Tektronix 2235 repair thread

[bd139]

Another Tektronix repair thread. Thought I better break this off the TEA stuff so it's easier to find if anyone else comes across it.

Bought a 2235 from a hamfest type event for a whopping £25. Described as working, which it mostly was for a bit so no complaints there. Powered it up, seemed ok bar a couple of issues. Decided to do a full refurb on it as these are known to be occasionally dodgy as hell.

Pretty good condition:



First step I did was empty it of the ageing electrolytic capacitors and mains filter:



New high quality Panasonic and Rubycon parts were substituted in. New schaffner filter installed. I replaced the RIFAs with NEW Kemet RIFAs. A sticker will go on the back of the unit saying these need to be replaced in 10 years. They are consumable, not dangerous.

Powered it up, all good. For a bit. About 30 minutes after power up, the IRF710 in the preregulator blew it's arse out.



Took the driver transistor out and a diode.



After this explosion, I tested the pre-regulator mostly worked by applying 22V across C925 to see if the SMPS controller was still working. It was. I applied 43V across TP940/950 to bring the scope up on DC and it came up fine.


Outstanding issues

1. Pre-regulator exploded

The pre-regulator MOSFET Q9070, an IRF710 exploded. This took out Q908 and CR907 too.

Q9070 is being replaced with an IRF820. Crap by modern standards but cheap and with a higher voltage headroom and similar characteristics to the original device without being too daring. Also known compatible as some units shipped with them in.
Q908 is being replaced with a PN2907A.
CR907 is being replaced with a STTH2R06 ultra-fast rectifier.
As a precautionary measure C907 is being replaced with a Vishay MKT1813 1uF 630V unit

I will lap solder the connections onto Q9070 and shrink it as the original connector was damaged by the explosion.

RESOLVED. Substituted above parts in and it worked.

2. Noisy 100V line

This is a cock up on my part. There is a lot of switching noise there, above min specification. I used a standard cheap ass 33uF cap on that instead of a low ESR type. I will replace that once I get the preregulator working.

RESOLVED. Replaced capacitor with low ESR one and it's fine now.

3. Bit of intensity flickering

Not sure what this is yet. Intensity flicks in and out a little bit. I have done some googling and looks like it may be the 510K resistors in the HT chain (known problematic) or the z-axis or the multiplier. If it's the latter I'm going to cry. This was occurring before I did any refurb work.

RS don't have any suitable replacements until mid December so I will shelve that problem until then.

If anyone has some advice or ideas here I'd appreciate it.

UPDATE: 510K resistors replaced. Unfortunately this made no difference. Have since isolated to the multiplier and have ordered a replacement one.

UPDATE: multiplier replaced. Improved but is still happening. Am going to inspect all remaining resistors in the HT section.

4. Won't show channel 1 on its own

Only shows both or channel 2. Likely switch issue. To debug...


Reference - power supply schematic:

[David Hess]

You are lucky that the failure of Q9070 did not propagate through the drive transistor and diode into the switching regulator controller.  I would have socketed and replaced the TL594 as well.  I probably would have replaced C907 and CR907 also.

It would be nice to know why Q9070 failed.  If you have the ability, taking a close look at the current and voltage waveforms may reveal something.

I think Tektronix stopped using those sockets for the TO-220 switching transistor and started soldering them to the wire extensions after they identified the socket as a reliability problem.

Something else I have wondered about in that design is the wire extension on the TO-220 power transistor.  Sufficient inductance could cause poor switching or high peak gate-to-source voltage.  Lead dressing would have a significant effect on this.  I would consider braiding the wires to minimize loop area but measurements should be made to find the best configuration.

Excluding the 2213/2215 models with the preregulator modification, the 2235 was the first model of this series of oscilloscopes.  One of my ongoing projects is to document the component changes between it and the last version of this design which was used in the 2235A, 2236A, and 2232.  Some of the changes were to support higher power, some were to replace obsolete parts, and I assume some were to improve reliability.

I do not have significant diagnostic data on this power supply design yet because none of my 22xx power supplies have failed or misbehaved and I am leery of messing with a working instrument.

[bd139]

Very lucky indeed it didn't take anything else out. Replacement controllers are available for cheap next day free delivery here so I figured it wasn't worth replacing until I know I've killed it. If it does go I will replace it and add a socket. It was outputting ramp and PWM outputs at around the correct frequency once the turn on voltage was exceeded so I figured it probably works.

I have my suspicions with Q9070 that I actually damaged the heatsink bond when I was replacing a capacitor if I'm honest. This is based on the amount of time it took to explode. The SOA rapidly shrinks as case temperature rises on that part which it does pretty sharpish due to RDS(on). I wasn't overly happy with the tightness of the SOA on it either. The thing needs to deliver quite a bit of energy to C940 to keep it charged, possibly close to the limit. I didn't measure it. I will this time.

Could be right with Vds which is a very tight 400V. One reason I picked an IRF820 as a replacement was the Rds(on) is about the same, higher Vds (500V) and gate charge is low enough and Ids is higher. I hope that will give it a little more headroom.

I hope yours don't explode. While this is fun I rather wanted to use the scope

I didn't include this before but here is a high resolution reference shot of the power supply before refurb. May be of use to you:



Date codes point to around 1985 I think for this unit.

Edit: also wondered about mains voltages. They have risen considerably here over the last few years. Mine peaks at around 255V RMS whereas it used to be around 235V RMS. One reason I refurbished the entire front end.

[David Hess]

I have my suspicions with Q9070 that I actually damaged the heatsink bond when I was replacing a capacitor if I'm honest. This is based on the amount of time it took to explode. The SOA rapidly shrinks as case temperature rises on that part which it does pretty sharpish due to RDS(on). I wasn't overly happy with the tightness of the SOA on it either. The thing needs to deliver quite a bit of energy to C940 to keep it charged, possibly close to the limit. I didn't measure it. I will this time.

That would explain it.  Many other but not all reported failures may be due to the same problem.

Could be right with Vds which is a very tight 400V. One reason I picked an IRF820 as a replacement was the Rds(on) is about the same, higher Vds (500V) and gate charge is low enough and Ids is higher. I hope that will give it a little more headroom.

I have some not quite reliable third party notes on what transistors were used as various times.  Eventually Tektronix used the MTP6N60E which is 600 volts, 1.2 ohms, 6 amps, and 125 watts for all models.  The transistor characteristics are not critical as far as I know.

The power supplies operate in continuous conduction mode so the peak switching current is only like 1.5 amps for the 40 watt 2235 and 3 amps for the 80 watt 2232.

I didn't include this before but here is a high resolution reference shot of the power supply before refurb. May be of use to you:

I only have 2230s and 2232s which is one of the reasons I have not messed with the power supplies as they are much more difficult to access with the DSO board covering everything up.

Edit: also wondered about mains voltages. They have risen considerably here over the last few years. Mine peaks at around 255V RMS whereas it used to be around 235V RMS. One reason I refurbished the entire front end.

I am not aware of more problems being reported when running on 240VAC instead of 120VAC except for the Rifa input filter capacitors.

[bd139]

Could be right with Vds which is a very tight 400V. One reason I picked an IRF820 as a replacement was the Rds(on) is about the same, higher Vds (500V) and gate charge is low enough and Ids is higher. I hope that will give it a little more headroom.

I have some not quite reliable third party notes on what transistors were used as various times.  Eventually Tektronix used the MTP6N60E which is 600 volts, 1.2 ohms, 6 amps, and 125 watts for all models.  The transistor characteristics are not critical as far as I know.

The power supplies operate in continuous conduction mode so the peak switching current is only like 1.5 amps for the 40 watt 2235 and 3 amps for the 80 watt 2232.

That's rather interesting actually. You can see roughly when that MOSFET gave up here: 

I only have 2230s and 2232s which is one of the reasons I have not messed with the power supplies as they are much more difficult to access with the DSO board covering everything up.

Do you know if they run without the DSO boards installed? I have seen a number of those going cheap but they scared me off (and I have a TDS210 that works fine anyway). Might be a possible debugging route.

[David Hess]

I only have 2230s and 2232s which is one of the reasons I have not messed with the power supplies as they are much more difficult to access with the DSO board covering everything up.

Do you know if they run without the DSO boards installed? I have seen a number of those going cheap but they scared me off (and I have a TDS210 that works fine anyway). Might be a possible debugging route.

Absolutely they work without the DSO board installed and that is one of the standard diagnostic procedures.  Without the DSO board, they function just like a 2235 as that is basically what they are.

The 2230 was a continuation of the 468 analog and digital storage design which use an analog channel switch and single digitizer but despite the limitations this imposes, it works really well.  The user interface is not what you would expect of a more modern DSO but it is usable and has a certain charm.  The two most annoying things about it are that the DSO configuration settings are not saved and the 20 MHz bandwidth limit does not apply in DSO mode.

The 2232 uses separate digitizers and has a recognizably modern DSO interface.

The 2230 is much more difficult to work on because removing the storage board requires mechanically disconnecting it from the front panel.  On the 2232, the storage board is trivial to remove for access.

[floobydust]

Recapping these, I also replace C944 1uF tant. and I count four Rifa X and Y-caps C904 (bastard behind on/off switch) and C900, C902, C903 (EMI filter PCB).
I upsize the primary capacitor to 100-120uF as modern parts are -15% actual and 75uF is the absolute minimum to begin with.

My 2235 V/DIV knob (clear) dials are fragile and broke, I couldn't find any so I made some new ones. I have to try perfect the text and laser kerf.

https://www.eevblog.com/forum/repair/tektronix-2215-scope-repair some discussion on the MOSFET Q9070 must be low capacitance, instead of low RDS on as the Holy Grail.

[David Hess]

One of the reasons I have been compiling the changes Tektronix made in the power supply is to eventually refurbish them to the last latest configuration and get some idea of what might have been changed to make the power supplies more reliable.  Mostly they increased the size of the output capacitors to handle higher current.

Tektronix had added the 39 ohm gate resistor and source ferrite bead by the time of the 2230 and they remained through the 2235A/2236A/2232 which were the last versions.  (1) Note that the 2235A is just a 2235 with all of the various improvements Tektronix made over the 6 years between the 2235 and 2235A.

The gate drive in this case is important because we know the power MOSFET has a reliability problem for whatever reason.  The gate resistor and source ferrite bead would be exactly the things needed to control the gate drive or suppress a parasitic oscillation but I would want some way to monitor the gate drive waveform at the MOSFET before doing more than adding a similar gate drive resistor and tuning it as described in that discussion thread.  Like I said before, I have not done this because none of my 22xx oscilloscopes have failed.

(1) There may have been an early and late version of the 2232; there was an updated service manual released but I have not found any differences between them or between the 2232 and 2235A/2236A.

[bd139]

Yes I’ve seen the oscillations in my attempts to build RF amps with switching MOSFETs. The main thing that improved what I was working on was keeping the source short and on a ground plane and adding 10 ohms or so in series with gate. Neither of these are happing with 10cm of wire between the board and the MOSFET. Not exactly a nice design.

If I can’t repair this preregulator I will probably buy a reasonable quality 48V SMPS module and bypass it. Should be able to turn it down to around 43V. Can be mounted on back of chassis.  Cheapest solution!

[David Hess]

Yes I’ve seen the oscillations in my attempts to build RF amps with switching MOSFETs. The main thing that improved what I was working on was keeping the source short and on a ground plane and adding 10 ohms or so in series with gate. Neither of these are happing with 10cm of wire between the board and the MOSFET. Not exactly a nice design.

It is usually not a problem in switching regulators because the transistor spends so little time in its linear range.  But the long lead lengths make this more of a possibility.

My suspicion is that the long lead lengths and lead dressing result in either poor switching or oscillation leading to excessive power dissipation or exceeding the maximum gate voltage.

If I can’t repair this preregulator I will probably buy a reasonable quality 48V SMPS module and bypass it. Should be able to turn it down to around 43V. Can be mounted on back of chassis.  Cheapest solution!

Someone on the TekScopes list did that and reported good results.

[lowimpedance]

I recently had both a 2235 and a 2235A open on the bench and one significant physical difference is the switching transistors heat sink on the non 'A' version is only around a third the size of the one in the 2235A. And when running these heat sinks get very hot indeed. Also noted while looking up info on these old timers that there is provision for a fan and some people have retrofitted one successfully. Although they did not indicate which direction the air flow would be  . I guess that would be inwards?.

[David Hess]

Tektronix used a couple of different configurations to power the fan when it was included.  It is 13.6 volts on the 2236 and 8.6 volts on the 2230 and 2232.  The older fans used the custom brushless DC motor shared with other older Tektronix instruments with a really impressive looking and performing impeller.  At some point they changed to using a standard tubeaxial fan.

The fans on my 2230s and 2232s blow outwards but if I installed one from scratch, I would probably configure it to blow inward with a dust filter on the outside. 

I think what happened is that Tektronix did a "final" design of the 2 channel 22xx series with the 2235A, 2236A, and early 2232 released in 1990 where as many assemblies were shared as possible.  The 2235 and 2236 were the oldest models of the 22xx series so they brought them up to the standard of the "new" 2232.  The 2213 and 2215 do not count as the oldest because they were more like prototypes and quickly replaced by the 2213A and 2215A which came out *after* the 2235.

[TERRA Operative]

I made my fan blow inwards, easy to filter and it blows air directly onto the hot bits.
Tek specify a 20ohm resistor, but I used a 10ohm IIRC to adjust the speed of the fan I used.

I made a thread about mine, and a video, I managed to cross reference the parts in the Tek service bulletin to modern parts and mine is working well.
I didn't swap the line filter module yet, but I found a shop selling TDK-Lambda brand filters for about $10, so I might grab one and stick it in.

Links below for reference:
https://www.eevblog.com/forum/repair/tektronix-2235-upgrades/msg1862394/#msg1862394




I'll be following this thread for further upgrades, I can't help tinkering where I shouldn't... 

[floobydust]

I don't think the Tek pre-reg PSU is properly understood. It looks like a flyback converter but has a resonant winding/cap C908, I think for low EMI.

The TL594 oscillator might have to be bang on. But the components differ in models:
2215: R920 20k + 1800pF C920 27.8kHz?
2235: R919 11k + 1800pF C919 50.5kHz?
2235A: R918 was hand-picked for frequency, not sure of the target.

I would rather have the supply working properly instead of adding a noisy fan. I like a silent scope.
I did make a mistake and replace C908 with a part 10X bigger 2,200pF, the PSU MOSFET ran much hotter.

[bd139]

Yeah I don’t want a fan in this. My main digital scope is a TDS210. Another silent item. The TDS210 actually replaced my DS1054Z because it was quieter and did the job fine! I was hoping to have the TDS210 and 2235 on the bench and leave it at that. Get rid of the other four scopes!

2235 targets 60Khz apparently. I was wondering if the frequency was related to loading. Higher loading, higher switch frequency.

I have all the parts in now to fix the power supply however this might not happen until the weekend due to other things I need to do.

[bd139]

Had a few minutes. I think we have a smoking gun!



Capacitance was fine but I noticed some browning on the bottom of the tape seal on the outside of C907. So I peeled it back about 20 turns and found some heat damage. Not sure if this was the cause or an outcome yet.

[bd139]

And another smoking gun. Someone has been at this. I noticed this, perhaps later than I should have. Some numpty has changed two of the 510K HV resistors for what looks like some 470K ones they had lying around. In the process of doing this, they left shit all over the board and managed to burn the power switch shaft.

The human race disappoints me sometimes. This wasn't hard to not fuck up.

[bd139]

Ok quick update. I have replaced capacitor, the driver transistor, MOSFET and the switching diode from the list above. I added the 39R resistor in series with the gate as per the service note.

Turned it on, blew the fuse straight away. Hmm what did I cock up?

Turns out I put CR907 in the wrong way.  Idiot! Swapped it round, brought it up on 46V DC this time across the preregulator filter cap and bingo, it powered up fine. Thank goodness that didn't break anything. Or did it?

Plugged mains in, boom fuse went again. Checked voltage at schaffner filter and all is good there, thus there is something wrong along the line. I reckon it's one of the diodes gone short. I will debug that later.

Good progress at least.

[bd139]

Ok fuse is my fault. I'm using fast blow where I should have slow blow in there    Inrush current to the main filter cap is blowing it before it gets to do anything.

I'd be dangerous if I knew what I was doing

[bd139]

Ok so after 20 minutes on 46V DC straight into the filter board it's running nicely. I have noticed another issue though which I will add to the list at the top.

IRF820 runs entirely cold at 46V. Lets see how it does on 350V or so later!

[TERRA Operative]

I found a little Deoxit went a long way on my unit.
A few of my switches were crunchy, but they came up like new after a bit of a clean.

[bd139]

Good to hear. I’ll have a go with that (I have a can). The flickering is the only thing that is worrying me. If it’s the multiplier or transformer then I’m going to need another unit.

[bd139]

Ok did some analysis on the flickering while it's open and safely connected to DC and I found that voltage at TP842 drops whenever it flickers. This is not on the driver side of things after hooking a scope to the collector of Q840, setting persistence to infinite and leaving it there for a few minutes so I assume the flickering is an HT issue. I will replace the known hooky HT resistors first and see if that solves it. If not, I will do all the capacitors. If not that I will cry into it and find a donor unit for multipler and transformer.

I will win if it kills me 

[David Hess]

I don't think the Tek pre-reg PSU is properly understood. It looks like a flyback converter but has a resonant winding/cap C908, I think for low EMI.

What I have been told is that C908 and the extra winding inject a counter current into the chassis cancelling radiated emissions and this is what the schematic shows with the separate power supply and chassis grounds.

2235 targets 60Khz apparently. I was wondering if the frequency was related to loading. Higher loading, higher switch frequency.

But the 2235 and 2235A have the lowest power requirements which is why they have no fan.

I would take a very close look at the operating frequency.  The TL594 clock is 60kHz but since only one of the outputs is used, why isn't the output frequency 30kHz?

[floobydust]

I don't think the Tek pre-reg PSU is properly understood. It looks like a flyback converter but has a resonant winding/cap C908, I think for low EMI.

What I have been told is that C908 and the extra winding inject a counter current into the chassis cancelling radiated emissions and this is what the schematic shows with the separate power supply and chassis grounds.

Thanks David for pointing out the two different grounds on C908. I'm not sure how T906 is wound but this extra winding surely is to lower EMI.
When I see care and attention like this, I wonder how using a 48VDC replacement SMPS can work without wrecking the scope's noise floor.