Rigol's New DHO800 Oscilloscope unbox & teardown

[Someone]

..
so that all makes sense. but what i am seeing here is also that there is this camp of people making the 10x generalization (broadly across purposes). and theres the people quoting the nyquist 2x upper aliasing limit (for some sine wave or whatever). then i suppose the 6x is meant to be a sorta 'in-between' those 2 extremes perhaps?..

Nope.. There are two arguments with 2 numbers basically:
a) the modern scopes require only 2.5xBW sample rate -  because the newest scopes have well made design

If you want to extra correct, that 2.5x figure was based on typical filters and 8 bit sampling. A better filter is needed for 12 bits while retaining the same noise floor as LSB counts. If a better filter existed why was it not used before?

[tautech]

  So it seems like a good comparison for the DHO800 would be the SDS1104X-U, and I'd have to say that if I was looking to purchase a scope and my budget was $400, this DHO800 would be a real contender.

Absolutely and therefore I´ll compare the DHO804 not only to the DS1054Z, but also with a SDS1104X-E.
Having 12 bit is not really a buying argument in this pricerange, it just makes things more complicated to handle.

Yep, 12 bit is just another feature within the full feature set.
Buyers would be unwise not to consider what any complete feature set offers and only focus on the 12 bit.

[2N3055]

ETS. Does the new DHO series support it?

No, it does not and not many scopes do nowadays..

[Fungus]

The Centaurus chipset should be 2GSa/s, afaik.. Why 1.25GSa/s in 800/900 then? Perhaps the lower grade FPGA in 800/900 limits the sample rate??

That's my theory - the cheapo FPGAs on the 800/900 series wouldn't be able to handle it.

[Muttley Snickers]

I'm wondering what the applications are for people in the target market of scopes like the DHO800 that even 100MHz+ BW would be a deciding factor? 

The cost of these gadgets also varies with region so it's not just a choice between one or the other for some people, particularly those of us on a budget. For example, down here there is a much larger difference in price between the Siglent SDS1104X-E at $693, the Rigol DHO-804 at $824 and the DHO-814 around $968. I really wanted a new scope to include an RTC for timestamps and at this point I have no clue what the Rigol does with saved files. The Siglent is able to provide timestamps and also includes a free Special Customer Support Package so the choice for me is made.

[Martin72]

The Siglent is able to provide timestamps and also includes a free Special Customer Support Package

LOL, but you´re damn right ! 

[Mr. Scram]

Not sure this has been posted yet (I checked the last handful of pages), but someone has already made a battery pack for the oscilloscope:



https://github.com/LorenceChen/DHO800_BatteryPack

[xrunner]

The Siglent is able to provide timestamps and also includes a free Special Customer Support Package

LOL, but you´re damn right ! 

Oh yea I've used that before - No Charge! 

[hubertyoung]

Today I tried to back up a 32G img. I bought a 64G A2V30 TF card and wrote the backed up img into it. Insert the TF card slot of the machine and turn it on to test the startup speed. But what is disappointing is that there is no visible speed increase in boot speed. So I guess that the boot speed is related to the Android virtual machine startup speed. This is determined by the performance of the soc and has nothing to do with IO.

[AVGresponding]

The Siglent is able to provide timestamps and also includes a free Special Customer Support Package

LOL, but you´re damn right ! 

Oh yea I've used that before - No Charge! 

Yeah, it's a very good free international service with a fast response time!   

[hubertyoung]

The good news is that I used the first batch of DHO804 img image packages to write to the TF card and successfully started the machine. Although there is an offset, it disappears after calibration. Then I replaced vendor.bin with 924 successfully. And there is no offset, even when calibrating again.

[dreamcat4]

The good news is that I used the first batch of DHO804 img image packages to write to the TF card and successfully started the machine. Although there is an offset, it disappears after calibration. Then I replaced vendor.bin with 924 successfully. And there is no offset, even when calibrating again.

Hi, can you please confirm? (for example in menus), that it's now has the CAN bus decode working on 804? Many thanks   

[hubertyoung]

The good news is that I used the first batch of DHO804 img image packages to write to the TF card and successfully started the machine. Although there is an offset, it disappears after calibration. Then I replaced vendor.bin with 924 successfully. And there is no offset, even when calibrating again.

Hi, can you please confirm? (for example in menus), that it's now has the CAN bus decode working on 804? Many thanks   

CAN bus decoding runs on 924, so as long as DHO924 is displayed inside, this function can be used. It is configured by software

[Fungus]

I predict the chances are very good that it can be done.

You mean, you're guessing / wishing.  Meaningful prediction requires data.  There's a difference.

Well, guess what...?

CAN bus decoding runs

[iMo]

..In some you'll see an LC filter between the front-end and ADC with at least a 3rd order response (and addition to the filtering that already happens inside the frontend). ..

There is an LC filter at the input of the ADC in each channel (see below).
Now, how it is tuned, in which BW..
The coils are rather large inductance, like several uH, imho, thus the low pass will be below 100MHz, my bet..

Edit: my wrong estimate as the coils are 0603 smd, so the inductance perhaps 15-50nH..

[Fungus]

Today I'm wondering if the 50M memory size could be raised.

I'm guessing they picked some round numbers that the FPGA could handle without bogging down but the RAM is all shared so maybe the size is completely configurable.

(...as long as we leave enough RAM left over to run the main app)

Maybe we could configure it for 100M and see what happens. 

We need SD card images from all the different models so we can diff them. Is it only "vendor.bin" that changes?

[Fungus]

The coils are rather large inductance, like several uH, imho, thus the low pass will be below 100MHz, my bet..

This is an 804 running the 924 firmware and getting an 800ps rise time:



Ref: https://www.eevblog.com/forum/testgear/rigols-new-dho800-oscilloscope-unbox-teardown/msg4978852/#msg4978852

[Blackflag]

Hello Fungus

This is an 804 running the 924 firmware and getting an 800ps rise time:

Question: which pulse generator did you use for this measurement?

Given that rise time would result in a BW(-3db) of ~440Mhz, right?

Ciao
U

[akkk44]

The good news is that I used the first batch of DHO804 img image packages to write to the TF card and successfully started the machine. Although there is an offset, it disappears after calibration. Then I replaced vendor.bin with 924 successfully. And there is no offset, even when calibrating again.

Could you please post the image somewhere? Is there a way to write the TF card without taking the scope apart? Thanks a lot.

[mawyatt]

If you want to extra correct, that 2.5x figure was based on typical filters and 8 bit sampling. A better filter is needed for 12 bits while retaining the same noise floor as LSB counts. If a better filter existed why was it not used before?

Agree, this gets well into the finer details of analog to digital signal processing, ADCs, and the effects of filter type/shape, aliasing, noise, # of bits, amplitude/phase linearity, ENOB and so on. You can't just slap on a 12 bit ADC in place of a 8 bit ADC, widen the data path and call it a day.

Best,

[hubertyoung]

The good news is that I used the first batch of DHO804 img image packages to write to the TF card and successfully started the machine. Although there is an offset, it disappears after calibration. Then I replaced vendor.bin with 924 successfully. And there is no offset, even when calibrating again.

Could you please post the image somewhere? Is there a way to write the TF card without taking the scope apart? Thanks a lot.

I haven't found anything that can be done without removing the TF card. But for fragile stickers, you can use an air gun to heat them at 100 degrees. Use a knife to remove without damage.

[nctnico]

If you want to extra correct, that 2.5x figure was based on typical filters and 8 bit sampling. A better filter is needed for 12 bits while retaining the same noise floor as LSB counts. If a better filter existed why was it not used before?

Agree, this gets well into the finer details of analog to digital signal processing, ADCs, and the effects of filter type/shape, aliasing, noise, # of bits, amplitude/phase linearity, ENOB and so on. You can't just slap on a 12 bit ADC in place of a 8 bit ADC, widen the data path and call it a day.

And not to forget sampling clock jitter. It would be interesting to have an ENOB versus frequency graph. Chances are the anti-aliasing filter roll-off doesn't even matter between 8 or 12 bit.

[Fungus]

Hello Fungus

This is an 804 running the 924 firmware and getting an 800ps rise time:

Question: which pulse generator did you use for this measurement?

It wasn't me who did it, follow the link under the image.

(My 'scope won't arrive until the end of the month...    )

[hubertyoung]

I haven't found anything that can be done without removing the TF card. But for fragile stickers, you can use an air gun to heat them at 100 degrees. Use a knife to remove without damage.

Yep. Learn to hack those stickers! 

On a related note: Has anybody tried using ADB over the USB connector after enabling debug mode? Does that work?

(hit the "about" button seven times...)

Instead of using USB to connect to adb, you can use the network to connect to adb. The port is 55555. eg. adb connect 192.168.10.193:55555

[Fungus]

I haven't found anything that can be done without removing the TF card. But for fragile stickers, you can use an air gun to heat them at 100 degrees. Use a knife to remove without damage.

Yep. Learn to hack those stickers! 

On a related note: Has anybody tried using ADB over the USB connector after enabling debug mode? Does that work?

If it does we can probably hack this to death without ever opening it up.

(hit the "about" button seven times...)