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

[alm]

EDIT: Actually, I found the "Component Level Service Manual"on an exotic website

tek.com is exotic?

but not really useful: Tek was already using hybrid modules for the analog front-end, thanks anyway 

That's a modern DSO for you . Lots of custom silicon, even more today than back then. This helps decreasing footprint and costs, and increasing performance. Some SMT passives on FR-4 just doesn't cut it for good HF performance.

Only exception is the low-end brands who just can't justify the investment.

[rf-loop]

Before I have tell this "bug" in design related to CH1 and CH2 as scope is booting.

I check it and:

R1CH1 go on just as scope is powered on.  Around +3,5V to coil as long as scope is booting.

So, scope input is directly connected to SELFCAL line (trough 248ohm resistor) as long as it is booting .
(what if I trust input is 1Mohm, but this is not so severe...how about high voltages on the input...)

Is it doing some kind of self test?

But  IMHO this is totally wrong!
Becouse there is no any protection.

Yest they tell that do not connect anything to input if do selfcal. Of course but they do not tell that scope is also connected to unprotected selfcal line as it is booting.

Front end read 300V CAT II.

I think someone need shame now.
How can tell Hantek that this can not do?

Scope may be connected to signal all time, powered or not.
There is no place what warning this.

Example if I have test system made... so that I'm continuously watching some signal with scope. But signal is always on... example some kind of transmitter.

I have made signal cabeling and matching.  Do I need every time take signal cable off if I want power (off) on this scope.
How about if power switch is on but shortly mains breaks.

Yes I understand these chinese boys who design and test these in they lab and only what they randomly think... and test just as they think use of scope. Well, maybe nobody think what all kind of situation scope may be in field.

I have speaked... Ugh

[alm]

I could only find the "Module Level Service Manual" there, I had to search for TDS520 "component level" service manual in order to find it somewhere else.

A search for tds520b on tek.com finds this.

The Tek 2232 was probably one of the latest
As hobbyists, we have no access to either Hybrids or analog ASICs, so both the "old" Tek 2232 and these cheap low-end DSOs are interesting to look at.

Agreed, you learn more from studying the older technology, but it's not how a modern DSO is build. Just like a modern consumer widget often consists of a blob of epoxy on a phenolic PCB.

[rf-loop]

I will still continue this HanTekway problem becouse this thread is for these scopes and specially not develop thread for open project front end.

This CH1 and CH2 bug is not at all so nice.

First as before. Both Channels are connected to Selfcal line trough 248 ohm resistor.
After these resistors both channels are connected together and after this they are connected trough 470ohm ressistor to TLV272IDGK output.
And this is not all. It is also connected to MAX II (cpld) pin 36 trough some small resistor and 220nF capacitor. (no dc but...)

You can test.

Measure resistance between CH1 center to CH2 center. Start up scope. Booting time it is around 500ohm.
Try with 300V fast rise ..  yes... after this you can buy new scope.

Someone ask... why ever some connect scope to signals before it have booted. It is good question.
But if it have done... first channel 1 signal and 2 are connected together as 500 ohm resistor. Nice.. yeah..
How about ~700 ohm to TLV272IDGK output?

If you have connected scope to some sensitive circuit... scope FW crash... do you remember connect probes or other signal wires out before boot...

I want go to qingdao and look this clever EE eyes when I show him what happend with his clever design.  After some smoke... yes put smoke back... it works agen.

Please correct if I am wrong!

[alm]

Popular 10x probes will add 9Mohm of series resistance. Even 1x scope probes add a few hundred ohms of series resistance. This may make the issue less severe. It does sound like bad design to me, though, inputs should either be 1Mohm or open-circuit when the system is booting/off.

[tinhead]

Well, that's confusing: for a single ADC, you need to meet was is known as the "Nyquist criterion", see Wikipedia ...

But what I am saying is only good when you have a single ADC...

As there are up to 10 of them, they can be clocked so that each one is sampling with a calibrated period that is 1/10 of the sampling period, thus achieving an effective sampling frequency of 1 GHz. Then, by digital reconstruction, you are able to get a 1 GSPS resolution digital signal, that you can THEN filter digitally to whatever you want below that.

And in this case, I really don't know if what I said for a single ADC is still valid, or if aliasing "magically" cancels itself by using several of them working with such a clock scheme, only requiring a max signal frequency of 1 GHz/2=500 MHz...

... don't take Wikipedia seriously ... 20+ yrs HF and later ARM/FPGA is better than pure theory,
but actually "nothing" in compare to ppl who design scopes since years, aynway.

For real time sampling, with no avarage and continuously signals you should apply Nyquist based on total sample rate,
(with avg. on or equ. sampling the sample rate didn't matter but the time of acq., think sampling scopes),
and for single shot mostly the single ADC clock with combination of random skew between
ADCs have more influence on the waveform than Nyquist (so total sample rate reduced by error factor coming
from skew/jitter + additional aliasing from single ADC clock). Finally don't forgett the non linearity of componets,
PCB itself and tons of other factors.

And of course there is aliasing on almost all DSO below 1GHz bw, some are better and have intelligent filter,
some really worse (especially due additional distortion caused by not properly synched ADCs).

For this particular design just take a look on the app notes and datasheets of the components which has been
used for these DSOs, they has been not chosed by price (only) but primary by reason. For up to 300MHz not bad at all,
sure there are some gaps around max bw, but you will have to pay much more to get something better.
The waveform distortion pure visualy is not worse than on TDS3000, and that's already good enough.

[rf-loop]

alm,  Yes, many times scope is used with 10:1 (or sometimes 1:1) probes, but not all time. 

Example I use lot of measurements directly from 50 ohm line or with some special other connection.
Scope have BNC vonnectors for signals, not only for probes. If it can safely use only with 10:1 probes they need be "fixed" probes...

I have use tens of years HP and Tektronix professional scopes, cheaper and more expensive.

I have never see this kind of design bug. Newer in any seriously bade test equipment for real working.

500 ohm directly from CH1 to CH2 center whole boot up time!

Direct unprotected connection ti deep inside scope sensitive electronix from inputs as long as scope is booting.

Do not go any serious real work enviroment for doing any measurements with this before you exactly know what may happend in system under test or what may happend inside scope with these signals what are going inside scope as it boots.

Sometimes scope may loose power accidentally and then power agen or it may do crash and you need boot...

Hantek probe 1x  center - center resistance 250ohm

Ok lets calculate... with probes... there is 1k between channels at boot time. 1V... 1mA. 10V 10mA. 100V, 100mA. Probe is accepted 200Vpk at 1x. Ok lets do it... -200 one probe. +200V other probe. No violations if look specs.

400V _difference_ between CH1 and CH2 center. But still in specs... Scope input. 300V CAT II, Probe 1x 200Vpk.

400mA 400V. Wau. 160W... how many milliseconds is enough...

Of course these are nearly as worst case. But who hell is designer who do not remember "worst case rules for designing"
Now if I do this fully documented...  some persons who can proof what is done and also example video.

How many scope I can destroy so that always Hantek send new ones.

So, I give my recommendation.
Just as in calibration state, DO NOT connect any signals to input this time when you shut off scope or when you start up scope. Of course always you know that powerline can not drop...
(low level signals are not danger to scope trough 1x and specially not at all trough 10x probe...

But remember also what it may affect in your circuit what are under testing. All circuits can not like 1kohm connection between to probe connection points.

I think, Hnatek need immediately solve this design bug. If there need boot time cal connection it need solve different way.


Why here is smoke. After this...

[rf-loop]

No any kind of real  antialising filter in front end. Front end is wide open for all signals over full freq span. And some peoples want it even more wide, but real advantage is make it more filtered but with better freq response shape.

8 (or 4+4) ADC clocked with FPGA PLL.

[ECL]

I think, Hnatek need immediately solve this design bug. If there need boot time cal connection it need solve different way.

I have not thought all of this through, so consider this idea "thinking out loud" 

Maybe 2/3 of the boot time is the OS coming up.  After that, the DSO app is run.
I have not observed any scope related state changes, prior to DSO running.  This
leads me to believe that the problem is caused by the initial state of the shift register,
controlling the relay.  Hantek should have added a tristate control, and pullup/down resistors
to all of the outputs, so the controls could be set to a knows state, prior to software initialization.
Looking at tinhead's (very handy) schematic, it appears that we may be able to force the
shift register to clear, by adding a RCD circuit to the SRCLR pin.  I have not checked if this would be
okay, for the rest of the signals, this register controls.  Another option would be to add an OE control
flop, tied to the RCLK term, and pullup/down resistors to the output lines.  The first write, to the SR
would enable the register outputs, until the next power cycle.  There is also the extra (open)
relay pole.  We may be able to do something with that, as well.

ECL  -K

[ECL]

Just curious if anyone else is experiencing this behavior?  If I power-on my
Hantek, with a USB stick installed (front port), the stick is not recognized by the
DSO application.  The OS, however, does see the stick, and auto-mounts it.
I need to unplug the stick, and re-insert, before the scope will flag it's presence.
It appears, at least on my unit, that the DSO app needs to see an insertion event,
before it detects the presence of a memory stick.

Does anyone else see this behavior?

Thanks,

ECL  -K

[tinhead]

500 ohm directly from CH1 to CH2 center whole boot up time!

are you sure ? When off, the self-calibration relay is not connected, the attenuator not active, so in principle 1.2M input
(the state of relays on my schematic is off mode)

Maybe 2/3 of the boot time is the OS coming up.  After that, the DSO app is run.
I have not observed any scope related state changes, prior to DSO running.  This
leads me to believe that the problem is caused by the initial state of the shift register,
controlling the relay.  Hantek should have added a tristate control, and pullup/down resistors
to all of the outputs, so the controls could be set to a knows state, prior to software initialization.
Looking at tinhead's (very handy) schematic, it appears that we may be able to force the
shift register to clear, by adding a RCD circuit to the SRCLR pin.

if there is not dedicated "enable" signal within CPLD design then the CPLD is controlling the
HC595 already during boot time, and after OS has been loaded switching to the last saved status.

This means if there is an issue then the dso.exe itself should (after the last state has been loaded) enable the HC595,
which is bad idea because we need then (or actually output) to be active to activate last state!
Even a simple solution like RC combination on /OE pin will break the functionality, a i/o controlled (from SoC and dso.exe)
will not work either. The only chance is (if not already done, which we don't know) to control the state of
attenuator/self-calibration control pins by the CPLD itself

However, even without knowledge what inside CPLD we can test it by not allowing the dso.exe and dn.rbf to
be loaded (just comment out these lines in /etc/init.d/rcS). I can't do it during next days, my both DSOs are now
disassembled (testing still a bit "ugly" clock domain hack).

[rf-loop]

Yes, all what I can now look is HW5 and HW7 with FW, ..225, 420, 427patch, 531.1... all have this problem and afaik problem was also before. Not nice but can live with it. If need lot of USB things all times, it is good to buy short cable.... no need wear scope USB connector. (specially becouse these quality is just as entertainment quality (all USB connectors not only hantek but who care... today electronic need only 1-3year lifetime and tomorrow we go more short "designed" lifetimes.)..

[tinhead]

Just curious if anyone else is experiencing this behavior?  If I power-on my
Hantek, with a USB stick installed (front port), the stick is not recognized by the
DSO application.  The OS, however, does see the stick, and auto-mounts it.
I need to unplug the stick, and re-insert, before the scope will flag it's presence.

yep, i know it. The funny part is, firmware update menu is even trying to access to stick, but
it fails because stick didn't exists (but the stick led is blinking ...) So it seems the firmware
just don't see a specific flag. Not sure if this is really firmware bug or linux hotplug delay bug,
it takes anyway a bit long to recognize usb devices. Maybe a delay before dso.exe will be loaded can help out.

[rf-loop]

500 ohm directly from CH1 to CH2 center whole boot up time!

are you sure ? When off, the self-calibration relay is not connected, the attenuator not active, so in principle 1.2M input
(the state of relays on my schematic is off mode)

No, not exactly 500 ohm... but 2x calibration line resistor
And after test long time, lot of shut off and on... sometimes it do not happend. Specially is very short time shut off and immediately on. But if litttlebit more time off... then looks like "always".

As scope is off relay contact is open, just as "rest" state.
After power on, both of  calibration relays get 3,5V to coil and contact close. After whole boot is over you can also hear relay go off. This time I have only HW7 open on my table and measured with it.

But I remeber HW5 scopes (you remember these special fail tests) I many times wonder in my mind why other scopes trace shift vertically momentarily as I shut one scope off..and agen on (for temperature realted things test)... signal was splitted simply self made multiport 50 ohm splitter and all scopes DC coupled. But this time I only write this note on the paper... as "check later what happend power up inside front end" becouse this time I have lot of busy.

Now I have check it with HW7. It is waterproof. Both CAL relays go on during bootup. (I mean bootup as time after power on to time when I can see trace.  Reistors are nearly 248ohm. Both relays on and... CH1 center to CH2 center just "500 ohm".
Btw, if exaxtly: CH2 relay get 3,58V and CH1 relay 3,49V
But why? have they design some kind of startup confidence test... and then forget what it can do in lab...

[ECL]

Yes, all what I can now look is HW5 and HW7 with FW, ..225, 420, 427patch, 531.1... all have this problem and afaik problem was also before. Not nice but can live with it. If need lot of USB things all times, it is good to buy short cable.... no need wear scope USB connector. (specially becouse these quality is just as entertainment quality (all USB connectors not only hantek but who care... today electronic need only 1-3year lifetime and tomorrow we go more short "designed" lifetimes.)..

yep, i know it. The funny part is, firmware update menu is even trying to access to stick, but
it fails because stick didn't exists (but the stick led is blinking ...) So it seems the firmware
just don't see a specific flag. Not sure if this is really firmware bug or linux hotplug delay bug,
it takes anyway a bit long to recognize usb devices. Maybe a delay before dso.exe will be loaded can help out.

Thanks to both of you   

Good point rf-loop, about an extender cable.  Might save having to replace
the connector, on the mobo, somewhere down the road.  Anyone ever check
if USB hubs are handled correctly?

@tinhead, yeah, funny that the firmware update utility does not force a check
of port/device mount status.  I'll play around with it, and see if I can figure out
what they're doing. 

ECL  -K

[colinbeeforth]

Hi tinhead,

You said: "Currently I'm testing Tonghui TDO3000 series, they have beautiful LC meters but actually nobody knows their DSOs."

Ah, nobody plus one!!!  *laughs*  Early this year I researched the low cost DSO market quite a bit.  Tonghui looked nice, for what little that is worth.  My summary of Tonghui was no local agent, display screen was only 320x240,  and acquisition memory was 4k and therefore too short for serious work.  10k is an absolute minimum to be able to see essential detail in a single line of video.  I used to work with video a bit so it matters to me.

I liked the deeper, more conventional shaped case with storage for probes!!!  One thing with lunch box DSOs, they might take up only a small space, but you can't put them at the bottom of a stack of meters... 

Do Tonghui have any new models out that improve the acquisition memory?

Cheers, Colin  Melbourne, Australia

[zeus]

Hi all,

I've been following this thread for a while now as i have a Hantek 5202B.

Wondering if people here can help me.

Well, channel 2 has stopped working and also the vertical position control (it jumps back down to 0). The oscilloscopes self calibration stops at step 13.
Channel 1 seems fine.
Anyone know whats the cause or how to fix.

Sorry if it's wrong place to post. just let me know.
Thanks.

[tinhead]

Well, channel 2 has stopped working and also the vertical position control (it jumps back down to 0). The oscilloscopes self calibration stops at step 13.
Channel 1 seems fine.
Anyone know whats the cause or how to fix.

this looks like broken ch2, the self-calibration is not checking hardware itself, but if the sampled data is wrong it will fail.

Where you bought it ? It could be that you have warranty case, so depends on where you bought it might be
easy to repair/replace due warranty or a disaster. If you from Europe, i can evt. ask Hantek.

[zeus]

Hi tinhead,

I bought it from www.youroscilloscope.com. I think they dropship from Hantek china.?
I'm in europe/cyprus.

So do you think i have to ship the oscilloscope back or is there an easier solution, maybe in europe?

Thanks.

[tinhead]

I bought it from www.youroscilloscope.com. I think they dropship from Hantek china.?

first of all try to contact them, for sure Hantek does have warranty, but it might be that the seller will not
accept it, e.g. user error so no warranty or you will have to pay shipment in both directions.

I'm in europe/cyprus.
So do you think i have to ship the oscilloscope back or is there an easier solution, maybe in europe?

"maybe", let's wait for the answer from "youroscilloscope"

[ECL]

I'll play around with it, and see if I can figure out
what they're doing. 

Interesting.  I did some experiments with DSO.EXE renamed, so it would
not be automatically executed at startup.  It appears that the OS driver
always detects the USB stick, from powerup, but it does not signal the
stick's presence, to DSO.EXE, until an insert event is observed.  Once an
insert event is observed, DSO.exe will always detect the stick.  DSO.EXE does
not need to be running, during the insert event, for the stick to be detected, the
next time it is run.  The problem does not appear to be a race-condition.  Delaying
the start of DSO.EXE does not help the problem.  Also, accessing the device, from
the OS prompt, prior to starting DSO.EXE, does not help. 

This leads to to believe that the problem can probably be fixed, by forcing a
refresh of the mount and/or interface status.  Unfortunately, I am not as familiar with
driver/kernal level Linux, as I would like to be   Perhaps one of the Linux guru's, on
the thread, can take a look, and see if they can figure out how to remount the drive, or
force a state refresh?  Near as I can tell, the driver is not creating a standard /dev.  Instead
they use SCSI emulation, and mount the flash stick directly to /mnt.  I do not see anything
in the fstab, that described the device.  There is an jffs2 entry, but I believe it is for the
onboard flash.   One curious note, when DSO.EXE detects the stick, it automatically does a
directory listing of the drive, which is displayed on the console, along with a note "get udisk first".
I have not found any sign of udisk in the OS image, except for a file called "udisk-flag", which is
always zero bytes.  One other interesting note, from powerup, the OS driver assigns the stick
"sda", but on most subsequent inserts, it assigns the device "sdb".

ECL  -K

[rf-loop]

These tests are all done with only one single unit what was originally virgin factory condition (DSO6062B).
Remember that this is only one single random unit.  HW version is 7. FW is (110531.1)
So if you buy later or have buy before situation may be different. So, do not thing that new 5062B is always 5102B with 5062B labels and logo.

All tests:

Sine, source HP8657B
50ohm good 2m length cable and "middle quality" tektronix  50 ohm feed trough terminator in scope BNC.  Input level, 50mV/div. Generator output signal so that scope display 6div p-p at 1MHz (around 107mVrms)
1MHz level  used  as reference level (0dB)
Example: 1MHz, level adjust for 6 div p-p.  Change freq to 100MHz.  Adjust generator level so that scope display agen 6div p-p.  If need add 1.5dB scope have -1.5dB at 100MHz
(here need also watch that there are not any more deep minimum between 1 and 100MHz)
Note about level:  HP flatness and level accuracy better than HP specs but may be example +-0,2dB inside this level and freq area.

Some notes: Cable and connectors attenuation unknown (but not bad). Better method is use accurate splitter in scope end and follow power with accurate level/power meter so that tst system signal flatness can correct)

This is also not good to measure real BW becouse input reactance. 50 ohm termination is not good method becouse it is not matced to 1M 20+-3pF input. It do not see scope input as pure resistive load. SWR is not 1.

A: one channel (CH1), real time, averaged and display so that more than one cycle in display (so mostly 8 - 2ns/div)

B: as A but 80ms/div (scan), mode peak, 4k mem (samplerate 2,5ks/s) and 40k mem (25ks/s) both same)

factory nev DSO5062B  (soft) modified to DSO5202B + filter 45

Note: this is not mistake. B: all looks same, but it is natural becouse I do not have real DSO5202B!
And becouse all done with exactly same scope and becouse scope whole front end construction.


__________A__________B
-1,5dB___100MHz____100MHz
-3,0dB___180MHz____190MHz
-6,0dB___230MHz____245MHz
-10dB ___275MHz____300MHz

Original factory new DSO5062B just as received  (after factory this unit was defined just as DSO5102B  exept: only logotype.dis have DSO5062B and front panel label DSO5062B but all other just as DSO5102B)

__________A__________B
-1,5dB___100MHz____100MHz
-3,0dB___152MHz____190MHz
-6,0dB___180MHz____245MHz
-10dB ___200MHz____300MHz


Original factory new DSO5062B just as received (soft) modified (defined) fully to DSO5062B (all as dst1062b exept hantek 60MHz logotype.dis.)
This is 60MHz model if factory have done it with exactly same main (HW7 and same "bandwidth" sub version) board as 100MHz model and defined it fully to 60MHz.


__________A__________B
-1,5dB____77MHz____100MHz
-3,0dB___102MHz____190MHz
-6,0dB___131MHz____245MHz
-10dB ___140MHz____300MHz

[rf-loop]

This is also not good to measure real BW becouse input reactance. 50 ohm termination is not good method becouse it is not matced to 1M 20+-3pF input. It do not see scope input as pure resistive load. SWR is not 1.

Only for information as: What it may be in practice?
Most of peoples of course know this, but many times this effect is underestimated in mind.

I just make small practical test with scope what have 1Mohm/15pF+-2pF input _and_ direct 50ohm imbedance input.
(CH1 and 2 mode selector AC/GND/DC/GND/DC_50ohm)
Scope is in good uncertified calibration and "perfect". Model Tektronix 2465A DM (including opt CT)

same as before: 0dB ref at 1MHz.
Signal gen as before with Hantek.

Two different measurements. In both cases signal level 6div p-p. 50mV/div  -3dB is point where signal level rised by 3dB give 6div p-p signal.

First: CH1: DC1Mohm + separate Tektronix 50ohm feedtrough terminator.
Second:  CH1:  DC_50ohm.
Flatness is extremely good to over 200MHz. Just "flat" but first measurement  give littlebit more variations. (still both are well inside +-0,3dB and there is also generator flatness inaccuracy included)

First: -3dB point  290MHz
Second -3dB point 375dBm

This show very clearly that this unmatching is not any marginal "perfectionism". It affect lot of as we go to higher frequencies.

Same scope, but model B, manual tell that using internal termination -3dB is same frequancy as -4,7dB point with input as 1Mohm and with external 50ohm termination. (typical)

Now Hantek input is 20 +-3pFand this Tek input is 15 +-2pF

So situation is littlebit more bad with Hantek if use external (normal)50ohm termination. This is always problem with all scopes what do not have real 50ohm inputs and need use external.

With 1Mohm input + 50ohm termination it is not well matched and terminated as real 50ohm! And with high frequencies it make markable effect. (This bad matching can also affect device under test do not function as normal)

[tinhead]

Now Hantek input is 20 +-3pF

i'm not sure if Hantek ever tested it, the user manual is based on Tekway user manual (v0.9) which
is really old (Dec 2009). There are other errors too in the maual, like the 2000wfrm/s which should
be btw 2500 wfrm/s (as announced, documented in presentations and website and confirmed later).

Anyway, that's only estimated value, Rigol CA or E does have identical input (CA except attenuator) and 15pF±3pF.
 
I have modified my attenuator too (now more or less like Rigol CA) which gives me 11pF when is enabled,
however when disabled (which is then the original curcuit) i can measure 15pF (with 100kHz LCR meter - maybe
not the best way but sufficient for estimated results).

Giving this i assume that Hantek/Tekway does have 15pF±3pF too (like Rigol CA or E) and not 20pF±3pF like in the manual.

Btw, the manual v.09, at time of writing Tekway was using different relais, some other components has
been changed as well.

[rf-loop]

Ok, I have not measured it.
So if it is like this Tek then it may affect just as Tek, exept that inductance is unknown.

So or so, but this 15 or 20pF make its effect with external terminator. This I want show.
If it is less than specs then it is littlebit better. Then it affect nearly as Tektronix.

Bad matching do not only difference in BW. It also affect so that BW curve shape looks wavy. (specially with upper frequencies)  (becouse unmatched and standing wave effect, bac trawel wave phase angle change as freq change and it add or subtract signal level)

So this need also remember if do tests that there is this effect.

How to make better external 50ohm temination if scope have not internal 50ohm impedance matched inputs?