General > General Technical Chat
Review: Hantek DDS 3X25. Anyone own one?
<< < (5/108) > >>
saturation:
As always, alm, a nice information packed post.  I'm still digesting all the articles you linked, and it does clear the air between what is a true AWG, which typically do not use the DDS approach, and what has best been called function generators with arbitrary capability, most which use DDS.

However,  the Hantek design is curious.  The 3x25 suggests it uses a variable frequency clock.  Since the output filter is vital to supression of additional frequencies generated with the DAC process, that are tied to the DAC clock frequency, how they manage to suppress it with a fixed frequency filter will be interesting to see.  Without it, I'd expect to see harmonics and images of the clock frequency contaminating the 3x25 output.

It thus, seems like a 'hybrid' providing some of true AWG capacities using a DDS process; but this is all theory, when I get the unit it will become more real!

3x25 Manual Specs:



In the updated DDS 3005 model, it states the DAC clock is a variable clock and the filters are programmable.





--- Quote from: alm on January 10, 2011, 09:04:34 pm ---
--- Quote from: saturation on January 10, 2011, 06:53:06 pm ---As I understand DDS, the final output stage is a reconstruction low filter.  How its made determines what harmonics and distortion will squeeze through.  As for arbitrary waves with different frequency characteristics, I'll have to play each as it occurs, doing what I draw and what I see comparisons ???.

--- End quote ---
That low-pass filter is for the harmonics from the DAC, which is usually a fixed frequency.

I believe the difference between DDS and 'true' arbitrary waveform generator (as defined by the manufacturers of those) is that a true AWG runs the DAC at a sample rate equal to number of samples * repetition rate. DDS runs the DAC at a fixed frequency, and adjusts the samples to match. It has to interpolate or round the samples if the sample rate doesn't happen to be a multiple of number of samples * repetition rate. The advantage of DDS is that it can easily vary the frequency, for example for sweeping. This makes it also easy to implement a function generator. I'm by no means an expert on DDS or AWG, most of my knowledge comes from reading appnotes and datasheets/manuals. Some links: Tegam AN-401: True Arbitary Waveform Generators Offer More Than Function Generators with Arbitary Capability, Tektronix XYZs of Signal Generators.

Most cheap arbitrary waveform generators seem to use DDS, even though true AWG seems simpler to me. Maybe implementing true AWG is harder than it looks, or it's because they want to bundle a function generator. My guess is that the Hantek unit also uses DDS.

I added two attachments about filters for DDS. The first is from an ELV design for a 30MHz DDS function generator (no arbitrary waveform capability, but DDS is DDS). The DAC is in IC5 (AD9834), the signal is outputted on IOUT (Sig Bit Out is used for high frequency square waves, just ignore that part). ADG736 is just an analog switch. Depending on the type of signal, the Chebyshev filter (knee frequency = 34MHz) is used for sines (up to 30MHz), but bypassed for square waves (rise time is <5nS IIRC). The sample rate is 67.109MHz, well above 34MHz. The filter doesn't change with frequency.

The second is part of the HP 33120A service manual, a function generator/arbitrary waveform generator (DDS, 15MHz, 40MS/s). Depeding on the type of signal, it either uses a very sharp filter (ideal would be a brick wall response slightly above 15MHz) for sines, or a shallow filter for all other types of signals. Both filters have a fixed frequency response.

--- End quote ---
saturation:
I received the unit yesterday and it does the job well, moreso its excellent for its price performance.  Just went through testing it at its extreme settings and it came through very well.  I think its well suited for providing function generator capacity with arbitrary waveform capability at very low cost.

I confirm dajones posts, there is a persistent jitter of 5-8ns most obvious in the square wave output.  I haven't tried mitigating it, or trying a whole suit of tests, and will post a more detailed review later as I find time.

Prelim testing was done with an unmodified Rigol 50 MHz 1052E and its FFT.  The limits below are written because they are  at the 1052E's bandwidth limits.  I did review the outputs at it maximum setting, 75 MHz sine, and 25 MHz square, and the additional instabilities can't be ascertained to be solely from the 3x25.


Pros

Low harmonic distortion stable sine wave in excess of 50 MHz, here's a sample screen shot at 1 MHz
Fast rise time and symmetric square, triangle, and sawtooths mostly clean [ as expected by a Fourier harmonic distribution] >= 5 MHz  
Instantaneous response from selection to output
Bug free software usage and installation; easy to use, manual not necessary, small size, 11 MB
Works in Windows OS: Vista 32, Win7 32, and XP 32 bit
Small hardware footprint, powered by USB, no external power needed
Interface USB and BNC cables included


Cons

Jitter noticeable in all but the sine waves, worst in square waves, but not exceeding 8ns
Self installing driver did not install in Win7; I manually installed it from the CD
Software could have used more ergonomics
No external power adapter included


EricF:
Thanks for the initial impressions saturation! Looking forward to hearing more.

EricF
DaJones:


--- Quote ---I confirm dajones posts, there is a persistent jitter of 5-8nS most obvious in the square wave output.  I haven't tried mitigating it, or trying a whole suit of tests, and will post a more detailed review later as I find time.
--- End quote ---

: /

I was HOPING that it was just 'my unit' that was having this problem.

You might want to take a look at the "Sync-Out" problem.

If I set my unit to 2.5MHz, and display the output signal on one trace, and the Sync-Out on another trace ( on the scope ), I see that
they are in step and not much jitter on the sync line. If I bump this up or down in frequency by a small factor ( 2.51MHz or 2.5000000.001 or 2.4999 )
the Sync-Out signal gets so jittery there doesn't seem to be any timing relationship between the two signals at all.
saturation:
Its a big problem with the Rigol's inherent noise and the potential for artifacts caused by the FFT itself, as we've discussed on eevblog in the past.  Its a reason I'll post scope images more than discuss it [ its also easy to do with the 1052E], you can see the signal vs harmonics+ noise for yourselves.  I also analyzed the maximum amplitude on FFT to see whatever harmonics are there and sometimes, not always, pick up a few distinct frequencies spiking over noise. So far the noise floor and/or specific frequencies are <= 1-5 mVrms for a fundamental at 1.4Vrms.  That's ~ <= 0.36%. 

The key issue is while one cannot be so sure what the specific harmonic components are without some sleuthing [they could be real or reflected aliases for example], if the harmonic outputs are near to zero [as expected in a sine response] its meaningful, once you also respect the limits of the 1052E's response and what FFT algorithms do.  Likewise, one can calibrate the FFT response by passing a reference wave to insure its anticipated harmonics, say of a square wave at the fundamental x 3,5,7,9,11,etc., happens as expected and compare it the frequency under test.  For example, if I see other blips at 1 MHz sine on the FFT using the 3x25 I'm not sure off, I examine a better 1 MHz sine from my Instek SFG-1003, which is calibrated and traceable [ so Instek says in the documentation].  Whatever difference shown is part of the distortion from the 3x25.

Right now for sine, the Instek and 3x25 are equal in quality, with the 3x25 better given it can go to 40 MHz +.

On square waves, the Instek is cleaner and jitter free.

Again, I'll post images and less theoretical discussion and would love to hear feedback or more ways to check my findings.

 


--- Quote from: shafri on January 12, 2011, 08:13:13 pm ---
--- Quote from: saturation on January 12, 2011, 07:51:49 pm ---Prelim testing was done with an unmodified Rigol 50 MHz 1052E and its FFT.

--- End quote ---
i'm no expert, but i have a feeling doubt to rigol's fft.


--- Quote from: Time on January 11, 2011, 07:21:40 pm ---
--- Quote from: BoredAtWork on January 11, 2011, 06:02:04 pm ---Kids, mhz means nothing.  mHz would be millihertz (10-3 Hz). But if you mean megahertz, then write MHz (106 Hz, or 1000000000 times larger than mHz). If that is to inconvenient for you, what about another hobby or profession?

--- End quote ---
Even if I have no interest in a thread topic and I see BoredAtWork has posted something on it I open it up anyways.  Just so I can enjoy golden comments like this one.

--- End quote ---
sometime me too lazy press the shift/caps button, i just rely on my extra terestial intelligent to read people's post, but a good reminder... with the butt kicking word "kids".


--- End quote ---
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod