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

[rhb]

Very interesting work.  It got me to wondering if one could load a predistorted waveform instead.  I'd like to have a go at it.

Can someone point me to the instructions for loading ARB waveforms?  The thread is so long now we need an index page.

[alexvg]

Here is the schematic of the first filter implemented (11th order filter without the compensation filter).


To design the filters, I use :
- Iowa Hills RF Filter Design v2.2 (http://www.iowahills.com/9RFFiltersPage.html)
- RF Tools - LC Filters Design Tool (https://rf-tools.com/lc-filter/)
- LTspice XVII (most of work is done with this software)

To measure the FY6600, I use :
- Rigol DS1054Z (oscilloscope)
- Siglent SSA3021X-TG (spectrum analyzer)
- UNI-T UT181A (multimeter)
- 50R terminators, 20dB attenuator, 2GHz+ BNC/BNC cable...

[alexvg]

I've done the new 5th order filter with sorted components : +/-1nH inductances and 0.3% capacitors.
I've changed the 50MHz oscillator with this model : Connor-Winfield D75J-050.0M (+/-1ppm)
https://www.digikey.fr/product-detail/fr/connor-winfield/D75J-050.0M/CW657CT-ND/1965865

The schematic is highly optimized with the components I have and PCB parasitics inductance/capacitor.



This curve result provides 60dB SPUR with the new 5th filter than 50dB with 11th order filter.
Yellow : New 5th order filter
Magenta : 11th order filter


The lineary between new 5th order filter and 11th order filter
The new 5th order filter provides very good linearity for sine signal (+/-0.3dB 0-30MHz) and multitone signal (+/-0.35dB 0-60MHz)
Yellow : New 5th order filter with sine sweep 1MHz to 30MHz
Magenta : 11th order filter with sine sweep 1MHz to 30MHz
Green : New 5th order filter with multitone sweep 1MHz to 12.45MHz
Blue : 11th order filter with multitone sweep 1MHz to 12.45MHz




The rejection of multiple points of this new 5th order filter
125MHz at -32dB
146MHz at -41dB
250MHz at -58dB


The same curve compared to old 11th order filter
Yellow : New 5th order filter
Magenta : Old 11th order filter


1MHz square signal comparison
Yellow : New 5th order filter
Magenta : Old 11th order filter


5MHz square signal comparison
Yellow : New 5th order filter
Magenta : Old 11th order filter


10MHz square signal comparison
Yellow : New 5th order filter
Magenta : Old 11th order filter


25MHz square signal comparison
Yellow : New 5th order filter
Magenta : Old 11th order filter



Conclusion of this new 5th order filter

Pro :
- Simple to implement
- Very good SPUR
- Very good rejection
- Very good linearity

Con :
- The square signal is awfull, probably due to lots of phase variations.

Note:
I use the term SPUR for SPUR, harmonics and artefacts (provides by sine sampling/phase error...)


Next ?
I'm going to remove the implemented new 5th order filter using 0805 SMD.
I'm working on a little more complex 7th order filter using 0603 SMD.
The goal is to provide more better features with a quasi-linear phase (useful for square signal).

[alexvg]

I've performed a measurement of the frequency response of provided cables.

Yellow : BNC/BNC short cable
Green : BNC/Banana => Banana/BNC cable.

[soundtec]

Theres a new version 6 software become available for the fy6800 from feeltech,looks like it has the firmware update option too ,might try it later on .

[soundtec]

So I tried the update on the 6600 ,no luck ,wanted to jump too version 3.4 but it wasnt happeneing ,
I did the update with the 6800 ,it worked ,its now at Version 1.7 , no visual difference from version 1.6 ,but there must be some difference somewhere .

the 6800 drops into update mode automatically, it seems to reboot a couple of times , its all done in a matter of a few minutes.

[bitseeker]

Sounds like a nice improvement.

[precaud]

Has anyone successfully downloaded an arb into the FY6x00 ?
I created .CSV files of my most-used arb waveforms, tried loading them into the FY6800, and the software does not like them, it terminates with "Run-time error 380:  Invalid property value".
The files are simple ASCII real numbers scaled from 0 to 1, which is what most arb software I've used wants to see.
There is nothing in the download from the Feeltech website that gives any details about the file format for arb waveforms.

BTW, this is the same error message generated when I click on the question mark icon for Help in the arb edit mode. There is no help...

[soundtec]

I believe you need to re-label the csv files into something else and its as simple as that ,there was a post by Dave R dealing with it somewhere way back if you can find it .Try renaming the files  .fy

[precaud]

Thanks soundtec. I think this is the post you were referring me to.

Back on waveform files, the FY6600 will accept CSV files, but ONLY if the file contains nothing but the data values up to a maximum of 8192 points.  Clearing out the other cells and info left by wafeform creation software only takes a few seconds in the case of those uploaded by soundtec, so it's just the same as copying the data values into a new text file.  In fact, the software accepts *.txt files as well, so there's no need even ro rename them to *.fy.

The CSV file I created fits that description exactly, but the FY6800 software will not accept it. I've attached the file I was trying to load in, 4096 time domain data points with voltages scaled from 1.0 to -1.0 . Renaming it to .TXT (as it is attached) made no difference.

EDIT : I just saw that your data files had X,Y data pairs per point. So I recreated mine that way. It still bombs with the same error.

(rant: requiring a time value for each data point is stupid, all it should need is the time per point...once...)

[soundtec]

you need a sample 8192 long for it to work ,I think there was a problem with other numbers .

[precaud]

OK, so it doesn't want X,Y pairs, only the data points.
I can open the file, it reads in 8192 points. But when I click "Load Data" it gives an overflow error and bombs.
This software is buggy as hell. What a waste of time.

[cybermaus]

I've uploaded several (to the FY6600), never a problem.

I also manually edited one, they are not CSV, they are normal TXT files!
Did you try making one with the tool?

The tool can produce a *.FYC file, which is a "descriptive form" or "a definition" of a wave form. Like "add 1 sine of amplitude 1 to 1 triangle of amplitude 1 and double frequency"
This *.FYC file is next used to calculate the actual points in the waveform numerically, which is in the *.FY file. The latter is just a list of 8192 floating point values between -1 and 1
I attached 1 set as example (4 Sine waveforms added together)


When I manually changed those points, it worked for me, but I admit, I always started with a FY file that was first generated, and then modded it using notepad++. I guess it may matter if you terminate with CR or CR+LF or LF+CR, and using another tool like normal notepad or XLS may mess up those line terminations. Especially if you think its a real CSV and you are using Excel. It's not, it's a flat txt. Just just notepad++

Edit: I checked: all values are suffixed by x20 x20 x0d x0a
meaning <space> <space> <cr> <lf>
The last line is padded with an extra empty line x0d x0a


When I run it through for example XLS, it indeed drops those extra 2 spaces, so that could be the reason
Just use a more flat editor, like notepad++

[bugi]

And if everything else fails, look at the file with a hex editor. Sometimes weird things creep in to files with some software, like unnecessary (invisible) BOM characters in the begin, or accidental "phantom spaces" (characters that are rendered as space, but are not considered whitespace by most parsers), etc.

[soundtec]

Maybe a more general search of the board is nessesary . Ive done a bit of fooling around with the various big name waveform generation software available ,easywave ultrawave etc . Which is the best of the lot and will make usable waveforms for the FY66,6800 is there any of these packages offer a library or pre-made waveforms that can be uploaded to the feeltech?

Is it possible to modify the name of this post to include the FY6800 version ,so that when someone browser searches Fy6800 this topic shows ?

[DaveR]

precaud - don't use the Feeltech software to upload the data files, use one of fremen67's versions of the control software, which gets rid of the unneccesary and confusing Chinglish options.

If you start with a csv file, open it in a spreadsheet and copy and paste JUST the 8192 data values into a blank text file with either a .txt or .fy extension.  Use Notepad so that you don't get any hidden formatting information saved in the file.  As bugi says, you've probably got an invisible character in the file which is causing the problem.

Try a couple of mine to make sure your software is working ok.

 

[precaud]

Thanks for chiming in, cybermaus.

I also manually edited one, they are not CSV, they are normal TXT files!

For those who may not know, CSV = "Comma Separated Values". It's just a list of data points, separated by a comma or other commonly-used separator. If there's only one data point per line, then CRLF substitutes for the comma.
Excel .CSV files are a little different, they have a few lines of descriptive header before the data points.
But the main feature is, they are simple text files of ASCII-encoded decimal values between +1.0 and -1.0
That is what my software generates. (Actually, its HP's software from the 90's...)
From the FY6x00 software POV, there is no difference between .CSV and .TXT or .SUX, it doesn't care what the file type designator is, as long as the data format is correct, it will read it.

Did you try making one with the tool?

No. Once I saw that the file format is the usual list of data points, I saw no point (pun intended!).

This *.FYC file is next used to calculate the actual points in the waveform numerically, which is in the *.FY file. The latter is just a list of 8192 floating point values between -1 and 1
I attached 1 set as example (4 Sine waveforms added together)

Thanks. My files are identical to yours, except a) yours has a blank line at the end (where point 8193 would be is just CRLF), and b) mine uses a small-case "e" for the engineering notation. I wonder if that's making a difference. The blank line isn't likely causing the problem, because the software reads the file in fine, the error comes when loading it into the waveform buffer.

I created a 4096-point version. It reads it in and loads it into the edit buffer. While doing so, it expands it to 8192 points by duplicating every data point. I sent that file to the 6800 and it plays back fine (if duplicated data points is what you want...)

I then edited off one data point from the 8192-point version. The software reads it in, counts the # of data points correctly (8191), and displays the waveform correctly in the graphics box to the right. Clicking "Load Data" loads the waveform into the edit buffer but imposes an ascending ramp modulation on the data and multiplies it way out of bounds. Wierd.

Edit: I checked: all values are suffixed by x20 x20 x0d x0a
meaning <space> <space> <cr> <lf>
The last line is padded with an extra empty line x0d x0a

Thanks. All mine are CRLF terminated too. No empty line at the end, though.

I'll try making the "e" an upper case and see what happens.

But this has raised two larger questions.
First: I bought this thing mostly for its arb capability. As I wrote in a thread started a couple months ago, I wanted to see if higher-resolution (14-bit) arb could make up for the lack of a real output filter and reduce the clock noise that is created when inevitable waveform delay shifts the samples and puts the stair-step at the center of the sampled point in the FFT. Some here opined that it would. My hunch was (and still is), no, it won't, there's no substitute for proper output filtering. This was a cheap way of testing it out, so I went for it.

Second: Only now have I learned that all waveform files have to be 8192 points. Can I live with that? I already know the answer: No. I need the flexibility of point-by-point waveform playback of any number of data points. Not all digitizers use power-of-two-based time record lengths.

So I'll try changing the "e"s to "E"s. But even if/when the data format issue is solved, it appears this unit is not going to do what I want.

[Philipp F.]

Hey,
I am a somewhat inexperienced electronics hobbyist and I'm in the market for a signal generator for testing various things and circuits because my cheap Chinese 3 knob signal generator kit isn't cutting it anymore.

I really want to get the FY6800 because of all its functions (the VCO and sweep stuff really interests me) and it costs about right what I can spend, but all this talk is kinda deterring me...

Is it worth getting or do I risk ending up with a 115€ brick?
I can get it shipped from a Germany warehouse, so I could at least hope for a basic 2 year warranty (and the seller also advertises 2 year warranty).

[DaveR]

precaud - I've just spotted the file you posted, and it loads into fremen67's 0.7 version of the control software, so it should upload to your 6800 ok (although what it will do to the 6800 when you run it is anyone's guess  ).

Regards,
Dave

[precaud]

@Phillip F., you can ignore all this talk, we're just getting into the weeds on some details.

So I changed the "e"s to "E"s, put a blank line at the end of the 8192-point file, and just for the halibut, renamed the file with a .FY extension. The software reads it in fine. Clicking "Load Data" generates the same error as before, "Run-time error 6 : Overflow".

There are no data points in the file larger than 1.0 or smaller than -1.0 . In fact, there are no data point equal to those values, either. The data conforms to all the rules. So the raw data is not the cause of the overflow.

This is the arb waveform I use more than any other. If it won't work with this, there's no point to me going further with it.

I've attached the file (with extension .ASC) if you're interested to see it.

DaveR - I just saw your post. That file is 4096 data points. It loads into the 6800 fine for me too, except every point is duplicated. Try the attached one and see if that works. Meanwhile, I'll look for fremen67's 0.7 version and try it.

[soundtec]

Im just having a look at bitscope data aquisition ,seems really good whats going on here .

http://bitscope.com/news/

I just downloaded the software to have a look , the idea of using a farm of rasberry pi's for data aquisition /processing looks nice
Theres a guy I know who runs a reseach project for one of the universities here ,he's invloved in siesmic data aquisition  from sea bottom tranducers around the coast ,I bet he could save massively on his overheads by creating his own supercomputer from PI's.

Thats great you had time to consider waveform generation as a way of improving the performance of the Fy66/6800 Precaud,
its a question we debated a bit over the time on this topic.

Theres a few small additions on the Feeltech version 6 software that would improve it  , one is the addition of a refresh button like Fremens put on so any changes in values on the instrument itself can update the software display values,in any case toggling the 'auto connect' option reloads the values from the front panel ,  would it be possible to make a push of the encoder wheel button to get  the FY  to transmit the data from the unit to the software or could it be made to do that automatically ? seamless integration between front panel and software would just make life a small bit easier and would make whats already a very intuitive machine  really sweet. Maybe Fremens could take another crack  at hot roding Feeltech Version 6 software ,the bluepill idea is nice and allows a lot more flexibillity ,but I dont see how front panel and software can live side by side in that scenario .

 I very much like the idea of using both my  66'&68' series as four channels of audio tone generation, run from a midi sequencer  tones could be layered up to  tracks and effects and gating on the analog end could be used to produce large musical 'swirlls'.
I have a pair of boss Ce-5 chorus's run in stereo, by flipping the phase of the clean and effect return a whole load of subtle to mad wobulation type effects are possible ,flange, phase, chorus and vibrato can all be had with different panning and phasing arrangements , add in tube driven spring reverb and speed controlled mk2 Watkins copicat for tempo synched delays ,hitting one key you can sound like Vangelis or Jean Michel Jarre ,same effects chain sounds absolutely great on an open tuned guitar set up for bottle neck slide ,Id be interested to hear from any others who are already using their FY for sound and music purposes ,the foundations of electronic music were born out of laboratory equipment after all .

That last one you sent loaded up fine for me in version 6 Precaud,

[DaveR]

Is it worth getting or do I risk ending up with a 115€ brick?
I can get it shipped from a Germany warehouse, so I could at least hope for a basic 2 year warranty (and the seller also advertises 2 year warranty).

No-one has reported a brick this year, as far as I can recall, so you can feel pretty safe with the latest versions of the 6600 or 6800.  Most of the talk on here for a long time has been about making the 6600 a better machine, which we've succeeded in doing, but the 6800 will be fine out of the box as it already includes some of the improvements which were required.  (And it will be in a different league compared with those you've used so far.)

Regards,
Dave

[precaud]

That last one you sent loaded up fine for me in version 6 Precaud,

Interesting, soundtec. That's the version I'm using...

[DaveR]

DaveR - I just saw your post. That file is 4096 data points. It loads into the 6800 fine for me too, except every point is duplicated. Try the attached one and see if that works. Meanwhile, I'll look for fremen67's 0.7 version and try it.

Loads fine for me - see attached.  (Just renamed it to chrp8k.txt and in it went!)

[precaud]

OK thanks guys, I guess I need to download and reinstall the software again.

BTW, the way to detect if data integrity is maintained is to load it into the text window first. That's where I saw it was duplicating points onthe 4096-point file.

Is there a link to the fremen67 version? I looked through this volumnous thread and found lots of talk but no link.