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

[cybermaus]

My device is now upgraded to version 3.2 and all the front panel buttons work swimmingly.

There was a problem with the buttons, that was resolved with V3.2?
Please explain further?

[SMB784]

There was a problem with the buttons, that was resolved with V3.2?
Please explain further?

No sorry there was no problem with the buttons in either version, I just mentioned that because I was proud of myself for not screwing up the soldering job (first time working with SMD)

I got the seller to send me a version 3.2 chip to replace my version 3.1 chip that supposedly could lose its ability to output a sine wave under certain circumstances.

Sorry for the confusion

[cybermaus]

Ah yes.

Indeed, Louis Rossman may make it seem like it's easy, but I also still think of successfully replacing a 48 pin SMD package as a an accomplishment.

[Insatman]

Ah yes.

Indeed, Louis Rossman may make it seem like it's easy, but I also still think of successfully replacing a 48 pin SMD package as a an accomplishment.

As would I...at least with normal hobbyist equipment/experience those large fine-pitch SMD packages are a B***H.

[Insatman]

Paulca, expanding on the suggestion of Cybermaus, you can get a triggered sweep for your scope by feeding a low frequency negative ramp from CH2 to the VCO terminal in back. Make sure your ramp goes from 5V to 0 V, using offset and amplitude adjustments. Set your start and stop frequencies, then use"Forth" in Sweep selection to get a forward frequency plot. Your scope can be triggered directly from the positive edge of the ramp using a second scope channel or ext trigger.

While this does work, it is cumbersome and fraught with possible errors. For more serious work with frequency response plots I suggest you get a low cost scalar network analyzer. There are plenty of low cost ones on eBay.

My FY6600 (V3.2) has quite significant variation of amplitude with frequency. While I have used the CH2 ramp to sweep, the result is approximate at best.

I thought a bit about using the FY6600 as a low-frequency network analyzer.  I have a decent network analyzer in my hacked Siglent SSA3021X  but at low frequencies, say below 1MHz it's performance isn't so great.   The SSA3021X tracking generator is not very stable at the low-end and the unit isn't even rated below 9KHz anyway.   I find myself using the FY6600 for sweeping the response from a few hertz to 10MHz for so.   It's output voltage is very stable through that range.  Only when I get above 30MHz or so the voltage drop off is significant.  Note: I have installed the THS3095 op-amps on my unit.

[spec123]

Yes, you can use the FY6600 as a scalar network analyzer. For fun, I swept a 14 MHz BPF and got nice response on the scope. But with only the 8 bit converters in the scope and the coarse ability to read out frequency and amplitude, there is much room for error.    Low cost scalar network analyzers like the NWT70 are on eBay for around US$70. I bought this particular version (there are many NWT versions on eBay) since it covered the 50 KHz to 80 MHz range. As I remember, I got it to go down to 1 KHz. Its more accurate than using a FY6600 and has 80 dB dynamic range with on screen plots.

BTW, after several days of experimenting with my new mods on FY6600, I am very pleased with the results. The new 50 MHz oscillator D75J you recommended is excellent. My sine wave FM modulation (jitter) was over 12 Hz at 10 MHz. Now it is way below 0.1 Hz... below what my equipment can measure.  It is stable and doesn't drift. The old one took 30 minutes to stabilize (within the specs in manual). This one is ready as soon as you click on the power.. I was very lucky as my D75J was spot on 50 MHz... did not have to apply the frequency correction to FY6600.

The two replacement amplifiers are working good too. Used heat sink pads under the heatsink. Barely gets warm. I noticed that below 20 dBm output (sine) distortion is below 60 dB. It rises much more as you approach full output of 30 dBm. Noticed that the display voltages on LCD are not correct in predicting dBm. My output was off by -4.1 dB... Ran it all day yesterday on a project and it performed very well. Its now becoming one of my favorite tools.

[SMB784]

Is there a summary of recommended mods (+part numbers) that could be uploaded to Der Kammi's GitHub? Also, is it possible to run the output opamps at a higher voltage (like +/- 15V) in order to avoid the distortion at high amplitude via a better power supply? Or is the high amplitude distortion baked in?

From what I understand, the timebase, the output opamps, the capacitors, and the power supply should be upgraded in various different ways. Is there something else that I have missed?

[Bob Sava]

My 30Mhz FY6600's square wave (1Vp-p, 10kHz) has a rise time of 20ns.  Would 60Mhz model have a faster rise time (I thought I saw 7ns screenshot here).

EDIT: I don't know what I did but now rise time measures 7.5ns.  Maybe there's something wrong with my scope or user error.

[bdivi]

Is there a summary of recommended mods (+part numbers) that could be uploaded to Der Kammi's GitHub? Also, is it possible to run the output opamps at a higher voltage (like +/- 15V) in order to avoid the distortion at high amplitude via a better power supply? Or is the high amplitude distortion baked in?

From what I understand, the timebase, the output opamps, the capacitors, and the power supply should be upgraded in various different ways. Is there something else that I have missed?

I am still waiting for my FY6600 and I have not implemented any of the improvements yet but reading trough the discussion here and also tests elsewhere here is what I have on my list:

1. Power supply
Classical transformer 14Vx2/20VA
Linear regulators 14V DC for some output peak headroom - possibly LM317/337
DC-DC to 6V and then linear to 5V

2. Output buffer
THS3091/95
Better option would be THS3491 but not yet available - it may be worth staying with the 3002 for some time
Thermal paste and heat sink for the THS pair - the 4558 does not need help and just sticking above the other two
Transil voltage protection for the output - something around 25V, low capacitance
Diode clipping to the power supply rails - BAV99

3. Reference clock
10MHz OCXO - the cheapest I could find on ebay
Clock multiplier x5 - NB3N502
Int/ext reference switch with BNC

4. Front panel
Would be nice to have the channels LEDs yellow and blue ;-)
The rotary knob is also mentioned by several people as being shallow

5. Altera
Someone mentioned heat sinking the Altera Cyclone - we'll need to check the temperatures, especially with linear power supply and OCXO all adding heat inside.

[cybermaus]

1. Power supply
Classical transformer 14Vx2/20VA
Linear regulators 14V DC for some output peak headroom - possibly LM317/337

Make sure the 14V has good good buffer caps, or else make the rail 12.5V instead of 14V. Because 14Vdc from 14Vac may cause some ripple if your caps are too small.
Especially since it was reported the 3095 on 11.5V already barely had clipping, so better a stable 12.5V then a ripply 14V

DC-DC to 6V and then linear to 5V

No need. That 5V is not used except to have a linear to 3.3V. So your DC-DC may as well go to 5V directly

Diode clipping to the power supply rails - BAV99

Ok, I missed that one. Why/How clip to rail with BAV99? Is that still needed if you replace the entire PSU?

Would be nice to have the channels LEDs yellow and blue ;-)
The rotary knob is also mentioned by several people as being shallow

Actually, colored LED would be a nice finishing touch. Completely not useful, but nice.
What is with the knob? I find the rotary know quite nice as is?

[bdivi]

1. Power supply
Classical transformer 14Vx2/20VA
Linear regulators 14V DC for some output peak headroom - possibly LM317/337

Make sure the 14V has good good buffer caps, or else make the rail 12.5V instead of 14V. Because 14Vdc from 14Vac may cause some ripple if your caps are too small.
Especially since it was reported the 3095 on 11.5V already barely had clipping, so better a stable 12.5V then a ripply 14V

DC-DC to 6V and then linear to 5V

No need. That 5V is not used except to have a linear to 3.3V. So your DC-DC may as well go to 5V directly

Diode clipping to the power supply rails - BAV99

Ok, I missed that one. Why/How clip to rail with BAV99? Is that still needed if you replace the entire PSU?

Would be nice to have the channels LEDs yellow and blue ;-)
The rotary knob is also mentioned by several people as being shallow

Actually, colored LED would be a nice finishing touch. Completely not useful, but nice.
What is with the knob? I find the rotary know quite nice as is?

Good point about the power supply - 12.5 to 13V seems to be the sweet spot with respect to clipping and minimal losses on the following regulators.

The 5V is regulated down to 2.5V which is the reference voltage for the offset that is superimposed on the output signal so I would rather have quiet 5V DC.

The diode clipping I mean is for protection from injection of high voltage to the output. The original implementation does not have any protection.

[DaveR]

What is with the knob? I find the rotary know quite nice as is?

It's because you haven't tried it with a proper knob, cybermaus.  That diddly little disc thing is ghastly compared to the real thing.

[ArthurDent]

On the encoder knob: I found my fingers had a tendency to slip off the thin knob because it didn't stick out far enough. Being lazy, I just pulled it out on the shaft to give me a better grip and I've had no further problems.

[fremen67]

Here is the v0.5 version of the blupill firmware, together with a modified version of the PC Software.

There is a new tab with calibration data. This allows to calibrate amplitude for low frequencies but more usefull, the DC Offset.
Once calibrated, the DC function becomes quite accurate. Of course it is better to do a manual calibration of offset and amplitude before doing the software calibration.
Calibration data are updated and stored after each calibration value input. They are stored at the end of the STM32 flash eeprom and loaded again on startup.
The Z output load mode is also available.

In the configuration tab you can now:
- activate/desactivate the HiZLoad mode
- modify the Load impedance value from 50 ohms to 1kOhm
- activate/desactivate the use of calibration data

As I also prepared resampled buffers of current waves in the bluepill for future LCD display, I modified the PC Software so that it could read and use those buffers to get a picture of the waves on screen. Square, CMOS, DC and AdjPulse waves pictures are calculated, the others are live resampled values. You should now be able to see on the PC the arbitrary waves that you download to the flash eeprom.

[cybermaus]

Well, bummer. I am no longer able to make it work.
Not to worry, it is bound to be me. Because I went back to v0.4 and that also does not work anymore.

So please remind me: if no PC is connected, and comms between CPU and FPGA is good, then it should start with a 10KHz squarewave right?
What I have is:
- Both LEDS turn on (red and green). If memory serves, this means initialization went OK, and there is comms between CPU and FPGA
- There is no wave output on either channel. If init is OK, why no output?
- The PC does see the CH340 driver on COM3, but the software does not see the FY6600. Not in auto, and not when I manually select COM3
- I tried all startup sequences: power first, USB first, press an extra reset later.

Any suggestions

[fremen67]

- Both LEDS turn on (red and green). If memory serves, this means initialization went OK, and there is comms between CPU and FPGA
Any suggestions

Red LED is ON by default when powering the bluepill. When Hardware initialization is done, I switch it OFF as soon as I recover control from the initialization routines.
If it stays ON, that means it is stucked during initialization or the bluepill may also be empty or corrupted.
I would first check the large ribbon connection and especially the FPGA Ready wire. Then I would also compare the bluepill program against hex file.
Otherwise, any backup bluepill?

[DaveR]

Nice work, fremen!  The calibration feature works well for DC, and offsets on mine are now within 5mV instead of 150mV.  Sine and square waves are also tightened up so that the two traces are now almost coincident, although the amplitudes don't change very much whether calibration data is switched on or off.  However, I didn't test over a range of frequencies, and certainly not much below 1MHz, so I probably shouldn't have expected to see much change anyway.  I'll do a more thorough test tomorrow.

[fremen67]

However, I didn't test over a range of frequencies, and certainly not much below 1MHz, so I probably shouldn't have expected to see much change anyway.  I'll do a more thorough test tomorrow.

This is only static calibration. For dynamic response modification, a filter in the FPGA might be the solution...
For offset calibration, we can compensate unbalanced -12v and +12v power rail for each range. For amplitude, we can only compensate the gain so that Vpp is calibrated but +Vtop and -Vtop could still be unbalanced.

[soundtec]

Welcome back Fremen,

I did end up adding a heatsink to the Fpga ,was one I got off an old computer board of some kind I think, about 30mmsq
I didnt bother doing any precise measurements on it ,but it seems from touch at least that now the chip sits between 5 and 10 C cooler ,
Either way its probably still even without it, opperating quite safely tempreture wise ,but shifting a little extra heat is always a good thing anyway . Ill pop the thermo couple onto it at some point later in any case ,just for reference purposes .

After re-seating the heatsink on the main outputs and leveling it with 2 layers of thermally conductive (grey) pad over each 3095 its definately running a bit warmer than before with the single op amp ,probably two reasons for this ,one the 2x3095 draw a bit more current than the single 30021, and two as the original contact between the op amp and the heatsink was marginal at best and it just wasnt removing heat as effectively as it should have been. Of course the amount of heat is proportional to output level too . As I said above I will get some figures for temp vs output drive very soon also.

[soundtec]

Well after about an hour and ten minutes the FPGA temp stabilised,
Ambient was 19.2c to begin with
So with a 30x30x8mm south bridge type heatsink ,with the lid placed on top loosely,but not clicked in , it leveled out at approx 41.5C
Ktype thermo couple was placed in the centre of the heatsink .
Be interesting if someone could check what temp was reached without FPGA heatsink ,but under similar conditions in other respects,just for comparative purposes.

I might try another similar test tomorrow involving the output stage heatsink ,maybe at 5,10 and 15 volts into 50 ohm load termination.

[DaveR]

Hi soundtec - I could have given you that temperature reading if I'd seen your message a bit earlier, but mine is switched off and cooled down now.  I'll check it tomorrow when I have it on again.  It's worth noting, though, that the rated temperature for the Cyclone IV is 125degC, and I don't think it's going to be anywhere near that (maybe 50C max?).  I did measure the other heat sink's temperature when I did the 3095s and I think it was only around 35C after a heavy session, but for interest's sake I'll check that again too.

Regards,
Dave

[soundtec]

Cheers Dave,
Well within spec so at 40-50C , as I said I didnt go to any extra trouble or expense with the FPGA heatsink ,it was just something I had lying around. I cant help thinking slightly better airflow in the case itself might make things more stable or improve the settling time ,
at any rate not worth hacking up the case and making it look like a dogs dinner over it.

[bitseeker]

FeelTech announced a little contest for a few pre-release FY6800 units. https://www.eevblog.com/forum/contests/feeltech-fy6800-signal-generator-free-trial/

I suggested that they get in touch with you hard working folks to get you units regardless.

[cybermaus]

I am in dire need of a new/better LA

Trying to figure out why it is not working anymore. It is weird, because the BP was never disconnected from the loom, so there is not much chance of me making a mistake with it.
Tried a new BP, and soldered it to a veroboard for better signal and connection quality, and same result. Both BP have the red LED stay on, and both BP nicely run the standard "blink" program from STM32

Using my LA at max 20MHz I am not able to decode what is happening, because the SPI is also at 20Mhz. Ergo, I need a LA of at least 40MHz, better 80MHz
Using my Scope, I was able to see that the SPI signals themselves are pretty clean and well defined, but with the startup delay and limited scope trace, I am not really seeing what is happening (good thing I never paid for the decoding, it is not so very useful)

So:
- Any chance Fremen you can, without too much effort, make a custom 5Mhz clocked version?
- Anyone has advice on what LA? Spec-wise I like the ScanaQuad SQ100, but its a bit pricey and I'd maybe like 8 instead of 4 channels.
  (maybe I should go and search a different thread for that, but here I am anyway)


PS: CHecked and re-checked, so pretty sure about the wiring. But if anyone wants to look, see pictures.
ALso, I keept the veroboard a little larger, so I can add buttons in the future...

[DC1MC]

@cybermaus: did you try with the original panel, is it working ?