Cossor CDU150 rescue

[MelJon66]

Hi All,

Cossor CDU150 - capacitor replacements on board H1 & H2 - Vertical Pre-amp.

Board H2. As reported in #59 replacing the Tantalum cap C43 stopped the crosstalk between channels on Ground input settings. I then replaced the ceramic caps C31 and C38 and the crosstalk was back at all input settings .

Board H1. I carried on regardless and replaced the ceramic decoupling caps C3, C8, C12, C19, C26 and C29. No change, the crosstalk was still there on Ground, DC and AC input settings. BUT, the flickering trace is very much reduced. I think the reason is that the crosstalk in the trace is now visibly sharper and the trigger circuits are finding this easier to spot the edges in a DC trace even at very low levels so it triggers successfully every time. Not an ideal solution but I can live with this as when both channel sensitivities are correctly set there is minimal crosstalk and no flickering and the traces are very sharp.

I was unable to replace caps C2 and C20 on board H1 as the ones I had in stock were the wrong values. I have ordered some in to complete the work. They seem to be in a critical position in the circuit so I am hopeful that they may have some impact.

The ceramics I used were modern MLCC types (Murata) and measured very close to the specified value. Some of the old caps I took out were well over the specified value so in effect I have replaced the old caps with lower values. Should I put higher value caps in? I read that the value of MLCC types vary a lot with voltage. Should I have used old single layer types that may have properties that are more compatible with the circuit as designed?

Any thoughts?

Cheers
Mel

[Cobaltblue]

Thread reopened.

Cheers

Mike T

[MelJon66]

Hi All,

After a very long absence the Cossor is back on the bench. The refurb was taking so much time that I had to put it aside so I could get on with other stuff. I was always planning on getting back to it but I didn’t think it would take so long. Hey Ho!

As a reminder the last problem I was grappling with was crosstalk from one channel to another. To try and remedy this all the decoupling caps on vertical pre-amp boards H1 and H2 were changed, as recommended by Chris Williams (Chris55000). I also changed C15 on H1 as it wasn’t mentioned but is also a decoupling cap. Unfortunately, the crosstalk remained so yet more time was required to investigate further………………

Anyway, my current investigations may have unearthed some information that might lead Chris and other resident experts to offer some guidance that will get me heading in the right direction again.

With reference to the attached photo:

Input from sig. gen. to the Cossor Ch 2 showing on the Cossor = 500Hz Sine wave @ 1V/cm = 8V P2P.

There is no input at all into the Cossor Ch1.

The Hitachi 'scope at the back of shot is showing the signal picked up when probing any one of many positions on Ch1 PCB H1, with no input to the channel. At the above Ch2 settings the crosstalk on Ch1 with the Hitachi set @ 5mV/cm = 12mV P2P. At these settings the crosstalk is a perfect sinewave and this happens at any input frequency. The crosstalk also occurs on Ch2 with an input on CH1 only. This level of crosstalk occurs irrespective of the vertical sensitivity set on Ch1.

If the vertical sensitivity on Ch2 is increased so that the Ch2 waveform on the Cossor goes far above the screen vertical limits the crosstalk on Ch1 starts to clip and becomes the square/sawtooth wave that I have shown in previous posts. I had not previously spotted that the crosstalk was a pure sine wave if I got the settings correct.

Additional comments:

Referring to PCB H1 and H2 schematic attached.

The +12.6V rail and anything on H1 supplied from it seem to be affected by what is going on.

Probing at pin 23 (top right of the schematic on PCB H2) the +12.6V rail measures +12.6V and is clean DC, with no AC component. R143 is a 22R resistor and measures correctly. The voltage after R143 reads +10.6V but is not clean and has 12mV P2P AC crosstalk riding on it. From that point on there is AC crosstalk on the +12.6V rail which can be picked up at any component on PCB H1 that is supplied from it.

C43 is a new Kemet tantalum cap and the polarity is clearly marked so I did get it in the right way around. This problem was happening before C43 was replaced anyway so the new cap isn’t likely to be the problem but it does seem odd that the crosstalk starts at that point on the +12.6V rail.

It strikes me as significant that the crosstalk is not present at pin 23 before resistor R143 but appears immediately after R143. This must be telling me something obvious but I don’t know what it is. R143 and C43 seem like obvious culprits but the resistor tests fine and the capacitor is new.

I have probed all the inputs to H1/H2 from the attenuators and channel mode/trigger switch and there is no crosstalk present so it appears to be generated on PCB H1 somewhere.

Any ideas on what my next move should be?

Cheers
Mel

[WME_bill]

Cossor CDU150.
Cross talk.
I wonder if it is dry joints or faulty earth connections. Have you checked every earth connection on the boards. I have trouble on a CDU150 once where the screening on a cable had become unsoldered at one end.
And on other scopes, particularly Telequipment with aluminium chassis, slight oxidation between the copper on a PCB and the Al chassis breaking the earth connection which was meant to be secured by bolting the panel down.
wme_bill

[MelJon66]

Thanks Bill,

Another member has reported trouble with grounding so it does seem to be an issue with these 'scopes. I think the grounding from the attenuators to chassis is fine but I can't recall seeing effective looking ground connection between the PCB ground tracks and chassis. It seems to rely on the PCB mounting screws but the screws and washers make contact with insulated areas of the PCB. I have cleaned all the screws and threads but I will pay more attention to ensuring there is metallic contact to ground next time I have the module out.

Cheers
Mel

[Chris55000]

Evening Mel!

I certainly will have to set up one of my CDU150s very soon and see if the same CH1/CH2 crosstalk fault occurs on mine!

I've not done anything more than run them up with a "wet finger" in the BNC's to see if the vertical does anything, as the Bradley 192 I have will need a major switch rebuild and so will the valve Tek Calibration Genny I acquired!

I'll try and get another H1 board sent to you as soon as I can get it out of the "parts–mule" I supplied you with the mains transformer from!

Chris Williams

[MelJon66]

Hi Chris,

A crosstalk check on one of your CDU150's would be very helpful. I sometimes think my expectations of what to expect from these 'scopes is unreasonable.

I have spent time improving the grounding around H1 and it has delivered good results, although not quite the results I was expecting. I will post the details soon.

Thanks for the offer of the H1 board, it will a great backup as these boards do seem particularly problematic. One of the position/set balance pots on mine continues to be very stiff to turn so a replacement would be great.

Cheers
Mel

[MelJon66]

Hi All,

Warning: Potentially fatally boring post unless you are into Cossor ‘scopes.

As suggested by forum members I have removed and stripped down the preamp module, cleaned up all the ground connections between H1 PCB and chassis and re-soldered all the ground side joints on the de-coupling caps. I have also added a grounding link from the module chassis to the adjacent Board G. I am confident the grounding on H1 is now as good as it can be.

The good news is that the traces are now sharper than ever, including the crosstalk traces! The great news is that the crosstalk is significantly reduced so when I probe the channel suffering crosstalk and view it on another ‘scope I now see 6mV P2P crosstalk with 8V P2P sinewave into the other channel (about 50% of what it was before). The sharper traces mean that I can now view the crosstalk clearly on the Cossor itself and can dispense with the second ‘scope. I can also leave the side panels on the Cossor so external noise pickup is minimised. Note: the crosstalk amplitude is much less when viewed on the Cossor, probably due to the way the channel selection is carried out on board H2.

With a 16V P2P 5kHz sinewave input to Ch2, 1V DC into Ch1 and the vertical settings at 1V/cm for Ch1 and 2V/cm for Ch2 there is no visible crosstalk on the Ch1 trace but the trace flickers badly. With the side off the ‘scope if I touch the preamp module chassis the flickering slows down a lot, sometimes stopping temporarily. If I remove my hand it starts again. If I connect a direct ground wire to the chassis from a mains socket it has no effect. I think this must show there is still a grounding problem but I don’t know how to resolve it. I plan to clean up the grounding on board H2 to see if this improves things further.

With inputs as above but both vertical sensitivities set to 5mV/cm the crosstalk is clearly visible as shown in the attached photo of the Ch1 trace. I realise this is a very unreasonable test as both traces extend well off the screen so the crosstalk is severely clipped but is does highlight the issue. The input to Ch1 has been removed for this photo as with input connected all you can see is a fuzzy fat trace full of ripple. If the input to Ch1 was a 1V P2P sinewave it would be unstable with out of phase ghost traces flickering alongside the main trace. This was very hard to photograph so I used a DC input for this photo.

What you see in the photo is severely clipped crosstalk with an amplitude of around 2mV on Ch1 picked up from Ch2. If I reduce the sensitivity of Ch2 to reduce the clipping the crosstalk on Ch1 does appear as a low level sinewave. I believe it is this crosstalk that is causing the flickering trace noted above as the triggering circuits struggle to latch on to the input signal plus crosstalk ripple.

The crosstalk does occur either way and the channel with the lowest input signal suffers the crosstalk. In this case we have 16V P2P sinewave to Ch2 and 1V DC to Ch1 so the difference isn’t as great as it would be in many circumstances. Trigger mode is set to Auto and trigger coupling is set to AC. If I switch the trigger mode to DC the flickering stops, which is acceptable in most circumstances, but is a bit of a nuisance having to do this when quickly switching between channels.

It could be that crosstalk between channels is normal on these old ‘scopes as I can see from the schematic that both channels are always powered and are connected together at the outputs of H1, so there is plenty of opportunity for crosstalk. My inexperienced interpretation of the schematic is that one channel is selected by blocking the other one but by this time the crosstalk is already affecting the selected channel so there may be no way of removing it. I am happy to be shot down on this by anyone in the know. As long as the crosstalk remains invisible at realistic vertical sensitivity settings it may be an acceptable design compromise but it does appear to cause instability on the channel with the lowest input level, which is harder to accept.

Questions:

• Any suggestions on how to improve the apparent grounding problem?

• If anyone with a CDU150 is willing to check if crosstalk occurs on their ‘scope as above, please let me know what they find. If it the same as mine I will be satisfied that mine is “normal” and get over it.

• Can anyone with knowledge of these circuits briefly describe how the channel selection circuitry works as set out in the H1 & H2 board schematic posted earlier?

Thanks in advance.

Mel

[Radio Wrangler]

Mine suffered crosstalk once you got the amplitude going off the screen on the driven channel. Many scopes do. The beam switches are usually diodes, balanced etc, but an overdriven channel can reach the bias voltage limits and start robbing current from the quiet channel. At high frequencies diode capacitance gets in on the act..

The manufacturers usually consider that one channel being far off screen for a lot of the time isn't 'a realistic measurement scenario', which is a load of cobblers. If you have some ripple on one level of a square wave and you want to focus in on it, you do drive the rest of the waveform offscreen.

Try a few other makes and models of scope and you'll find it.

David

[MelJon66]

Thanks David,

It's reassuring to know that crosstalk is not so unusual, particularly when both channels are driven at unrealistic levels.

I have a Hitachi that hardly displays any crosstalk at all, other than some transient spikes appearing on the DC trace. It does show up on my Iwatsu as a thick fuzzy trace on the DC line when both sensitivities are at maximum. Neither of them show crosstalk with such large amplitude and with such clearly defined traces but maybe that says something good about the Cossor's resolution ability. The Cossor is also a lot older and does not have the advantage of quieter IC based circuitry.

With the Hitachi and Iwatsu the trace for the low level input channel always remains stable and without flicker, despite the crosstalk. Maybe this raises questions about the Cossor's triggering ability in its current state.

I will have a look at the grounding on the triggering circuits on board C.

Thanks for your input.

Cheers
Mel

[Radio Wrangler]

Aye, triggering has to be taken off before the beam switch in a multi-channel scope.

However, the triggering needs to switch between channels sometimes trace-by trace in alternating mode, so trig channels sometimes contain a replica of the Y path switch!

David

[MelJon66]

Hi All,

Paula (frsimen) has very kindly carried out a series of crosstalk tests on her CDU150 based on the Cossor handbook recommendations. She has PM'd me the results and I have repeated the tests on my CDU150. In all but the most severe test, representing a 10x overload on the input channel, my results are about the same. On my CDU150 at 10x overload the crosstalk on the channel with no input was about double that of Paula's and some clipping was evident on Channel 1. At more realistic overload levels it is behaving much as it should. Paula also carried out the same tests with a Tek 465 and at 10x overload it fared worse than my Cossor so I think that sets a reasonable benchmark for 'scopes of that era. This also confirms that the crosstalk that I was reporting with a 400x overload did not represent a fair test at all . So, there may be a little more to investigate to try to resolve the slight channel imbalance but I can at last put my confusion and concerns about crosstalk to rest .

Paula has also provided an explanation of how the channel selection circuits work. This has already given me greater insight than I have ever had before but it will take some time for me to absorb the information in full.

The unstable flickering traces will be the focus of my attention now. I note that switching to DC trigger coupling stops the flickering and also that it gets less frequent as the 'scope gets hotter. I will start with checking the grounding on board H2 and around the trigger circuits on board C.

Many thanks to everyone for their support and particularly to Paula for going above and beyond what I could have expected from anyone.

I will report back soon.

Cheers
Mel

[Radio Wrangler]

Scope amplifiers are essentially push-pull right through with the input on one side and the shift pot on the other. This way, so long as the spot is on the screen, the amplifiers down the chain are all in their linear region. You can offset a trace rather a lot to look, say, at ringing on one flat level of a squarewave or pulse. with the rest of it being far off screen. While offscreen, the amplifier chain beyond some point will be driven onto clipping. This is inevitable. Trying to avoid it would require immense power supply voltages and force the use of slow, high voltage transistors if it could be done at all. Balanced circuits are only so balanced and there will be some common-mode caused, not all common mode can be rejected, and even though the amplifier chain may be balanced, right at the beginning it is fed an unbalanced signal.

The escape from this problem is time itself. The on-screen, high sensitivity plotting does not go on at the same time as the offscreen overload event. The scope amplifier must be designed to clip, to clip cleanly and to come almost immediately out of clipping with no slower recovery effects.

Have a two channel scope and one amplifier channel used to blow up a small detail on a waveform, with a big offscreen deflection while trying to see a much daintier waveform on the other channel and you hit limitations of the overdriven channel's imbalance modulating power supplies and bias voltages etc, and this getting into the quiet channel via its power supply rejection parameter. Diode switches leak. Big signals on one input will tend to pull the control voltages/currents around and some influence will get into the other channel. Scope designers normally try to keep things as good as they can with signals up to ten screen heights total amplitude.

Scopes like the Tek 465 are designed to do the best they can with input signals up to 200MHz (twice declared overall 3dB bandwidth) remember that part way down a '100MHz' amplifier, the roll off must be smaller than the overall specified allowance, so the bandwidth is greater. The competitor to the 465 is the HP 1740 which I know better and in this it does beam switching with dual differential amplifiers into shared collector load resistors, and gates the emitter constant current sources. This section is all integrated onto custom ICs using something like a 5GHz transistor process.

This crosstalk effect is one of the imperfections which can be minimised by careful balancing of the operating conditions in the differential amplifiers, and explains the DC balance trimmers sometimes to be found in such circuitry.

Due to the fully balanced nature and the shift pot voltage going in at the 'spare' input right at the front end, you'll see the same crosstalk with the effect of shift pots. Watch trace A while you use the B channel shift pot to move trace B offscreen upwards and downwards, and look at how much trace A moves in sympathy. No signal is needed! Then you try it the other way around.

In the frequency domain, there is a similar case where a user might want to drive things into overload. If you are looking at something small in the presence of large signals, you may have a trace that goes off the top of the screen by 10 or 20dB in order to get the small thing showing. You might be getting effectively a picture with 20dB more dynamic range than the height of the screen would show all at once. The logarithmic detector is easy about this, because while the analyser is scanning the frequency of the small signal, the big signal is outside the bandwidth of the IF filters and does not reach the logger. When scanning the frequency of the big devil, it doesn't need the finesse to show the small one.

The fly in the ointment is that though the signal levels turn up sequentially at the bulk of the IF amplifier and the logger, they all hit the earlier IFs and the RF sections simultaneously. So the big signals will create harmonics and intermodulation products which can turn up on you screen as spurii, looking like input signal components, but really created in the analyser. What you do is play around with the RF attenuator on the input and if you change it by 10dB, then any component showing on the screen that moves by any other amount is suspicious.

It takes some doing to teach a beginner that even though your analyser may be looking elsewhere in frequency, all frequency components put into the input socket travel part way into the analyser, until they reach filters which attenuate them severely. They still get to tenderise the front end mixers. It's just counter intuitive.

Moonbouncers, people who bounce vhf-microwave signals off of the face of the moon (for fun, really) have trouble explaining because most people instinctively assume if the moon isn't full, it isn't there for reflecting radio signals! Sounds daft? but it's a pitfall even supposedly intelligent people regularly fall into.

Ah, The gentle art of pushing test equipment beyond what on the surface you might expect it to do.

David

[MelJon66]

Thanks David for the very interesting overview of the compromises in 'scope amp design. I followed almost all of it which is refreshing.

I am still getting to grips with some of the basics, like your statement that 'scope amps are in push pull all through. I am familiar with complementary pairs in audio but not all the stages in the Cossor are complementary pairs so that throws me. It's push pull Jim, but not as we know it . It seems push pull has a wider definition than I thought and complementary pairs are just one possible arrangement.

I can recognise TR1/TR2 as jfet source follower current amplifiers but I am not sure how they share the signal with TR2 being referenced to ground. TR3/TR4 are an emitter follower pair but do they each get the same signal from the previous stage? TR5/TR6 is a common emitter voltage amplifier so I am happy with that. TR7/TR8 appear to be identical and not a complementary pair so is that still push pull (I can't be too sure as my schematic is so fuzzy the direction of the emitter arrows is unclear)? So, I recognise some sections but I am mulling over how it all links together. I do know that by the time the signal gets to pins 10 and 11 at the output of H1 the signal at the two pins has the same amplitude and phase (with AC coupling at the channel input) so I assume they are destined to be added together at some point. I introduced some DC offset on one channel expecting to see it appear on pin 11 but it didn't so I don't know where any DC component in the signal shows up.

Luckily, most of the pleasure is in the learning, which is just as well as that means I have a lot of pleasure to come .

Cheers
Mel

[MelJon66]

Edit,

Paula (frsimen) has just PM'd me to says that most of the transistors are NPN. I should have gone to Specsavers! In reality I just need a better schematic. I think I have a very good copy of the 1st Edition manual so I will see if that is any better.

Doh!

Mel

[MelJon66]

Hi All,

With a bit of research I discover that push pull simply means where pairs of transistors share the load, as opposed to single ended, and the pairs don't have to be complementary at all. You can even get push pull class A, which I have never encountered in practice.

With a better quality schematic in hand the fog clears a little and I see that there are no complementary pairs in the H1 circuit at all and most of the bjt transistors are NPN's except for TR9 and TR10 which are PNP. I am still a bit unsure about how much signal is being handled by TR2 but it seems the signals couple together over the early stages and the pairs are sharing equally by the time the signal reaches TR7/TR8.

The statement I made in post #74 has also proven to be incorrect and the signals at pins 10 and 11 are out of phase and I am picking up DC offset on both pins. Even my probing/scoping skills are in need of urgent attention!

It's just as well that persistence pays .

Cheers
Mel

[MelJon66]

Greetings All,

Advice needed again.

I am currently trying to set up the attenuators H.F. response on my CDU150. I did this last year and did struggle with one channel but I got there in the end. This time I am not getting anywhere at all with Ch2.

The first photo attached is Ch1 with 5kHz square wave input and it looks nice. The second is Ch2 with the same input. It looks like massive overshoot to me but adjusting the attenuator caps does nothing at all. These are old fashioned mechanical screw type variable caps and I imagine they are pretty tough and they can't all be broken so there must be something else awry.

I am following the procedure in the handbook but I fail at step 1.

It says:

Connect the LC Meter on 30 pF FSD range across INPUT socket and adjust
C28 on PCB J2 (for Ch2) for 25pF

My LC meter shows around 490 pF and adjusting C28 does nothing. When I go through the rest of the setup and adjust the relevant cap for each input sensitivity it has no effect on the trace.

Measuring the input capacitance on Ch1 doesn't work either but the trace does respond to adjustments to the caps on PCB J1.

I am baffled. Any insight to offer?

Cheers
Mel

[Radio Wrangler]

You may need to try with a higher frequency test signal, with a very fast rise time.

The trimmers may be having an effect, but the roll-off they are affecting may be at frequencies much higher than the harmonics of your signal source reach.

Your amplifiers have adjusters for their roll-offs, so the scope should be rather flat with all attenuators out at the input. (if there is a second switched gain later in the amp, then there will be separate flatness adjusters for each setting it has. The twiddlers on the inout attenuator are there just to twiddle the relative roll-offs of just the attenuator sections.

Tektronix made a special ultra-fast tisetime pulse generator for setting these things. It was good enough for 500MHz oscilloscopes and used a tunnel diode in a sort of microwace cavity.

It was so specialised that HP bought them to set up their own scopes and listed them in the calibration equipment in their manuals.

You don't quite need one of these, but a good fast pulse generator with a rise time spec better (shorter) than the spec of your scope is needed.

David

[MelJon66]

Hi David,

Thanks for the input.

Following the sequence given in the handbook I set the main Y amp gain, then the Ch1 and Ch2 gains, then the Y Amp Frequency Compensation and finally the attenuator adjustments.

The gain was set at 1kHz and both channels were the same after setting. Photo 1 shows the relative gains at 1kHz. Ch2 is the lower trace and leading edge overshoot transients are visible.

Photo 2 shows the traces at 20kHz where the channel gains are still similar but the Ch2 trace is now showing massive overshoot.

The Y Amp Frequency Compensation was adjusted at 1MHz. Photo 3 shows that the channel waveforms were similar but with Ch2 showing more overshoot. Notably the Ch2 gain appears to be higher at this this frequency and continues that way up to 5 MHz where the Ch2 gain is about 2X that of Ch1. The waveforms are rounded at these frequencies but my Iwatsu 100MHz 'scope does much the same at these frequencies so I don't find this surprising.

I followed the handbook for the attenuator adjustments at 5kHz but saw no changes in the trace.

So, when the frequency increases the waveform seems to improve but the gain increases. This might explain why the HF adjustments are doing nothing as the traces appear reasonable at MHz frequencies. The problem with massive overshoot seems to be at lower frequencies. Does this suggest a linearity problem in the Ch2 amp somewhere? It could be a fault in Ch1 but as this is not suffering the overshoot this is behaving more predictably so is less likely.

Any ideas would be appreciated.

Cheers
Mel

[Radio Wrangler]

I think you need to take several steps back.

Find out which range has NO front end attenuators engaged, use a pulse generator (not a general purpose function gen or an audio gen set to squarewave - they're not fast enough) and first check that the amplifier frequency response/risetime is OK. If not adjust/fix the amplifier. Don't change capacitors just on spec or in hope. Some dielectrics can have some wierd effects on edge shapes.... some ceramics are quite non-linear, others are fine.

ONLY once you're happy with the amplifier do you look at the front end attenuator sections.

Any other sequence just leads to confusion.

David