General > General Technical Chat

Review: Hantek DDS 3X25. Anyone own one?

<< < (93/108) > >>

torch:
I got it working -- sort of. Not really a sine wave, and a bit jittery, but I think it will be accepted by an old tube radio:



The superimposed audio frequency sine wave has similarly jittery amplitude, but should work nonetheless:



So it will work at 17KHz (for an RF of 17MHz)

However, I still have the original problem at 15.26KHz, which I need for aligning the 15,260KC point.

zibadun:
nice work. for some reason I could not get Arb Express to generate anything for me.  I select Basic Waveform/Sine, 40MHz and all I get  is a flat line. :(

Mechatrommer:
ok i saw your csv file that you emailed me. the problem...

"instead of sine waves, it appears to be an underlying modulated wave"
due to aliasing resulted from downsampling made by goltek from unsupported 10,000 pts to 2048 pts which is the managable value by 3x25

"They are groups of square waves that gradually average out into the overall shape"
due to you are asking 3x25 to run at very slow sampling rate (40KSps). coupled with the kind of downsampled points will generate that very bad (worst) stairsteps, its the nature of DAC inside the 3x25.

"I suspect those random drop-out glitches are something to do with the Rigol and timing"
did you activate peak detect mode? if not then this is common problem with discrete sampling DSO... signal aliasing, is when the dso is sampling at the rate less than 2 times of the signal frequency. thats why Mr Shannon worked hard on this and produced his theory long time ago.

hint 1...
refer to gui.png, for optimum result, make csv that has the same number of points as shown in the gui, other than that goltek will interpolate (upsampling) or select few points from the whole (downsampling). the number of points is not something that we can choose as we like, it is specified by the device/api, the value is returned from function call (the calculation is based on marmad's KSps formulation) and that is used by goltek to make the interpolation. so if you expect goltek to do the interpolation, probably something bad can happen (disclaimer: i will not held liable for any property or life damage ;)) such as highly distorted or aliased signal.


hint 2...
for those of you who dont know... square and sawtooth buttons in goltek use different interpolation method, ie nearest neighboor interpolation, specifically made for sudden cliff like signal (you can load arb data to those buttons as well, but be carefull to save or keep your original square and sawtooth file in case you want to reload the original) ie when the data got 2 consecutive points inside it that go from lowest to highest and you want to preserve it in interpolated data. not like other button which use linear calculation to create NEW points. for eq if we have 2 points valued 0 and 1, we want to upsample to 4 points. square or sawtooth button will interpolate it to 0011, but other signal buttons will interpolate it to 0,0.333,0.667,1 not a good idea for square wave like or high pulse signal.

hint3...
to investigate what kind of data that is produced (interpolated or downsampled) and sent to 3x25 by goltek, you can right click on the graph picture -> dump -> hantek volt data. goltek will produces a file inside its directory "volt.dump.csv" (make sure you dont lock this file if already exists while dumping, nasty unspecified thing can happen) i've attached the sample below (volt.dump.zip) which is downsampled from torch's original data.

using hint1, i tried to produce 2048 data based on torch's formula in excel (attached modulation2048.xls zipped) but guess what? this intermodulation is also there! so whether goltek interpolation or using theoritical formula, its still the same, pay attention to the data shown in gui.png (Ksps and number of points) and make your own calculation and senses, beyond that i cant help much, we are live in discrete sampling world now.

in compare.png, i plotted first 30 points of what is produced and sent by goltek interpolation (left) and compared to the same data count/set using math (right), the result are about the same, those sharp peaks (pointing up and inverted) represent each point, top and down consequtively. when sent to 3x25 at finite sampling rate you'll get the malevolent stairsteps. let me re-view it again here.





FWIW.
Regards,
Shafri.

torch:
Well that certainly explains a few things. I didn't realize there was a 2048 point limit. I thought as long as it was within the sample rate at a given frequency, I would be OK.

Getting a usable waveform at 17KHz was basically dumb luck and the jitter was an artifact of the interpolation process. But that does suggest it is possible to get something usable at 15.26KHz, given the right combination. I quickly tried 2048 points, but as one might expect, the "audio" component was over 75KHz -- well beyond my range of hearing. If I can figure out a way to get ArbExpress to use trianglular instead of sine waves I could get away with 2 points per wave, dropping the audio to 15KHz or so. Even I should be able to hear that.  Otherwise I guess I'll just have to fiddle around and try to find the right combination.

I realize there is no warranty, express or implied, on Goltek or it's fitness for a given purpose. ;-) But I appreciate your time and help and thank you for it.

Mechatrommer:

--- Quote from: torch on January 13, 2013, 09:09:11 pm ---I didn't realize there was a 2048 point limit.
--- End quote ---
marmad has made extensive study on this. the max points 3x25 can accept is 4096, but its by luck at certain freq and/or KSps config to get that, mostly 3x25 will use 2048 points. me myself have not figure out the combination. i'll need to dig this up if i need to, its somewhere here or in my HDD. please check your PM will ya? ;)

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod