Keysight 3000T scope 1 GHz(or more) hardware upgrade mod

[TheSteve]

My Keysight MSOX3024T was born with 200 MHz of bandwidth, it now has 1 GHz.

There have been plans in the works to upgrade a 3000T series scope to 1 GHz for a while. The 1 GHz upgrade follows in the footsteps of the 3000A series 500 MHz upgrade found here - https://www.eevblog.com/forum/testgear/dsox2000-and-3000-series-licence-have-anyone-tried-to-hack-that-scope/msg987126/#msg987126

Here are the main points of interest regarding the upgrade:

#1 - This upgrade only applies the "T" series as the 500 MHz and below 3000A series scopes have a different PCB that cannot be populated with the required parts
#2 - I think I was borderline crazy to perform this upgrade as there were quite a few unknowns. Parts list or not think long and hard before attempting this as a 3000T series scope is far from cheap and you'll lose any/all calibration and warranty on it. My scope channels didn't work on the first few attempts as every part that needed to be swapped was not known.
#3 - The mod requires installing/replacing/removing 168 components, the smallest being 0402 in size.
#4 - You will need decent soldering/rework gear and skills to perform the mod and should have a hot air station
#5 - I was able to purchase all of the needed parts for under $200.00 USD including shipping/taxes
#6 - There are no modifications to the firmware to perform this mod, no changes or patches to the flash memory are needed and
there are no hacked licenses etc. You don't require a network card.
#7 - Any licensed features you already have will still operate just fine, including the MSO feature
#8 - The scope's model # and serial # will not change, only the reported bandwidth is updated
#9 - The mod is reverseable if care is taken to save/bag each component(highly recommended)
#10 - You will need the required BNC cables and splitters to perform a User Cal.
#11 - The parts list was compiled from pictures and component removal/measurement. I can't say the measured values are 100% exact and am not responsible if they aren't
#12 - If your scope blows up, runs away, makes you bleed(mine did) etc from the mod I'm not responsible.
#13 - All channels have a -3dB point greater then 1 GHz as they should
#14 - The noise floor closely matches a factory 3104T scope
#15 - User Cal passes with no issues
#16 - The 20 MHz BW limit works just fine after the mod
#17 - The 50 ohm path is much improved with the updated front end and use of the Teledyne 20 dB attenuator relays

Now this upgrade wasn't performed without some struggles. There were some missing parts in the earlier research so nothing worked at first. I identified those and kept going.
I modified one channel at first to see if it looked like it would work. The trigger levels and such were way off but I expected that. The bigger worry was that on some volts/div with 50 ohms enabled the values were double what they should be. That seemed like a lot for a UserCal to compensate for. Turned out to be a non-issue as the front end attenuation switches in at 200mV/div for a 1 GHz model and 500mV/div for 100-500 MHz models. So a UserCal handles the problem just fine.

The biggest issue though is that the amplitudes of each of my channels was not matched as the frequency got higher. By the time I had reached 1 GHz the channels varied by 5-10%.
I spent considerable time attempting to determine why this is. I swapped many parts and experimented but was not able to pin it down to any one component. The difference does not appear to be related to passive component tolerance. It could be Keysight has additional calibration for higher frequencies, it could be the front end ASIC varies. It could be some mystery component that was not measured/changed. Ideally the schematic of the entire front end should be drawn up which would make for much easier analysis.

In my specific case channel 1 had the highest amplitude, channels 2 and 3 matched each other and channel 4 was in between. The scope was 100% usable and when you're looking at high frequencies it is doubtful you're trying to determine the exact amplitude using a scope with an 8 bit AtoD. It still bothered me though so I had a play with LT SPICE and determined that changing the value of C1 slightly would alter the low pass frequency the nicest. To that end I changed the value of C1 to 2.4 pF on channel 4 and 3.3 pF on channel 1. This gives me amplitudes within 10-15mV between all channels when looking at a 1 Vpp signal at 1 GHz. The 3 dB cutoff frequency of all channels is between 1060 and 1070 MHz. At 1 GHz the attenuation is about 1.5 dB. Seems like pretty darn acceptable performance to me.
Another forum member will be modifying his scope in the near future, so we will see how his does.


The following forum members helped make this mod possible:
memset
HardDrive
mikeselectricstuff
TopLoser
Howardlong
adranp
HighVoltage
EEVblog(Dave)

And thanks to Keysight for making a cool scope that can do so much - please don't hate me!

Here are some pictures:

1 GHz signal @ 1Vpp in normal mode:


The same 1 GHz signal with some averaging:


The about screen after the mod:


Rise time of a 10 MHz square wave before the mod:


Rise time of the same 10 MHz square wave after the mod:


A pulse:


Input SWR of the 50 ohm path for both attenuation paths before the mod:



Input SWR of the 50 ohm path for both attenuation paths after the mod:

[TheSteve]

400 MHz ECL internal clock from an Agilent 33250A AWG measured with a 1134A 7 GHz differential active probe:


Square waves:


1 GHz signal with 3Khz 100% AM:


My updated front panel label:

[TheSteve]

Here is the parts list and matching pictures

Note all part identifiers reflect designators used in the the previous 500 MHz mod and diagrams where applicable.

The following parts are listed per channel(so you'll need 4 of each for a 4 channel scope) I have included the part # of the exact parts I used, feel free to select your own.

R1 110 ohms 1206 CRCW1206110RFKEA
R2 162 ohms 0603 CRCW0603162RFKEA
R3 100K 0805 RC0805FR-07100KL
R4 23.7 ohms 0603 CRCW060323R7FKEA
R5 23.7 ohms 0603 CRCW060323R7FKEA
R6 50 ohms 3024T only(move to position R7)
R7 50 ohms 3104T only
R8 68.1 ohms 0805 CRCW080568R1FKEA
R9 51.1K 0603 CRCW060351K1FKEA
R10 51.1K 3104T only CRCW060351K1FKEA
R11 51.1K 3024T only CRCW060351K1FKEA
R12 0 ohm (10 ohms in MSOX3024T)

C1 2pF 0603 NPO GQM1885C2A2R0CB01D ***See my comments in the first post***
C2 2pF 0603 NPO GQM1885C2A2R0CB01D
C3 4pF 0603 NPO GQM1885C2A4R0CB01D
C4 4pF 0603 NPO GQM1885C2A4R0CB01D
C5 10pF 0603 NPO GQM1885C1H100GB01D
C6 10nF 0603 GRM188R71H103KA01D
C7 10nF 0603 GRM188R71H103KA01D
C8 100nF 0603 GRM188R71H104KA93D
C9 100nF 0603 GRM188R71H104KA93D
C10 100nF 0603 GRM188R71H104KA93D
C11 100nF 0603 GRM188R71H104KA93D
C12 100nF 3024T only
C13 10nF 0402 GRM155R71H103KA88D
C14 100nF 3024T only
C15 100nF GRM188R71H104KA93D


L1 15nH green dot Coilcraft 0603CS-15NXGEU
L2 6.8nH red dot Coilcraft 0603CS-6N8XGEU
L3 6.8nH red dot Coilcraft 0603CS-6N8XGEU
L4 47nH brown dot Coilcraft 0603CS-47NXGEU
L5 6.8nH red dot Coilcraft 0603CS-6N8XGEU

Q1 IRLML2402 SOT23 (original is MMBF0201NL but that is discontinued)
D1 HSMP-3892 SOT23 HSMP-3892-TR1G
CR1 BAV99 SOT23

BD1 - ferrite bead 0805 Wurth 742792040

K1 - Teledyne A150-20-12 - 3104T only - This is the money part. I bought mine from a trusted vendor in China and verified they were 100% legit before install.
K2 - AGQ210S4H 3024T only (I recommend carefully removing using hotair)
K3 - AGQ210S4H 3104T only
***I don't recommend planning to re-use the removed AGQ210S4H relays - buy new ones***


These parts only require 1 of each for the mod:

Frequency Strapping resistors
R13 (L1) - 64.9K 0603 for 1 GHz CRCW060364K9FKEA
R14 (H1) - 100K 0603 for 1 GHz CRCW0603100KFKEA

U1507 74ACT04 SOIC(3.90mm width)

[TheSteve]

Pictures of the PCB front and back with the parts swapped/installed:



I also want to show a picture that demonstrates just how epic the HP/Agilent 1152A 2.5 GHz active probe is.
In this screen shot there is a reference image in the background showing a 1 GHz signal being fed directly into the scope. The same signal was then fed into a SMA 50 ohm termination. I am probing the termination with the 1152A active probe. You have to love how close the waveforms match even though the frequency is 1 GHz.

This is the termination being probed by the 1152A. The termination is an SMA chassis mount jack with 3 0603 150 ohm 0.1% resistors soldered on the end. Using 3 resistors seems to give the best return loss figures.


UserCal hardware:

[Muxr]

How do you sit down to use that scope with balls that big? Epic mod. 

[mrpackethead]

Really good work there.

Are you intending to publish the hack? ... I think i just need to be paitent. I see more stuff beig posted above.

[BravoV]

Subbed ! Thanks for sharing !   

[TiN]

Holy batmans that's a lot of steps. Well done, I'm almost became a time-nut reading all this  .

 

[VK5RC]

Absolutely amazing, lost for words!

[Jwalling]

Absolutely amazing, lost for words!

Same here! An incredible amount of work and perseverance pays off BIG! Nice job, Steve!

[HighVoltage]

You made it !
Just amazing how you pulled this one through.
Congratulations.

[Zucca]

Impressive!!!!

Crap now I need to buy a 3000T, anyone want a 500MHz not T?

[cgroen]

Impressive!
I have a MSOX3024T, but I'm afraid I lack the balls to do the modification 

[Dubbie]

I really am keen to try this but I am not sure I have the courage. Can you do it to a single channel? Or is is all or none?

[EEVblog]

 

[HighVoltage]

#13 - All channels have a -3dB point greater then 1 GHz as they should

You asked me before, to measure the -3dB point of my 1 GHz MSOX3104A
Here are the pictures of the measurements
I am using my trusted 3.2 GHz Agilent 8648C for a sine wave output, 50 Ohm in 50 Ohm
My MSOX3104A has the -3dB point at 1.16 GHz

[GlowingGhoul]

Incredible work. A huge boon for the hobbyist, and I doubt this will impact commercial sales one iota.

Seeing as how this has a BOM of only $200 retail, it really makes the market segmentation pricing seem absurd at this point.

Next generation entry point of the 3000 level scope should be at least 500Mhz, with a path to 2Ghz or higher.

 

[Keysight DanielBogdanoff]

Incredible work. A huge boon for the hobbyist, and I doubt this will impact commercial sales one iota.

Seeing as how this has a BOM of only $200 retail, it really makes the market segmentation pricing seem absurd at this point.

Next generation entry point of the 3000 level scope should be at least 500Mhz, with a path to 2Ghz or higher.

 

Impressive mod! It probably won't impact commercial sales like you said, but it's a great pathway to 1 GHz for guts folks.

A 1 GHz scope price has as much to do with design and NRE as BOM, and that tech trickles down into the lower bandwidth models. So, if we dropped prices on the high bandwidth models, the entire line would stop being feasible. Designing a 200 MHz scope is much, much easier than designing a 1 GHz scope. And, as is made clear by the mod, the whole backend of the scope is capable to 1 GHz (or higher  )

I wouldn't bank on a 3000 level scope starting at 500 MHz for a long, long time. I won't go into sales details, but for basically every line of scopes all the way up to the top, we sell a lot more of the lower bandwidth models than the higher bandwidth models.

[mrpackethead]

I wouldn't bank on a 3000 level scope starting at 500 MHz for a long, long time. I won't go into sales details, but for basically every line of scopes all the way up to the top, we sell a lot more of the lower bandwidth models than the higher bandwidth models.

Maybe because everyones hacking them. :-)    Probably because in all reality your average workshop/lab doe'snt need too many 1Ghz Scopes.   We have 4 100Mhz scope and 1 500Mhz Scope. The 100Mhz scopes are the workhorses and do 95% of the work..  They are cheap enough that you can leave them sitting on a bench, and not have to move them around.    the 1 500Mhz scope is on a trolley and gets moved to where its needed.    Not sure if this model scales out, but it works for us.

[philpem]

I wouldn't bank on a 3000 level scope starting at 500 MHz for a long, long time. I won't go into sales details, but for basically every line of scopes all the way up to the top, we sell a lot more of the lower bandwidth models than the higher bandwidth models.

Bandwidth: if you need it, you really need it. If you don't? Well... save a bit of brass and maybe buy SPI/UART Decode and MSO 

[electrolust]

Seeing as how this has a BOM of only $200 retail, it really makes the market segmentation pricing seem absurd at this point.

Not at all.  The BOM cost is not in the noise, but it's one of the smaller factors for a product like this (sales numbers in the thousands per year, not millions, with high NRE costs).  As something of an example, I do high grade security stuff, and folks that make EAL4 and up secure elements won't even talk to you until you're in the 100k of volume and at that volume you're a one-off.  They can't sell you unit quantities of chips even if they wanted to, because you can't afford the NRE.  When you're doing 1M+ of volume, that's when your NRE is spread across significant units and when BOM cost starts to be of primary importance.

If anything, it reflects how much of a bargain it is to buy the smaller bandwidth version of the same scope.  All the same engineering had to go into the product.  All the support costs are the same.

[electrolust]

This is amazing!!  I wouldn't do this myself, but just wow!

[lukier]

I wouldn't bank on a 3000 level scope starting at 500 MHz for a long, long time. I won't go into sales details, but for basically every line of scopes all the way up to the top, we sell a lot more of the lower bandwidth models than the higher bandwidth models.

That is at the point of sale I guess, but what about the BW upgrades? How many of these lower bandwidth ones end up getting a BW option?

[electrolust]

And, as is made clear by the mod, the whole backend of the scope is capable to 1 GHz (or higher  )

The gauntlet has been thrown!  OK next level mod is to take this thing to 1.2 or 1.5GHz!!

[electrolust]

Can you characterize the response curve?  My memory might be bad, but as I recall from the literature, the sub-1GHz models have gaussian response and the 1GHz (as shipped from Keysight) has maximally flat response.  Apologies if the posted traces show that fact, I'm a bit excited and haven't read through in detail!