Tektronix 2440 300MHz Dual Channel Digital Storage Oscilloscope
Tektronix 2440 Digital Storage Oscilloscope
During a very unhappy period, I ended up buying a broken Tektronix 2440 DSO to hopefully cheer up... It was a "For Parts or Repair" unit.
Specifications:
Introduced 1988
300 MHz bandwidth, full digital sampled oscilloscope
Graticule is embedded in CRT face
6809 CPU, 32KB battery backed RAM for storage, 1024 point per channel CCD
storage (memory depth)
Custom 16 bit VLSI waveform processor
2 input channels plus two trigger inputs
Estimated screen resolution: 500x400, 8-bit ADC (256 levels).
Internal 50Ω termination selectable on Y axis inputs
X and Y axis measurement cursors, along with software computed measurements
like rise time, frequency, RMS voltage, period, etc.
X/Y mode (channel 1 is X axis) - there is no Z axis input so no raster pictures
possible like on the Tektronix 2465 :(
"Sweep" from 5s to 2ns/division. Slower divisions can be done via side scrolling (Roll) or overlay (Auto)
Automatic setup (press button and scope tries to figure out best settings to view signal attached to the probes)
Adjustable trigger level with AC/DC/HPF/LPF/auto coupling, ±slope,
trigger on selected channel
GPIB communication port
I acquired this from eBay as "For Parts or Repair." Not quite the freebie I hoped but it wasn't way too expensive - actually I ended up paying about as much as the parts cost to repair my Tektronix 2465. One of the main reasons for getting this scope was that the service manual was actually posted on Tektronix's website. This helped immensely as it shows how the scope should work. In any case, the scope showed some signs of life but worked pretty miserably. Waveforms can be shown but the knobs were doing something strange. Though I could see some waveforms, they were far from being correct.
How the selftest looked when I got it:
Fail! Fail! Scope indicates hardware problem :-(
(C) TEK, 1988, ALL RIGHTS RESERVED
14-JUL-89 V2.24 / 2.5
0000 EXTENDED-DIAGNOSTICS FAIL
1000 SYS-ROM PASS
2000 REG PASS
3000 SYS_RAM PASS
4000 FPP FAIL
5000 WP PASS
6000 CKSUM-NVRAM* FAIL
7000 CCD FAIL
8000 PA FAIL
9000 TRIGS FAIL
HARDWARE PROBLEM -- SEE SERVICE MANUAL
RUN ONCE MENU OFF TO EXIT
^ v RUN/SEL MODE HALT
0000 - All tests
1000 - System ROMs
2000 - Register read/write test
3000 - System RAM (nondestructive read write)
4000 - Front Panel Processor and battery
5000 - Waveform Processor
6000 - NVRAM Checksum for corruption
7000 - Charge Coupled Device acquisition memory
8000 - PreAmplifer and Attenuator
9000 - Trigger
Starred tests were not run during last diagnostic run.
Note: On a perfectly working 2440, it shouldn't show this diagnostic pass/fail screen at all on startup. It will only show the pass/fail screen on cold startup if it fails a selftest (and if the CRT works) or calibration is needed. You can always manually bring up selftest menu with the "menu off/extended functions" button and then hitting the soft "EXTD TEST" button. With all selftests passing, the oscilloscope is ready from power button push to traces shown in 10 seconds on 120VAC power.
I quickly found the easy part: The battery on this thing was clearly dead and truly is a hardware problem. It has a 3.5V 1.6Ah Lithium Keeper II LTC-16P-CO-F-32 primary battery soldered into the processor board. Based on the data sheets upon typical usage, I calculated about 45 years of battery life to power the two Sony CXK58257AP RAM chips - though apparently it's much worse as it's completely dead before 30 years. This is probably because the shelf life of the battery is much lower than 30 years.
Swapping the Lithium Keeper II with 3V from two AAA cells in a plastic battery
holder allowed the retention of the contents of the memory, alas, does not
tell the FPP self test that the battery is now capable of retaining memory
as the battery is still below the threshold of 3.6V. So then I tried
4.5V from three alkalines, alas it triggerred the "voltage too high" error
and continued to fail the selftest.
So, I will need to go look for a 3.6V battery. For now I'm
running from three NiCd cells to get past the battery error. Self discharge
of the NiCd cells means this is not a longterm solution.
I bought a 3.6V Lithium Thionyl Chloride 14500 cell to try as this is the same voltage as the Keeper II. I used a plastic single AA cell holder from Radio Shack (had to hack it a bit to get contact with the 14500 cell). I placed the battery and the holder in a small plastic bag (in case it leaks, but did not tightly seal the bag) and tucked it under a ribbon cable - this should be good for at least 5 years if the Keeper II lasts 3. After adding this new Lithium battery, the scope now PASSes the FPP battery test.
The battery wasn't the only problem: after testing the scope a bit more, I wanted it to display what was showing on my Eico 324 signal generator as proof it could at least deal with 150MHz signals on both channels. As I don't have a pulser to measure rise time to indicate bandwidth, this was the only way to check high frequencies. However upon switching into the faster time/division, the scope violently shrunk and blinked the contents of the display, dumped out garbage on the trace, clicked the relays to switch to GND coupling (it thought it was in the internal 50Ω termination mode and overloaded - and rush job to get it kicked out), and eventually the software crashed, blanking the screen and blinking the trigger LEDs! Sometimes the CCD self test would fail too on startup! Ouch. This may be a long road. This was probably why this scope was shelved and this let the battery run out.
However, there are a lot of things that do work, as long as I don't sweep faster than 200ns/division. This is 1/100 the fastest sweep speed on this scope of 2ns/division, making this scope slower than my Vu-Data as it can go down to 100ns/division.
No. This is bad, my Vu-Data is faster. I have to fix this problem...
Using my newly repaired 2465, I noticed that the +5V that's used for the analog circuitry (the +5V for the TTL/digital circuitry looks fine) had some really nasty spikes in it. In fact it drooped immensely when using 200ns per division or faster though 100ns/division tends to crash the scope really quickly. After a bunch of probing of the PSU circuit with the 2465 (ohhhh a DSO would have been so helpful! oh wait...that's what this scope was for, gosh darn it), I found a bunch of voltages that were out of spec. Looks like voltages drifted over time, possibly caused by a bad capacitor. I temporarily repaired it by changing the current limit voltage by adding a resistor to allow a bit higher current - and now the 2440 works perfectly. I will have to revisit this a bit later when I have other stuff to replace, namely the noise filter X caps that look kind of cracked and sad in the unit. In any case the CCD error that showed up from time to time now is showing as PASS!
Well, it wasn't quite all hunky-dory yet. That CCD failure showed its ugly head again after leaving the scope on for a few hours. After significant debug, it seems that one of the 470Ω resistors that pulls the clock down to -5V when not asserted is intermittent. It caused some weird behavior on sampling on channel 2, not to mention causing selftest fail. Wiggling the resistor seemed to have fixed it, though will have to replace it at some point. Once more I get all self tests to pass.
I guess my Tektronix 2465 has now fulfilled its duty to help repair something (and it was instrumental - I don't think I could have fixed the 2440 without it), and all this in return for the favor my ps941b scope did to repair the 2465. The 2440 now probably will end up being the workhorse as it is a true DSO - and also being the first DSO that I've ever acquired. I've had many other analog oscilloscopes in the past, the 2465 would be my fifth though only have two now.
Eieew. That's the scope probe calibration output at 20ns/div?
This scope is old by today's standards, but I was quite pleased with it. I've had even older scopes so this is actually very new technology to me. Despite having only 500M samples/sec this is on both channels simultaneously whereas on modern scopes it needs to divide between channels. The 1024 sample depth is however a real issue, though not completely unusable in any sense. It also has 32KB of nonvolatile (using battery) storage to store waveforms that can be used to compare with recently acquired waveforms. With the waveform processor, it can calculate the RMS value of the waveform, peak voltage, frequency, and a lot of other interesting facts of the signal presented. It however does not have FFT mode to do spectral analysis. The waveform processor I'd guess eats around 6 watts and has it's own small heatsink.
The Tektronix 2440 uses two CCDs - yes the same technology as cameras - to capture analog waveforms. The CCDs are internally 4-way interleaved to get the full 500Msamples/sec. A 500KHz clock is clock multiplied to 250MHz within each CCD, and a new sample is taken every phase, resulting in 500MSa/sec. Each CCD is associated with an input and are not shared between them. It would have been nice if Tektronix had an option to daisy chain the two CCDs to get a 2048 entry depth capture for one channel, though even at 2048 entries it's still kind of shallow. The two CCDs are on the bottom of the scope with the different kind of heatsink than the usual Tektronix "flower" heatsinks. These chips get quite hot during use, so do not operate the scope without the case for long, or without a fan - so deep warning - don't let them overheat. They eat around 15 watts a piece I'd guess by how fast they get hot without a fan.
The digitizer uses a bipolar flash ADC, a MC10319. It runs at 4MHz though newer versions of the same chip can run at 25MHz. It's located on the timebase board on top. At 4MHz, it samples the CCD outputs slower speed than capture in most modes (Fast In Slow Out), which is then transferred it into system RAM. The slower time/division selections will put the CCD in bypass mode and the ADC can sample in realtime. Anything slower than or equal to 2MSa/sec, since the same ADC is alternately used for both channels, can be sampled in real time. The plastic packaged ADC only dissipates 0.6 watts unlike the CCDs, so it does not get that warm.
The custom waveform processor can then grab this data to do calculations. Everything is processed by working with the ADC output, this scope does not have any analog mode like the 2200 series digital scopes, and does not even have a delay line! The CCDs function as the delay line instead, along with being the storage device.
One strange fact: the 2440 is physically larger than the 2465 by 2 inches depth. I thought they were the same size at first glance but it was clear when I put them on top each other. I first noticed something was weird when the scope handle reached back "further" on the 2465 when tilted out of the way to the back. Then again, stacking them have always been an issue with Tektronix owners with those handles. I guess sometimes I'll just leave them sticking out front, leaves a bit of a ledge one can let cables dangle over temporarily.
Tektronix 2465 (top) and 2440 (bottom) displaying a signal from my Eico 324
signal generator.
Yes, I should have used "envelope" acquisition and make it look like the
2465.
The signal is not a lie... or at least the 2440 and 2465 agree! (Running a
crystal oscillator on a 74F series IC on a solderless breadboard. See the
cycle is exactly 10 divisions, and it's 10ns/division, 1/100ns = 10MHz!)
The 2440 versus my 2465? Both have their advantages! Having 4 inputs is a blessing at times, and storage for one off events is also helpful. However one thing that really shows up as annoying on digital scopes like the 2440 is aliasing. Many times I try to show a fast signal, if I don't get the time per division right, it may show a waveform but it's way off its true waveform due to aliasing - the sample rate is not high enough to truly get a view for the signal. This makes the waveform processor get confused of the true frequency and may trick people to thinking it's an unstable signal when it's just an alias for the true signal. Analog scopes don't have this issue - you just get a blur that instantly tells one to decrease the time per division to see the signal properly.
One thing that confused me is that though the 2440 samples at 500M sample/sec, how could it be a 300MHz scope since it's not enough for Nyquist (need 600MSa/sec)? Well, yes this scope does have problems at 300MHz. The Sampling Rate (SR) indeed is not high enough - but it does try to do estimation of signals when the signal is beyond the SR by incorporating more samples in "REPET" mode. This slows down the signal acquisition a bit to get a stable waveform. So yes you need to take the acquired sample with a grain of salt when the signal bandwidth exceeds the Usable Sampling Bandwidth (annoyingly abbreviated "USB") that is indicated on the scope. It also does mention in the guide that 200MHz is the highest one shot storage mode bandwidth by Tektronix's formula "USB=SR/2.5", and since everything goes through the CCD storage, this is the technical limit without oversampling. In summary, the analog section of this scope most definitely will handle 300MHz (and higher) signals just like the 2465 but the digital section requires some trickery to capture full bandwidth.
And the screen filter I need to get/make someday:
looks like it's 104mm x 87mm (4⅛" x 3⅜").
thickness is 0.8mm (1/32")
As an aside, implosion shield: 113mm x 93mm (4⅜" x 3⅝"),
thickness is 1.6mm (1/16") laminated