So i have recently gotten my hands on a Agilent DSO9204H scope. Featuring 2GHz of bandwidth sampled at 10GS/s. Mine even came with a 512GB SSD built in to give Windows 7 Embeded that runs on it a nice quick boot time. But it was a bit of an adventure with the software on it.
The software that came on it was fairly old so it was begging to be upgraded. The Keysight website said the latest version works on all 9000 series models so i downloaded the 1.7GB and have it a shot. All went fine and it worked but then i noticed something odd, the scope suddenly went all the way up to 20 GS/s so thats pretty nice, but the memory depth only went to 40Mpts. As far as i know i should have 100M as standard.
So i go look at the about screen to see whats going on with the options and i see no options and its saying i have an DSO9254A. Wait what? I mess around with it some and its pretty determined this is a 9000A scope. Well thats nice since the 9000A series is much more expensive on ebay. When i got the scope i did make a backup image of its SSD so i load the backup on to it and the old software again is saying correctly that its a DSO9204H. I installed the new software again and again it starts thinking this is a DSO9254A. I have no idea why this happens but it is most likely some sort of bug. It does make sense that both the 9000H and 9000A use the same hardware and i have verified both use a mainboard codenamed "Raptor"
But the thing is that the DSO9204H 2GHz bandwidth while DSO9254A is 2.5GHz. Sure the 500MHz is not much of a difference but why not call both 2.5GHz if its the same hardware. I had a bit of a look around the datasheet for both to compare the two and i noticed something else.
Notice that the 2GHz 9000H has the same rise time as a 4GHz 9000A. So i started to wonder what is the actual bandwith of my scope. At first i looked at the noise floor. When bandwith limited to 2GHz it met the noise spec from the 9000H datasheet exactly. But when not limited it was noisier than the 2.5GHz model should be but not quite as noisy as the 4GHz model. Since i didn't have a 4GHz source i decided to build myself a Jim Williams picosecond pulse generator to measure the rise time and get the bandwidth from that. I half assed put together one on the back side of an SMA connector.
Turned out the MMBT2222 transistor rated for 40V needed 180V to break it down so the output signal was pretty massive at 60V so that had to be run trough an attenuator to bring it down and messing about with it i couldn't get a pulse faster than 160ps from it, but then i found that a guy built it on a fancy PCB and got 150ps measured on a 32GHz bandwidth scope. Clearly this circuit is not fast enough to do what i need. Finally i went and bought a cheep 4.4GHz synthesizer that runs from USB on ebay. It claims to be calibrated and it does come with a calibration table on a CD with the units serial number so it should be flat-ish over its entire output range. It also came with a nice rigid coax SMA cable that i made use of too.
This now worked pretty well and it did spit out a 4.4GHz signal as advertised.I set up my scope for FFT and turned on infinite persistence while i made the synthesizer do a sweep from 100MHz to 4.4GHz. The result is interesting. The scope seams to be perfectly flat up to 3GHz but then it starts doing a strange dance up and down. By definition it is not a 4GHz scope since it dips below -3dB well before that. I was expecting to see it roll off as scopes normally do but here it actually climbs back up a bit at 4GHz. At this point im not even sure what to call it. Is this normal to see in high bandwidth scopes and it perhaps just needs calibration?
I have also tried it using different cables and in to different channels of the scope to see if it makes any difference. Those can be found here:
imgur.com/a/8fWn7 Turns out channels 1 and 3 have a similar response while channels 2 and 4 have a different similar response. Changing the signal amplitude didnt seam to make a difference, but interestingly this scope does not loose any bandwidth on the 1mV/div scale.
And finally the new software owes me extra 60 megs of memory that i used to have, so i went got it back with some interest on top too, unlocking the full 1Gpts of memory. Along with that came a few other useful features...well actually all of the optional features.
As one would expect the unit now thinks its an MSO9254B ....wait...what B? There is only an A model. Right?
Well i have no idea why and how. Perhaps its a future product or something?
In any case if you are thinking of buying a 9000A scope STOP right now. Buy a cheaper 9000H and turn it in to a 9000A yourself, perhaps getting extra bandwidth along the way.