Review: Hantek DDS 3X25. Anyone own one?

[alm]

One advantage of USB scopes is that the raw data can be imported into software for presentation in any format, such as Bode plots.

Just to be clear: this is true for any DSO with any kind of PC connectivity, which includes all of the popular Rigol/Instek/Hantek/Tekway scopes. Even RS-232 or GPIB would work. When I think of USB scopes, I think of a small box with BNC connectors and a USB port (like the Hantek AFG), without any display or knobs. You definitely don't need one of those, you can do the same with a standard DSO with USB port and decent software, and they're usually inferior to stand-alone units unless you pay big bucks to someone like NI or Agilent.

You can actually do the same with a non-storage scope, but you need to manually measure the amplitude at all frequencies, so it's a lot of work. With a computer controlled signal source and oscilloscope, you can automate it.

[saturation]

Hantek's web site has been down for a while, but its up and all its operational software are online to be downloaded for trial as you please.

http://www.hantek.com.cn/english/produce.asp?classid=30

However, there is something wrong with the 3x25 rar file, its 22MB long but when unrared only drivers are viewable.  I've sampled a few of the other products and the packages are OK.  

Alm is right of course, what I meant was USB interfaced scopes.  Software to make plots of multiple complex data sets is unlikely to be included as part of a Rigol type DSO package; but student oriented knob&button free USB scopes like syscomp or Velleman will come with a bode plotter and some other tools.

Software that can make bode plots and more, using any scope data so long as drivers exists, are Labview or Mathlab; 2 industry standards.  However, you may need to program it whereas on Velleman and syscomp its a built in routine tied to their hardware.  However, if you are going to be a professional EE, chances are Labview or Mathlab is in you career future, so its good to get familar with them.

For students, Labview can be had for $22, its almost the same packages as Labview base which is priced for professionals at over $1500.

If you need to start with plotting now, you can download a trial package until you get your license.

http://www.ni.com/labviewse/labviewse_faq.htm

One advantage of USB scopes is that the raw data can be imported into software for presentation in any format, such as Bode plots.

Just to be clear: this is true for any DSO with any kind of PC connectivity, which includes all of the popular Rigol/Instek/Hantek/Tekway scopes. Even RS-232 or GPIB would work. When I think of USB scopes, I think of a small box with BNC connectors and a USB port (like the Hantek AFG), without any display or knobs. You definitely don't need one of those, you can do the same with a standard DSO with USB port and decent software, and they're usually inferior to stand-alone units unless you pay big bucks to someone like NI or Agilent.

You can actually do the same with a non-storage scope, but you need to manually measure the amplitude at all frequencies, so it's a lot of work. With a computer controlled signal source and oscilloscope, you can automate it.

[alm]

Yeah, you might have to do some basic programming in something like a free MATLAB clone, probably not a bad skill for an EE. There's also software like SignalExpress, which is a nicer shell around Labview (and less powerful, obviously). Should be fine for things like bode plots and FFT. But since it's professional software from NI, it's likely expensive.

[Mechatrommer]

LARGE Hi res PCB photos are here, I did not take them, just found it by accident:
http://biot.com/p/

nice simplistic porn photos. i'm looking forward to take apart my unit. i just wonder why there is so much pcb area (bottom part) compared to components population (about a quarter only), why dont/cant they make the unit/pcb smaller?

I also bought a Hantek DSO5202B 200MHz scope, which is not bad at all for the price.  Although this is off-topic, I had a wild dream about using the generator and scope together as a Network analyzer (would give me a nice graph of amplitude and phase response as a function of freq for my device-under-test) - anyone have any ideas how to do this?   I also have this scope-and-generator-in-one:
http://syscompdesign.com/CGR101.html
which produces beautiful bode plots - but only up to 2MHz - I need something up to 10MHz.  Any ideas?

Good you brought these up:
You can use a signal source and a scope in the X-Y mode of the scope, and evaluate Lissajou patterns.  It will provide  phase, amplitude and frequency relationship between 2 channels.
One advantage of USB scopes is that the raw data can be imported into software for presentation in any format, such as Bode plots. But if you were trying to test your hardware for its frequency and phase response for practical reasons, you now have basic tools to make that assessment.  For example, the syscomp image would appear as a very squashed line at 12kHz in phase, then a turn to equal proportions at X=Y amplitude, and begin to move from a circle to an ellipse as the phase shifted.
http://en.wikipedia.org/wiki/Lissajou#Practical_application

frogblender's point interests me about the network analyzer, and the possiblity to bode/lissajou plot etc in pc software. but i'm concern about the delay that will be introduced by pc->generator->output and input->oscilloscope->pc with the DUT in between. will it not give some effect to the graph plot or analysis? i need to study more on this.

[Bored@Work]

i just wonder why there is so much pcb area (bottom part) compared to components population (about a quarter only), why dont/cant they make the unit/pcb smaller?

Either they had an abundance of the enclosures (they use it for other products, too), or they just want to impress.

[Mechatrommer]

...or they just want to impress.

you confirmed what i had in mind
but since this unit is already a small form factor compared to bench generator, make it smaller will become more advantagous imo, to save more space, rather than merely to impress people. they should know who they dealing with (not the smart combed hair and nice tie guy).

[saturation]

I wonder if the basic design simply lacks the added tracings used in the other model DDS based generators?  Only buyers of the 3005 can tell if they took photos so we can compare PCBs.  It would be a design efficiency to sell 2 tiers of products, they simply don't need to redesign the PCB wholly, just append the missing traces from their routing software that currently is just a huge ground plane.

nice simplistic porn photos. i'm looking forward to take apart my unit. i just wonder why there is so much pcb area (bottom part)

but i'm concern about the delay that will be introduced by pc->generator->output and input->oscilloscope->pc with the DUT in between. will it not give some effect to the graph plot or analysis? i need to study more on this.

One can expect delay or shifts by adding cables, adapters, etc., dependent on frequency, and the Lissajou can show you the shift.  The question is is it significant for the DUT?  Only the DUTs performance can say, it depends on its engineered tolerance to variation.  Regardless, the 3x25 comes with a phase shift adjustment so you can try to compensate for a shift, to get the signals in sync between 2 devices. 

[bekir]

Hi,

I have been using DDS-3X25 and DSO-2250 USB for about two years. I must say that they are very cost effective but you have to be careful. Signal generator broken once and oscilloscope broke twice. I do not know the actual reason (no high voltage input/short circuit at probes), believe me I was very careful. Anyway just a reminder do not leave them powered up (connected to pc) while not using.
* Op-amp at the exit stage of DDS broke had to replace it (bought online directly from manufacturer as sample).
* Same op-amp that is used in the input stage of oscilloscope broke (thank gos i bought one as replacement ).
* A diode in the input stage of oscilloscope broke (salvaged a replacement smd diode from an old lcd projector).

Besides all these I am happy with them. Now to the main topic.

I am using these to as VNA. I am not EE, so I may be wrong on selecting words. Let me explain my situation. I generate sine wave from DDS. Connect it to my network and OSC1. Connect the output of the network to OSC2. Then get the magnitudes of input (not necessary in fact) and output. Then I find phase difference between OSC1 and OSC2 and draw bode plots.

Another method i tried to superimpose sine waves that are multiple in prime numbers (i.e. if my base freq is 1hz then I impose 2,3,5,7,11,13... hz frequency sine waves (total of 18 freqs)). Run it through the network then run fft on it. To have better accuracy, I curve fit to the signal in time domain with the magnitude and phase information from fft used as initial guess value.

First one seems to work better than second one when i look at the response graphs of RC networks. In fft there is a slight difference that I don't know why it happens. I can give some example graphs if someone interested in help.

Anyway, I did all this stuff in MatLAB 32-bit (I had(!) access in our campus), I believe it can be also done in octave as long as it can call dlls. The problem rised when our university switched to 64-bit MatLAB. My m-files and functions no longer work giving dll load error. The bad part is i didn't keep the original dlls and header files that I've used in 32-bit. So, I am asking help from someone have experience with 64bit matlab dll calls. At the end I can share my functions for signal generation and oscilloscope in a toolkit or as a sample gui. Besides that if you want any graph or data to see the performance of osc or dds just let me know. But please remember I am ME so briefly explain how should i do it.

Regards,

Bekir

Ps. I am working on Win7 x64

[tinhead]

Hantek's web site has been down for a while, but its up and all its operational software are online to be downloaded for trial as you please.

http://www.hantek.com.cn/english/produce.asp?classid=30

However, there is something wrong with the 3x25 rar file, its 22MB long but when unrared only drivers are viewable.  

just rename it to iso and burn or mount.

[Mechatrommer]

thanx bekir for your info (esp on opamp failure warning). i'm looking forward to to see how the 3x25 works as vna. my 3x25 arrived today, made a quick run and take apart for the "first" evening/night test. i'm not intending to do full performance test as saturation did, just a rough one. here are some photos.

a) my unit labeling differ from the pic in documentation (no 50ohm +-3.5V max label, 350mA max etc), so not sure why. construction material not so impressive, ie uneven plastic casing color, its like its been ages. and it dont come with cd and manual, as advertised by the seller. have to download manually the software from the net, but i dont mind, as seller included free gift with it (happy customer here)
b) how to take apart.
c-h) the porn.
i) this is how i read the signal, pls correct if i'm wrong. i just directly connect the bnc cable generator to rigol, no 50ohm termination or fancy stuff. i'm not really good at it.
j) while analyzing the vb code sample, this unit just generate sin wave using precalculated 2083 points sample using sine formula.

Code: [Select]


Function myCreateBuf()
    Dim i As Integer
    Dim tmp As Double
    Dim arBuffer(4095) As Integer     'waveform buffer

    For i = 0 To g_nWavePointNum

        'sample value
        tmp = g_offset / MAX_VOLT + g_amplitude / MAX_VOLT * Sin((i / g_nWavePointNum) * 2 * PI * g_nWavePeriodNum)
        
        tmp = 2048 - 2047 * tmp
        
        If tmp > 4095 Then
            tmp = 4095
        ElseIf tmp < 0 Then
            tmp = 0
        End If

        arBuffer(i) = tmp
    Next

    result = DDSDownload(g_iDevice, arBuffer(0), g_nWavePointNum)
End Function


so i just changed the sin to tan...

Code: [Select]

tmp = g_offset / MAX_VOLT + g_amplitude / MAX_VOLT * Tan((i / g_nWavePointNum) * 2 * PI * g_nWavePeriodNum) 'SOA MODIFIED

then i got a cropped tan signal... just for fun. and... i've broke the warranty seal on the side of the unit

[saturation]

great work!  mechatrommer, I was looking at your photos and now they are gone!  but no worries; look forward to what you have to post.  Much to study, will analyze and report back.

Great work too bakir, I have some questions for how the Hantek failed, will analyze your post shortly.

Thanks for tip, tinhead!

[Mechatrommer]

i keep getting broken upload (disconnected) just now... still attempting... continue here, sized down from fullsize pictures (c-h)

[Mechatrommer]

i noticed some last minute hand soldering on the pcb. my pcb ver is 1.4, so go figure. cosmetic-wise, its not so impressive, but from rough functional test, i think its good for money and can do me some usefull work. the usb chip, main mcu and another chip after it have been erased.

[saturation]

Thanks bekir, when you replaced the signal generator that broke, did you repair it yourself, or bought a new 3x25?

I see from mechatrommer's photos the op amp is a LMH6702 and the comparator is TI TL3016; that's a pretty good op amp by the way.

To prevent further damage in high output tests you can put another op amp infront of this op amp and isolate it, just configure it as a voltage follower.

From your tests, did you check the quality of the sine wave output on any higher end equipment you have in your lab?  Particularly, what is the harmonic output of the sine waves; I'm just curious.  I only have a Rigol 1052e but I get nothing at its noise floor which is about -50 dBc.  A more sensitive instrument might show that the sine wave output is hifi audio quality, i.e., -90dBc or more.

Yep, we discussed using the Hantek as a source for making VNA calculations, I use Lissajou figures to make qualitative measurements for phase delay,  we just discussed this, find our threads on measuring phase shifts and the Hantek.

Your prime number test is interesting, to superimpose them, do you have multiple sine wave sources and measure simultaneously or did you send each prime individually then merge the waveform data?

Sorry, I have no Mathlab, so can't assist in your need there.

Hi,

I have been using DDS-3X25 and DSO-2250 USB for about two years. I must say that they are very cost effective but you have to be careful. Signal generator broken once and oscilloscope broke twice. I do not know the actual reason (no high voltage input/short circuit at probes), believe me I was very careful. Anyway just a reminder do not leave them powered up (connected to pc) while not using.
* Op-amp at the exit stage of DDS broke had to replace it (bought online directly from manufacturer as sample).
* Same op-amp that is used in the input stage of oscilloscope broke (thank gos i bought one as replacement ).
* A diode in the input stage of oscilloscope broke (salvaged a replacement smd diode from an old lcd projector).

Besides all these I am happy with them. Now to the main topic.

I am using these to as VNA. I am not EE, so I may be wrong on selecting words. Let me explain my situation. I generate sine wave from DDS. Connect it to my network and OSC1. Connect the output of the network to OSC2. Then get the magnitudes of input (not necessary in fact) and output. Then I find phase difference between OSC1 and OSC2 and draw bode plots.

Another method i tried to superimpose sine waves that are multiple in prime numbers (i.e. if my base freq is 1hz then I impose 2,3,5,7,11,13... hz frequency sine waves (total of 18 freqs)). Run it through the network then run fft on it. To have better accuracy, I curve fit to the signal in time domain with the magnitude and phase information from fft used as initial guess value.

First one seems to work better than second one when i look at the response graphs of RC networks. In fft there is a slight difference that I don't know why it happens. I can give some example graphs if someone interested in help.

Anyway, I did all this stuff in MatLAB 32-bit (I had(!) access in our campus), I believe it can be also done in octave as long as it can call dlls. The problem rised when our university switched to 64-bit MatLAB. My m-files and functions no longer work giving dll load error. The bad part is i didn't keep the original dlls and header files that I've used in 32-bit. So, I am asking help from someone have experience with 64bit matlab dll calls. At the end I can share my functions for signal generation and oscilloscope in a toolkit or as a sample gui. Besides that if you want any graph or data to see the performance of osc or dds just let me know. But please remember I am ME so briefly explain how should i do it.

Regards,

Bekir

Ps. I am working on Win7 x64

[saturation]

Thanks so much, mecha, those photos will save me the trouble of looking at mine with virgin eyes.  I think knowing the op amps may fail is a good tip, and knowing what the part is ahead of time is a good info.  The op amp used is excellent, BTW, and pricey, about $5 each.  When I do my next digikey order I'll grab a few because it has superb specs to play with.

I notice what looks like hand soldering in your photos. compared to the centerfold photos I linked earlier that looked all machine soldered:

https://www.eevblog.com/forum/index.php?topic=2170.msg33908#msg33908

The good news, is the some chips that have their parts numbers erased, are now identified!


I think the Hantek is holding its own as a poor-man's arb and 25MHz [ at least] signal source.  For $150 its still sounding like a bang for buck.  Also, when you think that Agilent wants $400 for non-educational purchase of its FG module for its new Infiniivision scopes, you know its not too bad [ the Agilent specs that Dave showed aren't that much better than this Hantek, but probably doesn't have its quirks and faults too, but nothing an electronics person can't fix.]

i noticed some last minute hand soldering on the pcb. my pcb ver is 1.4, so go figure. cosmetic-wise, its not so impressive, but from rough functional test, i think its good for money and can do me some usefull work. the usb chip, main mcu and another chip after it have been erased.

[tinhead]

the usb chip, main mcu and another chip after it have been erased.

well, you don't have care to much about ... the usb chip is STM32F102Cx or STM32F103Cx
acting as USB->i/o controller (you can see in driver description and a clear ARM JTAG interface),
the one tsop chip no idea for now, probably just a buffer/voltage converter
and the big chip is Spartan 3AN. Originally it was designed to use 3A , there is still empty
place near J2 (FPGA JTAG) for config ROM.

Probably Hantek decided to use Spartan3 AN because of the integrated flash to prevent
clone-makers from cheap clones (the hardware costs only few bucks),
maybe they did it after someone already cloned (and then removed of course external config ROM) - if so
it would be cool to get older revision with config rom soldered (U1).

No matter now, just be carefull, if you shot this chip you can buy a new DDS.

EDIT:
@saturation
you don't have to ping the JTAG anylonger, we know now what soldered (well actually you can take a look inside, maybe you have old revision ...)

[Mechatrommer]

you are welcome saturation. anytime if you want the fullscale image, just PM me, or if you want specific part of interest to be photographed, just ask, i'll be glad to take it for you to the best of my lens and skill . in the mean time, i just put it back in its clothing. @tinhead. i'm not capable to tinker with jtag or spartan, so i think i just use it as is. thanx for the info though, who knows i might have idea in the future.

any idea whats most probable causing the jitter in square wave at higher freq? i confirmed the +-5ns jitter. my speculation, its in the software, but i dont know. my speculation is based on the fact that this unit is using sample data to generate signal, not true "hardware generator" such as sin generator chip etc, so jitter might come from inconsistent (codelength) sample feed to the buffer/voltage converter etc. but still, i've not look/analyze the circuit detail enough and i'm not master in hardware, i just assuming based on my experience/background.

any good reference on how to convert (or the theory) this to network analyzer? i think i want to do this as an exercise when i got time.

[saturation]

Hantek made this forum available with a new URL, for tech support:

http://www.hantek.org/asken/iasklist.aspx?cid=I00011&state=2


Tinhead@ : thanks for those quick tips, I'll use that to check into the chip eventually.

mechatrommer@: you're welcome, if you find other interesting things with the 3x25, let us all know.  Your idea on the jitter is possible.  You can import the output waveform into the arb section by first selecting a fixed waveform, like sine, then hit the arb button, and it will show you all the data points.  So far, I haven't been able to find anything wrong with their sample, but let us know if you can improve on it!

As a 'network analyzer' all you do is check the phase, amplitude versus frequency of a system at higher frequencies.  I described it earlier with the Lissajou figure. 

[bekir]

Thanks bekir, when you replaced the signal generator that broke, did you repair it yourself, or bought a new 3x25?
I see from mechatrommer's photos the op amp is a LMH6702 and the comparator is TI TL3016; that's a pretty good op amp by the way.
To prevent further damage in high output tests you can put another op amp infront of this op amp and isolate it, just configure it as a voltage follower.

I've replaced it (LMH6702) myself, due to our limited budget I had no other option by that time. Input of the op-amp was giving the right waveform but output was shorted to ground. Maybe it is not the best way to not the degrade the performance, but I had only soldering iron at hand (I work in mech.eng. lab so we don't have reflow station or as such), so I hand soldered it. As you mentioned adding buffer will be the best option.

From your tests, did you check the quality of the sine wave output on any higher end equipment you have in your lab?  Particularly, what is the harmonic output of the sine waves; I'm just curious.  I only have a Rigol 1052e but I get nothing at its noise floor which is about -50 dBc.  A more sensitive instrument might show that the sine wave output is hifi audio quality, i.e., -90dBc or more.

I checked the output with an OWON oscilloscope which can not be a real measure i think. Also I have checked with Hantek DSO-2250 (100 mhz bw). Up to now I was using single frequency sine wave. To get the response accurately, I was sampling at least 2 or 3 periods then using linefitting methods on matlab.That also decrease error induced by noise. But as I mentioned ? am not experienced in this stuff so I am open to any suggestion. As for the tests, I may ask for the technicians in EE labs for better equipment is you wish.

Yep, we discussed using the Hantek as a source for making VNA calculations, I use Lissajou figures to make qualitative measurements for phase delay,  we just discussed this, find our threads on measuring phase shifts and the Hantek.
Your prime number test is interesting, to superimpose them, do you have multiple sine wave sources and measure simultaneously or did you send each prime individually then merge the waveform data?

I will check the thread right away. I have measured simultaneously. I have calculated voltage vs time in matlab then transfer it to DDS buffer. Maybe I can better explain by including part of the code:

Code: [Select]

    freqs=primes(64);                              % get prime numbers up to 64, 18 in total
    for j=1:length(freqs);
        yy=yy+sin(freqs(j)*2*pi*i+freqs(j)); % add unit amplitude sine wave at each prime freq multiple
    end
    yy=yy./max(yy);    % normalize to avoid saturation on output

I have added constant phase shifts between signals, this way fft works better. (I fft'ed yy for both 0 phase and this case, amplitude values closer to each other in this case).

[grenert]

EDIT: well, i just quickly realized it. its not a bug, just my stupidity at using the dso. its called aliasing http://en.wikipedia.org/wiki/Aliasing so case closed.

Dave was able to cause aliasing on the Rigol in his review of the new Agilent scope.

[Mechatrommer]

Dave was able to cause aliasing on the Rigol in his review of the new Agilent scope.

the better workaround for rigol is to use peakdetect or long memory resolution.

[saturation]

All DSO have aliasing once your sampling rate is ~< 2x sampling rate used to view the signal.

With the 1052e, and the 3x25, generate a 1 MHz signal, AUTO the scope and look at the 1 MHz waveform.  Hit, MEASURE, select DISPLAY ALL.  You'll see 1 MHz on the frequency counter that is measured on the actual waveform shown on the scope.

Now, reduce the timebase to 20 ms.  Hit HORIZONTAL -MENU to see the sampling rate used by Rigol, 10ks/s, 100x less than the input signal.  You'll see a ~ wavery 15 Hz sine wave. 

EDIT: well, i just quickly realized it. its not a bug, just my stupidity at using the dso. its called aliasing http://en.wikipedia.org/wiki/Aliasing so case closed.

Dave was able to cause aliasing on the Rigol in his review of the new Agilent scope.

[saturation]

Thanks bekir, replacing the op amp and knowing the 3x25 works well thereafter suggests the op amp alone was defective, and no something else causing it to fail.

Sampling the same signal multiple times is a good way to cancel out the noise best, the scope's do the same via averaging.  To measure the quality of the sine wave best requires a spectrum analyzer or a scope with a noise floor below -90dBc, a scope using FFT and an ADC of 16 bits or more.



What are you trying to measure, the systems impedance, phase relationship and amplitude for a linear network?  Its been a while since I did anything with gross calculations with FFT, IIRC primes just allows mathcad to use DFT and make processing easier.

http://iopscience.iop.org/0957-0233/19/10/105102;jsessionid=A1AB9DB658381B0A964FEB3B70D6B429.c3

I've replaced it (LMH6702) myself,

To get the response accurately, I was sampling at least 2 or 3 periods then using linefitting methods on matlab.That also decrease error induced by noise. But as I mentioned ? am not experienced in this stuff so I am open to any suggestion. As for the tests, I may ask for the technicians in EE labs for better equipment is you wish.

Yep, we discussed using the Hantek as a source for making VNA calculations, I use Lissajou figures to make qualitative measurements for phase delay,  we just discussed this, find our threads on measuring phase shifts and the Hantek.
Your prime number test is interesting, to superimpose them, do you have multiple sine wave sources and measure simultaneously or did you send each prime individually then merge the waveform data?

I will check the thread right away. I have measured simultaneously. I have calculated voltage vs time in matlab then transfer it to DDS buffer. Maybe I can better explain by including part of the code:

Code: [Select]

   freqs=primes(64);                              % get prime numbers up to 64, 18 in total
    for j=1:length(freqs);
        yy=yy+sin(freqs(j)*2*pi*i+freqs(j)); % add unit amplitude sine wave at each prime freq multiple
    end
    yy=yy./max(yy);    % normalize to avoid saturation on output

I have added constant phase shifts between signals, this way fft works better. (I fft'ed yy for both 0 phase and this case, amplitude values closer to each other in this case).

[bekir]

Thanks for the info on measurement of quality evaluation. I will ask ee lab. tech. for a better scope to test it.

What are you trying to measure, the systems impedance, phase relationship and amplitude for a linear network?  Its been a while since I did anything with gross calculations with FFT, IIRC primes just allows mathcad to use DFT and make processing easier.

http://iopscience.iop.org/0957-0233/19/10/105102;jsessionid=A1AB9DB658381B0A964FEB3B70D6B429.c3

What I want to measure at the end of the day is the complex capacitance of a sample. The reference is quite what I am trying to do. The difference in my setup is the measurement method. I am currently using "auto-balancing bridge method" as described in 10th reference in the paper (http://cp.literature.agilent.com/litweb/pdf/5950-3000.pdf). From measured voltages (magnitude and phase), impedance is calculated. I want to see if a material can be modeled with existing relaxation models such as Debye.

[Mechatrommer]

am playing with synch out. tried the 6mhz and 7mhz as in picture. it seems like some software-like trick, or leap year style of trick, it will either shorten the valley, or longer the peak at some interval. i'm assuming it tried to reach the set freq as it only doing at some multiple of freq, when you set not in this multiple, then this leap year trick will happened.