Cheap analog scope with decent (~5 MHz) X bandwidth?

[ebastler]

I was planning to use my old Hitachi V222 scope in XY mode as a display in a vintage computing project. (Aiming to recreate a really old machine which used an array of spots on a CRT to display its memory contents.)

The spots are output at a rate of 5 MHz -- bright spot for a '1', dim spot for a '0', move between positions as fast as possible and stay in place for as much as possible of the theoretical 200 ns dwell time. The scope has two channels with 20 MHz bandwidth, which turns out to be quite adequate along the Y axis.

But along the X axis, the beam is positioned much slower: While the input amplifier supports 20 MHz, the beam deflection does not. It does not have to in normal YT mode, of course. I checked the scope's specs, and X deflection is indeed only specified at 500 kHz bandwidth -- which is probably a conservative spec.

Hence I was wondering: Is there any type/brand of vintage CRT scope which happens to be known for generous X bandwidth? I am aware of dedicated XY displays, but these are rare and expensive over here; can't justify that expense for this niche use.

This feels like a long shot -- who would have wanted to pay for excessive X bandwidth in a scope back in the days? But maybe someone has a lead for me. Thank you!

[techneut]

Try to find old tv vectorscoops, they are x-y scoops internaly.

[David Hess]

It is going to be difficult to find an analog oscilloscope with 5 MHz of X bandwidth.  With one exception, the highest bandwidth units that I am aware of are 3 MHz and most are slower.

[ebastler]

Thank you for your comments, David and techneut!

It is going to be difficult to find an analog oscilloscope with 5 MHz of X bandwidth.  With one exception, the highest bandwidth units that I am aware of are 3 MHz and most are slower.

Yes, I looked around a bit in the meantime, and that seems to be the typical limit. Maybe that's good enough? The Hameg 203 series, which was very popular in Germany, is specified at 3 MHz. I'll try to borrow one somewhere and see whether it gives a usable display.

Try to find old tv vectorscoops, they are x-y scoops internaly.

Hmm, I am a bit reluctant to embark on yet another project to convert a vectorscope. They are not really common or cheap over here. Philips and Leader units show up occasionally, at an asking price around $150 for working units. But I can't find schematics for either.

[techneut]

The Tek 453A has a x-bandwith of "5 MHz or greater" according to the manual

[David Hess]

I checked my notes and the Tektronix 7504 and 7514 apparently have a 5 MHz X bandwidth but given their age and complexity, these would be restoration projects rather than useful instruments.

Note however that the vertical delay line in these oscilloscopes, and the 453A, creates a significant phase error between the X and Y axis unless the phase correction option is installed.  On the 453A, this is 3 degrees at 50 kHz.

[bdunham7]

If you are recreating a Williams tube, how are you going to read the data? 

Do you intend to keep the scope whole, or are you willing to chop it up for this project?  The horizontal amplifier may lack bandwidth, but I suspect the CRT itself won't have a problem with it.

[ebastler]

It's not a Williams tube -- just a display of the memory contents, as used in Turing's ACE design. The actual data storage will be in ultrasonic delay lines. (Small 64 µs glass delay lines from PAL TVs, not mercury-filled tubes as in the original.) 

I would consider modifying a scope; a decent X/Y display can probably come in handy occasionally. But would prefer to have the flexibility of adjustable input ranges and offsets which a regular scope offers; modifying a Vectorscope would limit flexibility a bit.

[TrendX]

I understand only half what you are talking about, but I got a HM205 and two HM305 with 3Mhz of X-Bandwidth. If you tell me how I can test if it is good enough for your needs.

If you're close you could borrow one as well if needed.

[ebastler]

Cool -- thank you very much for the offer! 

If you have a function generator which can output a staircase function, you could do a quick test with the scope in XY mode. The frequency should be set such that each step lasts for 200 ns; the number of steps does not really matter. My Siglent SDG2042 has a predefined function with 8 or 10 equidistant steps up, then dropping all the way back down, which works nicely.

If you put the scope into XY mode and feed the staircase signal to the Y axis, you should see a vertical column of nicely defined dots. (The Y axis will definitely have enough bandwidth.) If you connect the signal to the X axis instead and the dots still look OK, then the scope is fine for this job! My old Hitachi shows very blurry dots in the X direction, and gets completely messed up by the large step back down at the end of the scan line.

I am based in Hamburg -- would be happy to go for a bit of a drive if the Hameg works out. Thank you very much for offering your help and a loaner!

[bdunham7]

If you have a function generator which can output a staircase function, you could do a quick test with the scope in XY mode. The frequency should be set such that each step lasts for 200 ns; the number of steps does not really matter. My Siglent SDG2042 has a predefined function with 8 or 10 equidistant steps up, then dropping all the way back down, which works nicely.

If you put the scope into XY mode and feed the staircase signal to the Y axis, you should see a vertical column of nicely defined dots. (The Y axis will definitely have enough bandwidth.) If you connect the signal to the X axis instead and the dots still look OK, then the scope is fine for this job! My old Hitachi shows very blurry dots in the X direction, and gets completely messed up by the large step back down at the end of the scan line.

OK, I see how you are doing this.  If you have 200ns dwell time and you want the trace to move quickly to the next one, a scope with a maximum sweep rate of 50ns/div may will not have enough horizontal amplifier bandwidth.  A scope with a 5ns/div or 10ns/div sweep rate might be better, but unless you blank the trace during the movement, you'll still see the lines between the dots.  If you are representing the data by having the dot there or not, are you doing this with Z-blanking or by skipping that step in the trace?

I tried what you are attempting with a Tek 2445, which is certainly at the high end of 'cheap analog scopes'.  It has a 10ns/div max sweep rate without magnification and a rated X bandwidth of 3MHz.  I used an 8-step stair repeated at 625kHz, for 200ns steps as shown.  It works, but not very well--there is a line between the dots.  The picture is worse than the reality, the dots are clearly visible, but the line is there.  This scope has the ability to select the Y channel in XY mode, and you can select the same as X/CH1 for an 'XX' mode.  You might think this could only yield a diagonal line, but the bandwidth difference between the horizontal amp and vertical amp actually gives you interesting information.  One problem is that your 200ns-per-step stair has a fundamental bandwidth of 5MHz but like representing a square wave, you will need 5-10X that in bandwidth to have a sharp, crisp move from one dot to the next.  The other issue is that the visibility of the trace is quite non-linear, so even if the trace spends 99% of the time at the dots and 1% in between, you still see the line quite clearly. 

Lowering the frequency to 62.5kHz w/ 2us steps gave brighter dots with a dimmer line, but still not perfect.  6.25kHz w/ 20us steps gave nice clear dots with no visible lines and the retrace in XX mode was only slightly visible at the highest intensity. The 2445 is visibly out of phase in XX mode with a 3 MHz sine wave input.  Since any scope with a delay line is likely to have its X-bandwidth limited to prevent an erroneous out-of-phase display, what you are doing is going to be tough with any ordinary scope.  I assume each scan line will be a row of data and then you will step more slowly the other way to display multiple rows in a matrix.  Can you scan the other way, so as to use the Y-axis' greater bandwidth?  I tried it and got a clear matrix, but still with a visible line between the dots. 

[srb1954]

I checked my notes and the Tektronix 7504 and 7514 apparently have a 5 MHz X bandwidth but given their age and complexity, these would be restoration projects rather than useful instruments.

Note however that the vertical delay line in these oscilloscopes, and the 453A, creates a significant phase error between the X and Y axis unless the phase correction option is installed.  On the 453A, this is 3 degrees at 50 kHz.

A Tektronix 7104 opt 02 would do the job and using 7A19 opt 04 vertical plug-ins will allow tweaking phase shifts for X bandwidths up to 50MHz.

With the 7104 you would have to ensure that fixed patterns are not displayed for extended periods as this could put burn marks on the delicate CRT in these scopes.

[David Hess]

The Tektronix 7104 is the exception I was thinking of.  It uses a horizontal amplifier design which is more like the vertical amplifier on other oscilloscopes and it is what Tektronix would recommend if you wanted high bandwidth X response unless a sampling oscilloscope was acceptable.  With an analog sampling oscilloscope, horizontal bandwidths above 1 GHZ are feasible but they have other limitations.

It would not be difficult to modify a horizontal CRT amplifier for high bandwidth without compromising normal operation but this is not something I would recommend unless you are familiar with transistor level design.  The major change would be to increase the output idle current and maybe add cascode transistors.

[bdunham7]

I'm pretty sure this problem can't be completely solved with greater bandwidth and the 7104 is going to be problematic because the MCP CRT, in addition to being delicate, will amplify the traces in between the dots. 

As an experiment, I set my scope and generator up with a 8-step 200ns/step (625kHz) signal on the Y channel (150MHz bandwidth) and a 8-step 1.6us/step (78.125kHz) signal on the X channel for an 8x8 matrix of dots.  As you can see, it showed up OK, but even with 150MHz bandwidth on the scope, you can see the lines between the dots in the vertical direction and of course the traces as the X-channel steps.  The bandwidth of whatever is providing the signal will be an issue as well, although in this case the sig-gen is 120MHz.

Next I put the 625kHz step signal back on CH1/X and set the second channel of the sig gen for a 5 MHz pulse, which I connected to the Z-blanking connector.  After some twiddling, I was able to get a pretty clear set of 8 dots on the X-channel.  I would need another channel on the sig-gen to make a matrix with Z-blanking, so that's the limit of the demonstration here.  The Z-blanking setup works even better on the Y-axis.  The last shot is of the two signals displayed normally so you you can see how they relate.  There's something strange that should catch your eye if you understand Z-axis blanking signals...

So I think more bandwidth than 3-5 MHz is neither necessary or sufficient.  To make this work with any sort of conventional scope is going to require blanking or else it will just look a little messy.

[TrendX]

Cool -- thank you very much for the offer! 

If you have a function generator which can output a staircase function, you could do a quick test with the scope in XY mode. The frequency should be set such that each step lasts for 200 ns; the number of steps does not really matter. My Siglent SDG2042 has a predefined function with 8 or 10 equidistant steps up, then dropping all the way back down, which works nicely.

If you put the scope into XY mode and feed the staircase signal to the Y axis, you should see a vertical column of nicely defined dots. (The Y axis will definitely have enough bandwidth.) If you connect the signal to the X axis instead and the dots still look OK, then the scope is fine for this job! My old Hitachi shows very blurry dots in the X direction, and gets completely messed up by the large step back down at the end of the scan line.

I am based in Hamburg -- would be happy to go for a bit of a drive if the Hameg works out. Thank you very much for offering your help and a loaner!

Unfortunately my old signal gen cannot output a staircase signal like bdunham7 showed. Only square, triangle and sine.

I'm based near Heilbronn so it would be a very long drive for you. But if you find no other option you can PM me if you want it as I don't need it in the foreseeable future because I want to concentrate on my professional school this year.

[srb1954]

I'm pretty sure this problem can't be completely solved with greater bandwidth and the 7104 is going to be problematic because the MCP CRT, in addition to being delicate, will amplify the traces in between the dots. 

I didn't really intend my suggestion of the 7104 as being an entirely serious solution to the problem. Quite apart from the problems of static displays on the MCP CRT using such a high-end scope to display a few digital patterns would be like using a sledgehammer to crack a walnut.

As you state getting greater bandwidth on the X-axis only partially solves the problem; the requirement to blank the trace between the dots might be more challenging than meeting the horizontal BW requirement. With an external Z-axiz blanking signal you would need a fast enough response to suppress the transition between dots. This also translates to a high BW requirement on the Z-axis and even the mighty 7104 might not be sufficient in this respect as its rated external Z-axis BW is only 1MHz.

[bdunham7]

using such a high-end scope to display a few digital patterns would be like using a sledgehammer to crack a walnut.

Just the sort of thing an EEVblogger would do, of course.  I just used a 7.5 digit DMM to check a 12V SLA UPS battery. 

even the mighty 7104 might not be sufficient in this respect as its rated external Z-axis BW is only 1MHz.

The 2445 is 20MHz, I believe, and I still struggled to get a decent pattern with it.  I used quite a bit of overlap and as one can see from my post above, a significant phase shift.  And high-rate Z-axis inputs seem to blur the focus.  Some sort of translation to a raster display would probably look better and be more reliable, although probably not what the OP wants to do.

[srb1954]

using such a high-end scope to display a few digital patterns would be like using a sledgehammer to crack a walnut.

Just the sort of thing an EEVblogger would do, of course.  I just used a 7.5 digit DMM to check a 12V SLA UPS battery. 

Sounds perfectly reasonable to me.

The extra resolution lets you determine the rate of self-discharge in a few seconds. Using a 3.5 digit meter means you have to waste several hours to make the same measurement.