be a sampling artifact.  Here's what an RSP2 says set to 5 MHz BW.  10 MHz BW is similar.  The display is *very* different with narrower BW to the point of not seeing a pronounced peak at 40 MHz.  Not sure what is going on.

I've added a shot with the RSP2 set to 4 MHz BW.  Very strange.

[Ebel0410]

It appears to be a sampling artifact.  Here's what an RSP2 says set to 5 MHz BW.  10 MHz BW is similar.  The display is *very* different with narrower BW to the point of not seeing a pronounced peak at 40 MHz.  Not sure what is going on.

I've added a shot with the RSP2 set to 4 MHz BW.  Very strange.

Actually it isn't so strange at all, a SDR is not a spectrum analyser with professional specifications.
I think that relevant spectrum analysis can only be done with a spectrum analyser, and fortunately I own one.
I'll send full detailed spectrum traces to that thread in the next coming days if someone is interested in (no time for doing that right now)

[rhb]

Turns out the front end was being overdriven and went into oscillation when the bandwidth changed.  Not yet sure why it does that.

An SA is the next item on my shopping list, but good ones are expensive.  The RSP2 does a pretty good job over a 10 MHz span.  So combined with the FY-6600 it should be usable for checking filters for HF.

The sidebands displayed by the RSP2 are very different from the picture presented by the scope FFT.  For close in, I'm much more inclined to believe the RSP2 as the RBW is much smaller.  There's really not a lot of control of the FFT parameters on the scope and the dynamic range is also smaller.

For "no-signal" I just disconnected the AWG.  It seems likely that the RSP2 close in spectrum is the more accurate.

It would be very helpful to have spectra from a calibrated commercial SA to compare to the RSP2 and scope FFT plots.  I'm trying to sort out which of these things I should believe and when.  So references are very helpful.

I fitted a grounded plug to my FY-6600, but still need to get Y caps for it.

[cybermaus]

I fitted a grounded plug to my FY-6600, but still need to get Y caps for it.

In my view:

Y-cap are safety. Because mains to any touchable item *must* be safe, even when it fails, they must be super safe, tested to 10KV as well as (I presume) lots of mechanical stresses. That is why the capacitor spanning the transformer must be a Y-cap, as it is potentially violating the galvanic isolation.


But you are planning to put a cap between an already safe item (the BNC is already touchable safe) and earth (also already touchable)
So while the habit for these things is to talk about y-caps, other caps should also be OK. I did put a Y cap, because I had one, but on my other 2 supplies that I modified, I put X-caps (because I had only 1 Y-cap)

Also, as I wrote earlier, the mains-to-0V Y-cap should be significantly higher then the transformer parasitic capacitance
And the cap from 0V-to-earth should be significantly higher than that to divide up the floating mid-voltage to null potential as much as possible.

But it should not be too high. At least, not if you really want to use the signal gen as floating.
Because it does have "probe loading" consequences. If/When you connect the signal to a DUT that is not 0V potential, it will first have to charge the cap to whatever potential the DUT is. And you do not want that to be too much of a jolt.


So the original Y-Cap is about 1nF (I measured)
The extra Y-cap I used is 4.7nF (that was the only one I had)
I would try to not make it higher then 10nF


Of course, I am new to this particular topic, so do take my view with some critical thinking.

It would be very helpful to have spectra from a calibrated commercial SA to compare to the RSP2 and scope FFT plots.  I'm trying to sort out which of these things I should believe and when.  So references are very helpful.

I was also playing with SDR based SA. But I did not get any result that I would think usable. Nice to locate FM radio stations.
But as electronics SA: when you know what peaks to expect, you can find them. But only being able to find what you already know is there is of limited use. And I have to many other peaks that I suspect are artifacts, clouding up the picture.

Of course, my SDR is one of those toy DVB-RTL sticks that I had lying around. That RSP2 does look like it is giving better screens, but are they true?
If not, it may be better to not spend $150 for a RSP2, and put that toward a real SA instead?

[beanflying]

Would like a SA but luckily I can 'borrow' some time on a nice HP if needed 

Toy Budget left this week -$500+ pushing four figures 

[rhb]

There are lots of spurs as there are in any radio,  SDR or analog.  I'm quite happy with the RSP2.  The software less so, but SDRuno seems to provide pretty usable displays.  Now that I have serviceable signal sources I think I'll start a thread on the RSP2 showing it's application to filter characterization, etc.

The thing that is needed most for using an SDR as an SA is the ability to store a reference trace and then divide by the reference trace.  That would allow correcting for internal self noise and source variations and is a standard feature on an SA. Hopefully someone with access to an SA and an FY-6600 will repeat my examples and post them.  I'm most interested in the "signal", "no-signal" example with the small RBW.  I'd really like to know if that's a proper result.  The software aspect of T&M kit makes me reluctant to trust anything very much.

The LimeSDRmini should be a good low cost, wide range SA-TG with appropriate packaging and software.  I've got a LimeSDR which I bought mostly for use as a test instrument.  Still working on suitable packaging.  They get hot and they need shielding.  The RTL and RSP are based on clever use of TV tuner chipsets.  There's a wave of much more general chips coming along which should provide substantially better performance.

Now that I've started using Windows, I'm considering a VNWA from SSDR-kits very seriously.  The support in the yahoo group is excellent.  For a long time I refused to use Windows, but I find Linux not all that much better and there is far more Windows EE software available.

[Ebel0410]

Hi folks
for those who are interested in some measures relatives to the HF behaviour of our beloved one, here it is.
SA is a Rohde & Schwarz, Professional reference in Spectrum analysis.

[technogeeky]

Hi folks
for those who are interested in some measures relatives to the HF behaviour of our beloved one, here it is.
SA is a Rohde & Schwarz, Professional reference in Spectrum analysis.

That's actually pretty damn impressive.

The two anomalies that I see are between picture 14 and picture 15 (why is there a 10dB rise in the noise floor, but only on the lower sideband?), and between pictures 16 and 17 (why would moving from 14.8 MHz to 14.9 MHz add some comb of spikes just above the noise floor?). I don't think either of these two things are deal breakers (after all, these two problems lie 70 dBc and 65 dBc down, respectively). But I would be curious to hear a hypothesis about them.


And just for reference, please edit the post to include the exact model of the specan and the R/S signal generator (not that I doubt their performance, but perhaps someone wants to get a sense of the instrument's noise floor and phase noise).

[cybermaus]

I also have the impression the source column in the table may be incorrect.
Or else I cannot explain 14,15 and 16,17, why measure the same thing twice on the same device?

[technogeeky]

I also have the impression the source column in the table may be incorrect.
Or else I cannot explain 14,15 and 16,17, why measure the same thing twice on the same device?

It's not the same. I think he must have noticed that a small difference in frequency (10.7M vs 10.8M, and 14.8M vs 14.9M) causes marked differences in output.

The center frequency is the same, but the actual main tone can be seen shifting in both cases.

[Ebel0410]

It's not the same. I think he must have noticed that a small difference in frequency (10.7M vs 10.8M, and 14.8M vs 14.9M) causes marked differences in output.
The center frequency is the same, but the actual main tone can be seen shifting in both cases.

Yes, that was my point, focusing on figures at particular frequencies.
For instance 10,7MHz, at this specific frequency, the noise floor raise suddenly 10db on the lower frequencies, really strange behaviour indeed.

The test equipments that I use are :
SMG = 801.0001.52 Rohde Schwarz signal generator 0.1-1000MHz
SA = FSC 3 Rohde Schwartz spectrum analyser 9k-3GHz

[hainjedaf]

Maybe a dub question, but I hope to learn:

I have a switchmode PSU (MeanWell PT-45B) kicking around. Would this be an improvement to replace the FY6600 psu?
Of course I'd add a grounded mains connector.

[cybermaus]

In my view: Yes, but No.

Its too big, physically. Its also a bit of a waste, because its unneededly big, electrically.

It will also not get you any better leakage result then simply adding a 3-prong and connecting to ground:
Note the specifications of that supply give you <0.5mA (or <500uA) leakage current, a lot worse than what is already in the FY6600. This is because the leakage is a function of the power a supply can deliver and as already stated the Meanwell is unneededly big.

It would really only improve the fact that the unknown Chinese PSU may blow its caps sooner rather then later and take the AWG with it.
Which is a big deal I suspect,

So while I am as yet choosing to keep the original supply, I cannot state that is a good idea. But that Meanwell is not a good choice.
Go with a smaller one. Or replace the 16 caps with 25V ones and add a 3-prong.

[hainjedaf]

So updating the caps in the original PSU would serve better purpose.

Where does one hook up the PE lead? Secondary ground?

[don.r]

So updating the caps in the original PSU would serve better purpose.

Where does one hook up the PE lead? Secondary ground?

That would be the easiest. The 25V caps don't fit in neatly as space is tight on the output side but they can be mounted a little off the board. Grounding should be on the secondary side, there is an empty pair of ground pads near the PSU output jumper that can be soldered to.

[xzswq21]

I wonder if those sidebands (at roughly 30KHz / 300pSec offset) are jitter from the oscillator.

Would be interesting to see if they still exists after replacing that oscillator with that NB3N502 PLL that supposedly only has 15pSec jitter
Either ArthurDent needs to check that, or I need to buy a decent Spectrum Analyser.

15ps is high
with 15ps of jitter and ideal analog output driver, we have:
@4.2MHz the ENOB<11Bits
@8.5MHz the ENOB<10Bits   ------>    SINAD<62
@15MHz the ENOB<9Bits      ------>     SINAD<60
@34MHz the ENOB<8Bits
@69MHz the ENOB<7Bits      ------>    SINAD<44
any bad layout, power supply noise, bad selection op-amp,... will ruin the performance.

[ArthurDent]

The array of modifications various posters have come up with to improve the FY6600 is interesting and depends a lot on what that particular user’s main requirements are. My first interest was mainly to make the FY6600 safer by including a real ground through a 3-wire power cord and not to rely on an iffy virtual ground.  Secondly I really wanted to make the timebase more stable and a LOT more accurate with the option of using an external 10 Mhz reference from GPS or a rubidium source. Along the way I discovered the existing power supply was crappy and not what I considered fixable so that had to be replaced as well. The grounding and timebase modifications I made worked as well as I hoped they would so that left the power supply and possibly the output amp that others considered lacking.

Recently I tried replacing the output amps as others had recommended and didn’t see that really improved anything for me and went back to the original single chip. I rarely go above 5 volts output so this modification wasn’t that high on my to-do list. Others have run the FY6600 main board on +/-15 to improve the output at higher levels with no reported problems but I was still worried about a possible failure. I had tried a +/-15 vdc supply I modified to put out about 14.2 vdc to perhaps give me a little more margin of safety. I was uneasy in running the main board at a full 15 volts with the caps on the board rated at 16 volts (as others have mentioned) plus I wasn’t so sure that some of the other parts wouldn’t be stressed by the higher voltage. Almost all SMPSs have one adjustment for the +5 and the other two supplies are ratios of that so to just use that adjustment to lower the +/-15 supplies to about 14.2 volts, which I thought might be safer, would mean that the +5 would be lowered as well by about .3 volts, and I didn’t find this acceptable. What I tried was to put a conventional 3-amp diode in series with the two + and – supplies which would drop those voltages by about .8 to .9 volts while leaving the +5 right on.

While that worked to lower the +/- voltages a little and made me feel a little better, the supply I used that just fit in the space in the case was a supply capable of supplying 60 watts. This meant that although the main board doesn’t require much power, the supply used 10 watts and most of that was given up as heat and the FY6600 ran warmer than I liked. The first 40 watt +/-12 vdc supply I tried was a little more efficient but still generated some heat and used about 8 watts. The supply I settled on was a 25 watt +/-12 vdc supply I have shown previously that uses only 6 watts when powering the FY6600 and 4-5 watts when the front switch is turned off but parts of the main board are still powered. Keep in mind that the OCXO that I used as a timebase stays powered when the back switch is on and does use about 1.5-1.8 watts when it is at temperature but that doesn’t contribute much internal heat.
 
Here are 2 photos. One shows how I added the diodes in series with the +/- supply lines and bypassed then with a couple of tantalum capacitors if anyone might want to try that. The second photo is of the 25 watt supply I showed before. I’m pretty sure these are all the modifications I’ll be making to the FY6600 but I’ll keep reading to see what other great ideas others come up with. In a way I spent more time with the FY6600 apart than with it together and doing what amounts to trying to make a Ford run like a Mercedes when I might have been better off buying the Mercedes in the first place! 

[ArthurDent]

Oh, and one thing I've wondered about and it might be worth checking into, is the failure rate on the FY6600. One reason a lot of people are concerned about the leakage voltage on the output is that it could zap whatever you're connecting the FY6600 to. Is it possible that if you have an ungrounded FY6600 that you could zap it by touching the keypad? Capacitive coupling of the leakage voltage through the keypad could possibly account for this. Maybe there should be a poll of those who have had failures, especially of the display being garbled, to see if there is a correlation between ungrounded units and number of failures, or if it's just random.

[beanflying]

Happy enough with mine with Earthing and extra Cap. Tweaks to the OpAmps were also well worth the effort IMO.

Disappointing that Feeltech haven't jumped back in to help out. The 'send it back to the factory' isn't adequate at this price point or day and age of 'easy' online firmware upgrades for plenty of other equipment.

If they chose to open source the firmware even and open themselves up to hackers and modders might see a boon in their sales numbers and be able to pick and choose mods by users to keep in the official firmware at little cost to themselves.

[rhb]

After I finished adding a ground and reassembled the unit, I ran it on the bench for an hour or so measuring the frequency with a counter and looking at the amplitude fall off at slightly over 50 MHz.  I put it aside (very cramped work space).  After supper I hooked it up to a scope to do some more analysis.  Borked display!

It works, but the display and UI are buggered, *exactly* the same as shown in the start of the thread on reflashing.  So it's obviously a pretty consistent defect.  I'm not clear on how cybermaus read the flash.  I chased some links, but didn't get anywhere.  I'll do some detailed testing to see if all the functions still work.

In looking at it, I notice that several chips next to the FPGA  have had the markings sanded off.  Has anyone else seen that?

Ironically, I had just given it a very positive review on eBay shortly before discovering it had borked itself.

 JK1 on the front panel board is the STM32 programming interface.  I just checked the pinout on the datasheet. Looks to be Spy-by-Wire.  I'll have to look at my collection of STMF32 boards to see if I can use one of them as a programmer/debugger.

There's been enough work done evaluating the unit, that I hope FeelTech will realize that they need to fix the bug and support the EEVblog community.  If it were reliable, it would easily be a very popular AWG.

[rhb]

All the waveforms seem to work except  SQUR, ADJ-PULSE and ARB33-64.  DC offset works. The other UI options seem a bit confused, but mostly seem to work.  A bit hard to test as I haven't had the unit long enough to get acquainted.

It is actually producing  NegRamp at the weird frequency that appears on the display at power up and the Vpp shown!

[cybermaus]

Damn. another one.

As to how I read the flash: I read the wrong flash.
Since my original happy post, we learned that I read the FPGA and waveform flash, but the GUI and control flash is in the STM32 and that is the one that fails. Which alas is protected from reading. So we are at a dead end there.

[jleg]

All the waveforms seem to work except  SQUR, ADJ-PULSE and ARB33-64.  DC offset works. The other UI options seem a bit confused, but mostly seem to work.  A bit hard to test as I haven't had the unit long enough to get acquainted.

at this very similar stage i did a "factory reset" (load factory defaults from system menu) - after that, also the last working wave forms vanished...

The interesting question is: what firmware version does your FY6600 have? The current "hypothesis" is, that it is propably a bug of V3.0 specifically..

[rhb]

Mine is version 3.0.  Thanks for the warning about doing a factory reset.

I think it's time to sniff the cables from the front panel.  Has anyone checked the USB - PC connection?

Based on Ebel0410's spectrum analyzer plots, the signal generation board is very good.  So finding an alternative UI is well worth while.

As FeelTech is t least aware of this thread I hope they're paying attention and will respond appropriately.  There is certainly nothing in the UI worth protecting.

I still want to know how cybermaus read the flash.  I was already aware it was just the signal generator flash, but not the details of how.

[cybermaus]

I did not sniff the USB connection, no, but I am not expecting anything of note in there, just the sending of control commands, but no way to read the flash.
Saying that out loud, I realize you may still be able to fully control the device, maybe even re-upload waveforms, using the PC program. But that is speculation.

I (and one other person) did hook up the STLINK to the STM32, and the flash is read protected, and unreadable, unless if you are a university level hacker.

As to how exactly I read the FPGA and waveform flash: Please study this post and its video, and if that is not enough, please at least have specific questions about what steps you have trouble with. There is also a follow up post on the flash data structure.