SOLD (EU/US): Keysight DSO6104L 1GHz scope (used)

[Neganur]

Up for sale is a very nice 1 GHz 4GSa/s 8MPts rack oscilloscope of the 6000-series.
I bought an Infiniium series scope and so this one has to go again.

Original shipping date 2010, from Kista Sweden (Nokia or Ericsson?)

Excellent condition (some scuffs), Options:
mem8M - 8 Mpts of sampling memory (8 for 2-ch pair, 4 all channels on)
MST - measurement statistics, cursor waveform tracking
E00 - enhanced trigger types (FlexRay), Lister

2500 EUR  2200 EUR incl. insured courier shipping within EU
(other places plx ask for shipping, US is ca 100EUR)

• 1 GHz, 4 analog channels, 4 GSa/s, 8 MPts memory, 100,000 wfms/s
• 1U-high 19-inch wide package (ca. 435x230x45 mm)
• Full connectivity with USB, LAN, GPIB interfaces and XGA out
• 16 digital channels with MSO option

Datasheet: http://literature.cdn.keysight.com/litweb/pdf/5989-5470EN.pdf?id=875632

no accessories, just the scope unit.

MSO, serial trigger&decode, Flexray etc. available as options.

pictures incoming in a moment
http://s7.images.keysight.com/s7viewers/flash/genericzoom.swf?logo2=false&serverUrl=/is/image/&contentRoot=/skins/&locale=en&config=Keysight/KEYSIGHT-IMGSET&image=Keysight/PROD-x202268-IS

[Neganur]

Update: added pictures

Edit: updated rise time pictures, the prior ones were with the channels set to 1 Mohm and thus not reflecting the bandwidth of the scope.

Additionally added an FFT with infinite persistance, swept with a signal generator from 10MHz to 2GHz at -7dBm (100mV rms)

[nctnico]

Perhaps also tell how this machine is controlled without buttons...

[Neganur]

Perhaps also tell how this machine is controlled without buttons...

I'd rather go the VNC route!

Well, since you posted in that other thread I guess you know. But here's a link to the manual and the programmer's guide for the InfiniiVision series 5000/6000/7000:

InfiniiVision 5000/6000/7000A Series Oscilloscopes User's Guide
6000 Series Oscilloscopes Programmer's Guide (PDF Version)


Edit:
It also seems you can send keyboard commands via the VNC interface, and I had plans to make a physical rotary encoder interface (or by chance get hands on the DSO6000A front panel).
This could be run by some RPi or Beagle bone or w.e. that can handle the VNC link. But then I got the other scope and don't have the energy to do another side project.

I could also imagine you can just swap out the acquisition board if you have e.g. a DSO6034A and want to upgrade to 1GHz.

[nctnico]

I know about VNC. I was just wondering if there where other local control options I'm unaware off which are only available on the headless 1U units.

[Ice-Tea]

For some reason (read: Agilent probably doesn't want to make it too easy  ) The unit has USB but won't accept a mouse. Bloody shame but still an awesome scope!

[Neganur]

Gee yeah, or having the USB recognise a touch screen board...that would be a winner with the XVGA output.
Best "pico scope" ever

[Keysight DanielBogdanoff]

For some reason (read: Agilent probably doesn't want to make it too easy  ) The unit has USB but won't accept a mouse. Bloody shame but still an awesome scope!

This is a common scope for manufacturing racks. Generally the use is 100% automated, but it functions just like the 6000A sans front panel & buttons. I *think* it has an intensity knob, FYI.

[linux-works]

they are now calling 'sensors' scopes?

this looks like a great sensor pod, a data collector and super duper a/d converter

but no display onboard, no buttons, needs a remote computer to be what we used to call a 'scope'?

interesting..

I like the idea of separating the capture from the UI.  don't get me wrong.  but is this headless box really ok to be called a scope?

[Neganur]

they are now calling 'sensors' scopes?

this looks like a great sensor pod, a data collector and super duper a/d converter

I would think yes!

Compared to an USB oscilloscope (except for the really expensive Pico scope 6000 series), this one has the high sampling rate and the bandwidth of a 'real' scope, and an embedded controller so that it doesn't need the PC to run and will show a waveform with just the monitor attached. Granted, you need something to send commands to it so you may argue that it's a handicap rather than a something good. Also the advantage of having all waveform math, serial decoding and mask testing done by the Megazoom III ASIC is maybe offset by it being cumbersome to set up.

It features a lot of connectivity (USB, LAN, GPIB), the 16-ch MSO, external trigger in/out and 10MHz reference in/out. I would be willing to say that this is something a USB scope could easily do too but they generally don't. Also this one supports tons of nice Keysight probes with the AutoProbe interface.

It has the high resolution mode if you have spare sample rate. It also does equivalent time sampling up to 400 GSa/s if you turn off real time sampling by the way!

So I would not call this a sensor pod or data collector (digitizer), since it really is lacking the bit depth, memory and above all the fast data link compared to dedicated PXI solutions.

As Daniel and others noted this is a full blown 6000-series scope without the display and the user interface knocked off. I would agree that's a disadvantage for a professional using it for R&D on his desk but for a clever electronics person might overcome that by creating a small PCB with rotary encoders, buttons and an MCU that sends a mix of SCPI and VNC commands to control it.

This way it fills very little bench space (mount it under the shelf on your desk like a kitchen radio) and I at least have several monitors around my lab and they all have dual-input capability.


By the way, fire sale, now for 2300EUR incl. (EU) DHL shipping

[jpb]

It is a little off topic, and sadly I can't afford a couple of £k for a new scope, but I'm puzzled by the rise time.

Rise time should be approx 0.35/BW. So my 350MHz scope has a rise time of around 1 nsec when used with the same pulse generator (I recognize the Jim Williams one).

So I'd expect the rise time for a 1GHz scope to be about 1/3 of a nsec, the screen shot seems to show something over 1 nsec?

As a sanity check I just dug out my pulse generator and attach a photo. I think the BW of the 1GHz scope must have been set to a limited mode perhaps?

[Neganur]

Yes I've been thinking about that too. I've come to the conclusion that I had left the channels in 1M mode.

I'll update the pics in a moment

[Neganur]

Pics updated, channels were indeed set to 1M so no wonder the rise time was limited.
Also added a sweep with an sig gen with an FFT set to infinite persistance. Step size is a bit coarse but I guess it works.

The www.keysightoscilloscopeblog.com has a fantastic blog post from Taku Furata, explaining how you can measure your scope's bandwidth if it supports multiple math functions (FFT of d/dt of a fast rise pulse):

https://keysightoscilloscopeblog.com/2016/05/04/how-to-measure-your-oscilloscope-and-probes-bandwidth/

[jpb]

I prefer to do things analytically.

If you make the approximation that the scope input looks like a single pole RC filter (which for low frequency scopes below 1GHz isn't bad) then BW = (ln 9/2.pi)/trise where trise is 10% to 90% as per my attached notes. (I claim no originality of course, this is a very well known approximation but often it is given without any real explanation so I like to work things out for myself.)

Of course, for high bandwidth scopes the approximation of a single pole breaks down and the FFT method presumably is more accurate but then if you can afford a multi GHz scope you probably have the equipment to directly measure the BW if you want to!

[linux-works]

does this run standard old vnc, into an IP addr that this box gets via DHCP?  I am guessing yes.  (I love vnc and use it at home from a windows desktop to my linux back-end boxes).

the funny part about this scope is that the UI really is not expensive in terms of parts.  a display is cheap, even hdmi displays are cheap.  rotaries and controllers are cheap.  a college-hire could design and build one (maker-style) in a summer or less and probably for $100 or less. 

I do like the concept of the data capture box being small and not having to include the UI baggage.  makes great sense.  but for a desktop user, asking them to use a mouse and pc keyboard - that part I'm not in love with.  this is clearly meant for scripting and automation, of course.

[Neganur]

Of course, for high bandwidth scopes the approximation of a single pole breaks down and the FFT method presumably is more accurate but then if you can afford a multi GHz scope you probably have the equipment to directly measure the BW if you want to!

It was more a band-aid actually since my sig gen sweeps so painfully slow. But you got me thinking...I should be able to use the sweep of my VNA.
Unfortunately, it seems the source bands are split at 748 MHz and again at 1.5GHz so I can't do a nice continuous sweep from 10 MHz...2GHz but this was quite fun to set up:

Input is -10dBm = 100mV peak
-3dB bandwidth at 70.7mV peak ca 1160 MHz.

In the second sweep I matched the 750 MHz wide sweep (750...1500MHz) with the full 10 divisions of the screen and used the cursors to estimate the -3dB point.
That's  750 MHz sweep / (2.75ms per div) = 27.273 MHz per mili second
-3dB point is at roughly 15ms so that's 15*27.273 = 409.1 HMz.
Adding the starting point of the sweep gives roughly 750+410MHz = 1160 MHz bandwidth.

Not very accurate either since who knows how good the power level of the VNA is etc. but this was fun I think.

[Neganur]

does this run standard old vnc, into an IP addr that this box gets via DHCP?

Yep, I use mainly the VNC started by the java browser web control but RealVNC works too.
The difference is that the java gui has additional clickable side-buttons (see rise time screenshots) and takes keyboard shortcuts as command input (CTRL-s for single shot etc).
The Real VNC software seems a bit limited with just providing the mouse and I'm not sure if it's just this particular software application. I have not tried any others.

[jpb]

Of course, for high bandwidth scopes the approximation of a single pole breaks down and the FFT method presumably is more accurate but then if you can afford a multi GHz scope you probably have the equipment to directly measure the BW if you want to!

It was more a band-aid actually since my sig gen sweeps so painfully slow. But you got me thinking...I should be able to use the sweep of my VNA.
Unfortunately, it seems the source bands are split at 748 MHz and again at 1.5GHz so I can't do a nice continuous sweep from 10 MHz...2GHz but this was quite fun to set up:

Input is -10dBm = 100mV peak
-3dB bandwidth at 70.7mV peak ca 1160 MHz.

In the second sweep I matched the 750 MHz wide sweep (750...1500MHz) with the full 10 divisions of the screen and used the cursors to estimate the -3dB point.
That's  750 MHz sweep / (2.75ms per div) = 27.273 MHz per mili second
-3dB point is at roughly 15ms so that's 15*27.273 = 409.1 HMz.
Adding the starting point of the sweep gives roughly 750+410MHz = 1160 MHz bandwidth.

Not very accurate either since who knows how good the power level of the VNA is etc. but this was fun I think.

Nice measurements. I find this sort of thing fascinating - in fact I probably waste far too much time and effort measuring my equipment (like getting standard resistors to check my DVMs and so on) than I actually spend using the equipment to do anything useful!

Anyway, there is fairly close agreement with the 0.35/332psecs (=1.05GHz) and your direct measurement of 1.16GHz. If the 1.16GHz represents the intrinsic BW of the slope which gives a rise time of 302psecs and the combined rise time of the pulse generator and the scope is the measured 332 psecs then this implies the pulse generator has a rise time of sqrt(332^2 - 302^2) = 138 psecs. This is probably a bit low (I think my one was measured at 170 psecs) but is in the right ball park.

[linux-works]

quit visionary (lol) of them to use vnc.  that's actually a really smart move.  who made that design decision?  and why is he still working there?  (grin)

I would have guessed some stupid proprietary client that you have to license.  instead, its regular old standards-based (sort of) vnc?  wow, very very cool indeed.

you could easily create 'macros' IF there are keycodes that, when received by the scope, do simple atomic things.  otoh, how would you query it?  I guess the vnc is a 'push only' kind of interface in terms of buttons; meaning, you could automate a button press over vnc but unless you wanted to screen scrape, there's no VNC way to do a GET or READ on a button or knob state.  for that, I guess you'd query the box via scpi and not vnc.  vnc really would be just for user control.

also, regular vnc can run multiple viewers in sync or read-only mode.  so I could see a single 'scope' being displayed at many places at once, with the vnc server handling all the unicast display copying.

[mikeselectricstuff]

quit visionary (lol) of them to use vnc.  that's actually a really smart move.  who made that design decision?  and why is he still working there?  (grin)

I would have guessed some stupid proprietary client that you have to license.  instead, its regular old standards-based (sort of) vnc?  wow, very very cool indeed.

you could easily create 'macros' IF there are keycodes that, when received by the scope, do simple atomic things.  otoh, how would you query it?  I guess the vnc is a 'push only' kind of interface in terms of buttons; meaning, you could automate a button press over vnc but unless you wanted to screen scrape, there's no VNC way to do a GET or READ on a button or knob state.  for that, I guess you'd query the box via scpi and not vnc.  vnc really would be just for user control.

also, regular vnc can run multiple viewers in sync or read-only mode.  so I could see a single 'scope' being displayed at many places at once, with the vnc server handling all the unicast display copying.

Remember you still have a real XVGA display, so you don't need to read anything back, unless you wanted to replicate the illuminated buttons on the original front panel, which you could easily live without.

[Omicron]

This scope also works with the new mobile app Keysight recently released, although the functionality of the app is still pretty basic. You can also use it in conjunction with a 169XX series logic analyser, but curiously if you want to do that you can't activate the MSO option as that will cause the software to say it's an "unsupported model".

Interestingly, according to the service manual the main board in this scope is the same part number as the main board in the regular DSO/MSO6104. The regular models just seem to add the front panel board. So if you can pick one of these up cheaply you could presumably use it to bring back to life or even upgrade a regular 6000 series.

[Neganur]

Has to go.
Price drop

[Howardlong]

Pics updated, channels were indeed set to 1M so no wonder the rise time was limited.
Also added a sweep with an sig gen with an FFT set to infinite persistance. Step size is a bit coarse but I guess it works.

The www.keysightoscilloscopeblog.com has a fantastic blog post from Taku Furata, explaining how you can measure your scope's bandwidth if it supports multiple math functions (FFT of d/dt of a fast rise pulse):

https://keysightoscilloscopeblog.com/2016/05/04/how-to-measure-your-oscilloscope-and-probes-bandwidth/

If you set it to equivalent time sampling, it's even better. This is my 7104B (same as a 6104 but with a big telly)

[jpb]

If you set it to equivalent time sampling, it's even better. This is my 7104B (same as a 6104 but with a big telly)

It may just be that you're using a faster rise time pulse - you're peak-to-peak is much lower which implies a different pulse source. Though equivalent time sampling does give very nice smooth curves.

[Howardlong]

If you set it to equivalent time sampling, it's even better. This is my 7104B (same as a 6104 but with a big telly)

It may just be that you're using a faster rise time pulse - you're peak-to-peak is much lower which implies a different pulse source. Though equivalent time sampling does give very nice smooth curves.

With a <40ps system (generator plus cable) rise time pulse from an HP 54121T using its TDR pulse generator I get the same as you, about 340ps on real time sampling. Switching to equivalent time squeezes out that bit more, 280ps. Real time is only 4GSa/s so in real time there'll be plenty of interpolation artefacts.

With a Tek demo board QuickStart8 which has a <200ps rise time signal, I get almost identical results using the same cable (slightly bizarrely a tiny bit better, Tr=270ps).

I don't know what signal source you're using, it looks like it may be a Jim Williams with a transmission line pulse stretcher at a guess.