Hacking the Rigol DHO800/900 Scope

[stj]

the LA on the ui is not a problem.
lets be honest - your gonna try to add it sooner or later anyway! 

[norbert.kiszka]

I found some another thing. This Jitter thing (along with other hidden functionality) wasn't working before - because there was no switch case of plot function and it crashed when tried to execute plotting from null pointer. So for a quick test, I changed few instructions to nops (NOP stands for no operation, instruction requesting from CPU to do nothing and go to the next instruction). This will prevent from crash (and from rendering normal waveform). This is how it looks:

[GingerGeek]

WITHOUT having a useless LA component in the UI?

Is that such big problem? Without the source code, it's quite time consuming to do a smallest change. Also every change makes a risk of doing some mistake, which will end up as a bug.

I do get your point, and my experience with correcting intel ME regions and CRC correction on EFI boot ROMs is child's play by comparison to your work I am sure... Even developing WITH source it can be hard enough to debug sometimes IME. Since I noticed you have is a custom UI with many unlocked features, it was just a question.... I guess I have the answer And thank you for your input 

[norbert.kiszka]

About half of the app is on the .so lib, so Im making changes in Assembly. But about other half is compiled Java. Decompiled form if this is Smali. Working with the Smali code is a horror. It's 100x more difficult than Assembly. Whole UI is mostly Smali, XML layouts and some "pseudo SVG" format (Android thing).

With all my knowledge about communication with AFE and FPGA I can make, and I have plans to make new scope app (in C) without sh*tty Android. But this is a quite long way to make. As for now, there is working modified Debian with a two kernels - each one has drivers for a Rigol things (AFE, FPGA, etc). But as for now, I did only initialization of FPGA and ADC by linking mentioned .so lib and calling one of its functions.

[mrisco]

If You want to have it not only hacked, but to have much more memory depth (125 M) and many other features, You can install my firmware mod.

Ok so I do now have 100Mhz and 50Mpts, but umm... "much more memory depth (125 M) and many other features2 you say...?
Would this include unlocking BW and memory depth WITHOUT having a useless LA component in the UI?

Thanks for the help people

You can unlock all the features of the device, forget SCPI activations codes or changes in the device series simply by installing this:

Click here for more info.

[jerrymk]

You can run Python scripts like here:

Extended GUI user here, how to get the launcher/shortcuts utility in your video that allows for exiting an application or split screening?

Also, any idea why DHOTools show Remote Cmd Execute Error, even on the scope itself? (How does this app work anyway? How does it know what IP address to connect to, for example?)

[mrisco]

Extended GUI user here, how to get the launcher/shortcuts utility in your video that allows for exiting an application or split screening?

Also, any idea why DHOTools show Remote Cmd Execute Error, even on the scope itself? (How does this app work anyway? How does it know what IP address to connect to, for example?)

Hi, I use Zonelauncher. It allows you to keep the stock launcher and create a zone for calling other applications. You can configure it to show an icon for split the screen and call other applications.

DHOTools must be started after the Oscilloscope application is full connected to the network. I saw sometimes the DHOTools application can't connect if the oscilloscope application is not loaded side-by-side like in the video, maybe there is a issue with the application priority in Android.

DHOTools look for the oscilloscope application connected in the same sub-net by using Py-Visa.

[jerrymk]

DHOTools must be started after the Oscilloscope application is full connected to the network. I saw sometimes the DHOTools application can't connect if the oscilloscope application is not loaded side-by-side like in the video, maybe there is a issue with the application priority in Android.

I just found out how you're supposed to use this app, apparently because Math1 is not enabled, but the app defaults to Math1 for some reason, it didn't work out of the box. I'm not sure if it's supposed to do "remote cmds" to auto enable the channels though, if so then I guess it's still broken. Changing it to an enabled channel worked. Running from another phone does work too, but not on the Windows app (Can't connect to a device). I don't think I need it at the moment though.
Thanks for the reply.

[mrisco]

I just found out how you're supposed to use this app, apparently because Math1 is not enabled, but the app defaults to Math1 for some reason, it didn't work out of the box. I'm not sure if it's supposed to do "remote cmds" to auto enable the channels though, if so then I guess it's still broken. Changing it to an enabled channel worked. Running from another phone does work too, but not on the Windows app (Can't connect to a device). I don't think I need it at the moment though.
Thanks for the reply.

Oh yes, I forgot to mention that you need first to enable the relevant channel on the oscilloscope. But in that case the error message should be "Can't select Wav source"

What error message did you receive when using the Windows app? Make sure that only one app is running at a time in the DHO, on your phone or PC. No other DHO tools must be running in the background.

[jerrymk]

What error did you get when using the Windows app?

Can't connect to a device.
The app never asked for network permission, even if I manually allow it, it still shows Can't connect to a device.
(The computer can access the web control with local IP)

[mrisco]

Can't connect to a device.
The app never asked for network permission, even if I manually allow it, it still shows Can't connect to a device.
(The computer can access the web control with local IP)

I just tested it on a virtual machine with Windows 10 that was recently installed. The virtual PC and the oscilloscope are in the same subnet (192.168.0.X) and it is working.
Please check that your PC and the oscilloscope are on the same subnet. If you want we can continue in the Github discussion for the DHOTools.

[Randy222]

Glad to see the 800/900 is still a popular item and has not faded away.
I still have a couple of mods to do, and after that I will try and hold back on additional changes, unless absolutely needed.

[Lathe26]

i realise the ram is a buffer that is filled and then processed,
i just thought that as i dont need huge speed or packet size it would not fill the ram so fast that it would need 3x the space before processing it.
infact if i use a clock input to sync the capture then i probably only need maybe 20-50 capture frames to spot a bad logic signal.

yes, i know i could just buy a cypress-based usb board for €5 on ali for that, but it then needs another laptop on the desk!

---------------------
answering my own question for others.
according to this thread: https://www.eevblog.com/forum/testgear/low-cost-compatible-rigol-pla2216-logic-probe-for-dho900-(and-hacked-dho800)/
the LA will run without extra ram.
 

If you're only looking to get a logic analyzer to monitor 8-bit and 16-bit systems, the DHO924S will likely barely meet your needs.  Unfortunately, the LA functionality is full of bugs.  It does ok with recording data and displaying it on the screen.  However, most methods of simply saving or exporting the data are either broken or compromised (though it can be done).  Further, you will quickly feel constrained by only having 20 channels (4 analog + 16 digital), especially if you are working with CPU with an address bus (likely 16-bit) that is separate from the data bus (8 or 16-bit).  Even if the bus is 16-bits wide and is time-multiplexed, you'll likely need to use a few extra channels for control signals to differentiate when the address vs data is valid and knowing a read vs a write.  If, on top of all of this, you also need to check other signals in another part of the system, then you'll likely not have enough channels left over for that.

One huge advantage of the LA on the DHO800/900 is that 4 of the channels are analog, which can be useful at times.

As someone who works with similar retro systems, I would recommend using the Sigrok/Pulseview software (easier to use than the Rigol touch interface) with a Hantek 6022L which offers 32 channels of digital input.  Sigrok supports other LAs with 32+ channels as well.  It also supports a huge number of protocol decoders that you yourself can add or extend (I massively overhauled Sigrok's MIDI protocol decoder).

I own an DHO924S and paid for the LA (PLA2216) and frankly feel a little ripped off by the cost and the bugs.  My Hantek 6022L with Sigrok works far better as an LA.

[norbert.kiszka]

Glad to see the 800/900 is still a popular item and has not faded away.
I still have a couple of mods to do, and after that I will try and hold back on additional changes, unless absolutely needed.

DHO800/900 has huge adventages over MHO900 and other scopes. It's more lightweight and less power hungry than MHO900.

Also it have very low noise. In my upcoming release of my mod, there will be possibility to limit bandwidth with 72 options to chose (one is auto). Lowest possible cutoff frequency is 5 MHz, which gives measured noise ~9 uV RMS.

After my hardware and software experiments, I realized the most of the noise comes from the first stage(s) in the AFE chip and almost nothing from the ADC. I have plans to design and make two simple hardware changes, to make noise even lower (with every bandwidth setting).

Speaking of the AFE bandwidth filter, it's possible to change all LC filters in order to give better pulse response and more bandwidth (900 MHz or even more) at the same time. So normally bandwidth can be limited to ~300 MHz in AFE (this filter in AFE chip has a near perfect pulse response) and if somebody at some moment will need one channel with much higher bandwidth (i.e. to see square wave at 50-100 MHz), just open this channel settings and change auto (or something else) to 600 MHz or OFF.

Making changes for upcoming release already took a lot of time and I need to spend little more to make few more changes to make it works as everybody would want to see (like auto bandwidth working properly in every single case, no matter what). Stupid and classic race condition caused to make me think I did something wrong for many days, while the cause was extremely simple and extremely simple to fix. Exact same problem is with the stock app, but it was much less often, since sending data to AFE was less often. Instead of sending like 10 bytes to one AFE chip, 5 bytes was sent and other 5 was sent to a random one, which resulted with very strange things. Those AFE chips are using SPI bus. App together with the driver (kernel module) was made in this way: send data to chose CS and later in a separate call, send one byte and repeat this process for every byte if needed. With this and multiple threads trying to send something to the two separate AFE chips, it created a mayhem.

I own an DHO924S and paid for the LA (PLA2216) and frankly feel a little ripped off by the cost and the bugs.  My Hantek 6022L with Sigrok works far better as an LA.

I already fixed some problems with LA in my mod. Likely I will take a look into problems related to exporting data, but in the later release instead of upcoming one, because people already are waiting quite long for it.

[scopeman]

Looking forward to it Norbert. I know that you put a lot of time, heart and soul into the project and the wait for the next edition will be worth it.

Happy New Year,

Sam
W3OHM

[stj]

the guy who cloned the awg said a lot of the noise comes from the buck convertors using unshielded inductors.
something to think about.
maybe build some cans or get some mu-metal sheeting. ??

[norbert.kiszka]

the guy who cloned the awg said a lot of the noise comes from the buck convertors using unshielded inductors.
something to think about.
maybe build some cans or get some mu-metal sheeting. ??

I forget to say about my one accidental discovery, which I shared here quite long time ago, but probably a lot of people missed it or just forget.

At that time I had bypassed LC filters in the CH4 (and added 50 ohm termination soldered into BNC socket pads on PCB), so naturally I had more high frequency noise. When I was trying to lower the noise with adding more decoupling capacitors (in parallel - either in missing spots or soldered one of top of another), temporarily I didn't put all of the screws on the heatsink. After some experiments with it, I found which one screw should be missed to have easiest and in the same time, permanent result of lower noise. In the attachments I used Dave photos, because it will take some more time to find my own photos.

I didn't made (yet) full rev. eng. of the power supply lines of AFE chips, but lately I found something when I was doing experiments with different designs of LC filters (right below the ADC chip). Below the heatsink on the right side of CH1, there are four inductors. For a test I replaced those with 5.6 nH, which I just used for ADC and I don't have too much SMD coils to chose from. It resulted with noise bigger by about 20-30%. Before switching back to original ones, I did some continuity tests. Two upper ones are powered from upper DC-DC converter and probably (it was a quick measurements) are powering two AFE chips. Other two coils are powered from this second DC-DC converter. I think it may be worth to try to change it to some low noise linear regulator (small separate PCB and separate transistor instead of the one used to switch on/off) and maybe change this inductors to something else.

On the second attachment on the left side, there is a one decoupling capacitor (0402). I changed it to to 1 uF/10 V and looks like there is not much difference. However I didn't made too much comparisons with different settings - I just did this change when I actually had PCB on the operating table.

Right now I did a comparison between CH1 and CH2. CH1 have definitely much bigger noise on the lower frequencies and it's closer to this two DC-DC converters.

[Despotes]

 I'm considering ordering the DHO802 model as I don't need 4 channels. I was debating between this model and the Uni-T UPO1082HD, but leaning in the Rigol direction because of the mods.
Will the mods (Sparrow Extended Home etc) work with the DHO802 2 channel and if so, will the UI show 4 channels or two?
Are there any mods yet for the Uni-T UPO1082HD? Is it still worth considering or is the DHO800 series a better value? Quality?

[shapirus]

I'm considering ordering the DHO802 model as I don't need 4 channels.

I suggest reconsidering this with the question: "am I absolutely positive that I will never need more than two channels?"

[Despotes]

I'm considering ordering the DHO802 model as I don't need 4 channels.

I suggest reconsidering this with the question: "am I absolutely positive that I will never need more than two channels?"

Absolutely.

[norbert.kiszka]

Personally when Im using scope for measurements (instead of testing my mod), mostly Im using one channel. Sometimes two and in a very rare situations three. In the last channel (CH4) I have added 50 ohm termination (4x50 Ω), which is more convenient and it gives better pulse response than any external terminator.

Price difference between two and four channels is very small. On the hardware side, two channel (DHO802 vs DHO804) scope has three AFE chips instead of four and exact same thing with BNC sockets. In both there is the same four channel ADC chip.

Let we say, You will have 2 years warranty. For three years You will be using one or two channels only (somehow). If one AFE will fail or You will decide to have one separate channel with some hardware modifications (usually either internal permanent termination or modified LC filters or both), You have some options. If one channel will fail (those cheap scopes are not 100% reliable for sure) in two channel scope, You are left with single channel scope and You will need to spend a lot of money on another one - so in the end, it will cost more.

[Despotes]

So the mods require a 4 channel scope?
I'm also considering the DHO-1072, but the 4 channel DHO1074 is $170 more. I like the bigger screen for these older eyes.

[norbert.kiszka]

So the mods require a 4 channel scope?

I would say, it does not require. But for sure all of them was more tested on four channel scopes. So problems are less likely to happen on four channel scope.

I'm also considering the DHO-1072, but the 4 channel DHO1074 is $170 more. I like the bigger screen for these older eyes.

Both DHO800/900 and DHO1000/4000 have HDMI output with mirrored screen. You can buy cheap 10-12 '' screen with touchscreen and even with Vesa. Actually I have such idea to mechanically connect my DHO924S with a used industrial screen - either by some DIY or some existing double holder after small modification (because of the replaced fan with one which goes through the middle of the back case).

Technically (speaking about the hardware) it's possible (with other OS kernel - which I started to work on it) to have two screens without mirroring, which will allow to have more user friendly UI (like some buttons on this separate screen instead of taking Your time to do things like open some menu, change tab and then actually change what You needed). Two screens without mirroring will also allow to use some other software without the need of turning on separate device (I would say computer, because technically even Nokia 3310 is a another computer). Anyway, this is future.

Edit: this seems to be nice: https://www.ceneo.pl/167263501

[Despotes]

Do your mods affect the rise times at all?

[mrisco]

So the mods require a 4 channel scope?
I'm also considering the DHO-1072, but the 4 channel DHO1074 is $170 more. I like the bigger screen for these older eyes.

This is a DHO802 UI: