Hantek - Tekway - DSO hack - get 200MHz bw for free

[marmad]

today we know Rigol E have 800 wfrm/s (you remember maybe Dave's video, 50000events, after ~60second first event was shown on Rigol).

No, no, no!!  Dave's conclusions were wrong.  I don't think it's likely that the Rigol DS1052E (or E series in general) does 800 wfrms/s - if it did, they would advertise it in their specs because that is a VERY good figure (look at the figures in the Agilent literature for much more expensive scopes).  Please, let's put an end to this unverified information right now and not spread it any further.

When talking about the ability of a scope to capture glitches or randomly ocurring events, we talk about the probability - a 99.9% probability.  Probability means something is [not] LIKELY to happen with a certain regularity - but it certainly CAN [not] - without breaking any rule.  Like rolling 6x the number 2 on a die - not LIKELY, but it can happen. The 99.9% probability that something will happen in 1000x amount of time does not exclude, in any way, the possibility that it occurs in 1x.

You CAN NOT find out how fast a scope is (wfrms/s) by running a glitch test a few times - you can only prove how fast it isn't.  This might sound confusing or the same thing written two ways, but there is an important distinction and worth understanding:
Using the probability formula for a 10/s glitch, with a timebase of 100ns (from the R&S literature):
If my scope captures the glitch within 7 seconds - every now and then, or even a few times in a row - it does NOT prove it is a 1,000,000 wfms/s scope.  But if, one time, it takes my scope over 10 minutes to capture the glitch, it DOES prove (at least beyond a reasonable doubt) that it isn't a 1,000,000 wfrms/s or 100,000 wfrms/s scope, because the probability of a scope that fast taking that long to capture the glitch is EXTREMELY low (99.999999.... etc).

If you really want to try to calculate actual wfrms/s by testing the probability of capturing a random glitch, you have to run MANY tests - not just a few.  It's no coincidence that probability theory has the law of large numbers ("If a fair coin is tossed many times, then roughly half of the time it will turn up heads, and the other half it will turn up tails. Furthermore, the more often the coin is tossed, the more likely it should be that the ratio of the number of heads to the number of tails will approach unity."); in other words, the MORE tests run, the greater the chance of working out the correct probability (or in this case, the actual wfrms/s).

[tinhead]

today we know Rigol E have 800 wfrm/s (you remember maybe Dave's video, 50000events, after ~60second first event was shown on Rigol).

No, no, no!!  Dave's conclusions were wrong.  I don't think it's likely that the Rigol DS1052E (or E series in general) does 800 wfrms/s - if it did, they would advertise it in their specs because that is a VERY good figure (look at the figures in the Agilent literature for much more expensive scopes).  Please, let's put an end to this unverified information right now and not spread it any further.

yes, Dave's statement is wrong. You have actually to run for hours (at the best on multiple devices in multiple
time differences - to not catch the blind time for example or to not get fooled by positive event).
Agilent FG/test signal is 500kHz square signal, with one glitch every 50000 clocks .. so in principle 10 glitches
per second (now please don't start to divide these 800wfrm/s becasue this is not that simple).
This is why you can see on Dave's video multiple glitches catched by DSOX2000 (and not "single one").

Anyway, i build yesterday glitch generator (similar to what Agilent doing) , 500kHz square,
glitch every 500, 1k, 2.5k, 5k, 10k and 50k clocks.  I just don't liked the code which i posted before,
it does the work too but every 1.3 second and one single event ever 65535 - for a scopes with
low wfrm/s the event takes forever and the chance to not catch them i very high.

Based on luck i can start the acq. in right moment, so getting waveform captured with calculated
1000-2000wfrm/s, or i can have no luck and start perfectly in blind time - after 30minutes no
single frame captured. Now this is of course when i set to 20ns/div, so only one rising edge visible.
Of course when i set to 2us/div you can see the gltich randomly on every xxx edge (if you can't imagin
how this looks like check Dave's DSOX2000 review).
Of course the probability to see the glitch increas with 500events per clock and decears with
50k eventsper clock - this is why i build all these versions to show the probability.

The interessting thing on HanTekway is that each xxx (random) glitch even all edges are triggering
- this is when the scope capture thread is synced with the glitch - followed by few ms nothing.

I will see maybe i shot a video of all these glitch snapshots with 500/1k/2.5k/5k and 10k.

Attached the code, works on every Atmega with 8k (for lower rates) to 16k (for higher rates), clocked with 20MHz. 
You might wonder why so many nops, well i don't like the delay function because it si producing additional glitches,
a clock-exact nop is better. Even with nop's to 50k code was producing 50ns additional glitch every
10ms, so i corrected it to match too.

To trigger only on gltich you can setup your DSO to trigger on pos. 200ns pulse.

And yeah, you can of course output the 200ns pulse of different port and add it into the base signal
-  for base signal it makes sense to add series termination to slow down rise time - for glitch leave unchanged
- and you will get very similar signal to what on Agilent (except the glitch width, it is 200ns and Agilent 100ns).

[marmad]

Anyway, i build yesterday glitch generator (similar to what Agilent doing) , 500kHz square,
glitch every 500, 1k, 2.5k, 5k, 10k and 50k clocks.  I just don't liked the code which i posted before,
it does the work too but every 1.3 second and one single event ever 65535 - for a scopes with
low wfrm/s the event takes forever and the chance to not catch them i very high.

Ha, ha... I also wrote a glitch generator yesterday, using the Hantek DDS3X25, for testing the Owon's wfrms/s    After starting to test, I quickly realized I could easily prove how fast the Owon wasn't - but it wasn't so easy to prove EXACTLY how fast it is.  You need A LOT of tests!

[marmad]

....or i can have no luck and start perfectly in blind time - after 30minutes no single frame captured.

Sorry, tinhead [to be again poking at your figures with a stick ] but this doesn't make sense.  The acquire time/blind time has to shift slightly during capturing, even at the same div/sec setting (I can actually see it doing just that on the Owon with simple tests) - it HAS to, otherwise ALL scopes would be completely blind all the time at certain precise frequencies at each sec/div - and that wouldn't be logical because it's so easy to change that with firmware.

Blind time, acquire time, etc, are all calculated into probability ratios.  If you don't capture a single frame in 30 minutes, it means that your scope CAN'T be a xxxxx wfrms/s scope - because probability theory says there is a 99.9999....% chance that it would have captured the frame otherwise.  As I said in my previous post, it's easy to prove how fast your scope ISN'T with glitches - not how fast it IS.

[tinhead]

Sorry, tinhead [to be again poking at your figures with a stick  ]

don't sorry if you poke, i don't mind

If you don't capture a single frame in 30 minutes, it means that your scope CAN'T be a xxxxx wfrms/s scope - because probability theory says there is a 99.9999....% chance that it would have captured the frame otherwise.  As I said in my previous post, it's easy to prove how fast your scope ISN'T with glitches - not how fast it IS.

if a scope capable of capturing whatever value we now take for Hantek was not able to capture in 30 minutes a single glitch
then i don't have any other explanation (or well, maybe just the firmware i'm using freezed whatever functionalty -
however i saw DSO is operational and capturing data - on screen, uart and debug port).
After a next Stop/Run, without any changes to test/setup condition, glitches got captured as expected.
So just don't try to understand this, these are just observation which i made during tests of HanTekway DSO. 
I wasn't able to replicate this, but it didn't means that i or others will not get them too (again).


And as for the capture process shift, of course it is there, you can see it with the code posted
above really good (e.g. in 2us/div where multiple periods are visible).

[marmad]

Anyway, i build yesterday glitch generator (similar to what Agilent doing) , 500kHz square, glitch every 500, 1k, 2.5k, 5k, 10k and 50k clocks....after 30minutes no single frame captured.

Which specific test were you running when you had the 30 minute gap?

Lets take each case one by one - using your glitch rates and sec/div of 20ns (* 20 for the Hantek - but I realize this might be slightly different in your test) - and figure out what the wfrms/s have to be in order to have a 99.9% probability of capturing within 30 minutes:

1000 glitches/s  -  20ns/div *20  -  30 minutes  =   ~10 wfrms/s  -  the statistical formula is:  wfrms/s = ( log ( 1 - ( 99.9 / 100 )) / 1800 {seconds} ) / log ( 1 - ( 1000 {glitches/s} * 0.0000004 {sec/div * div}))
  500 glitches/s  -  20ns/div *20  -  30 minutes  =   ~19 wfrms/s
  250 glitches/s  -  20ns/div *20  -  30 minutes  =   ~38 wfrms/s
  100 glitches/s  -  20ns/div *20  -  30 minutes  =   ~96 wfrms/s
    50 glitches/s  -  20ns/div *20  -  30 minutes  = ~192 wfrms/s
    10 glitches/s  -  20ns/div *20  -  30 minutes  = ~959 wfrms/s

Here they are again, for a 99.9% probability of capturing within 1 minute:

1000 glitches/s  -  20ns/div *20  -  1 minute  =     ~288 wfrms/s  (    ~576 wfrms/s on 10 div screen)
  500 glitches/s  -  20ns/div *20  -  1 minute  =     ~576 wfrms/s  (  ~1151 wfrms/s on 10 div screen)
  250 glitches/s  -  20ns/div *20  -  1 minute  =   ~1151 wfrms/s  (  ~2302 wfrms/s on 10 div screen)
  100 glitches/s  -  20ns/div *20  -  1 minute  =   ~2878 wfrms/s  (  ~5756 wfrms/s on 10 div screen)
    50 glitches/s  -  20ns/div *20  -  1 minute  =   ~5756 wfrms/s  (~11512 wfrms/s on 10 div screen)
    10 glitches/s  -  20ns/div *20  -  1 minute  = ~28782 wfrms/s  (~57564 wfrms/s on 10 div screen)

Again, this does NOT mean that if your scope captures a 10 per second glitch within 1 minute, once in awhile, that it does 28782 wfrms/s - ANY scope can do that.  It means the opposite - if it takes your scope, at least one time, 30 minutes to capture a 10 per second glitch - then it's speed is  <=  959 wfrms/s.  If your scope NEVER takes more than 1 minute maximum to capture a 10 per second glitch (@ 20ns/div - or - 40ns/div on 10 div screens) - congratulations! - you have a scope that is = 28782 wfrms/s [and you're pretty damn lucky  ]

P.S. Both the Hantek and the Owon appear to use 20 divisions for a full-screen capture - which is double the 10 division scopes - so you can basically double their rates in comparison (or halve the competition with 10 division screens).

[marmad]

BTW, I've been using a the glitch generator I wrote at 1 glitch per second - with a sec/div setting of 1ms.  While clocking the Owon over several minutes, the longest time it took it to capture the glitch was ~32 seconds.

So, wfrms/s =  ( log ( 1 - ( 99.9 / 100 )) / 32 ) / log ( 1 - ( 1 * 0.02)) - and that means that the Owon's wfrms/s (at least at that sec/div setting) <= 10.68  (or ~21 compared to a 10 division scope).

I will check a faster timebase now.

[tinhead]

Which specific test were you running when you had the 30 minute gap?

don't remember, i think it was with 2000.

Anyway, recorded two small vids:

2us/div view of glitch once every 1000 clock cycles (so 500 glitches/s), screen refresh is 50,
The video looked good on cam but flash is not showing anymore the real speed (sorry, i'm not video junkie)




and the same zoomed 80ns/div (actually it didn't matter that much when in 20ns/div from speed point of view)
EDIT: this one is with peristency on and auto refresh - note, in persistency mode these scopes are slower.




Now feel free to run calculations of the wfrm/s ^^

P.S. Both the Hantek and the Owon appear to use 20 divisions for a full-screen capture - which is double the 10 division scopes - so you can basically double their rates in comparison (or halve the competition with 10 division screens).

yes, in 19.5 div mode HanTekway is in real using 20divs, but the 16divs mode is not just a cut,
the firmware is really switching over to 16 (with all calculations, you can see it in the asm code which i posted).

[tinhead]

what a surprise, obd2motor.com have already listed the BM and BMV models, the first person
who can provide firmware dump and pictures of PCB will get from me a nice gift!

[pgup62]

Yes, surprising as these DSO are not full ready , Chinese websites used to promote products ahead of time. None on the market before beginning of september and prices should be a glitch higher, what doesn't seem to be the case of this chinese retailer...
BMV series will come first with chinese video help, better is to wait all languages video help. We wonder how will be the spoken translation...
Whatever, Tinhead, what kind of gift you can propose? 

[tinhead]

Yes, surprising as these DSO are not full ready ,

right, Hantek confirmed this once again last night.

BMV series will come first with chinese video help, better is to wait all languages video help.
We wonder how will be the spoken translation...

didn't matter as long it is nice looking young woman with sexy voice

Whatever, Tinhead, what kind of gift you can propose? 

hmm, you a dealer, what can i give a deal (maybe schematics of hw1007?)

For non-dealers i have :
- 50USD (winner to provide paypal acc.)
or
- nice young girl fresh imported from xxx (winner to pay shippment costs)
or
- nice young man fresh imported from yyy (winner to pay shippment costs)

so depends on location "option 2 and 3" might be not a gift anymore.

EDIT: sry, option 2 and 3 sold out.

[A Hellene]

What about all those poor SOBs that take pleasure in abusing, torturing and dismembering under-aged minors? Won't they deserve some kind of reward for providing any dumps and pictures of the brand new models they take apart?

Mind you, ever the pedophiles have rights in this world...

[tinhead]

Mind you, ever the pedophiles have rights in this world...

definitely, a "white" pedo would say "we just different, like black ppl, but still human beings"

Anyway, how far are you with Rigol schematics?

[A Hellene]

Rigol schematics? What is this thing?
Being extremely busy these days, I have not made any significant progress... Plus the fact that the continuity testers of my DMMs do not really help with those large capacitances the power lines have plus the very low resistance value of the differential line terminators.

Now that I have some time in my hands, I am constructing a special continuity tester based on a tiny45 that is able to distinguish resistances down to mohms if needed and to source up to 4mA test current when shorted. Another thing about this tester is that it can distinguish three programmable resistance ranges using a test voltage of 400mV and report them with three different audible tones. These are very useful features when the PCB is full of large capacitances and small resistances, like the DS1000 PCB is. For example, testing a large capacitance line will sound all the three resistance range tones in sequence and will stop within a time of half a second!

Anyway, by this time the power management and the resistance measuring and reporting modules are already working exactly as designed! Right now, I am adding a setup module in order to define custom resistance ranges and hibernation times without the need of reprogramming the FLASH. The final step, before resuming the DS1000 schematics, will be to put the t45 PCB with a CR2032 cell on board in a small plastic box; I have some spare car-alarm fob remote control enclosures that seem to be perfect to host this little device.


-George

[vlindos]

Greetings from Bulgaria!

First - Big Thanks for this thread and tinhead! And then... I finally bought my Hantek - Tekway - DS5201B! From this seller on eBay. Its was the cheapest place I found out there - 299 GBP. I don't care about the warranty since China is far way and there is no local dealers here. Yet more - there is option for returning the purchase within 30 days. Hope that helps the people still wondering how and where to purchase.
The parcel came in within a week. Package was much better than the Tekway package I have seen on this thread (you will see the pictures). I guess the mother company invests more in the marketing.
It is hardware version 10070x55553e8, still hope I would be able to upgrade to 200 mhz even though the package came with 150Mhz probes set. I ordered this USB TTL. Do you think it will work?

The bad thing... For the whole this thread I have read I didn't found some one to complain about the fan. I saw that most people are saying there is no fan. Yeah - but this has one. And it is loud, very very noisy one. Like most of you I disssembled the scope and after reviewing it - choose the most easy thing to fix the noise. I had added 150 Ohm /1W resitor. The current without the resistor was 40 mAmps and with it ~ 17mAmps. I did tried 0.5W resistors but they gets warm so 1W is better. Using the resistor, the fan is spinning at reasonable speed but at much lower speed than the original. I guess that would be enough but I don't really understand the purpose of that fan. There are holes over the PSU and at the same time the fan is far way from it and it is blowing to outside of the box. What do you think about this ?

Also did self calibration several times and tuned the probes but there is noise you will see on the attached pictures. They use the scope's oscillator (peak 5.2V). It looks like the noise is taking 1/15 the voltage, and so the noise would be ~0.3V. Is only me finding this disturbing? I see some of you had discussed it, but there is no solutions, right ?

In the end I want to say even though the above problems the scope is fantastic for its money and so far it serves well. It has lots of features that some of the low class scopes from the high end brands don't have. Also I would say that the UI is good very good. For example the the large display scope from Agilent that Dave reviewed don't have the option to high the side menus while here these menus are poping out only on certain context in the rest of the time you will be enjoying the big 800x480 display.

See the pics attached. Some of them are a bit blurred. Its due to my camera and photo skills quality. Don't judge me too much.

[tinhead]

It is hardware version 10070x55553e8, still hope I would be able to upgrade to 200 mhz even though the package came with 150Mhz probes set.

yes it will.

I ordered this USB TTL. Do you think it will work?

yeah, this one is ok.

The bad thing... For the whole this thread I have read I didn't found some one to complain about the fan. I saw that most people are saying there is no fan. Yeah - but this has one. And it is loud, very very noisy one. Like most of you I disssembled the scope and after reviewing it - choose the most easy thing to fix the noise. I had added 150 Ohm /1W resitor. The current without the resistor was 40 mAmps and with it ~ 17mAmps. I did tried 0.5W resistors but they gets warm so 1W is better. Using the resistor, the fan is spinning at reasonable speed but at much lower speed than the original. I guess that would be enough but I don't really understand the purpose of that fan. There are holes over the PSU and at the same time the fan is far way from it and it is blowing to outside of the box. What do you think about this ?

right, since few weeks Tekway and Hantek is installing fan by default.

I asked Hantek what is the reason (because previously it was only as option for high temp environment/countrys),
it seems that some ppl was not happy with the enclosure temperature. In principle the device was designed to work
without fan (and without any heatsink on FPGA/ADCs), later to imporve stability and signal quality heatsink has been
added to ADCs and FPGA. This is actually enough for main PCB, however the PSU itself is still a biggest heat source
- without small air flow the enclosure is getting hot - not really tragic in typical environment, but confusing.
A slow running fan is sufficient, the best solution is to replace the heat source - the 3.3V LDO by DC/DC converter.

If you don't want to replace the 3.3V LDO then just reduce speed of fan, but a resistor ehm yes it work too
but a 7805 (or even 7808) instead of the 7812 on PSU PCB is better.

Also did self calibration several times and tuned the probes but there is noise you will see on the attached pictures. They use the scope's oscillator (peak 5.2V). It looks like the noise is taking 1/15 the voltage, and so the noise would be ~0.3V. Is only me finding this disturbing? I see some of you had discussed it, but there is no solutions, right ?

first of all before you run self calibration do once "default setup".

The noise, hard to see from your pictures (but it seems to be a bit high), just don't shot pictures with your cam,
make a screenshot to USB - set both channels on, coupling GND, 1V/DIV, and 20ms/div, 2us/div and 200ns/div -
then i can compare to what here.

See the pics attached. Some of them are a bit blurred. Its due to my camera and photo skills quality. Don't judge me too much.

i can see your serial ... not that i really care about, but is maybe not wise to talk about haking the device
and to publishing own serial number ...

[pgup62]

In fact the fan was installed to lower temperature stresses on components and also to insure long term stability. We use to operate new units over 24 hours tests to avoid forwarding bad units to customers, which can happen because we believe Hantek have no tiime to perform deep tests.
After 24 hours burn-in in room temperature 20-22 deg, we can find the units are really cold. But as Thomas said, replacing the 12V DC regulator with a 5Vdc one should be sufficient (we should try and qualify it). Anyway for warranty purposes, China is far away and not  convenient and it will cost you a lot to send the unit back and to get if again ! This is a cost and a lottery. Elec3i had very few returns for one year, just for the ones who considered warranty, from Italy or Germany, or Finland, they were happy to get their machines repaired or exchanged in less than one week...

[rf-loop]

Only side note.

Thermal design is not always so easy.
In this special case I have not look how it go exactly inside HanTekway.
But it rise in my head one "de ja vu". (I really do not know thermal data inside scope after some changes - but I have some data without and with slow fan (not public))

This de ja vu:

Inside case is hotter air than outside. Holes are where they are. Inside box there are also high thermal gradiendts.
Free air convection is related ti dimensions, holes, temp gradients etc. There is network of pipe effects. I remember one industrial case where I drop one power losss (reduce heating power) and one hotpot inside cabin go more low temp. But what happend. There start thermal origin problems. One memory go littlebit more slow and enough for problems. Why.

This card was nearly bottom but just over bottom air intake. (cooling was designed with natural convection)
After lot of head pain we find problem. This component temperature have rised. Why? Becouse I reduce other place power loss - for make unit  more cool. Wow. Why this happend? Becouse air flow trough bottom intake was now less. Cooling this memory IC was reduced. (yes timing was also designet too critical but it was not my problem becouse it still have worked well just as it was designed. Design was well made and tested as industrial machines need. (becouse there need extremely high reliability)  I make problem reducing one heat source and then other component go more high.   Think how it can also be (littlebit) inside HanTekway. (I have not measured exactly so I do not know - only think tah there is also this posibility.

So, imho, it is of course clever to reduce this power unit some components power loss for reduce temp. (If think these components it may rise longevity and reliability)
Now there are many hot komponents on the main board inside faraday box. ADC's FPGA mainly. FPGA is BGA. High thermal stress and also variations may after time affect soldering reliability also + component itself age. Now with small (stupid selected) heat sinks they may rise surface temp easy over 60 celsius in around 20 celsius enviroment. If machine is not in ideal free air and enviroment temp is example 40 celsius we may find some komponents surface as high as around 80 - 100C.
Yes as factory tell they are still in specs. hehe. Yes nearly abs max area.

There is two posibilities. Design free convection cooling better. It do not cost anything but littlebit design work and calculations on the table. And some lab tests.  But becouse they do not know how to do - becouse lack of enough experience for this work they go easy way. Add fan. And agen. without thinking. Without any labbing, for finding optimum with noise and enough air flow.

Air flow is just right direction.  If think this unit box and layout. If it can design different then air flow can turn to intake and it make also possible to put some dust filter to air intake.

But now need take fresh air in mainly or markable amount from bottom holes! So I understand fan direction.

Also this unit drift around 1 - 2 hours after power on without fan.. Becouse thermal stabilization need long time - becouse construction is what it is.
Fan can reduce markable this.  If construction have not changed and only add fan. It need only slow airflow.
Also fan and other thermal design must be better than old Tektronix. It need be so that if fan fails machine do not die with overtemp. Watching or thermal designed so that worst case do not kill machine,
Of course I can be wrong.

[pgup62]

Rf_loop, hehe déjà vu ! we can practise more french  together
Your point of view is pertinent. We believe air flow noise can be reduced if fan is not against the case or something can let the air coming without less stress (number of holes in the case, etc.) or another place in the design.... Speed fan is the characteristic that affects the most the noise disturbance.
Whatever, for people like us surrounded with other noisy machines in our laboratory, the DSO fan is a piece of cake...

[rf-loop]

Whatever, for people like us surrounded with other noisy machines in our laboratory, the DSO fan is a piece of cake...

Yes, example if I turn on my TEK TM5006 frame includin CG5011 and CG5001 scope calibrators I feel like takeoff with J-11B.
But maybe most heavy noise is in one old one HP test gear what have 3 "turbines" together inside and case. 

In lab there is always noise but it is not very nice. It was -80's when nobody think noise.  Todays peoples play heavy rock with earphiones but then they are extremely critical if some equipment give small noise. But yes - I hope manufacturers keep this noise levels down as possible without super high price.

[marmad]

In lab there is always noise but it is not very nice. It was -80's when nobody think noise.  Todays peoples play heavy rock with earphiones but then they are extremely critical if some equipment give small noise. But yes - I hope manufacturers keep this noise levels down as possible without super high price.

But also, with modern low-noise fans, there is no excuse (except a couple of dollars) to have a lot of noise.  I paid € 299 for the Rigol DS1052E and it was noisier than 2 PCs running full-time (with a few quiet fans each) in my lab.  Would I have paid, for example, € 302 or € 303 for the same Rigol with a quiet fan?  Absolutely.

[vlindos]

@tinhead. Thanks for the responses man. You really rule and should be getting sallary from Hantek

I have updated the post with the usb short of the noise. Could please take a look and compare ?
Also could you give more info about the 3.3 V LDO to DC/DC converting. I have read your post about the capacitors changing in the PSU board but didn't understand if only the DC/DC replacement would be enough. Could publish more shots ?

[walt]

Is only me finding this disturbing? I see some of you had discussed it, but there is no solutions, right ?

This is the standard noise for this model.

To find out the true nature of his need to send the unit to vivisections. And no one can say beforehand that no one oscilloscope were harmed.

There may be a dirty power ADC, bad layout, noise on the input, even from the relay coil, switching 1 or 2 channels.
Maybe this is a mathematical artifact, or whatever.

to identify the source of the noise The easiest way its unsolder the relay RL04 by tinhead scheme and connect input pins ADC  with him  reference voltage.

[vlindos]

Hi walt,
Have you tried actually that? Is there any more pictures?
Could you give us more information about the element - like picture of the main board?

[tinhead]

@tinhead. Thanks for the responses man. You really rule and should be getting sallary from Hantek

I have updated the post with the usb short of the noise. Could please take a look and compare ?
Also could you give more info about the 3.3 V LDO to DC/DC converting. I have read your post about the capacitors changing but didn't understand the part of the DC/DC (i'm ee novice).

the noise look "good" or better said like on typical HanTekway DSO.

Now regards the LDO replcament, i did posted schematics look for it and i did posted a list of part which can be used to replace the LDO.
https://www.eevblog.com/forum/index.php?topic=1571.msg42363#msg42363

In principle you need on each side of DC/DC converter one GRM32ER71A476KE15L, on 3.3V output additionaly two EEFCD0J470R. Then you can replace the
org. caps in 5V path by bigger types, EEUFM1A152L  and EEUFM1A472L. Important is of course to solder these two GRM32ER71A476KE15L as close as
possible to DC/DC converter (PTH04070WAH) pins. this can be a small separate PCB like i and other did.
The noise level will be bit lower, max. the same as org. PSU, but the unnecessary heat from PSU will be gone (maybe someone should produce such replacement kit ...)