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

[aimc]

This is all well known, but did you notice that (with proper termination of course) the generator's output declines from 10-60 MHz by about half (-6dBv)? I am talking about the FY6800 actually and want to use it as VCO expecting a flat response. I didn't go inside yet to measure directly at the DAC but I suspect there is something fishy going on at the amplifiers. Has this topic been discussed here or anywhere? Is there a "hack" available to fix this?

Thanks in advance
Lutz

[CDaniel]

I can't talk about FY6800 but an FY6600 is flat from 0 to 60Mhz ... maybe you did a simple mistake in your setup , it wouldn't be the first time in this thread 

[rhb]

This is all well known, but did you notice that (with proper termination of course) the generator's output declines from 10-60 MHz by about half (-6dBv)? I am talking about the FY6800 actually and want to use it as VCO expecting a flat response. I didn't go inside yet to measure directly at the DAC but I suspect there is something fishy going on at the amplifiers. Has this topic been discussed here or anywhere? Is there a "hack" available to fix this?

Thanks in advance
Lutz

What was the output voltage?  Check it at 1 Vpp.

[aimc]

Thank you for the reply. I did measure it again just now (see attachment) and yeah - well it has some ripple reaching around 2.5dBv (~0.75) at 60MHz (my former statement was a bit higher due to a mistake, sorry). Not sure if a 0.75 drop is acceptable for a function generator, but then again its a cheap product.

[rhb]

I don't understand what you're showing.

Is the sweep synchronized with the scope trigger?

Also please take a look at the PCB.  Is there any flux residue visible?  Flux absorbs moisture out of the air.  This results in an environmentally dependent parasitic capacitance.  It would not take much capacitance to  cause the drop in output you're seeing.

If you see any flux, clean it off with isopropyl alcohol and an old toothbrush and retest.  I've done quite a few repairs, including an HP 34401A which had been repaired previously that consisted of nothing but cleaning flux residue off the board.  The 34401A was completely non-functional until I cleaned it.  But after cleaning meets the factory spec.  It's the more accurate of my two 34401As.  The other one was a closet queen and saw very little use.

Have Fun!
Reg

[aimc]

I don't understand what you're showing.

Is the sweep synchronized with the scope trigger?

Also please take a look at the PCB.  Is there any flux residue visible?  Flux absorbs moisture out of the air.  This results in an environmentally dependent parasitic capacitance.  It would not take much capacitance to  cause the drop in output you're seeing.

If you see any flux, clean it off with isopropyl alcohol and an old toothbrush and retest.  I've done quite a few repairs, including an HP 34401A which had been repaired previously that consisted of nothing but cleaning flux residue off the board.  The 34401A was completely non-functional until I cleaned it.  But after cleaning meets the factory spec.  It's the more accurate of my two 34401As.  The other one was a closet queen and saw very little use.

Have Fun!
Reg

Sorry, I did not describe the attachment. Well, essentially both images show a periodic ~0.5s sweep from 0.1-60MHz (left to right), one in the time and the other in the frequency domain. An yeah I triggered on the edge where 60Mhz jumps back to the higher amplitude 100KHz to keep the presentation stable (although I stopped it anyways later). The interesting part is in the envelope, which gets kind of jaggy, likely due to the sampling lag between the much finer FY6800 sweep increments compared to the scope acquisition speed. I can show that along the time axis the amplitude decline becomes visibly discretized into stair steps when you crank up the scope's front end amplification (not sure, but this also may saturate the amp, that's why I didn't show it).
The other screen shot shows the FFT of the FY6800 waveform during the same frequency sweep at much faster time base. Time base has to accommodate sufficient sampling to provide a decent time resolved FFT window, small enough to approximate it as a single tone at each time base sweep. Then, as the tone frequency slowly sweeps from acquisition to acquisition, with the FFT set up to keep only the maximum of the peaks you see a decent magnitude over frequency representation. This will take some time and acquisition sweeps until all maxima form a solid line. With the cursors set to the math output, you can quantify the roll of. When you compare both screen shots you see the expected similarity only different by the different log scale from the FFT (which I chose to show dBv units). The two representations also compare quantitatively pretty good within a reasonable margin of error.
Hope this explains my measurement steps.

I will have a look inside the AWG and look at the solder connections. Thanks for this advise

Regards
Lutz

[CDaniel]

I supose you have another generator that is linear as you want , otherwise you don't know where the problem is .

[rhb]

I'll post a quick comparison of a Keysight 33622A and an F***Tech FY6600 with V 3.1 FW once I clean up a bit.

The 34401A is the most spectacular  example of the evils of flux residue. When it arrived, it just displayed random digits on all functions and ranges.  I opened it up and saw a 1 cm brown spot around where an LF357 had been replaced.  I sprayed it with 91% isopropyl alcohol and scrubbed it with an old tooth brush, checking with a magnifying glass to make sure I got everything off.  I then dried the board with a hairdryer.

When I powered it up it worked flawlessly and still meets factory spec on the 10 Vdc range checked against two very high quality LTZ1000 voltage references.

I've "repaired" several TV remotes, a CD player, DVD/VCR player and other things this way.  In the case of the CD player, I didn't get all of the residue off the first time.  About 6 months later it started acting up again ("NO DISC").  I cleaned it again using lots of alcohol.  Haven't had a problem with it again in several years.  I think the flux I missed was underneath the MCU chip.  The only residue on the board had been a small spot near one corner of the MCU.

The 91% IPA from the supermarket is not very pure, so blotting excess off with a towel is important.  Otherwise you will have s white film of something left behind.  You can see minute amounts of flux with a 5-7x magnifier by looking at the smoothness of a reflection from the board.  It's not uncommon for me to find that after an initial cleaning and drying.  If I see it, I do it over.

Have Fun!
Reg

[aimc]

I supose you have another generator that is linear as you want , otherwise you don't know where the problem is .

No, not really - only a cheapo AD VCO development board that goes from 35M-4.4GHz. I do know the response of the scope (with and without probe) though and hope to find the problem with the scope only. Frequency range is rather low, so will try to tap in before the amp of the FY6800 if I find a good location to do so.

[rhb]

Here are a couple of screen shots.  The FY6600 V 3.1 sync output when sweeping it being swept :-(  Makes it  useless.  As a result, the only way to trigger on the sweep is the drop in output.

Sweep is 10 kHz to 60 MHz for the FY6600 and 1 kHz to 60 MHz for the 33622A.  Scope is an Instek MSO2204EA with 50 ohm thru terminators.  Signal amplitude is 200 mVpp.

First the 33622A.  Note that it has some signal drop as the frequency goes up.  Not sure what the culprit is.  I'll investigate.



Then the FY6600 which has much more drop.



I think the biggest FAIL is the sync signal not being the start of a sweep.  I spent some time trying to get proper synchronization on the F***Tech, but could not find a way to do it.  So I set a rising edge trigger on the signal at the 10 kHz end above the level at the 60 MHz level.

I bought the 33622A used from Keysight on ebay for 30x what the F***Tech cost. Older HPAK AWGs were so expensive on ebay that they didn't seem to make sense.  The 33622A goes to 120 MHz and is  quite amazing.  I bought it because my F***Tech was dead. F***Tech made promises but did not deliver.  My choices were more Chinese which did not fill me with great confidence or a 15 MHz HP for $500.  So I went with the 33622A.  I'm glad I did.  And glad I could afford it.

I've got an HP 8560A and a 438A, so I have other test options.  I'll see if I can use some of the other F***Tech rear panel inputs to synchronize the two AWGs and overlay the sweeps.

Have  Fun!
Reg

After I posted this, I noticed that the sample rates imply aliasing, so I'll need to examine that.  I suspect that sweeping over such a wide range into a DSO is not a good test.  So I'll try repeating with my Tek 485.

[aimc]

That's interesting! Yeah the trig out of the FY does not make much sense for any sweep or VCO operation which kind of sucks. Otherwise it looks like you are getting similar results for the F...T brand while the Keysight seems much better. Your Keysight manual boasts with +-0.2dB amplitude flatness, so it better be. Looks like your time base and VCO sweep was faster than I did because your jaggies are bigger and you can't see that the response has a slight s-shape. A bit concerning as you say is the sampling rate in both our cases @60MHz. You used 5MSa/s and I used 20MSa/s which is grossly under sampled. However, it should not impact in the amplitude per say. If you have enough periods and non-integer multiple sampling points (you don't because you sweep) the lower frequency alias amplitude will at least be as high as the signal. This is covered by sampling theory. To demonstrate I repeated the experiment with 100MSa/s (I know its not >120MSa/s as it should be but close enough when using sinc interpolation, see aliasing in attachment "..no_SINC")  and found the drop being almost the same as with my former under-sampled post. For this reason alone I think the FFT method is likely better as you are using well oversampled small sweep-time pieces for each FFT. But there you have imprecisions with the windowing. So nothing seems better than the good old simple peak detector (or any new-fangled detector you can buy from Analog Devices).

If you don't trust me, you can try a higher sampling rate - but I think at the end of the day you measure the same drop...

Thanks and cheers
Lutz

[CDaniel]

I have v3.1 , the firmware will work ?

Yes it will. I tested it with flash eeproms v3.1 and V3.2.
It doesn't work with a V3.3 eeprom but as I don't have a V3.3 front panel, I can't see what changed in the FP-FPGA protocol for v3.3.

Hello , I flashed a spare microcontroller with your firmware ... quick bug , in sinewave the output is badly distorted if I go up in frequency up to 197KHz . If I go back down from here is OK untill 1KHz or something when the distortion appears again . Something in comunication with the FPGA is not quite right .

[fremen67]

I have v3.1 , the firmware will work ?

Yes it will. I tested it with flash eeproms v3.1 and V3.2.
It doesn't work with a V3.3 eeprom but as I don't have a V3.3 front panel, I can't see what changed in the FP-FPGA protocol for v3.3.

Hello , I flashed a spare microcontroller with your firmware ... quick bug , in sinewave the output is badly distorted if I go up in frequency up to 197KHz . If I go back down from here is OK untill 1KHz or something when the distortion appears again . Something in comunication with the FPGA is not quite right .

Hello, did you swap the microcontrollers?
I tested with my 2 FY6600s. One has a swapped MCU, the other has a bluepill… Sorry but I can't reproduce your problem 
The sines are normal till 62.5 Mhz…
The MCU is not supposed to have an impact on sine output or shape. It just send the parameters to the FPGA when you change them but then no communication anymore with the FPGA.
Could you also test with an ARB sine (ARB1 if you did not change it otherwise you can create one and send it with PC Software).

Do others have the same problem? DaveR, Soundtec, jleg??

[fremen67]

Hi Fremen67 and welcome back! I surely hope that you were just busy and nothing tragic happened.

Thank you! Well... death is part of life… time eases things...

How may I help to figure out V3.3 and what do I need? My FY6600 is V3.3. I don't have a logic analyzer but I do have a Rigol DS1054Z. Probably not enough, eh?

You can do things with a DS1054Z but I am afraid the only reasonable way is the logic analyzer.

[CDaniel]

OK , sorry about what I said , no bug , I forgot to connect a wire and the analog part wasn't properly powered . My FY6600 power supply is modified and uses a PWM signal from the microcontroller .
Yes , the original microcontroller is swapped with another one .

Just a sugestion , the frequency displayed is still hard to follow with just one decimal point when it is in MHz range for example ... I have to count allways how many digits are before the dot  . Would be nice to display MHz , KHz and to move the point accordingly .

The sub-hertz many many digits I think shouldn't be displayed when are not used ( set to 0 ) .

If you go to the max with the tens of MHz digit ( by error or not ) , the rest of the display goes to max also  and you lose what you have set , maybe you want 59.123456MHz . I think would be better not to let you go beyond 5 if 6 and the rest of the digits would be an invalid frequency .
When you go down in frequency the same thing happens if you reach 0 and turn the encoder one more time , the rest of the digits are erased to zero , I don't think this is desirable . Zeroing the display could be usefull , but with a long press on the encoder or another combination when you are in frequency adjustment .

[rhb]

I discovered that by generating a 2 kHz ramp on CH2 and feeding that to the VCO input on the back I can get a quasi-stable sweep by triggering off the ramp.  However, there is a *lot* of jitter in the FY6600 sweep relative to the ramp.

Here are the ramps produced by the 33622A and the FY6600.



Next the sweep using the ramp input to the VCO input on the back of the FY6600.  The 33622A is the down ramp.



Same thing but using the ramp from the 33622A



The jaggies and banding are aliasing effects.

It may be that I can make some useful measurements of the F***Tech output using the nanoVNA.  I've never tried to measure the output impedance of an amplifier, so significant research needed.  But it certainly appears as if there is a shunt capacitance across the output.

[aimc]

Hmmm interesting, but now the drop is even deeper from the looks of it... Maybe the VCO mode has a different can of worms. Also 2KHz seems a bit fast for a sweep. Maybe the VCO can't handle it and/or it can but then the waveforms gets distorted within a single sine period depending on frequency? Haven't done the math on that... I know if you do it slower, the scope's sampling rate drops and you get aliasing, but as I said before the peak values of the aliases remain the same.

[CDaniel]

I suggest checking Feeltech linearity manually in 100KHz increments or how do you want , only with the generator connected to the scope .
For 0-60MHz linearity , the probe proper compensation is very important , not just that basic visual 1KHz sqarewave no overshoot , undershoot procedure . Unfortunatly a scope and probe is not that linear as we would want , it is not a RF voltmeter . A judgement could be made only if you have another generator you are certain that is linear .

[soundtec]

I noticed I needed a version of the original Feeltech software installed on the machine before Fremens would run  ,I had been trying out the latest FY6900 version pc software ,maybe this caused the issues I am seeing , I wonder is there a particular version of the Feeltech pc  software I should be using for maximum compatibillity to Fremen's FP code ,

I also saw the software Yves created for his Fy3224s ,theres several features  that looks very usefull , both to the radio enthusiast which its designed for ,but also would be very helpful for programming test tones or sweeps for audio or any other test purpose that comes to mind . The programable buttons to trigger arb waveforms from memory could almost turn the unit into the equivalent of a sampler ,might be of use to those who are looking to incorporate an arb gen in a musical context .

I tried a trial version of Ni labview to display the .VI file included with some of the Feeltech software packages , seems to make an attempt to copy the controls of the unit on screen but the  labels are garbled charachters ,there appears to be a control element missing from the Feeltech VI file as well.

I did also see RHB's post about the Fy6600 WSPR transmitter , the author of the article did create a basic VB code for control of the FY6600, so thats another thing Id like to try out.

I did have another think about the double button press being required  to select channel and mute/unmute  , maybe a nice idea would be when you initially depress the channel 1 or 2 button  have it select that channel to top if its not there already ,then let the unlatching of the button send the command to mute or unmute as required , now only a single press (and release) covers both functions , it  only logical  captain  Fremen    


Many thanks for all the efforts .


 

[Miti]

Can anybody post a csv file that is accepted by FT software, or fremen's software? I can't figure out the format. I'm trying to do a simple ASK modulation on CH1 from CH2 and anything I try will just return an error message and shut down the program.

Thanks,
Miti

[radiolistener]

It would be really nice to see FOSS code that would allow other Chinese OEMs to clone the F***Techs and improve them.  It could not happen to a more deserving company.

yeah, it will be great. If it will be released I will buy some FY6600 clone

[soundtec]

I used Wavemanager plus to create a few compatible wave forms ,

 version 4.13 of the  program is included in the download for the 3224s here

https://drive.google.com/drive/folders/1hqqHqvfk1FKhzYSjJ-S6_SUTr_zx0fhp?usp=sharing

Ive been using version 5.8 of the Feeltech software , and so far find Fremens 0.8 is unusable with a 'run time error 380'
invalid property value, V 0.7 is usable but just for the basic channel functions .

[fremen67]

Can anybody post a csv file that is accepted by FT software, or fremen's software? I can't figure out the format. I'm trying to do a simple ASK modulation on CH1 from CH2 and anything I try will just return an error message and shut down the program.

Thanks,
Miti

Find attached a sine wave created with the modified PC Software 0.8.
This is the original FT format.
You can download it to a standard FY6600 or a modified FY6600 with PC Software 0.8.

[fremen67]

Just a sugestion , the frequency displayed is still hard to follow with just one decimal point when it is in MHz range for example ... I have to count allways how many digits are before the dot  . Would be nice to display MHz , KHz and to move the point accordingly .

Thank you for the feedback.
Yes, as already written to sountec, this is planned. I will also add a thousands separator to ease the reading. This just a matter of prority.

The sub-hertz many many digits I think shouldn't be displayed when are not used ( set to 0 ) .

Not sure this will look nice but why not...

If you go to the max with the tens of MHz digit ( by error or not ) , the rest of the display goes to max also  and you lose what you have set , maybe you want 59.123456MHz . I think would be better not to let you go beyond 5 if 6 and the rest of the digits would be an invalid frequency .
When you go down in frequency the same thing happens if you reach 0 and turn the encoder one more time , the rest of the digits are erased to zero , I don't think this is desirable . Zeroing the display could be usefull , but with a long press on the encoder or another combination when you are in frequency adjustment .

The behaviour your are describing is what I have on my RIGOL DP832 but there are much less digits on it… and I already find it painfull...
In your example, with 59.123456MHz, this would be a pain to set 60.000000MHz...
The behaviour I programmed is the one that was existing on the FY6600. It is also the behaviour of my RK8511 electronic load and my SIGLENT SDG2042X for the max limit.
I don't know if there is a standard on that matter …
What do others think?

[fremen67]

I noticed I needed a version of the original Feeltech software installed on the machine before Fremens would run  ,I had been trying out the latest FY6900 version pc software ,maybe this caused the issues I am seeing , I wonder is there a particular version of the Feeltech pc  software I should be using for maximum compatibillity to Fremen's FP code ,

I tried to stay compatible at the beginning with a standard FY6600 but I had to modify the original protocol for the new functionnalities. I don't know if Feeltech changed the protocol for the FY6800 or FY6900 but logically the FY6600 version should be the most compatible with the new FP. Not sure why you would want to use it with the new FP anyway...

I did have another think about the double button press being required  to select channel and mute/unmute  , maybe a nice idea would be when you initially depress the channel 1 or 2 button  have it select that channel to top if its not there already ,then let the unlatching of the button send the command to mute or unmute as required , now only a single press (and release) covers both functions , it  only logical  captain  Fremen    

If I get you right, this would than be impossible to show CH1 for example if not displayed without changing CH1 output state …. the same for CH2.
The problem is that there are few buttons so we have to compromise … The problem is exactly the same with measure, sweep and modulation functions. One push on the button will show the view if not already displayed, another push will activate the function.
This is also now the standard way on oscilloscopes with common buttons for several channels: one push to select the channel or deactivate it if already selected. I don't say this is easy at the beginning but you get used to it