开云体育

As already mentioned, I have a new-to-me 5S14N. Yay.

So far it seems that the horizontal section and triggering work as expected but it has a few problems in the vertical section.

I certainly haven't done all my troubleshooting homework yet, but I thought I'd post up some early findings to see if anyone already knows what they indicate.

As received, it had good bias cells in channel 2 and dead cells in channel 1. Not 0.0V dead but sub-0.1V, iirc. Pretty sure they were under 0.5V. I didn't record the numbers.

Channel 2 looked imperfect but basically functional -- square wave tops not flat but low noise, about expected amplitude allowing for feeding both ch2 & ext trigger, and <500ps rise/fall from a source that's spec'd not quite that fast. Using a 1x probe to snag the ext trigger input probably contributed to distortion.

Channel 1 was a flat line, which made sense for dead bias cells, but responsive to DC offset control, and close to display center with the control centered.

Intermittently the inputs appear to self-oscillate, even with nothing connected. The last time that happened a power cycle fixed it. About 0.5-1V amplitude iirc - similar but not the same amplitude between channels - looks like it might be a sine but not stable enough to distinguish a clear waveform vs aliasing - suggesting a frequency close enough to some harmonic of the scan rate to drift in and out of looking confusingly almost like signal. While typing this, I suppose that indicates that it's happening after the trigger pick-off -- which I suppose couldn't be otherwise because it's all passive until after the trigger pick-off. ?. Anyhow, I mention it because it's there, but seems like a secondary issue at this point because most of the time it doesn't do that.

I tried swapping the sampler boards to put the good cells on channel 1. Internal triggering worked fine. Square waves looked more distorted than before, even without the 1x probe in play, but I figured that could be due to mismatch between the ch2 sampler and the rest of the ch1 circuit until re-adjusted. ?. I've since un-swapped the samplers.

On to more disabling faults:

As described elsewhere[1], I replaced the bias cells in ch1 with silver oxide cells. Nominally 1.55V but measuring 1.59V fresh from the blister pack. That did not magically fix ch1, and now ch1 behaves differently. With the cells removed, which I'd expect to be like it was with dead cells, the ch1 trace is pinned high off screen and unresponsive to the offset control -- as revealed by the beam-finder button. With the cells in place, the trace shifts off the bottom of the display, but responds to the upper range of DC offset and can be shifted up to about 1/3 up the display at maximum DC offset and 0.5V/div. At higher sensitivity (0.2V, higher?) the beam-finder shows the trace responding to the upper range of DC offset but not enough to bring it into the display area. When the trace is visible (0.5V/div, max offset), the unit can trigger and show a stable but noisy and attenuated signal.

I've read that vertical trace shift can indicate fault with the sampling diode pair. If indeed operation with no cells is like operation with dead cells, then this wasn't a fault before I replaced the cells. I'll be a little disappointed if I killed the diode(s) myself, after trying to not do that.

And then a couple things seem kind of strange and maybe specific enough to suggest specific faults.

When ch1 trace is visible (0.5V/div, max offset) and no input connected, the trace shows ~50mV (1/2 minor division) "noise". But not really noise. It appears to alternate between two distinct levels about 50mV apart, at irregular intervals that look like a few 10s of transitions/sec i.e. a random-ish distribution of intervals clustering around a few 10s of msec. Noise on ch2, visible at high sensitivities, looks like ordinary noise.

The next strange-seeming thing is that the "LO NOISE" function shifts the ch1 trace up the display and amplifies it about 2x, in addition to reducing noise. The up-shift looks like less than doubling the large DC offset needed to shift the trace into view. The ch2 trace does not get shifted or amplified. (for completeness: at high sensitivity such that the no-input trace shows noise, the ch2 trace shifts up a fraction of the width of the noisy trace, i.e. noise-reduced trace runs above the center of the normal trace blur, but I assume that's within normal adjustment)

The ch1 up-shift brings the top of the offset range for a no-input trace at 0.2V/div into the bottom half of the screen and the 0.1V/div trace into the bottom division at max offset. Shifting the higher sensitivities into view, together with the slow sweep and mysterious ~2x amplification the LO NOISE mode, gives a better look at the mysterious bi-level "noise" that isn't noise.
Single-sweep traces clearly show irregular alternation between two distinct levels that appear about 100mV apart (about 2x the apparent difference in non-lo-noise view). In LO NOISE operation, it looks like sweeps run about 3/sec, making each horizontal division represent about 1sec/30. The distribution of irregular intervals of alternation appear to cluster around 1 div (i.e. shortest intervals probably not visible and longest intervals <10div), which would be consistent with the appearance of ~tens of level-switches/second. I've posted a photo of three single-sweep LO NOISE traces at 0.2V/div separated by different DC offsets: /g/TekScopes/photo/270747/3351116. It looks like level-switches happen at ~random intervals with a distribution around ~1/div. Assuming the pictured sweeps show about ~1s/30 per horizontal division, that would be consistent with the subjective impression of a few 10s of alternations/second. At the same time, e.g. dual trace display, ch2 at high sensitivity shows what looks like actual random noise.

Having only one scope complicates troubleshooting. I have access to other scopes but I have to take myself and stuff to them with some inconvenience. I'm thinking of making up a partial Y extender (I have edge connectors) to power to the 5S14N from a single vertical slot while using the scope in normal Yt operation with two plugins. Assuming it won't be hard to persuade the unit to operate like it had normal signal connections also. But I don't know how much power the plugins draw vs what the mainframe or a single slot connector can deliver. Seeing that the unit uses the HV supply also ups the ante for build quality if I attempt that :-/

I've written this from memory, so I might end up correcting details later.

Any insights appreciated!

[1] /g/TekScopes/topic/7s14_5s14n_bias_supplies/87551264 ; /g/TekScopes/album?id=270597

keyword: 7S14

On Sun, Dec 12, 2021 at 05:21 PM, Paul McClay wrote:

Having only one scope complicates troubleshooting. ...
I'm thinking of making up a partial Y extender (I have edge connectors) to
power to the 5S14N from a single vertical slot while using the scope in normal
Yt operation with two plugins.

And if that works.... maybe connect the sampler horizontal and vertical outputs to the respective plugin inputs - then the mainframe could selectively show either the sampler display or normal Yt operation. Assuming a 2-channel vertical amp. Or maybe send the vertical output back thru the power Y extender and add a "DISPLAY" switch for that slot. With attention to routing vert plugin/channel select logic thru the plugin connector. ??. Anyone know what traps I'm not considering?

I am absolutely not knowledgeable on your unit. What happens when you switch the channels with good cells in both channels? Does your bad channel symptoms follow the card/assembly that had the dead cells? If the assembly that had the good cells in the opposite channel position look reasonably correct, then pull both assemblies you are swapping and for any visible potentiometers and adjustable caps compare the adjustment positions between both assemblies - good channel vs. bad channel. Someone might have tried to adjust it with the bad cells in place. I would note any position differences for now, particularly if any of the adjustments are at the ends of their adjust range.
Once you are convinced that there are no other failures on the bad assembly, then you could measure pot resistance from wiper to both ends on each pot on the good assembly and match it to the readings on the good assembly and then adjust the bad assembly pot to match. This works reasonably well on multi-turn pots and is not necessary on single turn pots. Vari-caps are almost always single turn. This is not a full adjust but if the assembly has been misadjusted with bad parts in place it should get you close. Another design premise to consider is that anytime a vari-cap or potentiometer is designed into a circuit, it is normally centered in it’s adjustment range, not always but usually.
Don Bitters

On Sun, Dec 12, 2021 at 05:21 PM, Paul McClay wrote:

If indeed operation with no cells is like operation with dead cells, then this wasn't a fault before I replaced the cells

Dead cells would leave their two terminals effectively shorted, while removed cells leave the terminals open. I have no idea if that would make a difference here.

If you look at the 7S14 schematic <1> Samplers and Compensation, it shows that the batteries are actually part of the signal path downstream of the sampler. Removing a battery leaves the signal path open and completely unbalances the sampling gates.

With the age of these units it is also possible that the 2k resistors R1, R2 may have drifted and this also unbalances the sampling and means that you need to apply the DC offset to centre the trace. The originals are carbon composition, probably for low inductance.

Regards,

Roger

tl;dr: ch1 at least 'ok'; ch2 cr2 on order

Welp, as said to begin with I hadn't done all my homework before posting some early results.

On Sun, Dec 12, 2021 at 09:35 PM, Don Bitters wrote:

. . . What happens when you switch
the channels with good cells in both channels? Does your bad channel symptoms
follow the card/assembly that had the dead cells? . . .

The various peculiarities of both channels follow the sampler boards.

On Mon, Dec 13, 2021 at 08:36 AM, Jim Adney wrote:

Dead cells would leave their two terminals effectively shorted, while removed
cells leave the terminals open. I have no idea if that would make a difference
here.

Right. Don't know what I (wasn't) thinking. And yes, it makes a difference because...

On Tue, Dec 14, 2021 at 01:48 AM, Roger Evans wrote:

If you look at the 7S14 schematic <1> Samplers and Compensation, it shows that
the batteries are actually part of the signal path downstream of the sampler.

Oh. And it couldn't be any more clear in the schematic. I looked right at that and didn't see it.

Maybe because my brain simply vetoed the possibility. It's really counter-intuitive to me that batteries pass signal with >1GHz bandwidth. I can't say why I think they wouldn't -- it just seems like if someone had asked me out of the blue if chemistry belongs in a GHz signal path it would have been easy to answer 'no' and believe myself. And there's no 'a-ha moment' yet -- I've had a little think about it and it's still skew in my head.

But there it is and it works. Thank you both for kicking me in the right direction. So, given that the cells pass GHz signal, I've since spent some time staring at the schematic, and at the board, and re-reading that bit of the manual and about avalanche transistors, and back to the schematic, and the board, .... etc. ..., and maybe I have the strobe generator/sampling gate/preamplifier about halfway figured out. I suppose that's why I was fascinated to get one of these and all the more to have one in hand: pretty much everything about it makes my head hurt.

Anyhow...

On Tue, Dec 14, 2021 at 01:48 AM, Roger Evans wrote:

With the age of these units it is also possible that the 2k resistors R1, R2
may have drifted

R1 & R2 measure 2.15 & 2.17 kΩ. In circuit but at least one end looks isolated by caps so I _think_ that's a fair read. For now I'll assume that's not enough difference to overwhelm the offset control. (on the healthier sampler they measure 2.10 & 2.10)

However both sides of the CR1 diode package measured hundreds of kΩ both ways, which sounds unlike diodes. I've just ordered some of the HSMS-8202 diodes pairs mentioned in other threads here. They're out of production and out of stock among reputable sellers but google found some for low-risk money and estimated delivery about when I expect to get another swing at this next year.

...So that's the more broken sampler, which was ch1. I've kept that one out for more surgery later.

I put the better working sampler from ch2 in ch1 to make at least a single channel unit for now. The short story of that is that working through the LF compensation, loop gain, DC balance & memory balance has it working fairly well. At least as far as I can tell with a 5kHz-200MHz square wave source (Adafruit Si5351A module). I'd like to go back for another round of comparing between the sampler and normal Yt operation, which seems like a meaningful comparison up to about 10MHz with the 50/60MHz 5441 scope. The sampler is obviously faster, but within limits of the slower scope they don't always agree. I'll have to check again since getting it better dialed in, but one strange thing that persisted was for the longest square wave periods, ~100-200μs, the sampler showed a small step-up in the last ~10μs of the high state while the normal Yt display showed very flat tops. That seems impossible... how does the sampler know that an effectively steady-state signal is going to transition in 10μs?

Some of what had seemed strange may have been the intermittent self-oscillation syncing up with signals around 5MHz.

The odd two-level "noise" persisted on either channel with the 'bad' sampler. No idea what that could be - but first the diodes.

Until next year-

Merry Christmas, Blessed New Year, and all the best of whatever else you celebrate around this part of the year.

Hoping this finds you all well with a good start into 2022.

Back at this. I've replaced the dual diode with an HSMS-8202. That solved DC offset, with the no-input trace right in the center of the display with the offset control centered, and the mysterious bi-level "noise" went away. DC offset can shift the trace off top or bottom of the screen. That gets me back to showing an attenuated and somewhat slow representation of the input on that channel. Attenuated to ~1/4 expected amplitude with rise/fall time up closer to 1 ns vs. ~400 ps on the healthier channel. The LO NOISE still amplifies ~2x and shifts the displayed signal, but with less shift. The attenuation far exceed the small range of the front panel gain adjustment.

I tried adjusting the loop gain, dc balance, and memory balance on the sampler for lack of a better idea. The procedure includes adjusting DC BALance to minimize trace shift while twiddling LOOP GAIN, then setting the loop gain just below where the displayed response to a square edge starts to overshoot. The DC balance control was able to reduce but not eliminate trace shift while twiddling loop gain.

Increasing loop gain *decreased* displayed square wave amplitude (increase attenuation) down to zero -- i.e. reduce wave height above baseline -- then *continued* to depress the wave tops below baseline, inverting the wave shape. Further increasing loop gain produced an inverted signal with greater amplitude than the maximum at zero gain, but still less than input amplitude, before going into overshoot. That sounds like maybe a clue to what's wrong, but I haven't decoded it. For all of the unity-or-less loop gain range, the LO NOISE mode displays the same upright, ~2x amplified trace regardless of loop gain. The MEM BALance control was able to affect but not eliminate trace shift with LO NOISE mode.

For slow signals, the square wave tops appeared to follow a sine wave shape. Adjusting L. F. COMP appeared to alter the phase of that sine pattern relative to the square wave without affecting it's amplitude. The amplitude of the sine increased with lower frequencies. Pictures will likely help with describing that...

I haven't messed with the AVALANCHE control for lack of a lower lever (~250mV) square edge source which the calibration procedure calls for. Needing an adjustable source voltage or attenuator. My current thought is to solder a calculated SMD resistor across the back of the input from my fast(ish) square wave source -- which is stuck directly on the scope input BNC. The manual doesn't suggest that will influence amplitude, but it's another knob to twist to see what it reveals.

The behavior described here follows the sampler board when swapped between channels.

My small lot of HSMS-8202s, an obsolete part, came from a marginal source. They are marked as described in a found datasheet: 2R{datecode}. At DC the one I've used functions as two diodes with Schottky-ish forward voltages matched to 3 decimals. Does anyone re-mark or fake SOT23s and package them in tape? Seems that would have to be a *very* efficient operation.

I added a photo of the installed SOT23 packaged HSMS-8202 to the album [/g/TekScopes/album?id=270747]. Had to undo/redo the battery receiver to make room for that. Could say more about that; maybe later; maybe in that topic vs here [/g/TekScopes/topic/87551264].

So far it seems the reduction/inversion of sampler output with increasing loop gain is the best clue. Dunno what to make of it yet. Happy to hear any suggestions!