Hacking the Rigol DHO800/900 Scope

[Fungus]

FYI - my DHO-804 I've got today shows the fw v 1.03..

I'd upgrade it, ASAP.

[iMo]

..upgraded to 1.04.00.02, now doing the selfcal..

[norbert.kiszka]

Don't forget to calibrate all probes. All channels should be practically identical (unless factory screw up another scope), so You can do it using same channel.

[iMo]

..upgraded to 1.04.00.02, now doing the selfcal..

The calibration took 36minutes, while the last 1% took aprox 10minutes..
So something similar to the windoze install

[norbert.kiszka]

I don't remember exact time, but in my case it was more like 10-15 minutes. However my model is DHO924S.

I'm not using Windows for almost two decades, but I remember Win 2K on PII 266 MHz (SDR, not DDR rubbish) when last 10% took couple hours.

[shapirus]

Don't forget to calibrate all probes.

What do you mean by this? I don't remember anything related to calibrating probes in this scope. It runs calibration with all channels left open (nothing connected).

[norbert.kiszka]

Don't forget to calibrate all probes.

What do you mean by this? I don't remember anything related to calibrating probes in this scope. It runs calibration with all channels left open (nothing connected).

I mean manual probes calibration (or adjust if You prefer), not scope self-calibration which is something completely different.



Or if You want to be very precise for some reason (not necessary that slow and imprecise DHO800 rubbish): https://cdn.teledynelecroy.com/files/tutorials/ten_minute_tutorial_probe_calibration.pdf

[shapirus]

Ah, that. I thought of something like maybe frequency response calibration or things like that :)

[Fungus]

Don't forget to calibrate all probes.

What do you mean by this? I don't remember anything related to calibrating probes in this scope. It runs calibration with all channels left open (nothing connected).

"Compensate" the probes. Connect the probe to the compensation signal on the front (1kHz square wave) and adjust the little screw on the probe connector until the top/bottom is flat.

[Fungus]

All channels should be practically identical (unless factory screw up another scope), so You can do it using same channel.

Nope. You're supposed to match each probe to its channel. That's why they put little colored rings for the probes in the bags.

(There's probably not much in it, but that's what you're supposed to do)

[norbert.kiszka]

Nope. You're supposed to match each probe to its channel. That's why they put little colored rings for the probes in the bags.

Sometimes I cant understand If You trolling again or something else.

Colored rings are because You can use all four probes at the same time and it's easy to make a mistake with swapping channels - that will create bad conclusions.

There's probably not much in it

Looks like a user of a magic glass ball. All four AFD are equal. Only one possible cause of difference is flux residues after cleaning - personally I had very small amounts, but still with it, I have exact same response on all four channels.

[tautech]

Nope. You're supposed to match each probe to its channel. That's why they put little colored rings for the probes in the bags.

Sometimes I cant understand If You trolling again or something else.

Colored rings are because You can use all four probes at the same time and it's easy to make a mistake with swapping channels - that will create bad conclusions.

They ensure the same probe is used on the same channel it has been compensated on.
This is basic probe use and ensures measurement repeatability.

[sandor626]

FYI - my DHO-804 I've got today shows the fw v 1.03..

I'd upgrade it, ASAP.

I recently got (with the discount) the DHO804, it arrived yesterday, with the FW version 1.03
. So, I don't think I'll upgrade and I don't think I'll "upgrade" to DHO824 (or 924).
This decision of mine was born from these considerations:
I already have an original DHO924s and I was able to compare the operation of the two oscilloscopes. In use, the DHO804 heats up much less than the 924S (I think due to the absence of the frequency generator and because it has a 70 MHz bandwidth).  These two characteristics (being colder, and having a lower bandwidth) mean that it has less noisy inputs (which is highlighted by the Averange Vpp measurement) and the trace also appears cleaner than the 924S (for ranges of 1 mV and lower) .
To make the behavior of the two oscilloscopes almost uniform I must select the bandwidth limit at 20 MHz. Even if the trace of the 804 is still a little preferable (after all, the 924S is much warmer than the first).

[sandor626]

furthermore I won't go from 1.03 to 1.04 because I also believe that one of the reasons for the cleaner track depends on the factory calibration, which I would lose if I upgraded the FW

[Fungus]

furthermore I won't go from 1.03 to 1.04 because I also believe that one of the reasons for the cleaner track depends on the factory calibration, which I would lose if I upgraded the FW

Calibration has no effect on noise levels.

There's really no such thing as "factory calibration". All they do at the factory is press the "self cal" button which you can do too.

You can use 1.03 if you want but be very careful with XY mode because it can make your 'scope unbootable if you power-off in that mode.

[iMo]

What we should see as the noise background?
When I terminate ch4, for example, with 50ohm terminator and I look at the noise I see 340uVpp BW OFF, and 225uVpp BW 20MHz, DC at 500kSa/s and 1Mpts (after upgrade to 1.04).

[gf]

What we should see as the noise background?
When I terminate ch4, for example, with 50ohm terminator and I look at the noise I see 340uVpp BW OFF, and 225uVpp BW 20MHz, DC at 500kSa/s and 1Mpts (after upgrade to 1.04).

Peak-to-peak is not a very meaningful or comparable metric for Gaussian-like noise because the probability distribution is unbounded. Instead, measure the standard deviation (some scopes call it AC RMS).

[iMo]

What we should see as the noise background?
When I terminate ch4, for example, with 50ohm terminator and I look at the noise I see 340uVpp BW OFF, and 225uVpp BW 20MHz, DC at 500kSa/s and 1Mpts (after upgrade to 1.04).

Peak-to-peak is not a very meaningful or comparable metric for Gaussian-like noise because the probability distribution is unbounded. Instead, measure the standard deviation (some scopes call it AC RMS).

BW 20MHz     Avg. 39uV AC RMS, drops down to 33uV after some time (due to Avg math?)

BW OFF         Avg. 60uV AC RMS, drops down to 50uV after some time (due to Avg math?)

[sandor626]

furthermore I won't go from 1.03 to 1.04 because I also believe that one of the reasons for the cleaner track depends on the factory calibration, which I would lose if I upgraded the FW

Calibration has no effect on noise levels.

There's really no such thing as "factory calibration". All they do at the factory is press the "self cal" button which you can do too.

You can use 1.03 if you want but be very careful with XY mode because it can make your 'scope unbootable if you power-off in that mode.

I stick to the facts. I have two oscilloscopes, one from about a year ago (dho924s) and one from a few days ago (DHO804).  The DHO804 has better tracking than the DHO924S. Furthermore, the DHO804 is much less hot (almost cold).  Since my considerations, in your opinion, are not valid, tell me your considerations

Ps: if the dho804 1.03 no longer works, the Germans who sold it to me would replace it immediately

[sandor626]

What we should see as the noise background?
When I terminate ch4, for example, with 50ohm terminator and I look at the noise I see 340uVpp BW OFF, and 225uVpp BW 20MHz, DC at 500kSa/s and 1Mpts (after upgrade to 1.04).

this shows that by limiting the passband the noise is reduced, so the opposite is also true, with the same front-end the wider the passband the greater the noise.
The dho 804, which has less bandwidth than the 924, is less noisy. So if you are looking for a very quiet oscilloscope, I think you won't make any bandwidth upgrades.

Then there would be the matter of cleaning the track shown on the video... on this point I await Fungus' response.... and in any case, until I am convinced of the step to take, I have no intention of ruining my DHO 804 virgin

[iMo]

..I have no intention of ruining my DHO 804 virgin

That is ok, indeed. Could you look at your both scopes and tell us what the noise background noise you see? Best when you put the termination on the channel's BNC such it does not pick up a noise from the outside.
That would be an interesting comparison. And we do prefer the numbers here..

[norbert.kiszka]

Maybe the DHO800/900 was indeed designed with a higher sample rate in mind. Then either some technical challenge got in the way late during development, or product management intervened and wanted to protect the more expensive product lines for the time being. The latter is a more likely explanation in my opinion, and might mean that we will see faster sampling a year from now, maybe even software-upgradable in the current scopes... 

The latter is indeed the most likely explaination. But it does also have a different FPGA which could potentially impact that.

If You take a look at bootom side of heatsink, it becomes clear, that they designed it at first with another FPGA, but changed their mind for a cheaper one, changed placement of FPGA and modifed heatsink without removing old staff for old FPGA. At least mine DHO924S has it.

[norbert.kiszka]

I think what has been confirmed is that (a) the DHO900 models have these chips populated, and (b) the DHO800 models can be upgraded to 50 Mpts memory via just a software hack, i.e. the extra memory chips are not required for sample storage. That's what led to the assumption that the extra memory supports the logic analyser functionality.

True, but there's something that doesn't make much sense: Why does Rigol add two 256MBits * 16 RAM chips to sample 16 digital channels while one of these chips would have been more than sufficient for this job. Considering the ADC delivers a 12bit datastream at 1.25GSa/s and a single such RAM chip, interfaced via the Zync FPGA/SoC is well fast enough to cope with this data rate, we may in future DHO900 units only see one of the additional footprints populated with a RAM or .... there may be hacking possibilities that as yet nobody has on their list. At least, all the front end hardware including ADC and ADC sampling clock PLL are capable of 2GSa/s. And two such RAM chips, if you swap the ADC / MSO data streams, and the FPGA section of the Zync has enough I/O connectivity ... well I guess you know what I mean

Of course that's something that isn't done at home easily, but maybe even Rigol has plans for a further uprated model based on this hardware. Pure speculation... 

Maybe Im not RAM memory specialist, but I guess we can't use theoretical bandwidth in memory. 1.25GSa/s and 12 bit data gives 2.5 GB/s which is a lot. CL latency in DDR3 is around 10-15 ns. So changing address can take that long (not everytime). So they used more chips.

So why not smaller ones? Answer for that is most likely the same as for AFD chip capable of more than 1 GHz in 250 MHz scope - because it's better to reuse same parts across many models, instead of storing huge amount of parts and to lose money because of it. Also in machines You need to put those parts and take more time to take care, not speaking of human errors, how costly they can be.

1 GB of DDR3 is very cheap, so it's not worth of redesigning hardware and software (or FPGA firmware) with huge risk of getting into a big f*cking wall after couple months, because somebody wanted 10$ more income per month.

[norbert.kiszka]

What we should see as the noise background?
When I terminate ch4, for example, with 50ohm terminator and I look at the noise I see 340uVpp BW OFF, and 225uVpp BW 20MHz, DC at 500kSa/s and 1Mpts (after upgrade to 1.04).

Peak-to-peak is not a very meaningful or comparable metric for Gaussian-like noise because the probability distribution is unbounded. Instead, measure the standard deviation (some scopes call it AC RMS).

BW 20MHz     Avg. 39uV AC RMS, drops down to 33uV after some time (due to Avg math?)

BW OFF         Avg. 60uV AC RMS, drops down to 50uV after some time (due to Avg math?)

Noise caught internally when You switch configuration most likely - just look how PCB looks and You will know. Other reason can be software bug, but I doubt it.

[norbert.kiszka]

Nope. You're supposed to match each probe to its channel. That's why they put little colored rings for the probes in the bags.

Sometimes I cant understand If You trolling again or something else.

Colored rings are because You can use all four probes at the same time and it's easy to make a mistake with swapping channels - that will create bad conclusions.

They ensure the same probe is used on the same channel it has been compensated on.
This is basic probe use and ensures measurement repeatability.

Noise between channels in my scope has much more differences than impedance on inputs. If You wan't to measure very fast signals from something very precise, then DHO800 and DHO900 series at very best can be used as a paper-weight to hold 100x more expensive probes (and other scope) than this scope.