Tektronix 2440 PSU Repair

[grtyvr]

I was able to pick up a 'for parts not working' Tek 2440 off of ebay for a good price so, off I go on the repair train.

When I got it home and unpacked and plugged in, there was nothing.  The listing said that it was previously working before it had been put away but they tried to start it up again after storage and no go.  There was impact damage on the case so I am not sure that is the full story.  Anyway, I downloaded the manual and started to walk through the troubleshooting flowchart for the 'no signs of life'.

Checked CR245 and it was fine.  Since the scope is 'of a certain age' I decided to pull C244 180uf since it is early in the startup of the scope.  It had a bulge on the bottom, but that might have been due to the manufacturing method, with an epoxy base.  I decided to pull a few of the others of the same type.  Sure enough one of them was starting to leak.  So I replaced all of them.

But still there is no power coming out on the regulated rails.  So, I connect a variable psu to CR245 and start injecting voltages.  The emitter of Q836 is not behaving like it should so I disconnect the jumper W462 to see if I put in a dummy load then that will make a difference.  Success!  The primary supply is producing ~ 11V.  So what is up with the rest of the supply?

I took a look at the gates of the PWM FET's Q423 and Q421.  Sure enough it is a mess.  Not oscilating.  That means we need to check some IC's and related parts.  Logic is not high on pin 10 of U233 - SG3525A.  But it is warm.  So off to the parts store to pick one up.  After cleaning up after the desoldering of the original, the new one goes in.

Not the FET's gate waveform is well behaved and square like, as well as test waveforms at 189, 190 and 191, which is enough for me to try powering it from line voltage again.  Lo, and behold!  It lives.  All the test points produce the requisite voltages.

So I button it back together and power it up to see what else I have to work on.

It goes through the boot up and self test to report the 4700 errors about NVRAM having low voltage.  I will dig in and check to see if the 5V rail is off, but I doubt it.  I doubt it so much that I ordered some DS1230AB NVRAM.  They should be here by the end of the week.  Then I can try to bring it into rough calibration and see if that clears off some of the higher numberd errors.

Stay tuned.

[David Hess]

The external calibration procedure is pretty easy compared to other oscilloscopes of that vintage.  You could do it just to clear all of the errors while leaving the DSO in an uncalibrated state.

[grtyvr]

As I understand it, I need a few voltage references and a pulse generator to do the external calibration.  I have a square wave generator with an adjustable duty cycle, so i plan to use that as a pulse generator.  The voltage "references" will be via an adjustable power supply.  Am I wasting time trying that?

[David Hess]

That is how I did it the first time.  I used attenuators to get enough low voltage resolution from my power supply and the sync output from my function generator for the fast pulse edge.

[BradC]

The only critical one for a full calibration is the fast rise pulse. I bought a 2440 a few years ago, and then picked up all the gear I might need to calibrate it for the day the nvram dies. It hasn't yet, but I'm sure it will.

I still prefer the 2440 to most of the budget Chinglish scopes I've tried, but my mate who has a Rigol just says that's because I'm a dinosaur.

[tautech]

As I understand it, I need a few voltage references and a pulse generator to do the external calibration.  I have a square wave generator with an adjustable duty cycle, so i plan to use that as a pulse generator.  The voltage "references" will be via an adjustable power supply.  Am I wasting time trying that?

You don't need a lot of gear to do a basic alignment check. Some scopes have a very detailed calibration process that need be followed to the letter and to do it properly you need known accurate gear.

Often near enough is good enough for a hobbyist.

Here's a Tek troubleshooting guide that might have some bits in it that can help you:
http://www.sphere.bc.ca/test/tek-parts/troubleshooting-scopes.pdf

[David Hess]

The only critical one for a full calibration is the fast rise pulse. I bought a 2440 a few years ago, and then picked up all the gear I might need to calibrate it for the day the nvram dies. It hasn't yet, but I'm sure it will.

The fast rise pulse is not very critical; it just needs to be fast enough and the output from a TTL gate is good enough.  The pulse is applied to both channels through a splitter and equal length delay lines and the DSO automatically measures the delay between channels which is used as the calibration constant to align them from that point on.

I still prefer the 2440 to most of the budget Chinglish scopes I've tried, but my mate who has a Rigol just says that's because I'm a dinosaur.

Eh, the 2440 is real time, has a 16 bit processing record for accurate measurements, supports peak detection, and has a high resolution display.  The Rigols are real time anyway.

[BradC]

Eh, the 2440 is real time

Not really. On any of the faster timebases it samples each channel into a pair of CCDs and then clocks those analogue samples out into the 8 bit A/D at a much reduced rate.

[David Hess]

Eh, the 2440 is real time

Not really. On any of the faster timebases it samples each channel into a pair of CCDs and then clocks those analogue samples out into the 8 bit A/D at a much reduced rate.

"Real time" in the traditional sense means bounded time; most DSOs meet this requirement but some like the ancient 7854 do not.  In the oscilloscope sense, it means a high enough sample rate to support its bandwidth without equivalent time sampling although the 2440 does not strictly meet that requirement at its highest bandwidth.  It does however compared to a Rigol where the sample rate is divided between the input channels.

You could make the same argument about many modern DSOs which sample each channel into acquisition memory and then clock the samples out slowly into processing or display memory; the 2440 and entire TDS600 series just happen to use analog acquisition memory.  Tektronix used the term more for DSOs where sample rate was independent of the number of channels used and high enough to support the maximum bandwidth so the 2440 was the ancestor of their real time DSOs for a long time.

[BradC]

Eh, the 2440 is real time

Not really. On any of the faster timebases it samples each channel into a pair of CCDs and then clocks those analogue samples out into the 8 bit A/D at a much reduced rate.

"Real time" in the traditional sense means bounded time; most DSOs meet this requirement but some like the ancient 7854 do not.  In the oscilloscope sense, it means a high enough sample rate to support its bandwidth without equivalent time sampling although the 2440 does not strictly meet that requirement at its highest bandwidth.  It does however compared to a Rigol where the sample rate is divided between the input channels.

Fair enough. I've never heard it described in that way before. Still a beaut little (well, relatively little) scope.