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

[JohnPen]

I have just tried the 1 khz 2khz check on my FY3224S using my DS1054Z scope on CH1 and Ch2.  Observing the display in XY there is a slow drift in the lissajou display.  A complete cycle of the the difference takes just under 3 minutes (about 2 secs under).  Also the scope frequency meter (on CH1) seems to cycle between 999.999 hz to a minimum of 999.997 hz. Just to be sure I checked with both sources set to 1khz but one channel set to a 45 degree phase difference to make it easier to watch.  There was no noticeable drift under this circumstance although the frequency counter on the scope CH1 still showed a drift as mentioned earlier.

Interesting.

John

[JohnPen]

I repeated this test for 10khz 20khz and could not see any drift on the lissajou.  However at 100hz 200hz it was the same drift rate as found at 1khz 2khz.

John

[gby]

Whoops, I missed the suggestion of doing two different frequencies, e.g. the suggested 10 kHz and 20 kHz.  I just tried 1 and 2 KHz and see what you mean.  Given the simple nature of these generators and given how the frequencies are made I can see how two different frequencies would be an issue.  Probably some kind of truncation of the math used to set the frequencies.

If you need to have two different simple integer ratio frequencies that stay absolutely locked, then I would use one as wave type sine and one as wave type arb with multiple sine cycles in the arb.  For example, set CH1=CH2 frequency mode, program Arb for 2 sine cycles per 4096 pints (for the FY3200).  Then select sine wave for one and the 2 cycle Arb sine for the other.  Now you have perfectly locked signal with 2:1 frequency ratio.  I just tried this on the FY3200 and it works fine.  You can even change the phase value to change the relative lock.

Of course, if this mode is just a test of the device's capabilities, then yes yet another weakness has been found.  But, what do you expect for a generator costing 10-40x less than quality ones  .

[dadler]

If the DDS is implemented in the textbook fashion, off of a common clock, I can't think a good reason for the two channels to not be phase locked if one frequency is an integer multiple of the other.

https://en.m.wikipedia.org/wiki/Numerically_controlled_oscillator

[gby]

You would that textbook math for DDS when both frequencies are integer divisors of the main clock and are in an integer ratio to each other would yield the expected ideal result.  However, many (most?) actual generators use a fixed binary length to the phase word to cover one complete/360 Deg waveform cycle.  This means that if you are not careful there will be non-integer values required to get the specific frequency you ask for.  In addition, the FW that sets the digital logic accumulation rate is finite precision arithmetic and rounding/truncation is involved.  So, you might end up off by a small amount.

Consider a made up example similar to the FY3200.  I say made up because we don't know the exact FY3200 inside details.

   clock = 250 MHz = accumulation clock
   minimum frequency = (250 MHz)/(2^36) = 0.003638....  Hz

     >>> 36 bit accumulator clocked at 250 MHz with the top bit representing 180 electrical degrees.

To DDS 1.000 kHz the input word to the accumulator would need to be:

    (2^36)*(1 KHz/250 MHz) = 274877.907 which rounded to an integer = 274878

To DDS 2.000 kHz the input word to the accumulator would need to be twice that or 549755.814 rounded to 549756

The ratio of 549756/274878 is exactly 2 and should work as desired.  But, if the FW did truncation of the result instead of rounding you get 549755/274877 = 2.000003638 which will yield a lissajou that will precess and take about 5 minutes to make a complete cycle.

Note that this particular example does have an exact 2:1 solution.  But, there are numbers where the proper rounded value for the higher frequency would be an odd integer and when you cut it in half and round you will not get an exact integer ratio.  Basically, I am saying that even in a well implemented textbook DDS system it depends on the detailed number on whether two frequencies that you want to be 2:1 or 3:1 or whatever ratio can actually be exactly that ratio.  Basically I am not convinced that this test proves there is any design flaw in the FY3200.

[dadler]

Yeah, I worded that wrong--I can't see a good reason for there to be a non-cyclical phase difference between the two waveforms.

However, it does depend on the implementation, and how they are handling truncation of the phase accumulator.

I would expect the waveforms to sync back up with a predictable period.

However, as you say, there should be frequencies where this is not the case and they remain phase locked, if done properly.

[gby]

Dadler, that is exactly what the FY3200 generator does.  When you select 1.00 kHz and 2.00 kHz the Lissajous pattern very slowly precesses and eventually repeats in a cyclic fashion.  As JohnPen reported 10.0 khz and 20 kHz does not precess and stays phase locked.  Locked or precessing just depends on the details of the frequencies selected.

So, I conclude that there is no bug in the FY3200 for setting frequencies and holding them.  The issue is inherent in all DDS generators because of the finite precision frequency input to the phase accumulator.  As I mentioned before, if you really need 2:1 or 3:1, etc exact frequency ratios that are reliably phase and freqeuncy locked put a 2 or 3 or whatever cycle waveform in the FY3200 4096 point Arb wave shape.  Then no matter what frequency you select they stay frequency and phase locked.

[jcrubin]

Maybe Im missing something..... Does anyone know where or how to set the trigger output on the sweep function to go to the trigger on the scope?

[JohnPen]

I haven't found a way to do it as there is no trigger output. I tried to persuade CH2 to do something and it will trigger on say 10 sec intervals etc but the output unfortunately doesn't relate the timing of the sweep output.  Presumably one could reverse engineer to add a trigger output but currently I am leaving well alone as I have only recently purchased my unit.
John

[quantalume]

• Set the frequency to 1999.29994 KHz
• Press the blue knob in (changes display from KHz to MHz)
• Power down the unit and wait 15 seconds
• Profit! (Sine wave capability extends to 24 MHz)

[dadler]

• Set the frequency to 1999.29994 KHz
• Press the blue knob in (changes display from KHz to MHz)
• Power down the unit and wait 15 seconds
• Profit! (Sine wave capability extends to 24 MHz)

How did you ever figure this out?

[SaabFAN]

• Set the frequency to 1999.29994 KHz
• Press the blue knob in (changes display from KHz to MHz)
• Power down the unit and wait 15 seconds
• Profit! (Sine wave capability extends to 24 MHz)

Do you know of a similar way to "hack" the MHS-5200A?

[quantalume]

How did you ever figure this out?

I wish I could take credit for it, but actually I heard about it from a friend of a friend. 

Do you know of a similar way to "hack" the MHS-5200A?

Nope, sorry mate, but I'll ask around. 

[quantalume]

I figured out the riddle regarding amplitude modulation first discussed in reply #40. There is no way to amplitude modulate one signal with another on this generator. All of the presets, PRE1-PRE10, are simply stored waveforms. PRE9 happens to be a waveform where the modulating frequency is 1/16 the carrier frequency. The only thing you can control is how often this waveform is "played out."  So, setting the frequency to 10 KHz gives you a 160 KHz carrier modulated with a 10 KHz signal.

In the good news department, the TTL outputs on the back are driven by a 74LVC14 Schmitt trigger operating on 3.3 V power. This means you have a signal with sub-nanosecond rise times that can drive 3.3 or 5 V logic. Furthermore, if you set the waveform type to square, the duty cycle on the TTL outputs will follow what you have set on the front panel. So, it's a lot more useful than just a sync output.

[jcrubin]

What I really needed was a trigger signal for the sweep generator........

[gby]

After applying the sequence provided in quantalume's post yesterday to my FY3206S 6 MHz generator I see that ALL the wave shapes can now be turned up to 24 MHz.  The official manual/specs list 6 MHz max for square, triangle, arbitrary no matter which version.  So, this hack has unlocked the frequency setting for everything up to 24 MHz 

Of course, the generator does not have enough bandwidth to do a good job for square or triangle at 24 MHz.  However it is definitely usable beyond 6 MHz.  I tried a square at 6.6 Vp-p (with 50 Ohm termination 3.3V p-p for a logic signal) and found usable results beyond 6 MHz.

Square:
       Set 6.6 Vp-p and scope has 50 Ohm termination
       0.1 MHz    3.26 Vpp  (Reference)  great square shape
       1.0 MHz    3.24 Vpp  (- 0.6%)      good square shape 26 nSec rise time
       6.0 MHz    3.22 Vpp  (-1.2%)       OK square shape 24 nSec rise time flat top for 90 degrees (ideal would be 180)
      10.0 MHz   3.14 Vpp  (-3.7%)       Usable square shape 24 nSec rise time barely any flat top
      15.0 MHz   2.68 Vpp  (-18%)        Barely usable, basically a sine wave shape
      24.0 MHz   1.68 Vpp  (-49%)        Unusable, basically a sine wave

See below oscilloscope shot of 10 MHz square.  This scope shot shows the trigger wave and CH1 out.  The phase lag from trigger to CH1 out is 14 nSec or 50 Deg and probably all of this is from the output low pass filter.  As quantalume commented the trigger signal is quite fast with about 2 nSec rise time (at least on my 200 MHz analog BW scope).

I also tried 10 MHz square with duty at 33% and the amplitude was 2.91 Vpp which is still quite usable.

I quickly checked the sine shape at 24 MHz and found the amplitude to be down by 1.58/3.24 = -6.2 dB and it lags the trigger by 25.6 nSec/220 Degrees.  This data matches the results when putting 4 sine cycles in an Arb waveform and running at 6 Mhz rep rate for 24 MHz out that I reported earlier.  This confirms to me that this FY3206S generator output filter is causing the limit.

[jcrubin]

I did a video on this to test my new scope.....   while i did not read the maximum frequency limit of 6mhz for non-sine waves, I still thought the output was crappy even a 6mhz

[jcrubin]

I would like to isolate the circuit that sends the signal for the sweep to reset and make a usable pulse from it for a trigger for my scope.

[usagi]

Do you know of a similar way to "hack" the MHS-5200A?

Nope, sorry mate, but I'll ask around. 

an mhs-5200a hack would be ace 

[jcrubin]

Looks like the sweep resets in the processor..........
I searched for any artifacts on the board that I could use as a trigger, but no dice...


anyway, with the 5072P at least theres an FFT.

[gby]

jcrubin,

Your 6 MHz square wave signal looks significantly worse than on my FY3200.  The picture you show has a fair bit of ringing and overshoot while on mine the signal has essentially none.  See attached scope photo from a 200 MHz Tektronix connected via BNC cable with scope set for proper 50 Ohm termination at 6 MHz at FY3200 setting of 4 Vpp.  Essentially no overshoot, no ringing, and nice smooth rise and fall times within the 21 nSec rise time limit.  I also attached a 10 Vpp 6 MHz result which is almost as good but has a little more tilt on the flat bottom.

The scope photo you show measures 10 Vp-p ignoring the overshoot/ring.  If you have the scope set for 50 Ohm termination then the FY3200 is set for 20 Vpp and the output buffer is current limiting and going non-linear which can explain the poor wave shape.  See my earlier test results post on this generator to see the output waveform current limit issue in more detail.

If you don't have the scope input at 50 Ohm termination, then you will get bad ringing over the cable which could cause overshoot and distortion.

I suggest trying your test again with the FY3200 set to 10 Vpp, BNC cable from generator to scope, scope has 50 Ohm termination right at it.  If you don't have 50 Ohm termination available with a good BNC cable, then try the oscilloscope probe right at the FY3200 BNC output with no cable.  I think you will find the waveform quality much better than you show and like the ones I attached.

On your quest to find a trigger signal for the sweep function I think you are going to be out of luck.  The sweep function is done entirely in the software, FPGA in these generators and thus unless they specifically design it to have a sweep trigger signal then it will only exist inside the chips and not on a pin.

[jcrubin]

A substantial amount of testing was done yesterday, and the signal did improve with the correct 50 termination using bnc cable

ive posted my results in the 5072P thread below


https://www.eevblog.com/forum/testgear/hantek-dso5072p-thread-not-the-bandwidth-hack-thread/new/?topicseen#new

[pdereus]

• Set the frequency to 1999.29994 KHz
• Press the blue knob in (changes display from KHz to MHz)
• Power down the unit and wait 15 seconds
• Profit! (Sine wave capability extends to 24 MHz)

...and of course you can downgrade it again (for whatever reason):
Set the frequency to 1999.09996 kHz (or 1999.19992 kHz)
and follow the rest of quantalume's steps and you have a different model

[jcrubin]

Some FFT testing of the waveforms

[Pete69Walton]

I have recently bought an FY3224S frequency generator and it arrived without any documentation or software. Although I was able to download the manual from the web I have been unable to find any software. Does anyone have any such software or know where I can get hold of it?
Many thanks!