FeelTech FY3224S 24MHz 2-Channel DDS AW Function Signal Generator

[bitseeker]

@masterx81: The impact of the various on resistances makes a lot of sense. Interesting what you found about the loading on the Altera. Hopefully, resistor values can be found that will yield a happy medium between load and noise.

[masterx81]

Yes, it's a tradeoff...
I think that on dac side i'll go with the increased value resistors way. If i want to reduce the opamp error a second opamp can be added, with low bias current. In this way we free the second opamp of the THS3002 chip, allowing to put in parallel the two opamp and get more output current without changing it with the more powerful THS3092 (paralleling them in this way: https://e2e.ti.com/blogs_/archives/b/thesignal/archive/2013/03/26/paralleling-op-amps-is-it-possible , as a normal parallel need 2 pot for setting amplification the same on both amps). For the attenuation side i think that the best thing to do is use the adg733 analog muxes. Need to try to find them cheap. After all this remain to tune the output filter.
I think that with this mods the device will be quite accurate...

Also the ADG1433 with 4ohm Ron is nice, but it's not pin to pin compatible with the 4053...

[bitseeker]

Sounds like a good recipe. You've become quite the guru on this generator! Looking forward to seeing your new-and-improved version (Masterx8124S  ).

[masterx81]

A guru is someone that already know what's doing, i'm only learning and doing tests to made this product a bit better without spending too much. A good function generator costs too much for me, and i prefer to increase my knowhow a bit...
Someone with good skills can do all this in few minutes

In the meantime i was searching other triple analog muxes with ron<5ohm, and the only ones seem adg1633, adg1433 (almost identical, with different bandwidth) and the pin to pin compatible adg733. The first two have 4ohm Ron, the supply can be unchanged (works with +5/-5v), but the pinout is a lot different, and not seem pratical the conversion.
The adg733 is almost compatible, but it have a suggested supply of +2.5/-2.5v. This is not good as the control lines from the altera chip are 3.3v and the digital lines allow a maximum of vcc+0.3v (so 2.8v), so at least the +v must be +3.0v (but we have already a 3.3v regulator so i'm using it), giving a total voltage of 5.8v, a bit out the reccomended 5.5v max but a lot less than the absolute maximum limit of 7v.
The things to do for have it working in the signal generator is lift the +vcc pin and connect it to the +3.3 voltage regulator (including coupling capacitor), change the onboard -5v voltage regulator to a -2.5v one (the -5v seem used only by the analog switches), change the feedback resistor of the opamp after the dac for have a max 5vpp (better a bit less) signal instead of the near original 7.5vpp, and change the gain of the power output opamp. the 2.5ohm ron resistance of the adg733 vs the 50~80ohm ron of the 4053 will increase a lot the accuracy of the amplitude setting. if it's not enough, it's possible to increase the value of the r-2r attenuation network.

The big question about all this mods is the output filter, i not know almost nothing about filter design  Plus at least one hundred of things that for sure i'm missing as i have no experience in this field... But i want to learn 

[So-Hm]

Upgrade of FY3224S(opamp THS3002): linear psu and the amplifier on 1 channel from opamp LT1210CT7.

[masterx81]

Upgrade of FY3224S(opamp THS3002): linear psu and the amplifier on 1 channel from opamp LT1210CT7.

Interesting, did you have scope traces of ch1 vs ch2 at various frequencies?

[So-Hm]

Interesting, did you have scope traces of ch1 vs ch2 at various frequencies?

Ch1(out1 LT1210) vs ch2(out1 THS3002), amplitude on the generator 10V:

[masterx81]

Thanks!
Have you placed the LT1210 as a buffer after the ths3002 or you have bypassed and substituted the output opamp? As seem that the ths3002 is present (for amplificating the signal after the DAC)

[So-Hm]

Thanks!
Have you placed the LT1210 as a buffer after the ths3002 or you have bypassed and substituted the output opamp? As seem that the ths3002 is present (for amplificating the signal after the DAC)

I did this scheme, LT1210 as AV=2 after a resistive divider 1:2 that after ths3002(50 Ohm out)

[masterx81]

Good for have only the benefit of higher current without have to recalibrate the output filter as i have to do... But the high slew rate of the ths3092 that i'm using extend a bit the frequency range of the square wave (you can see that the edges on my scope traces are more "steep" than the untouched channel), and the 260ma seem enough for not have attenuation at higher frequencies.
As soon arrive all the components i'll update my research....
Thanks for sharing!

[So-Hm]

I too wanted to establish in the beginning ths3092 instead of ths3002, but is more optimum (more powerfully several times) for my tasks there was LT1210.

[masterx81]

Yes, if you remove the 50ohm output resistors (actually 2x 100 in parallel) you can use all that output power,  else with +/-10v output you get 200ma max (while the ths 3002 have only 100ma output)...in your schematics i see 1ohm output resistor so i think that you need to drive huge loads (and the psu was designed with that in mind)...

[So-Hm]

Yes, if you remove the 50ohm output resistors (actually 2x 100 in parallel) you can use all that output power,  else with +/-10v output you get 200ma max (while the ths 3002 have only 100ma output)...in your schematics i see 1ohm output resistor so i think that you need to drive huge loads (and the psu was designed with that in mind)...

Yes, capacitive load really big. For example comparing, the oscillogram 1 - THS3002 (a maximum load 6V amplitude), 2 - LT1210(7V amplitude on the generator). The upper channel - U, lower - I (via the resistor of 1 ohms).

[Dwaine]

Upgrade of FY3224S(opamp THS3002): linear psu and the amplifier on 1 channel from opamp LT1210CT7.

Wow.....  That's look great...

[So-Hm]

Wow.....  That's look great...

Thanks.
Heating of radiators under good loading in the closed case isn't higher than 46C(radiator LT1210/ 40C psu) at the room temperature 22C. For improvement of ventilation in the case I have made several cuts.

[masterx81]

Hi! I've done some other mods to the signal generator.
Added a 40x40 12v fan, temperature controlled via a tc648b controller ic at 30hz pwm (nice toy )
I've also changed the R-2R DAC resistors to 1k/2k 0.1% insted of 500/1k, no noticeable increase in noise, the output of the first opamp is near 3.3vpp (quite ok for the maximum 5vpp that i'll have to use with the adg733 mux, maybe i'll try to icrease it at 4/4.5v for reduce at minimum possible the amplification of the power output opamp).
The error at the change of the MSB is reduced by a lot (see attached image), and i think that it's a good tradeoff. Can be almost totally eliminated by increasing the R-2R values (next step with common resistor values is 5k/10k), but i not think that is the case as for me it's ok 1 or 2 bit step error.

[Kleinstein]

The use of 0.1 % resistors at 1:2 ratio does not help much with the R2R DAC. There is also the output resistance of the digital outputs that is effectively in series to the 2R resistors. So suitable resistors would be at something like 1 K and 1920-1980 Ohms depending on the Outputs. So 1 K and 499 Ohms was bad. 510 Ohms instead of the 499 would be likely the better solution.
Even worse the output impedance can be different for high and low state and temperature dependent.

To reduce the error one might need to adjust the resistor(s) at the most significant bit(s). As a pot will add too much capacitance this can get a little tricky.  The 2 K resistors in series to the bit lines are slightly on the high side. So one could use a resistor and series pot to GND to reduce the amplitude of the lower bits contribution (connect where the last 1 K resistor meets the second last 2 K). Assuming about 30-50 Ohms for output resistance, something like a 27 K to 50 K resistor and a 10 K Pot could be about right.

[masterx81]

Yes, i know that using perfect 2:1 ratio isn't the best choice, in any case digital output On resistence differ from Off resistance, and change with temperature, so it's difficult to spot the right value. Also "strange" resistor values in 0.1% are not easy to find (and not cheap...). I'm trying to do the best as i can with cheap components...
Increasing from 499/1k to 1k/2k has made a better result (tested with oscilloscope) because with higher value resistor the On and Off resistance made less difference to the output network. Say on resistance of 2ohm on a 500ohm resistor is 0.4% error, 2ohm on a 1k resistor is only 0.2%.
For the impedance of the output, i'm not sure if made difference, as the output enter directly in the opamp (really high impedance).
In any case, thanks for the suggestion, as i'm not really confident with this type of circuits, and i'm learning playing with it...

[CCB]

Hi Everyone,

Just received mine from Banggood today (26/10/2016) and thought I'd post some pics/version numbers as it may help someone. 

Can't read the supplied cd, will try with another drive tomorrow.

Firmware version 3.5
Board version 2.1
Power Supply - still has leakage the same as other versions.

[masterx81]

The HW is the always the same, but they have updated the firmware... Would be interesting to see what changes they have made...

[StillTrying]

Have they fixed the leading 0s yet, I can't stand strings of 0000s which look like 8888s to me.

[bitseeker]

Wow, version 3.5. Mine is only 2.0. Is there a changelog somewhere? A quick search online wasn't fruitful.

[masterx81]

Have they fixed the leading 0s yet, I can't stand strings of 0000s which look like 8888s to me.

Not easy to fix/almost impossible.... And i bet that it's present on most 12bit signal generators. In any case the error is only few mv, and you can half it by using 1k/2k resistors instead of 499/1k.

[Kleinstein]

Changing the resistors to 1K/2K will only reduce influence of output impedance. With a maybe 20 Ohms output resistance they are still about 1% off. So only good for 6 to 7 Bits, not even 8 Bits. There is also a downside in going to higher impedance: parasitic capacitance gets more important and there can thus be larger glitches on switching.

So even 8 Bits only makes really sense if they go to more suitable resistors like about 510 Ohms instead of the 499 Ohms. As 510 ohms is a standard value chances are to find better values in newer units.The 510 Ohms / 1 K or 499 / 980 Ohms combination is likely better than the 1K/2 K pairs.

So instead of changing all the resistors, the more useful way would be adding suitable resistors (e.g. 50 K for a 20 Ohms driver resistance) in parallel to the 1 K ones for the more significant bits this way changing the 1 K to maybe 970-990 Ohms. The best value depends on the FPGA used and the actual value of the resistors. So it would take a few testes, possibly with temporary trimmers find the right value. Already changing the resistors for the 2 highest bits could be a significant improvement.

Getting true 12 Bit resolution is likely impossible even with well trimmed resistors, as the output impedance is quite temperature dependent and different for H and L state. Unless they have changed the resistor values, adding 4 more bits is not helping much.

[CCB]

Have they fixed the leading 0s yet, I can't stand strings of 0000s which look like 8888s to me.

It's still the same.

I found this feedback on Banggood about the 3.5 firmware, I guess this is listed on the supplied cd. (My CD doesn't see to work need to try a different drive)

"The latest (v3.5) firmware has nice features - tone bursts (start & end cleanly on zero-crossing) - triggered by manual, external or channel 2, ASK & FSK modes (manual or external); 'manual' for these modes is the rotary encoder button. The counter can count pulses as well as measure input frequency"