Hantek DSO5102B - weird effects

[TinkeringSteve]

[UPDATE]

Appears to be fixed with newest firmware !

http://www.hantek.com/english/down_list.asp?unid=30



----------------------
Hi there,

I bought a DSO5102B scope a few weeks ago, now trying it out a bit, I encountered weird effects when both channels are on.
I'm aware that both channel have only half the sample rate then, but the scope will switch between resolution of measurement in ways that don't seem to make any sense, and you can't be sure what you're looking at, which makes the device useless.

(details in YT description boxes)

Look at this:
Here it just switches wildly between two settings, without me doing anything, as soon as CH2 is on (~ 2nd half of the vid).


and this:
Here, it apparently chooses lower rates for a certain time base range.


I wonder if anyone can reprouce this. I have the apparently newest firmware, 2.06.3 (110531.1).
If no one has encounetred this, and it#s not the firmware - could any sort of hardware fault cause this?? The box it came in showed heavy hits in one corner, no marks on the device, though.


EDIT:
-----
Some images to show setup:
http://img267.imageshack.us/img267/189/hantek15triggersetup.gif
http://img840.imageshack.us/img840/8282/hantek15ch1setup.gif
http://img802.imageshack.us/img802/9534/hantek15ch2setup.gif

Here's how it looks with bandlimit 20 MHz on:
http://img17.imageshack.us/img17/9216/hantek15bandlimited20mh.gif

...which is not exactly like what it's doing when the error occurs:
http://img257.imageshack.us/img257/9968/hantek15devicescrewupti.gif


Do any of you have experience with Hantek in cases such as this?
I mailed them about it perhaps a week ago, but haven't heard of them yet.


I hope someone can make sense of this
- TS

[rf-loop]

" 2.06.3 (110531.1). " <-- this is far away from newest FW.

New FW is 110909.0

But can not fast imagine what is your problem becouse so fast you move something (or not) in video.

Do you mean it randomly changes itself  so that you do not touch anything and so that signal itself is routed also reliable to scope without any other possible error source.

(yes it is possible there is also HW problem. (bad soldering, bad relay or what ever, and also probe may fail if you use probe)

If you do selfcalibration, do it go ok?
If you use only single channel CH1 or CH2..  Same problem both channels.
If you use double channel mode. Look signal with CH1 or CH2. Both show same problem.
Look also that can you see any differencies between channels pulse shape.

What is your HW version?

Mailed to China (Hantek)? If so, please be patient becouse they just have been holiday. National Day (Oct. 1)
 3 days (Legal holiday = "minimum") Oct. 1 - 3
many business and offices follow this: Oct. 1  - 7
 

[TinkeringSteve]

New FW is 110909.0

Oh, I'll look into that! On their site they just stated "2.06.3" under nrewest firmware, I didn't look for the longer number until somebody today asked me for it in another forum.

But can not fast imagine what is your problem becouse so fast you move something (or not) in video.
Do you mean it randomly changes itself  so that you do not touch anything and so that signal itself is routet also reliable to scope without any other possible error source.

In the part #1 video, I do not touch anything! the scope is doing this weird stuff on its own.
Only in part #2, I switch the time base every second or so.

If you do selfcalibration, do it go ok?

Self-cal was reported as successful, but the problem persists.

[TinkeringSteve]

I updated to FW 110909.0 - problem still there.

[Spiro]

Looks like filter is kicking in randomly.

[amspire]

Looks like filter is kicking in randomly.

It does look like a 20MHz filter switching in and out randomly, doesn't it.

Something like an intermittent dry joint on a capacitor could cause a problem like this, but it would be all the time - not just dual channel mode.

Is it only Channel 1 doing this or both channels?

If you switch on your 20MHz bandwidth filter, does it stay in the filtered mode, or does that switch in and out as well?

Richard

[TinkeringSteve]

Looks like filter is kicking in randomly.

It does look like a 20MHz filter switching in and out randomly, doesn't it.

Something like an intermittent dry joint on a capacitor could cause a problem like this, but it would be all the time - not just dual channel mode.

Is it only Channel 1 doing this or both channels?

If you switch on your 20MHz bandwidth filter, does it stay in the filtered mode, or does that switch in and out as well?

Richard

Hey,

well it does that also if I flip the channels, i.e. which one is on GND and which one measuring.
Here is how it looks when the filter is set active manually:

Looks quite different from the videos?

I have not tried to activate the filter when the device is "jumpy" and look if it stays stable then - I'll try that when I'm at home later.

Also, a Hantek guy just emailed me, and sent me a newer firmware than what I found on the site yesterday, which I'll try out, too.
But if no one of you guys ever experienced this behavior, I doubt that it will help.

- TS

[tinhead]

there is nothing wired*, the overshot/ringing is much faster freq. component than the actual (or visable on that DSO) rise time
of the test signal. Additionaly when you switch second channel on the sampling rate is 2 x 500MSs (and not 1GSs like
with single channel on). Therefore while in some timebase settings (think amount of sampled point and resolution)
the interpolation can not always reconstruct properly the dirty component of your test signal.

I can perfectly simulate it on my DSO using dirty signal, and can also get properly sampled/reconstruced signal
with "clean" sqaure. When you e.g. switch to long memeory (40k) you will see these effects with even one channel enabled,
just because the sample rate for one channel is already 400MSs (and the RF component freq. much higher than sample rate/10/7)


*wired --  maybe except interpolation filter sensivity, which can be reduced to show more clean - but not true - signal.
Of course the firmware can do it a bit better, with less sharp filter (like in early firmware versions) the signal was looking
better (RF not filtered out) when switching into two channles, but with only single chan enabled the waveform was bit
more distorted. With bit sharper filter (while undersampling like current fw on your example) the signal is looking
sometimes switching between real and filtered condition, with more sharp filter the firmware will be for me useless
(as i wish to see overshots and ringings because they are there).

So you can see the firmare can be changed in two directions, the right balance is the trick, known manufacturers
are doing this since years and you will still get sometimes fancy artifacts or nice looking waveforms (where the signal
is completly dirty crap).

I've edited my posting, so you can see how it looks on older fw (2.3.x), with less filtering and on current (well, since 2.5.0) firmware.
Feel free to send these pictures (and your videos) to Hantek, a "perfect" filter is something between fw 2.3.0 (where dual chan was better looking)
and 2.5.x/2.6.x (where single chan is better looking).

[TinkeringSteve]

[...]
I can perfectly simulate it on my DSO using dirty signal, and can also get properly sampled/reconstruced signal
with "clean" sqaure. When you e.g. switch to long memeory (40k) you will see these effects with even one channel enabled,
just because the sample rate for one channel is already 400MSs (and the RF component freq. much higher than sample rate/10/7)

What do you mean by "clean" vs. "dirty" square? What I'd call clean square has much more overtones than the scope can handle, therefore, the anti-aliazing filters take away some of it which causes the squares to look like they have an sine waveform modulated on them.

On a e.g. 10 MHz signal, if I adjust the timebase to show a few periods of that waveform, each of that period should be made up of 500M / 10M = 50 Samples, right? So in any case, there should be a lot more detail visible than in the parts of the video where the displayed waveform looks flattened.

On a tektronix 2GS/s machine, it looks a lot more like in the first half of the part#1 video, than anything else. So I consider this "the real signal".
Now in video part#2, you can see that on lower timebase settings there is more detail, on very high ones too, but somewhere in the middle, it switches to lower detail.
I'd call this weird.

Also, the fact that in video part#1, second half, the scope switches, all by itself, wildly between settings, I'd call super weird!

[TinkeringSteve]

I've heard another interesting theory now: Processor Load.

A colleague suggested that the computations the scope does might take almost all of the processor power there is, and, at some points, more - which makes the software, which detects this, switch to lower resolution.

But for this to be true, all of those scopes must have these effects, right?

Tinhead:
Would be interesting what exactly you were able to reproduce.
Also, I still don't understand what's normal about video part#1, the wild jumping between settings.
My colleague, used to expensive gear, doesn't agree at all that this is normal for an intact oscilloscope with the stated capabilities.

[TinkeringSteve]

Update:

I have installed the newest firmware I was given by Hantek, the problem persists.

I think Tinhead's threory does not explain it sufficiently.
If I feed the 12 MHz crystal signal, and set the time base to 40ns, all looks good. As soon as I set it to 20ns, the scope will randomly switch between two different "resolutions".
If Tinhead was right, the lower detail version should display *all of the time*, not just a few seconds, and without me doing anything, shortly jump to the higher res setting, or am I missing something?

- TS

[tinhead]

I've heard another interesting theory now: Processor Load.
A colleague suggested that the computations the scope does might take almost all of the processor power there is, and, at some points, more - which makes the software, which detects this, switch to lower resolution.

ehm, what i he, rocket science engineer?

It does not have anything with processor load (and to be very honest the current version is creating less load),
when the calculation task is ready then it is ready and data will be displayed - when it is too complex
to get processed within one waveform cycle then the waveform rate will drop - but the signal remains the same.
It does not have anythign to do with resolution (and no, the software is not switching down resolution, this is not a PC game)
as the sampling is real time on ADC and definitely fast enough pre-postprocessed on FPGA.
The data is then in memory and will be bw filtered (remember - fw is calculating corner freq. based on model and applying it to
stored waveform), then alligned and filtered again, parallely FFT and measurments are running, and finaly displayed on screen.

But for this to be true, all of those scopes must have these effects, right?

all these or all DSOs?

all DSOs are doing many math things to filter/dither/interleave/restore the signal from sampled data.
There are many parameters to be watched, many error sources.

"all these" is not exactly truth, because it depends on your setup (btw,why you care about while measuring
repetitive signals? Use avg and don't undesample and you will be fine. With random signals these "effects"
are almost not given as the filter is from what i can see skipping first 4 waveforms).

Tinhead:
Would be interesting what exactly you were able to reproduce.
Also, I still don't understand what's normal about video part#1, the wild jumping between settings.

the jumping signal like on your video as said above, depends on resolution, signal quality and sampling rate
not a big deal.

My colleague, used to expensive gear, doesn't agree at all that this is normal for an intact
oscilloscope with the stated capabilities.

what is normal ? DSO is always trying to show a smooth waveform from a bunch of dots,
especially on interleaved DSOs not a big deal to get distortion due the interleaving process - page 9 Agilent 5989-5732EN
app note. You can do many tricks to try to allign waveforms from MANY ADCs to one smooth waveform,
some of the tricks are just pure filtering. When the signal (in your case the RF component of your signal) is on the
filter border the waveform might switch from not filtered to too much filtered anly because the filter don't know
and will never know what is now distortion coming from interpolation and what is RF component of the signal.

As i said above, don't care that much about repetetive signals as avg sampling is what you should use for such signals,
for random signals there is no issue because the firmware is skipping 4 waveforms

[tinhead]

Update:

I have installed the newest firmware I was given by Hantek, the problem persists.

aha, give me version number ...

I think Tinhead's threory does not explain it sufficiently.

well, it does. It didn't matter what timebase you setup, i can get these effects with 2ns as well as 200ns/div.
What metter is as already said sampling rate, signal timebase as a combination of factors where the firmware
can't (based on sampled data) recognize if the RF part of the signal is coming from interleaving errors/distortion
or from the signal itself.

However, as you can see from pictures above, they CAN change the filter to be less sensitive,
this will remove these jumps between filter on/off but it will then show more distortion created
by the - or well - not really expensive DSO.

I can tell you something, probably every day they getting emails with questions like "why this or why that, change pls",
and (as i'm decompiling all firmware versions and checking what inside) they changing always something
to make ppl happy. Sometimes it would be maybe better to say " are you crazy? what did you paid for that DSO?"
as the hardware have some restrictions and you can't just fix everything in software.

But who knows, maybe some day Hantek will mange to write perfect firmware which is able to detect
based on actually nothing what is the real signal and what distortion

When you really don't like it and don't wan't to use avg. mode for repetitive signals then please continue to
sending them tons of emails - use my pictures too when you wish - and maybe some day Hantek will
implement a user selectable filter. This will be nothign new, there are DSOs with such option.

[TinkeringSteve]

I've heard another interesting theory now: Processor Load.
A colleague suggested that the computations the scope does might take almost all of the processor power there is, and, at some points, more - which makes the software, which detects this, switch to lower resolution.

ehm, what i he, rocket science engineer?

No, just an electrical engineer.
It was a guess to explain the effects which he never experienced on any DSO he operated - and he has many years of experience with those.

But neither he nor I have closely looked at how exactly the DSO5000 devices are designed and how they working,
so thank's for your explanations so far, I'll think about it when I'm more awake again ;-)

[...]
It does not have anythign to do with resolution (and no, the software is not switching down resolution, this is not a PC game)

By resolution I was referring to the really used resolution (e.g. number of samples) put through the calculation pipeline to, in the end, display something on the screen - which is not bound to the sample rate of the ADCs.

as the sampling is real time on ADC and definitely fast enough pre-postprocessed on FPGA.

Ok, if you say that :-)

"all these" is not exactly truth, because it depends on your setup

Yeah, well I meant, on all DSO5000B series scopes, using my setup.
If not most/all other people can reproduce this, i have to suspect that my device is defective, and could not trust what it is telling me, and better should get it repaired.

(btw,why you care about while measuring
repetitive signals?

Why shouldn't I? I'd like to see whether they look how they are supposed to?

Use avg and don't undesample and you will be fine.

How can I make sure to not undersample, when the device selects the sample rate automatically? (Maybe stupid question, I never had such a device before, and just worked a bit with DSOs at work and never had such experiences as now. I'm mainly a software guy, so my experience here is limited.)

One thing I find strange still, though, is the fact that I know someone with a scope that is supposed to work in a similar way (many lower-samplerate ADCs used interleaved to get a high sample rate), namely a Rigol DSO. And he never experienced this.

With random signals these "effects"
are almost not given as the filter is from what i can see skipping first 4 waveforms).

I'll try at work the next days, they have way better equipment (function generators etc) there to try such things.

aha, give me version number ...

It is  110923.1


Lol about sending them tons of emails :-D I'm not sure, I feel already guilty for flooding them when they had a huge stack of emails to be read after their holiday that you mentioned, hehe...

We'll do some more thourough testing at my work place within the next few days, with a guy who knows much more about these things than I do, let's see what he says. I'll post it here then.

- TS

[TinkeringSteve]

Maybe interesting extra side note:

I just asked a friend to try a setup close to mine (basically, just a crystal oscillator, in his case. 8 MHz, like I had in one of the videos), with his Rigol 100Mhz scope, with pretty much the same specs as my hantek one, and, from what I've heard, similar electrical design.
he wheeled through the entire time base range of his scope, but nowhere he found such effects as my Hantek scope is displaying.

Oh jeez, now it's runnign really wild, when I cycle through the mem settings (perhaps unrelated bug). If I then turn the time base knob forth and back a step, it loks normal again:


- TS

[tinhead]

Yeah, well I meant, on all DSO5000B series scopes, using my setup.
If not most/all other people can reproduce this, i have to suspect that my device is defective, and could not trust what it is telling me, and better should get it repaired.

i can get perfectly similar effects on my working hardware (jumping filter on/off) ...

However, your last video is confirmation that there is something worng on hardware side too.
From what i can see you switching only memeory depth, there are no bugs in 110923.1, so must be something in hw.

Do default setup, do self calibration and try again - when still strange crazy signal when switching memory depth then send you DSO back to seller.

(btw,why you care about while measuring
repetitive signals?

Why shouldn't I? I'd like to see whether they look how they are supposed to?

Use avg and don't undesample and you will be fine.

How can I make sure to not undersample, when the device selects the sample rate automatically? (Maybe stupid question, I never had such a device before, and just worked a bit with DSOs at work and never had such experiences as now. I'm mainly a software guy, so my experience here is limited.)

Repetetive signals have one big advantage - they repetetive. So you can easily remove all random artifacts/glitches just
by switching avg on - the result is "exact" how the signal looks like.

When you look for glitches of course this will not work, then you have to switch avg off and sample fast.
Sure, the scope i changing the sample rate automatically, so all you can do is to sample with one channel
to be able to sample as fast as possible to catch fast signal rising/changes/glitches.

One thing I find strange still, though, is the fact that I know someone with a scope that is supposed to work in a similar way (many lower-samplerate ADCs used interleaved to get a high sample rate), namely a Rigol DSO. And he never experienced this.

right, Rigol managed to implement (any kind of) filters much better. You can actually see (when you enable to user selectable low/high pass filter on Hantek) that these
guys have some issues to implement such functions perfect. Of course it is a bit easier on Rigol due slower waveform/s rate, but still nothing what can't be done therfore ...

Lol about sending them tons of emails :-D I'm not sure, I feel already guilty for flooding them when they had a huge stack of emails to be read after their holiday that you mentioned, hehe...

... don't think too much about email flood. When you, as customer, think something is wrong and the manufacturer (like Tekway or Hantek) is
responding to customer request/question why not to email?

Tekway told me once that they wish the feedback, as this will help them to continuously improve their products. Maybe not always everythign
can be fixed in given firmware, but still it helps out. When you look on companys like UNI-T, well, they would tell you (when you ever get answer)
that "the new hardware will fix errors", which is like in next 2 yrs - where you asked to fix scalling errors on current firmware, so not an good answer at all.

[TinkeringSteve]

@ Tinhead

hey, I just realized... you said why I care to get a repetetive signal with real-time, I could use average to look at how the waveform looks really like.

BUT: even if I turn average on, the waveform looks still the same - with the details missing, on the same time base settings as with no averaging!

Does that change anything?

(not yet @ work & tested with other gear, just got this on my mind when waking up)

- TS

[TinkeringSteve]

However, your last video is confirmation that there is something worng on hardware side too.
From what i can see you switching only memeory depth, there are no bugs in 110923.1, so must be something in hw.

What makes you think that? Could it not just be an unrelated bug?
(not that, although not visible in the last video, when switching mem depth, I have Acquire type = peak, not normal! Only then, the bug of the messy screen occurs, when having switched through all mem depths and gettingback to 4k)

I haven't gotten another reply by Hantek so far, I asked them if they think he hardware has damage...

I mean, I wouldn't mind that much if this is all a software issue, and I know I can get full detail if needed when using just one channel, and using theh 2-channel setup for porposes not needing all detail (although the wild switching is a bit annoying).

But... I really need to make sure my hardware is intact & reliable ... who knows how long it will work as more or less good as it works now, if it's the hardware...

On the other hand, I don't want the hassle to send the scope back to china and wait for who knows how long before I get something back, meanwhile not having a scope at all...

Damn! It's not easy ;-)

[TinkeringSteve]

Appears to be fixed with newest firmware !

http://www.hantek.com/english/down_list.asp?unid=30