Hacking the Rigol MHO900 Scope

[Hydron]

That's encouraging at least - lets keep the bug reports going in while there seems to be attention being paid!

Unfortunately I haven't had much time to use mine yet (so haven't found any bugs myself, though I'm sure I will).

[norbert.kiszka]

Again, might be just my experience, but I tend to disagree. I reported a bug in early October (LA threshold set and LA threshold used off by a factor of two), received  some additional questions about this and finally a test firmware on last Friday. I just tested this today and the issue is fixed now. It will certainly not always be like this, but in my specific case, I have nothing to complain about.
There are obviously people working for Rigol who are capable and willing to fix existing issues. Well, at least regarding blatant bugs which are easy to reproduce...

Bugs in DHO1000/4000 and DHO800/900 are not fixed for years.

[Vadimchik1973]

Hi.

Is there a hardware difference between the RIGOL MHO934, MHO954, and MHO984 oscilloscope models?

If I buy a RIGOL MHO934, will it become the highest possible (1GHz) model after applying your patch?

This patch: https://buymeacoffee.com/norbert.kiszka/e/479030

[norbert.kiszka]

Hi.

MHO934, MHO954 and MHO984 are identical in hardware but can differ in accessories. In my local supplier MHO934 has added 350 MHz passive probes, while other models has 500 MHz probes.

In my mod (in the current version 0.1.2) Rigol license system is bypassed in multiple different ways at the same time. In case of bandwidth, You can select bandwidth limit separately for each channel - almost like in original, but with additional 350, 500 and 800 MHz options. It works on all of those models and on the MHO98.

This bandwidth selection (in 0.1.2) currently use original AFE binary flags to set low pass filter, that's why 800 MHz option gives 1 GHz as "OFF" (default) option, because flags are identical in both of them.

Speaking about the bandwidth, currently Im working on DHO800/900 mod (same AFE chip as in MHO900) - I have finished reverse engineering of this binary flags and I did exactly 72 bandwidth options (also separately for each channel), including auto, which depends on the current sample rate. Later, likely in the next year I will implement same thing in MHO900 and it will be published as a free update.

[LesWright]

Sweet, this worked out great on my MHO954, excellent.

Incidentally, when you pull the memory dump, its much faster to do an ASCII search for the string "CH4_CHIP_TEMP", that string appears right after the key.

[pizzigri]

Getting a lot more interested in your mod. Especially for the bug corrections, and early next year - I actually am away from home and will not be able to play with my MHO98 for some time - I will donate and download your mod!

[gsobol]

Sweet, this worked out great on my MHO954, excellent.

Incidentally, when you pull the memory dump, its much faster to do an ASCII search for the string "CH4_CHIP_TEMP", that string appears right after the key.

Yeah, I didn't want to anchor the search to this particular string since I wasn't sure it will be the same for everyone, and who knows, it may change in the future with a different firmware version.

[zbyr]

Again, this is bad design.

Rigol sells a 100MHz signal generator, not a 500MHz signal generator.

Really? I didn't know that before. Thanks to letting me know.

In that case, explain why this is AFG and original software allows to make 500 MHz with custom samples?

Rigol also cannot stop you from putting 10kV into input, yet you should not.
That is why there is somethic called specifications which you should follow, you cannot make everything idiot proof.

What is the point of choosing such fast and expensive DAC and using it's (presumably) full sample rate, when other parts can't keep up with it?

It is called oversampling and increase signal fidelity, but other reason could be that they just used what met minimal specs and some of them are surplus or simply what they got available.


I guess you are software guy (reverse engineering, making mods), but not very good in analog domain.
Not everything that can be "unlocked" in software is backed up by proper hardware, chaning software won't magicaly improve components used in the scope, there are limits in software for a reason.

[norbert.kiszka]

That is why there is somethic called specifications which you should follow, you cannot make everything idiot proof.

Actually, as I noticed in my first days on this forum, this forum is also not idiot-proof.

If You are smarter than me, You should know better specifications of the AWG in MHO900. Did You read it at all? Do You know what is the sample rate of the DAC in this AWG?

What is the point of choosing such fast and expensive DAC and using it's (presumably) full sample rate, when other parts can't keep up with it?

It is called oversampling and increase signal fidelity, but other reason could be that they just used what met minimal specs and some of them are surplus or simply what they got available.

Ever heard about Nyquist theorem? Go check sample rate in Your phone or computer, when it plays audio. After that, check on Your oscilloscope if it can play 22.05 kHz sine wave. And don't tell me it's impossible, before checking it by Yourself.

I guess you are software guy (reverse engineering, making mods), but not very good in analog domain.
Not everything that can be "unlocked" in software is backed up by proper hardware, chaning software won't magicaly improve components used in the scope, there are limits in software for a reason.

"I guess" You are trying to start a another fight here. Actually analog domain is my strongest knowledge and experience. Guess who did a reverse engineering of analog path in DHO800/900 and reduced overshot and ringing couple times by using parts for less than 1 $. In same time I increased maximum bandwidth (which can be reduced in UI which enables selected low pass filter in AFE).

If Your assumptions were correct, I wouldn't be able to calculate total system bandwidth after hacking AFE chip, which is used in MHO900, DHO800, DHO900 and few other series.

There are two explanations for AWG in MHO being unable to make anything above 100 MHz. First possible explanation is: extremely bad design of the analog path. Second and more probable explanation is: Rigol did a another scam by writing 5 times faster sample rate that it is in reality.

In DHO900 AWG officially can output up to 25 MHz. I did 80 MHz without any issue and with very clean sine wave. Can You explain this? "I guess" You can't.

If You are so much smart in the analog world, tell me how to increase bandwidth in the scope and in same time make it to have better pulse response. All by passive parts only. Do it without checking my other posts, where I explained exactly how to do this.

[zbyr]

You totally ignored my point about oversampling, Nyquist theorem only said that MINIMUM requirments for recreating signal of limitted bandwith you need to sample as twice the speed of the highest componnet in the signal, but there is more about signal than just recreating approximate of it, oversampling increase dynamic range and SNR, but you need to first read about it before based on your limitted knowledge.
In case of your audio example, guess why sound cards allow you to change sampling rate to 96k, not because they could produce sound above 20kHz.

As for increasing bandwidth in the scope you can tune it in some way to have better pulse response, but did you maintain flatness in the whole range?

[norbert.kiszka]

You totally ignored my point about oversampling, Nyquist theorem only said that MINIMUM requirments for recreating signal of limitted bandwith you need to sample as twice the speed of the highest componnet in the signal, but there is more about signal than just recreating approximate of it, oversampling increase dynamic range and SNR, but you need to first read about it before based on your limitted knowledge.
In case of your audio example, guess why sound cards allow you to change sampling rate to 96k, not because they could produce sound above 20kHz.

As for increasing bandwidth in the scope you can tune it in some way to have better pulse response, but did you maintain flatness in the whole range?

Sounds like a excuse. Rigol claimed AWG in MHO900 to have 1 GSa/s. Try to explain why it can do 100 MHz properly and 101 MHz looks like a mess. Only good explanation for such thing is sample rate 200 MSa/s, which is much lower than Rigol claimed.

Please explain me, how it can happen when DAC running at 1 GSa/s, can't make anything more than exact 100 MHz at its output? Please be scientific, instead of using capslock - because only one thing that You proven is Your lack of knowledge about sample rate in MHO900 AWG.

[SiliconWizard]

What's the reference of this DAC?

[norbert.kiszka]

What's the reference of this DAC?

https://www.ti.com/product/DAC38J82

In this datasheet (available under above link), there is a graph in which max output frequency was 1.2 GHz...

[gf]

Then I guess the DAC is operated with 4x interpolation, 1 GSa/s output sample rate and 250 MSa/s input sample rate. According to the datasheet, the filters are halfband filters with a stopband corner of 0.6x input sample rate which implies a max. reproducible frequency of 0.4x input sample or 100 MHz. In this case it is to be expected that the attenuation of images already begins to suffer noticeably at 110 or 120 MHz.

I'm surprised that they spent a quite expensive DAC.

[washley]

MHO900s have one of those? That by itself is a ~$100 chip

[norbert.kiszka]

Then I guess the DAC is operated with 4x interpolation, 1 GSa/s output sample rate and 250 MSa/s input sample rate. According to the datasheet, the filters are halfband filters with a stopband corner of 0.6x input sample rate which implies a max. reproducible frequency of 0.4x input sample or 100 MHz. In this case it is to be expected that the attenuation of images already begins to suffer noticeably at 110 or 120 MHz.

I'm surprised that they spent a quite expensive DAC.

It goes bad at 101 MHz. So IMHO sample rate is more like 200 MSa/s.

In case of DHO800/900, it also has low pass filters (LC after the DAC) and officially it can output up to 25 MHz with 156.25 MSa/s. Without changing sample rate, it works properly up to 50 MHz (even higher with higher THD) with reduced amplitude because of filters. After changing PLL clock from 1.25 GHZ to 2 GHz, sample rate goes up to 250 MSa/s and 80 MHz instead of 50 MHz.

With such DAC chip, Rigol definitely did something wrong - either very bad design of AWG board or sample rate is 200 MSa/s.

[gf]

It goes bad at 101 MHz. So IMHO sample rate is more like 200 MSa/s.

In case of DHO800/900, it also has low pass filters (LC after the DAC) and officially it can output up to 25 MHz with 156.25 MSa/s. Without changing sample rate, it works properly up to 50 MHz (even higher with higher THD) with reduced amplitude because of filters. After changing PLL clock from 1.25 GHZ to 2 GHz, sample rate goes up to 250 MSa/s and 80 MHz instead of 50 MHz.

With such DAC chip, Rigol definitely did something wrong - either very bad design of AWG board or sample rate is 200 MSa/s.

In practice, you cannot realize an ideal brickwall filter. So it is an illusion that you can generate 100 MHz with 200 MSa/s. In practice you'll need a factor of at least 2.5.

Should be easy to find the true sample rate with a spectrum analyzer (or FFT at full BW and max. scope sample rate). If you generate e.g. "bad" 110 MHz and see the first image appearing at 140 MHz, then the "true" AWG sample rate is 110+140=250 MHz.

The reason for using an interpolating/oversampling DAC is to relax the requirement for the analog reconstruction filter after the DAC. Passing 100MHz while eliminating a 150MHz image requires a lowpass filter with a very high selectivity. A high order elliptic filter can possibly do that, but it has a significant phase and group delay distortion in the passband. OTOH, the digital FIR filter in an interpolating DAC is linear phase and can easily achieve a high selectivity and stop band attenuation. Due to the higher output sample rate of the interpolating DAC, the analog filter after the DAC has relaxed requirements and can be designed for better group delay flatness instead of maximum selectivity - at 1GSa/s the stopband corner can be at 900 MHz, allowing a wide 800MHz transition band.

Other AWGs use this approach as well. E.g. SDG2000x has 1.2GSa/s interpolated sample rate at the DAC output, while the true (input) sample rate is only 300MSa/s, and it can generate up to 120MHz.

Operating the MHO900 DAC at 1GSa/s interpolated, 250MSa/s real, and 100MHz max. would be quite comparable to that (same ratios). As said, that's just my guess, it would need to be verified with a SA. I can't as I don't own a MHO.

Does the DHO900 have the same DAC, or is it a plain, non-interpolating one?

[mawyatt]

That TI DAC is impressive part!!

Not sure if some of these type DACs support what was called Bi-Phase mode way back where the DAC output is flipped in sign, thus Bi-Phase modulated. This produced a result (frequency) centered (symmetrical flipped) around the Bi-Phase clocking rate. Also another DAC technique used was operation at multiple Nyquist Zones. For these modes to operate properly the DAC output and follow on circuitry (Filters & Amps) needed to support the higher frequencies with appropriate fidelity.

Recall back in ~2009 the 16GSPS Keysight Griffin 14b DAC had these modes and was a spectacular performer back then, remember producing better than -76dBc NPR at 1GHz!!

Wonder what DAC the Siglent SDG6000X utilizes?

Best

[0xdeadbeef]

Then I guess the DAC is operated with 4x interpolation, 1 GSa/s output sample rate and 250 MSa/s input sample rate.

As a side note: I recently used the AWG to create a 975kHz square wave signal and was a bit puzzled that it jittered considerably between ~972kHz and ~976kHz.
A bit of jitter was to be expected but with an actual 1 GSa/s sample rate,  the jitter should be in the few hundred Hz range, not in the several kHz range.
Actually what I saw would fit a 4ns resolution pretty well (1/(256*4e-9) = 976562.5, 1/(257*4e-9) = 972762.65).
So, 250MSa/s actual sampling rate (without interpolation) seems pretty likely.

[SiliconWizard]

MHO900s have one of those? That by itself is a ~$100 chip

Obviously that's not the price they get and using the same chip for all their scopes further lowers costs.

As talked about earlier, the "entry-level" scopes are crippled on purpose in order not to compete with their higher-end scopes - or for that matter, with their standalone AWGs as well.

Now, even if the DAC itself could be used at higher sample rates, the scope may not have the resources otherwise to leverage that (fast enough DDS and appropriate output stage).

[mawyatt]

Recall that some AWGs in the past created a squarewave by "zero crossing detecting" a DDS created sinewave.

Best

[Dagger]

Hi. I saw that Rigol released a new firmware for the MHO900 (v00.01.00.00.25)
Is it possible to have the hack along with the new version by replacing the firmware files in the "files_v0.1.2\firmware_00_01_00_00_24_mod_v0.1.2" folder and using the posix script intallation or should we wait for the next release of the hack?

Thanks a lot for your work!

[norbert.kiszka]

Hi. I saw that Rigol released a new firmware for the MHO900 (v00.01.00.00.25)
Is it possible to have the hack along with the new version by replacing the firmware files in the "files_v0.1.2\firmware_00_01_00_00_24_mod_v0.1.2" folder and using the posix script intallation or should we wait for the next release of the hack?

Thanks a lot for your work!

Hi. I noticed their release, which was a surprise (updates are extremely rare). I have already downloaded it to make a review. After comparing changes I will need to decide either to merge changes or to make everything almost from scratch.

As for now, Im maintaining three mods and my current work is with DHO800/900, which already took much longer than I expected. It's based on the same code (also similar hardware) and good thing is I can do same fixes and improvements for other two series.

Making mentioned review and merging changes or making it almost from scratch will take at least one-two weeks.

Before that, I have plan to release quick update with changed AFE bandwidth binary flags, based on my findings from a reverse engineering - mainly this will allow to have more than 1 GHz bandwidth (around 1.4 GHz - 1.5 GHz).

[rteodor]

MHO900s have one of those? That by itself is a ~$100 chip

Obviously that's not the price they get and using the same chip for all their scopes further lowers costs.

As talked about earlier, the "entry-level" scopes are crippled on purpose in order not to compete with their higher-end scopes - or for that matter, with their standalone AWGs as well.

Now, even if the DAC itself could be used at higher sample rates, the scope may not have the resources otherwise to leverage that (fast enough DDS and appropriate output stage).

Would it be possible that Rigol got an underperforming batch of DAC chips for cheap ?

[Fungus]

Would it be possible that Rigol got an underperforming batch of DAC chips for cheap ?

No.