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

[soundtec]

Page 35 of the THS 3491 data sheet explains things clearly in relation to rF and rG, layout and incidental capacitance etc .
An adjustment pot could cause extra pF's to be added to the feedback network and hamper high frequency stability .
Maybe a bag of 50 each value 1.5k and 499ohm 1% parts could be selected down into matched pairs with very close to 1:3
ratio or I do see some 0.1% in batches of ten that might be worth a try . Im unsure of the wattage rating and size of the resistors needed also.

[DPA31]

What tollerance are you guys seeing on the 499, 1.5k ,10kand 50(2x100ohm)  parts ?

Hi Soundtec,

Here are the resistor values measured in situ on channels 2 & 1:

498 & 497
1492 & 1485
10.15 K & 9.99 k
50.4 both

Dom.

[soundtec]

Those figures are interesting Dom ,thanks .
I think it shows why the output from the final stage is off by a margin , slightly low compared to where it should be .
To be honest Im not very wall practised in SMD rework , I look at the size of those resistors and the density of components I the area ,and get a bit scared by thinking about how easy it would be to mess up the board . Still I very much would like more accuracy from the amplitude display ,so Im strongly cosiddering upgrading all the resistors to 0.1% or better if I can through selection of components .

In the mean time ,I put together an enclosure for my post attenuator , it has a toggle switch with  three positions , -10db, -20db, -30db , I also have  a rotary two pole switch with 12 positions I'll set up as 10 single steps of 1db , a unity gain position ,and an off position that shorts the DUT input to ground through a 150 ohm resistor for noise testing . As far as precision is concerned ,if I can make the 10 db ranges accurate ,I dont really mind some 'slop in the gears' on the 1db range switch .

As you will see from the photos Im all done with the casework and constructional side ,she's looking good  ,I have an idea the form the attenuators will take in my head but Im still not sure which is the most appropriate style of attenuation ,ladder or H network etc .
I also found a very neat stepped attenuator calculator ,you fill the blanks in terms of charachertistic impedences and attenuation  you want and the software calculates the values , I think you can model several different kinds of attenuator topologies with it too . I'm unsure about the exact form both attenuation circuits need to be to function properly in series , and any advice people can offer would be welcome .

Its all a bit chicken and egg  as far as choosing impedences , I need it for audio signals primarily ,but Id also like the attenuator to be usable at least up to a couple of hundred khz sine, I'll need to be able to feed single ended line level audio inputs of 10 or 20 kohms as well as high impedences  directly from the output ,in the case of balanced connection  being required a mumetal transformer can be used across the outputs ,I have a 1:1 10k-10k sowter transformer thats good up to about +10dbu and 50khz of bandwidth feeding 20 kohms input impedence. For mic level inputs which are around 3kohms normally, maybe a mic input transformer used in reverse(10k-600ohm) might work well.


I need more specific advice before I start building the attenuators really  ,so I dont end up having to re-do it again ,you know the old saying 'measure twice cut once' , but my head is a little fuzzy on the maths   

Pictures to follow soon , coffee time . I'l post too another forum about the specifics of the impedences and see if anyone likes the idea enough to fill in the blanks for me .

[DPA31]

Hi Soundtec,

You have an interesting project. Could you please put the link for your attenuator calculator.

You mentioned attenation of -10dB, -20db, -30db : are they dBV (20lov V2/V1) or dBW (10log V2/V1).
I understand you keep 50 ohms as output impedance, am I right ?

To continue on THS3095 gain adjustment :
I set 5.0V sine 50Hz and I measure 1.768Vrms that is 5V pp
Then I set 5.1V and I got 1.7655 V but at OPA 686N output I got only 0.446V rms which multiply by gain 4 gives only 1.784Vrms instead of 1.803V rms. Before to

Dom.

[soundtec]

 A little passive test Jig set for audio signals comprising the attenuator ,a transformer , switchable low pass filter  (to improve matters when making THD measurments) and some patch cables makes a nice addition to the FY6800  .

Maybe a simple LC low pass could be integrated into the 10,20,30db attenuator network ,then either switched to one of an handfull of known frequencies or out of circuit as required ,might afford a reduction in the harmonics of 40-50 db or more from the raw sig gen output. If we're at .03%THD at full output from the gen  ,if we could mange 60 db of attenuation say at 2 khz ,driving a 1khz signal ,wouldn't that give us .00003%THD (with some attendant insertion loss in amplitude terms of course).

Im not sure  100% sure about setting the characteristic impedence of the attenuator ,50 ohms ,its always handy to have low -Z output capabillity ,but this will come at some penalty in load on the output IC's and distortion produced  , it will also cause -3db output due to the series output resistor ,on the plus side with a 50 ohm characteristic bandwidth will probably be in the many hundreds of khz. Maybe 600 ohm might be a good compromise for general audio work , I do see an implementation of the 3491 with 2,3 and 4 op amps per channel in the spec ,at high frequencies the thd is reduced by more than 30-40 db over the case of a single chip , thats at 50 ohms ,if your opperating into higher impedences at audio frequencies THD dissappears off the scale to -100db or more down from fundamental .

dbu or dbv might be the correct expression,  I think dbu relating to a given impedance and reference level,dbv relating purely to voltage  ,Im a bit sketchy about that ,so double check for yourself in anycase .

I think the effect your seeing on the output levels is the purely down to 1% components on the network surounding the output ic  Dom ,broadly the results of the voltage checks agree with the ohms your measuring on the individual components . Im just looking for a bit of software I found last night which allowed allowed statistical evaluation of rF and rG and its outcome on the accuracy of op amp gain setting. It looked really usefull for us ,I'll find it again dont worry ,and the attenuator step calc 

Good to see a few old regulars turning up for a look at the topic again ,no need for names , ,saluts .

[soundtec]

Im just trying to visualise the circuit ,
Its essentially three -10db sections ,with the 1db range always in circuit , seeing as the rotary switch is 2x12pos I was thinking h pad type network on the 1db range ,then the three way toggle switch offers gain ranges of, unity to -10 db , -10 to -20db and -20 to -30db in one db steps.
Maybe frequency switchable  low pass LC sections could be cut in and out across the three attenuation  ranges ,that way the more attenuation you can live with, the better the out of band rejection of the harmonics .

Little passives gadgets that 'do exactly what they say on the tin', are uncomplicated and instantly feel intuitive to use , Im a huge fan of .
I could think about making a small smd board kit including toggle switch coarse range ,1db fine range control ,frequency cutoff control ,enclosure ,bnc's and maybe selected or even paralelled 0.1% resistors throughout  . Might not be too difficult to to end up with a very close tollerance network that fits on a postage stamp ,good layout with low incidental capacitance would ensure flat frequency responce to very high frequencies .It would  be possible  to use Lorlin rotary 1x12way for fine and 2x6 for frequency cutoff controls ,
hardwired 0.1% throughhole resistors  could be used without having much effect at multiples of any audio frequency of interest .An analog attenuator post output definately beats any form of digital control of output level ,if levels less than -30 db are  required you could always add an inline attenuator with a given loss ahead of the unit, and you'd still have your precise 1db steps .

[DPA31]

Hi Soundtec,

I see your project is progressing. Nice cabinets.

On my side I have finalised THS3091 with 18k and 27 K in piggy back mods on R54 and R45. see attached picture.

I am now looking at the output attenuator for 0-0.5 V range.

Looking forward to seeing your progress Soundtec.

Dom.

 

[soundtec]

Added a  scale ,poor quality print as you can see,and there are some mistakes ,but for now it will do .

[soundtec]

Just looking at the schematic , I see the gain and dc adjust pot are symetrical on the OPA686 ,this means the extra capacitance should balance out and have little effect on the signal .

Simply putting a pot in place of our 1.5k feedback resistor on the 3095 or 3491 would cause extra capicitance that might upset our output stage . What if we   replace the 499ohms with 470 +  a 50 ohm preset to ground, now any extra capacitance caused by the pot is 'hidden'   by the ratio of resitance  of  30:470 ohms  and it allow us precisely trim the output stage gain so whats coming out of the generator on both ranges corresponds exactly to the numbers on  the display.

Heres a nice attenuator calc that allow you specify impedence and it works out the values in the chain, you can rejig the impedence to get it to coincide better preferred resistance values.Shunt calc by the same author. 

http://rssconsultancy.co.uk/atten.html

I found an good op amp gain calculator here ,  its quite comprehensive compared to most ,

https://www.allaboutcircuits.com/tools/op-amp-voltage-and-gain-calculator/

You can easily add in the measured values from your Fy6800 and calculate the discrepency to expect . This software computes in the effects  of series and input to ground resistor too ,in our case 50 ohms and 10k , even 1% tollerance components in this position  will have  a very little effect on the circuit calibration, as pointed out earlier . A perfect 499:1500 ohms resistor set will have a margin of error ,but as little as a few ohms either way in the resistors can get you an exact gain of four . maybe if you had a batch of 1.5k and 499 ohm resistors to choose from you could find  closely matched pairs ,so that if you have a high 499 ohm you could match it with a slightly high 1.5k to maintain our 4:1  gain ratio.

[soundtec]

I found some very high spec resistors at 1.5k ohms , .02% Beyschlag mini melf,
The op amp calculator shows with exactly 1.5k in rF position with 50ohm series resistor and 10k input loading a 497 ohm is needed in rG for a  perfect x4 gain .
Maybe a batch of 499 1% could be selected from to find 497 ohms +or- a margin that works with our fine tollerance component  to produce a very tight calibration on the output of the FY66/6800 . Set and forget with very low drift components (15-50ppm)beats even a preset ten turn adjustment ,Its probably less that 10 euros worth of parts and a little work to make the FY into something approaching a voltage reference , so Im gonna give it a try,  as someone way back at the start of this forum page  asked,
'Whats the point in a screen displaying  10.0000 volts if the last few digits are are garbage?'

I found an op amp noise calculator last night ,it allowed tuning in of the circuit values and noise parameters of the op amp ,I tried a THS 3491 in the circuit of the FY66/68 , I got a whopping noise figure of around  -128 db with x 4 gain ,thats within a few db of the theroretical limits imposed by the laws of physics itself , it also means any noise contribution from the op amp is very much below whats produced by the dacs ,always a good thing .

[kirys]

is the ghost voltage gone on the 6800?
I read from this review (see the comments) https://tech.scargill.net/fy6800-signal-generator/ that the 6800 still has the ghost voltage, can someone confirm this?
Thank You
K.

[soundtec]

I did read Ted's blog and wondered if the mains ground was good in his holiday home where he did the initial test .
The y cap and 1meg resistor were removed from the powersupply of the 6800 and probably newer 6600's too ,and with it all grounding through the plug I see no safety issue. I own both 66 and 68 hundred's
 

[soundtec]

This guy does a good review and tear down of the 6800 ,he also does a tutorial of each of the functions of the Fy6800  and measures the leakage voltage too .

[chrisjn52]

I am interested in purchasing an FY6600 or FY6800 Function Generator with a view to PC control via the NI-VISA drivers and the SCPI command language using Python & PyVisa.

As one might expect for such a low cost instrument, it does not conform to the SCPI or IEEE 488.2 specifications but I would appreciate
it if someone with more experience than I could confirm whether there would be any major issues?

[GregDunn]

I may have mentioned it in an earlier post, but my 6800 has a nice IEC inlet, a fuse holder, and I measured no leakage on the outputs.  I've used it on numerous occasions and never experienced any issues with other test gear or equipment under test.  I do wish it was a little heavier so it doesn't slide when you press the buttons.   

[Supernick]

Hi all, please help

i buyed fy6800 on ali, but on the all frequencies (excl 5.000Mhz) i have a static jitter.

1) checked oscillator - all ok. good sine, freq 50.0007 on oscilloscope. Amplitude is good.
2) checked direct lines from ALTERA - jitter is present.. and on the control point (T1.. T2..) signal with static jitter

from ch1 and ch2 signal is splited


what is it? how to fix? what the check?

ps. i tryed rewrite my flash (winbond). Dumps from this thread. Device or does not work or non stable and without signal on chanels


MODEL: FY6800-60
BOARD: 1.51
FW: 1.7

[SMB784]

That jitter is both normal and expected. In fact you will find that for non-sinusoidal waveforms, all frequencies above 5MHz that aren't an integer multiple of the sample frequency of 250MHz will have a 4 ns jitter (4ns=1/250MHz).

At the moment there is no known way around this issue, and it appears to simply be a feature of the generator.

Maybe in the future it will be possible to correct it with software, but for the moment it's just something you have to deal with. Also, according to the spec sheet for other similar generators like the more expensive and less capable models from siglent, rigol, and gw instek, their jitter is as bad or worse for non-sinusoidal waveforms

[Supernick]

thank you for your reply!

but in fy6600-30Mhz all supergood. any signal forms without jitter on any freq
and other FY6800 owners (I asked) confirmed that they have a signal without a jitter

why?

[Supernick]

also.. jitter is present on all range. from 10KHz.. Mhz
only 5MHz without jitter 

[SMB784]

Try the following square wave frequencies: 60 MHz, 50MHz, 30MHz, 25MHz, 15MHz, 12.5 MHz, etc. Then try them with sine waves, and see where the jitter appears

[Supernick]

Sine is cool - on the full range.
square is bad from start range..

please look

PS. jitter is 4ns, it's true. But why in fy6600-30M jitter is not present?

[rf-loop]

Also, according to the spec sheet for other similar generators like the more expensive and less capable models from siglent, rigol, and gw instek, their jitter is as bad or worse for non-sinusoidal waveforms

As far as I know square wave and rectangle wave is not sinusoidal.
What Siglent model have this kind of jitter with Square/Rectangle (other than 50% square)

But then yes, if look Triangle/saw, pulse, arb, there is jitter related to clock. (and example cheaper Siglent models may have example 150MSa/s clock what mean 6.7ns cycle to cycle jitter exept some "golden" frequencies.

But as seen here some previous example images about FY6600 using square have lot of jitter, cheap Siglent model 1000X do not have this type of jitter ancluding also rectangle waves = square duty other than 50%  (not even with 0.001% duty what can also call as "pulse" but in this mode rise and fall time or pulse width can not adjust using 0.1ns adjust steps) My opinion is that Square and Sine need do so that they do not have 1 clock jitter.  Other waveforms can perhaps accept more in cheap simple equipments.

Siglent do not have this "1 clock" square jitter in real life and also -- not in specs sheet (exept pulse mode, triangle and arb)
Is it better to stay with facts and not trumpths.

[SMB784]

Also, according to the spec sheet for other similar generators like the more expensive and less capable models from siglent, rigol, and gw instek, their jitter is as bad or worse for non-sinusoidal waveforms

As far as I know square wave and rectangle wave is not sinusoidal.
What Siglent model have this kind of jitter with Square/Rectangle (other than 50% square)

But then yes, if look Triangle/saw, pulse, arb, there is jitter related to clock. (and example cheaper Siglent models may have example 150MSa/s clock what mean 6.7ns cycle to cycle jitter exept some "golden" frequencies.

But as seen here some previous example images about FY6600 using square have lot of jitter, cheap Siglent model 1000X do not have this type of jitter ancluding also rectangle waves = square duty other than 50%  (not even with 0.001% duty what can also call as "pulse" but in this mode rise and fall time or pulse width can not adjust using 0.1ns adjust steps) My opinion is that Square and Sine need do so that they do not have 1 clock jitter.  Other waveforms can perhaps accept more in cheap simple equipments.

Siglent do not have this "1 clock" square jitter in real life and also -- not in specs sheet (exept pulse mode, triangle and arb)
Is it better to stay with facts and not trumpths.

I never specified which models I was referring to.  In fact for the SDG830 (>$300) there is 1 clock cycle (8ns) for arbitrary wave forms, which are non sinusoidal. Nothing that I said was untrue.

The main point of what I was saying is that the 4ns jitter in the FY6600&6800 is to be expected, and is actually less than that of competing generators in some modes.

[rf-loop]

Also, according to the spec sheet for other similar generators like the more expensive and less capable models from siglent, rigol, and gw instek, their jitter is as bad or worse for non-sinusoidal waveforms

As far as I know square wave and rectangle wave is not sinusoidal.
What Siglent model have this kind of jitter with Square/Rectangle (other than 50% square)

But then yes, if look other non sinusoidal as example triangle/saw, pulse, arb, there is jitter related to clock. (and example cheaper Siglent models may have example 150MSa/s clock what mean 6.7ns cycle to cycle jitter exept some "golden" frequencies.

But as seen here some previous example images about FY6600 using square have lot of jitter, cheap Siglent model 1000X do not have this type of jitter ancluding also rectangle waves = square duty other than 50%  (not even with 0.001% duty what can also call as "pulse" but in this mode rise and fall time or pulse width can not adjust using 0.1ns adjust steps) My opinion is that Square and Sine need do so that they do not have 1 clock jitter.  Other waveforms can perhaps accept more in cheap simple equipments.

Siglent do not have this "1 clock" square jitter in real life and also -- not in specs sheet (exept , triangle and arb)
Is it better to stay with facts and not trumpths.

I never specified which models I was referring to.  In fact for the SDG830 (>$300) there is 1 clock cycle (8ns) for arbitrary wave forms, which are non sinusoidal. Nothing that I said was untrue.

The main point of what I was saying is that the 4ns jitter in the FY6600&6800 is to be expected, and is actually less than that of competing generators in some modes.

As I told Square and rectangle  are not sinusoidal and Siglent (not even SDG800 serie or SDG1000X serie) do NOT have this jitter as FY have. And not even in Pulse mode. You told that not sinusoidal have. So it was wrong.  It is surprice that FY have this one clock jitter in Square or rectangle in square mode. And btw, also SDG800 series do not have it in square/rectangle mode. Square is one typically most important to be low timing jitter after sinewave. 
Totally different things are other wfms like arb, triangle, pulse in fixed clock system. There also in Siglent is one clock jitter and in 125MSa/s models it is 8ns and in 150MSa/s it is 6.7ns and so on as every school kid can calculate ans also told in datasheets.

Also, according to the spec sheet for other similar generators like the more expensive and less capable models from siglent, rigol, and gw instek, their jitter is as bad or worse for non-sinusoidal waveforms

Simplified:
If you talk non sinusoidal wfms it include all but not sine. (sentence logic!)
The argument is false, even if you try explain you did not say or mean wrong.

It is true only with FY

If look square, rectangle and pulse mode in Siglent SDG800 and SDG1000X they both outperfom FY just hands down and specially with this jitter. Siglent do not have one clock jitter in Square, rectangle and pulse mode, using what ever user-defined pulse width and or with what ever user-defined pulse rise and fall times or duty% if talk Square function. Including all frequencies.

Square(including rectangle)  and  Pulse
800 series is worst but  even it have specified max 500ps +0.001% (10ppm) period
1000X series max 300ps +0.05ppm period
And typically much less.

This is why they use own developed "EasyPulse" technology what make possible to do these without 1 clock jitter in this kind of simple DDS generator and what FY can not do.

Even oldest SDG1000 serie do not have 1 clock jitter with square/rectangle what is derived by hardware from internal sine.

Arb is different, it is simple way produced from arb memory.
Also triangle have 1 clock jitter.

---------
(and even in this el cheapo FY it is really strange when they still can produce normal almost ok sine and then they can not do it with square without one clock jitter.  Other waveforms than sine and square are bit different things and in simple principle DDS generators it can understand these have one clock jitter (+ small amount other origin random jitter))
NNR

[thielj]

I bought the FY6800-60 and the display stopped working after the 3rd use (backlight lights up, but display simply stays white/blank, while the unit, all buttons, signal generator etc is still working). After some discussions, the seller promised to send me a display board to exchange myself but nothing has arrived yet.

I've now opened the case and checked all the connectors etc. Is there anything obvious that could cause the display to stay blank that I would be able to fix myself?