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

[fremen67]

This is a lot of work, and I had a lot of distraction today (my boss/colleagues/clients kept bothering me)

Attached the XLS with -10~+10 hard range and both High-Z and 50 Ohm.

A few notes:

- At 50 Ohm above 10V, the OpAmp is just not cutting it. Rather then trying to correct this (it will not be able to anyway) this needs to be corrected by a better OpAmp and maybe 15V rails in stead of a 12V rails that is to weak as well. Or at the very least a 12V rail with more power.

- Around 10V - 60Hz there is a weird additional negative offset. It also attenuates more, but I think both are caused by the OpAmp cutting of the high crest of the sine more then the low crest of the sine.

- Likely bad sine to the right and bottom is caused by the OpAmp attenuating, toward the left by a weak DAC signal, amplified noise and DAC resolution.

Perfect 
A lot of work indeed but essential as we need to analyze what can be done by software and what cannot be. And it will also be the reference for future hardware modifications.
With those results, it already seems obvious that some OpAmp upgrade will be a must. A more powerfull power supply could also be usefull.

To be honest, Fremen67, I will provide you with the info requested to my best ability, but I wonder if level correction should not be a lower priority. Sweep/Mod/Counter/LCD may be more rewarding features. Will work on 0.275 range next.

That is exactly why it is important to have data and think before doing any useless things.
At least static calibration can be done without much effort. A digital filter in the FPGA might be the real solution for bandwith upgrade.

[fremen67]

And for some reason it won't generate wave forms this evening Flashed the 0.3 and 0.4 and it will connect fine. But no output. 

Was going to fill out Cybermaus his excel to see some diffs maybe. But it is not happening yet 

Lot of chances that it is a wiring problem. I also wasted 1 hour last week because I swapped 2 wires when switching configuration...

[fremen67]

Small intermediate result:

The 5.5Vpp in a hard 2.75V range gives such a perfect 6Vpp that I am wondering if you have a coding error in the software.

Also: apparently  the 2.75 range is able to go up to 3V

I have a perfect 6.12V on mine for 5.5V 
I think it is now obvious why feeltech did not use the full range. A mistake is always possible but the formula is the same for the whole range... At least you should have the same results with the original firmware in the -2.5V/+2.5V range. In fact based on the formula, it should even go up to 2.778V  for full range (4095 points)
Results should be very similar between the Low and Medium range as there is only a voltage divider to go from one to the other

[cybermaus]

Ok, uploaded new XLS above, calling it a day.
Also changed the color formulas a bit, if you are using this template.

My intermediate conclusions:

- If you want above 10Vpp with 50 Ohm load, you must hardware mod OpAmp and PSU (at any frequency) (or maybe a cap close to the OpAmp)
- the factory 20MHz drop out of high range seems justified (for autorange)
- the factory switch to low range below 0.5V and to high range above 5V seems justified (for autorange)
- any optional 50 Ohm readout mode should double voltages on native and high range, but 2.65x in low range, due to bad attenuator values.

Edit: Additional thoughts:
- A 75 Ohm readout mode seems tempting, but may falsely lead people to think it can be set to 75 Ohm output impedance.
- Also a true 75 Ohm impedance is easy to make. Just add a 25 (22) Ohm series resistor at the BNC, its close enough to to 50 Ohm resistor to not have relevant reflections.
- Thinking about 75 Ohm I suddenly realize this device is missing a build-in TV (PAL-NTSC) waveform. Not that I need it, but for nostalgia reasons.

[Insatman]

Re: FY6600 Jitter

Insatman... Excellent results!!!  Would you do a simple FM modulation test?
Set CH1 to sine wave with freq as before, 10.1398 MHz.
Set CH2 to Square wave with freq 0.25 Hz. (Don't worry about Ampl and Offset as they are not used for internal FM).
Set modulation source to FM from CH2. Use BIAS 00'000'.001'500'000KHz. This will be set modulation depth to 1.5 Hz.
Your SA FFT waterfall should show the 1.5 Hz square wave modulation of the sine wave.

I'm considering your FY6600 modifications. My smd skills are poor. Probably can make the Power Supply mods. But removing the smd oscillator seems beyond me at the moment. If I can get it removed was thinking of tack soldering wires to the new osc module and tacking these wires to PCB. Any comments?

I did the test you requested with one change.  The CH2 squarewave frequency was reduced to .02 Hz.  This was necessary as the screen/waterfall update rate is about once every 2.5 to 3 seconds on my SA at the RBW of 3Hz.   Running at the faster frequency resulted in "fuzz" rather than the "notched" waterfall pattern.   Note that the spikes on the waterfall plot only occur at frequency transition times.  Also not the lack of average frequency drift on both plots as the measurement took several minutes at approx 2.8 seconds per measurement line.

This was the first time I used the VCO function on my FY6600, so it took me awhile to get it working.

I removed the smd oscillator easily with a hot-air gun.  This was after carefully masking surrounding components with two layers of Kapton tape.   Soldering the new oscillator was more difficult as my smd skills are not great, but I do have a microscope to work under.   As for attaching wires to remote the oscillator...yes this is possible for a short distance as the 50Mhz frequency does not tolerate much inductance.  Twist the wires if at all possible to minimize inductance.   

[DerKammi]

Back in business. It was the wiring, but not a fault in which goes where, but a faulty connector. Used a different one in the ribbon and all is well.

And for some reason it won't generate wave forms this evening Flashed the 0.3 and 0.4 and it will connect fine. But no output. 

Was going to fill out Cybermaus his excel to see some diffs maybe. But it is not happening yet 

Lot of chances that it is a wiring problem. I also wasted 1 hour last week because I swapped 2 wires when switching configuration...

[Gandalf_Sr]

Which output are you referring to? Front panel output could benefit from different op-amps if really needed.

Back output won't be able to keep up with that high frequency due to a 74HC245 buffer which runs at typical 55 MHz max. so 60 is maxing it out pretty much.

I haven't read/followed the whole thread but was aware of the FY6600's existence, seems like you guys are kicking it into shape - good work

Depending on the voltage levels it needs to run at, there are several CMOS family alternatives to the 74HC245 buffer that would have much lower gate propagation times.

1. Is there a schematic available?
2. Who is the preferred vendor to buy an FY6600 from?

[cybermaus]

1. yes, please go back a couple of pages and/or search DerKammi's posts.
2. I think they are all equally bad/good. Pick one far away and cheap if you can wait, or local if you want it fast and without added customs tax.

[DerKammi]

Here you go...

Which output are you referring to? Front panel output could benefit from different op-amps if really needed.

Back output won't be able to keep up with that high frequency due to a 74HC245 buffer which runs at typical 55 MHz max. so 60 is maxing it out pretty much.

I haven't read/followed the whole thread but was aware of the FY6600's existence, seems like you guys are kicking it into shape - good work

Depending on the voltage levels it needs to run at, there are several CMOS family alternatives to the 74HC245 buffer that would have much lower gate propagation times.

1. Is there a schematic available?
2. Who is the preferred vendor to buy an FY6600 from?

[spec123]

Thanks Insatman for all the information. Plots are very instructive. Glad to see that your plots are close to other ones from Kaku. This means that this SG should allow some precise work with narrow band FM. Excellent performance for such a low cost SG.

I'm leaning in the direction of making the modifications you describe. Now feel more confident in replacing the original 50 MHz oscillator. Checked and found the TCXO Model D75J in stock at DigiKey. Thanks for all the tips.

Right now I am contacting the vendor of my FY6600 M60, FW 3.2 to see if it can be exchanged. The residual FM with my unit is too much for my work. They now have the FW 3.2.1 version in stock. I'm wondering if they upgraded the oscillator. I noticed earlier in this long thread (photos by DerKammi) that two units had different 50 MHz oscillator chips. Kaku's unmodified unit did not have the jitter problem. Maybe I have the poor quality chip in my unit.

[Kaku]

My FY6600 died  . Works yesterday, but today will not start, hangs up at startup at "Feeltech logo", nothing more. I also tested with BP, same result, not work. I'm trying to figure out the problem later.

[cybermaus]

Not sure if it will help you but:

Hanging with the logo displayed means the STM32 CPU is not getting (correct) responses from the FPGA.
For example if you disable FPGA because you are programming Winbond, this is shown. Also some people have reported stuck logo because the cable came loose during transport.

[Kaku]

Not sure if it will help you but:

Hanging with the logo displayed means the STM32 CPU is not getting (correct) responses from the FPGA.
For example if you disable FPGA because you are programming Winbond, this is shown. Also some people have reported stuck logo because the cable came loose during transport.

I know the effect of fpga disabling. I'll check the cable and other things later, now 

[DerKammi]

Had my fun testing ranges according the sheet of cybermaus, Did the upper 2 ranges with same values of the sheet.

I have modded the opamps for 2 separate ones. Does yield some differences perhaps.

Connected the AWG directly to my scope via coax cables and switched on/off the builtin terminators of the inputs.

Suggestion from my side. maybe we can have an automated step function with adjustable step time for these tests? So then we can just write down values and not switching back and forth programs and edit values. I'm half dead right now

[cybermaus]

Yes, it does get boring after a while, doesn't it?

If ever we build a per value / per device calibration matrix, yes, we should have such auto-set macro function. Does not even have to be in the BP, can be in the PC.
However, I think for now, all we need is some rough guidelines on what autoranges to select, and maybe some coarse global correct. So maybe one or two more person, then we are done for now.

Unless Fremen67 has different idea's.


Mind you, I did see something interesting in your data: It does not have the weird breakdown in the 10V-60MHz area, where mine does an extra cripple on the top half of each sine, causing negative offset and extra attenuation. I am suspecting the OpAmp for that. Not sure if the entire OpAmp range, or just my specific chip. Anyway, you replaced those, and have clearly better results in that area.

At the same time though, the 15V and above at 50 Ohm load range, your improved OpAmps are (slightly) worse. Only able to drive 6.6V instead of mine 7.4V.
I wonder if that is because my PSU is rather "uncalibrated" (iow bad), and giving out 15V on the 12V rail, or maybe your improved caps on the PSU are not that much better

What we / I should do is hook up the 12V rail completely external from a pair of bench power supplies, I have some capable of 5A each, should be enough.

PS: what OpAmps did you use again? And what voltage is your 12V rail?

[cybermaus]

Also, you made a couple of likely typo's on:

B61; C58; D63; F61; B43; C48; C33; B30;

I15 is also suspicious

Regarding block line 53 to 63: You filled in a 2.75 50Ohm range there, but I had High-Z on that position. So the dB calculation should have an extra /2 (or *.5) that is why it is so red, basically all dB values should be +3.

PS: apologies if this sounds a little pedantic, not really intended, but unsure how else to bring the info.

[Candid]

I'm leaning in the direction of making the modifications you describe. Now feel more confident in replacing the original 50 MHz oscillator. Checked and found the TCXO Model D75J in stock at DigiKey. Thanks for all the tips.

Does anyone know another source for the TCXO than digikey? In germany you pay 34 EUR ($42) including postage when you order from digikey.

[Scratch.HTF]

I believe that at least one PLL (which is synchronized for DAC data against the DAC clock) is most likely set to 1000 MHz.

[DerKammi]

PS: apologies if this sounds a little pedantic, not really intended, but unsure how else to bring the info.

Don't worry one bit, I'm all about content and not about 2nd guessing peoples meaning. If you think it might be wrong I'll check it asap this weekend, no problem. It was becoming very boring in the second half of the measurements so a mistake is easily made.

I did all the measurements in AC mode as I have a DC offset in all measurements. Didn't want to look at it all the time. So I cannot comment on the neg offset of you. And I'll update the sheet for the LowZ of the mid range.

I did not modify the PS actually, only the opamps and the oscillator. I want to replace it with a full linear +5 +- 14V and also +-5V for replacement of the 7805 and 7905 which get to hot for my liking. But need to measure to current draw first.

All in all I think I am a lit bit lower in gain on both output ranges. My 2nd channel is way lower even. Needs to be adjusted first. Also want to change the 10k for the input of the opamp to 33k as suggested earlier to correct the gain for a better x4 gain.

[cybermaus]

I looked a bit further into the 20V into 50 Ohm challange. Basically hooked up the +5V +12V and -12V to linear lab-supplies limited to 1Amp
tl;dr; It *barely* matters. Maybe even makes it a tiny bit worse.

• 15Vpp into 50 Ohm with 12V rail: 6.72Vpp  from -3.36V to +3.36V
  baseline. Flattened, but still balanced
  
• 15Vpp into 50 Ohm with 15V rail: 6.8Vpp from -3.52V to +3.28V
  Slightly more signal, but looses balance. I wonder if that is my device only.
  
• 20Vpp into 50 Ohm with 12V rail: 7.68Vpp  from -3.92V to +3.76V
  Now 12Vrail too the top half looses more then the bottom half
  
• 20Vpp into 50 Ohm with 15V rail: 7.68Vpp from -4.00V to +3.68V
  Does not even address Vpp, just more unbalanced
  

So what I conclude is that only upping the power from the PSU is not going to help. Not on my device anyway.
In fact, in my case it brings a tiny bit more imbalance, not sure if that is my device only.

Yellow is 50 Ohm loaded wave, Blue is non-loaded sine, for shape compare

I also measured current with *both* channels 50 Ohm loaded on 20Vpp: -15V was 101mA and +15 was 99mA and +5V was 394mA
Current does not vary from 10KHz to 20Mhz and above that Vpp drops anyway.
However, when going down to 0.01Hz (practically DC) we do get an increase whenever the crest is moving by:  -15V at 130mA and +15 was 128mA

----

Regarding a new PSU: I was thinking: a dual 15V or 18V transformer may be enough.
Two good + and - 15V regulators, as well as a cheap 5V DC to DC switched regulator.

Because if you look at the schematic, you will see the 5V from the PSU is *never* used as rail. Instead it is used to make the lower voltages 3.3V, 1.8V etc.
And the actual system board 5V rail is taken from the +12V supply.

So a small switched 5V module at 400mA would only need 150mA from an unregulated rectified 18V (so hooked up before the 15V regulator)
A simple dual 15V (or 18V) transformer capable of 250mA may be smaller, cheaper, easier then a triple one. Make it 500mA for some headroom.

Of course, I have no proof more power actually helps. Not with the stock OpAmp at least.

[DerKammi]

Could you check the input of 2 stages of opamps? Is it going wrong before or after the THS3001 or even earlier?

[cybermaus]

Well, since the non-loaded output is perfect, I did not see how it could be the input of the OpAmp, but I checked anyway.
As expected, the input of the THS3002 is still a good looking sine with curves in all the right places.

Also, I ran the wave all the way down to 0.01Hz, and the shape of the flattened top stayed the same. Meaning it is not a response time thing. Not even a current thing, because if the local capacitor would drain out, surely it would fall off more toward the trailing edge.

I wonder if it is something to do with the fact this OpAmp is described in the datasheet as a "CURRENT-FEEDBACK AMPLIFIER". I do not know what that means, normally when I mess with OpAmp I consider voltages. Need to read up on that.

Also, I wonder if it can be resolve by, instead of replacing, *adding* an OpAmp.
Just leave the THS3002 in place, and add a pair of THS3001. They have the same behavior, the PCB traces are in place, de-soldering is harder then soldering.
All we need to do is add to the unpopulated footprint.

[cybermaus]

A different topic: I found a bodge and a PCB variance
If you look at the images, my V1.5 board has some missing traces which they fixed with a solder blob.
But also, even if the traces were there, it would be slightly different to DerKammi's V1.501 board.

There is also a lot of unpopulated stuff around that chip. Seems to be a place for another tuning pot.
That area is not yet traced into the schematic is it?

[cybermaus]

While I was studying your schematic, also I think I spotted an error (sorry again).
The DAC's for offset output. VoutA and VoutB seems swapped between the A channel and the B channel. Is that correct?

[DerKammi]

Well, since the non-loaded output is perfect, I did not see how it could be the input of the OpAmp, but I checked anyway.
As expected, the input of the THS3002 is still a good looking sine with curves in all the right places.

Also, I ran the wave all the way down to 0.01Hz, and the shape of the flattened top stayed the same. Meaning it is not a response time thing. Not even a current thing, because if the local capacitor would drain out, surely it would fall off more toward the trailing edge.

I wonder if it is something to do with the fact this OpAmp is described in the datasheet as a "CURRENT-FEEDBACK AMPLIFIER". I do not know what that means, normally when I mess with OpAmp I consider voltages. Need to read up on that.

Also, I wonder if it can be resolve by, instead of replacing, *adding* an OpAmp.
Just leave the THS3002 in place, and add a pair of THS3001. They have the same behavior, the PCB traces are in place, de-soldering is harder then soldering.
All we need to do is add to the unpopulated footprint.

Didn't except to see otherwise. Just a sanity check. Current feedback opamps normally have a higher slew rate. Don't know the details which I can explain in person.

Don't connect 2 opamps with the output directly to each other, need some sort of balance resistor normally.

I'll do some LTspice sims on this to try to understand why this clipping is happening. Current is also not a problem as the lower freqs are fine. The combo voltage/current at the 15..20M is something to look at.

A different topic: I found a bodge and a PCB variance
If you look at the images, my V1.5 board has some missing traces which they fixed with a solder blob.
But also, even if the traces were there, it would be slightly different to DerKammi's V1.501 board.

There is also a lot of unpopulated stuff around that chip. Seems to be a place for another tuning pot.
That area is not yet traced into the schematic is it?

Indeed, but the blob is fine. It's a short anyway for the VCO input buffer which is on the schematic. The other side is unknown to me and looks to be a part of the counter input. This is on the planning of next week.

While I was studying your schematic, also I think I spotted an error (sorry again).
The DAC's for offset output. VoutA and VoutB seems swapped between the A channel and the B channel. Is that correct?

Is correct, checked it a couple of times myself. Board routing is more simple in the current state.