Analog versus digital oscilloscopes 2

[bdunham7]

I just know I've tried several different Tektronix DSOs connected to a vector arcade game board and none of them have displayed a usable picture, my 465B on the other hand displays a beautiful sharp image that looks just like it does on a proper vector monitor. Testing vector game boards is definitely a niche application though.

Yes, any actual vector graphics or anything requiring a Z-axis input, an old Tek is your best friend, almost as good as an actual XY monitor. 

[james_s]

I just know I've tried several different Tektronix DSOs connected to a vector arcade game board and none of them have displayed a usable picture, my 465B on the other hand displays a beautiful sharp image that looks just like it does on a proper vector monitor. Testing vector game boards is definitely a niche application though.

Yes, any actual vector graphics or anything requiring a Z-axis input, an old Tek is your best friend, almost as good as an actual XY monitor.

I actually don't even use the Z input most of the time, the deflection amps are fast enough that the line between the objects isn't visible, the only artifact is a dot in the middle of the asteroids.

[magic]

tggzzz:
I remember one of those, a true "potentiometer", where the wire was supplied mounted on a meter stick.  1 mm resolution was 0.1% of full-scale.
Here's one, made by Central Scientific Co. of Chicago, now in Constantinople:  http://www.hasi.gr/instruments/ele66
We also had some of the original-style Wheatstone bridges, where the resistive legs were varied by placing tapered metal pegs into mating sockets between the wire-wound resistors inside the box.

That's it I suspect the school made their own rather than buy one, though.

Did you verify linearity of this fine instrument, though?

[PA4TIM]

I bought a 100 MHz Rigol (>10 years ago, it was the first version) I also have a nice collection analog scopes, even more back then, upto 250 MHz and 1GHz sampling scopes/plugins. I preferred the analog scopes so much more as that crap Rigol that I gave it away to a student. I almost never used it because measurement results were way off and mechanically it sucked even more (knobs falling off, switches that needed to be pressed several times before it worked.

Then, a year later, I bought a 350 MHz Hameg. (the first models designed by R&S and fabricated in Europe) And that was the end of using my Analog CROs other then for some very rare cases or just for fun. Replaced it this year with a 300 MHz full featured R&S (4 ch, 2 LA 8 ch pods, protocol decoders, bodeplot, arb generator and pattern generator). OK, it costs a lot more as a 40 year old CRO but the CRO scopes that I have where high end versions and far out of reach of a hobbyist when new.  The CRO that you could buy on a hobby budget back then, was not great in quality and performance.  Old analog scopes will die sooner or later. I have no problem repairing them (repaired my 7704 several times, the plugins even more) most of the others came in defect (and for free) as long as I can find parts and that can be a double problem for a beginner (no repair skills and no parts) 7000 or 24XX scopes are not very repair friendly. 465 and alike are a lot more repairable.

My MSO is a swiss knife, great quality and  so much more functions as an analog scope. I have a pristine 100 MHz late Hameg analog scope that has a component tester still above my bench for the rare cases I want a ACRT.

I also have 2 LA's (+80 channel Tek and a 16 ch Zero) but bv far the most used is the LA from my MSO's. But I am in the repair business, not a firmware developer. The scope LA is fast, real time and can deal with higher voltages. For decoding the zero is more convenient (it knows +100 protocols)

[tggzzz]

On eBay you can find a (for example) 4CH Tektronix 2465A 350MHz for ~$150, and at least, bring more quality to your waveform analysis, since it performs well even at 750MHz.

In summary: I think these CROs fill this gap between $350 and $1,350 scopes, as a 2nd scope.

If you can find (or have found) a good condition 2565/A/B for $150 and you have room on the bench and a specific need for its capabilities, then go for it.  Most of the ones I've seen for $150 on eBay are broken or even incomplete.  I have several (2445, 2465B, 2221A, and more) that I have fixed up and the specific reasons I might drag one out are usually related to vector graphics or a specific need for the Z-axis input.  Also, on one occasion I needed something with better input overload recovery, but I also have two other DSOs that would suffice. If I had room to keep one on the bench, I might use it sometimes just because it was there.  IMO the gap that an old high-performance CRO would fill is small and waning, but YMMV.  What do you use yours for?

Undoubtedly many Tek 24x5 and similar machines on fleabay are suspect, and need a little work. But not all. 
Undoubtedly digitising scopes will be the way of the future
But...
A good working analogue scope is a joy to use and is very capable, and can be used for many purposes.
100MHz was fine for digital logic 40 years ago, but modern jellybean logic requires far more than that.

Anybody that declares that only digitising scopes should be considered and that working analogue scopes should be ignored is, IMHO, a twit.

For most of my purposes, a sufficient combination is a 2 channel >300MHz analogue scope (assure signal integrity), plus cheap logic/protocol analyser and printf statements. If those aren't sufficient then you need serious thought or money.

Just about the *only* reason to consider an analog scope outside of a few specialized niche uses like XY mode is price. If you find a good deal on a *working* one or just want the experience of using one then by all means go for it, they are still useful instruments. The key there is finding a working one that is a good deal, these are getting harder to find all the time and they're only going to keep getting harder to find as time goes on. Nobody is making them anymore and every one that gets worn out, blown up, destroyed in shipping, or otherwise scrapped is one less available for someone to buy. A working >300MHz analog scope is a specialized item, they were always expensive and never all that common, I've never even seen one in the flesh, much less actually owned one. Most people are not suggesting ignoring working analog scopes, but it is not really realistic to assume that most people can just go out and find a good deal on one. It's like suggesting someone go out and buy a 30 year old car, I have one, I love it, it gets me around just fine and I wouldn't trade it for anything, but it's still a 30 year old car, some of the parts are hard to come by these days, and if I didn't know how to work on it myself and didn't enjoy working on cars it would be a hassle. A 30 year old scope is the same way.

Basically I agree, except...

Too many people on this forum do explicitly and unequivocably state that working analogue scopes should be ignored - often in favour of a 1054z! Such people also deny that analogue scopes do have some advantages over low-end digitising scopes, and that analogue scopes' relative simplicity is an advantage for a beginner.

[tggzzz]

tggzzz:
I remember one of those, a true "potentiometer", where the wire was supplied mounted on a meter stick.  1 mm resolution was 0.1% of full-scale.
Here's one, made by Central Scientific Co. of Chicago, now in Constantinople:  http://www.hasi.gr/instruments/ele66
We also had some of the original-style Wheatstone bridges, where the resistive legs were varied by placing tapered metal pegs into mating sockets between the wire-wound resistors inside the box.

That's it I suspect the school made their own rather than buy one, though.

Did you verify linearity of this fine instrument, though?

Which do you estimate would be more linear: a 2% moving coil meter or a length of piano wire

[rsjsouza]

On eBay you can find a (for example) 4CH Tektronix 2465A 350MHz for ~$150, and at least, bring more quality to your waveform analysis, since it performs well even at 750MHz.

In summary: I think these CROs fill this gap between $350 and $1,350 scopes, as a 2nd scope.

If you can find (or have found) a good condition 2565/A/B for $150 and you have room on the bench and a specific need for its capabilities, then go for it.  Most of the ones I've seen for $150 on eBay are broken or even incomplete.  I have several (2445, 2465B, 2221A, and more) that I have fixed up and the specific reasons I might drag one out are usually related to vector graphics or a specific need for the Z-axis input.  Also, on one occasion I needed something with better input overload recovery, but I also have two other DSOs that would suffice. If I had room to keep one on the bench, I might use it sometimes just because it was there.  IMO the gap that an old high-performance CRO would fill is small and waning, but YMMV.  What do you use yours for?

Undoubtedly many Tek 24x5 and similar machines on fleabay are suspect, and need a little work. But not all. 
Undoubtedly digitising scopes will be the way of the future
But...
A good working analogue scope is a joy to use and is very capable, and can be used for many purposes.
100MHz was fine for digital logic 40 years ago, but modern jellybean logic requires far more than that.

Anybody that declares that only digitising scopes should be considered and that working analogue scopes should be ignored is, IMHO, a twit.

For most of my purposes, a sufficient combination is a 2 channel >300MHz analogue scope (assure signal integrity), plus cheap logic/protocol analyser and printf statements. If those aren't sufficient then you need serious thought or money.

Just about the *only* reason to consider an analog scope outside of a few specialized niche uses like XY mode is price. If you find a good deal on a *working* one or just want the experience of using one then by all means go for it, they are still useful instruments. The key there is finding a working one that is a good deal, these are getting harder to find all the time and they're only going to keep getting harder to find as time goes on. Nobody is making them anymore and every one that gets worn out, blown up, destroyed in shipping, or otherwise scrapped is one less available for someone to buy. A working >300MHz analog scope is a specialized item, they were always expensive and never all that common, I've never even seen one in the flesh, much less actually owned one. Most people are not suggesting ignoring working analog scopes, but it is not really realistic to assume that most people can just go out and find a good deal on one. It's like suggesting someone go out and buy a 30 year old car, I have one, I love it, it gets me around just fine and I wouldn't trade it for anything, but it's still a 30 year old car, some of the parts are hard to come by these days, and if I didn't know how to work on it myself and didn't enjoy working on cars it would be a hassle. A 30 year old scope is the same way.

Basically I agree, except...

Too many people on this forum do explicitly and unequivocally state that working analogue scopes should be ignored - often in favour of a 1054z! Such people also deny that analogue scopes do have some advantages over low-end digitizing scopes, and that analogue scopes' relative simplicity is an advantage for a beginner.

The problem is in the absolutes indeed. The test equipment is what you make of it and analog can still be useful and beat digital in some applications but it certainly beats having "no scope". So many started with just a multimeter and the leap to an oscilloscope immensely enriches the experience and ability to develop oneself in this field. Obviously that an used analog will require spare time and expertise (and frequently another oscilloscope) to buy, repair and maintain an older very high bandwidth "clunker".

This topic is not new and many long and winded threads were posted already. The one below is already five years old and in our exchange we were already talking about the use cases and pitfalls of buying and owning older gear.

https://www.eevblog.com/forum/testgear/question-about-analog-oscilloscopes-vs-digital-oscilloscopes/msg968187/#msg968187
1
Another issue is locality: IME the wear and tear of the average used oscilloscope available for sale in certain places/countries tends to be proportional to the average price of test gear in that place. After all, the effort to buy one can be so high that you will use it until it runs aground.

[magic]

Which do you estimate would be more linear: a 2% moving coil meter or a length of piano wire

Probably the wire. But even assuming uniform material, just 3% difference in diameter between the two sections is enough to eat 90% of you 0.1% error budget

[TimFox]

One could test this by assembling two identical units (using wire from the same roll) and measuring the error between them at different points using a galvanometer, were it important.

[David Hess]

By 'tangential measurements' are you referring to tangential noise measurement, similar to this but applied more generally to things other than amplifier input noise?

https://w140.com/tekwiki/images/c/c8/Measuring_conventional_oscilloscope_noise_garuts_samuel.pdf

That is right.

I'm not sure why this technique would not work on any DSO, intensity grading or not.

It depends on the display device faithfully reproducing the sum of two probably density functions, which is interesting because as the article you linked shows, it also works on sampling oscilloscopes that do not produce an index graded display.

As for the accuracy of the DSOs built in RMS measurements, there is a significant limitation when measuring input noise because the RMS measurement will include any DC offset, which shows up at very low levels (500uV/div, e.g.) even when AC coupled.  At higher levels of Gaussian noise, the RMS and StDEV match what my AWG says they are.

I am not referring to those errors which do need to be accounted for.  The limitation I am referring to comes from measuring the display record *after* it has been processed for the display.  This destroys the statistical properties of the signal.

I have wondered if it is the fact that so much of the electronics industry has moved overseas that all the used gear is in China now, or perhaps it's a generational thing where gear gets replaced in cycles as new waves of technology appear. For a while the market was flooded with stuff like analog scopes and 70s-80s gear as businesses upgraded to DSOs and other 90s-2000's gear. Then there was a wave of that gear as businesses upgraded and now we are at a point where the last round of high end gear that was purchased is still working. Or maybe the pandemic has resulted in enough people taking up hobbies or getting more into the hobbies they have and buying up supplies of gear leading to shortages. Either way it has been well over a year since I've seen anything resembling a deal on any sort of used test equipment I watch for. Thankfully I have just about everything I really need at this point.

I think it simply took that long for a majority of the existing old test equipment to get sold through Ebay.

Everything on eBay is suspicious, even a new product.

Even if an analog oscilloscope offered on Ebay is not suspicious, shipping can render it suspicious.  One of my Tektronix 7834s has a CRT which was likely damaged in shipping.

[james_s]

I think it simply took that long for a majority of the existing old test equipment to get sold through Ebay.

There should always be old test equipment though, companies are making new equipment constantly and as time marches on that equipment becomes old and in theory should end up on the surplus market. It may come in waves as new technology appears but I still would have thought there would be a steady supply.

[Trader]

I think it simply took that long for a majority of the existing old test equipment to get sold through Ebay.

There should always be old test equipment though, companies are making new equipment constantly and as time marches on that equipment becomes old and in theory should end up on the surplus market. It may come in waves as new technology appears but I still would have thought there would be a steady supply.

Exactly, a newly released device started to become old the next day.

Just a fun fact, a high-end scope released in the 80s, early-90s, discounting the US inflation, is costing around 1% of the launch price, and a high-end scope today, costs roughly the same*.

* based on this info: "1990: $13,045 (2467B) = $24,400 equivalent in 2017" (https://w140.com/tekwiki/wiki/2465B)

[David Hess]

I think it simply took that long for a majority of the existing old test equipment to get sold through Ebay.

There should always be old test equipment though, companies are making new equipment constantly and as time marches on that equipment becomes old and in theory should end up on the surplus market. It may come in waves as new technology appears but I still would have thought there would be a steady supply.

Exactly, a newly released device started to become old the next day.

But more and more used test equipment produced since the 1990s just becomes trash because it is unrepairable and unmaintainable due to lack of service documentation.

[Kleinstein]

It is not just the lack of service manuals that make the repair of newer test instruments difficult. The tend to use FPGAs / CPLDs with fine pitch. So soldering and measurements can be difficult and without the programmit does not help to replace the chip.
The modern electronics got quite reliable, but if it fails a repair is difficult.

In the old days failure was more common and parts were build for repairs - at least many instruments and especially higher cost ones.
Building for easy repairs adds to the costs and this may no longer be worth it. Especially for professional use a less reliable repaired instrument is not a good thing.

There seem to be not many good analog scopes left - at least the offers at ebay are often crap or broken or at least sold as defect to avoid returns. Occasionall there may be a good catch, but that is rare. So an analog scope is increasingly less of a realistic option.
There was a time they build combined analog and digital scopes, but this time is gone since the digital part got better and use of LCD screens.

[gf]

There seem to be not many good analog scopes left - at least the offers at ebay are often crap or broken or at least sold as defect to avoid returns.

Or even deliberately cannibalized (e.g. many/all valves missing), and then sold as defect (w/o specifying what the "defect" is).

Occasionally I see tested/restored units from trusted professional sellers, but then they quickly become expensive, too (say €500+ for a Tek 2465 in working condition).
But everything else is IMO rather a lottery, unless you get the opportuinity to test it before buying. It can be a bargain, but it also can be a total loss of investment at the end.

[nfmax]

The modern electronics got quite reliable, but if it fails a repair is difficult.

In the old days failure was more common and parts were build for repairs - at least many instruments and especially higher cost ones.

This change happened a bit earlier, IMHO, when CRT oscilloscopes started using cutom ICs and hybrids. Older CRT scopes are more repairable, but more likely to be broken. At the low end, the Hameg CRT oscilloscopes are made with standard parts - but even things like dual JFETs are difficult to find these days.

A CRT scope is getting to be a bit like a vintage steam engine - it can certainly 'work', but it's a labour of love (and a money pit) for an expert to keep it in working order. My advice to new electronics hobbyists, and certainly to students, is to forget that CRT scopes ever existed, except as a historical curiosity

[RoGeorge]

I think it simply took that long for a majority of the existing old test equipment to get sold through Ebay.

There should always be old test equipment though, companies are making new equipment constantly and as time marches on that equipment becomes old and in theory should end up on the surplus market. It may come in waves as new technology appears but I still would have thought there would be a steady supply.

I wouldn't count on that, and I'll guess the availability will be less in the future:

  - instruments are not build like tanks any more, nobody designs lab instruments with survivability to a nuke attack in mind
  - less repairable because of the lack of documentation and because of the lack of software/configuration tools
  - we don't know yet how reliable will be a flash memory in 50 years from now, or a FRAM chip (that has a limited number of reads, too, so it will wear out simply by running)
  - since the frenziness of data leakage there is a tendency from big companies to simply destroy the old instruments just to be sure no password or project info will leave the company together with the old instrument
  - there is a fashion to recycle old objects at all cost, even when the old object still has users and usage
  - big manufacturers of instruments realized their old instruments are competing over their own market of cheaper instruments
  - there is a tendency to put unique IDs inside chips, to encrypt the firmware, to link everything with a call home or a unique registration, therefore re-flashing a modern instrument is almost impossible without software tools specific to that line of instruments and specific to that company

Therefore I'll say the abundance of instruments that are affordable, second-hand, but still gorgeous in performance (the kind of boat anchors we see today on ebay) is a unique feature of our times, something that was caused by the cold war era combined with a very specific level of technology.

The instruments back then were seen as standalone and self-contained objects.  Now it is not like that any more.  Now the tendency is to have a box with no screen and no buttons, box that is usually controlled remotely by a computer.  So now the instrument-box suddenly will need a very complex environment around it, an external computer, with external software, and external operating system, and an external network to transfer data, and so on.

Good luck restoring and activating a Windows 10 in the year 2070, so to re-flash or simply to use that cool Tektronix spectrum analyzer box-instrument from 2020, with no buttons and no indicators on it!   

With the current trend, I see everything will be closed and locked-down and encrypted, first to a certain brand and environment, and very soon (as in a few decades away from now) we will see everything (everything as in objects, services, etc.), even your faithful DMM or your faithful handgun, everything will require a personally encrypted key and a valid personally authorization, or else that object/service won't be available to you.

This will happen whether we like it or not, simply because the actual level of technology allows to have that.

[Trader]

With the current trend, I see everything will be closed and locked-down and encrypted, first to a certain brand and environment, and very soon (as in a few decades away from now) we will see everything (everything as in objects, services, etc.), even your faithful DMM or your faithful handgun, everything will require a personally encrypted key and a valid personally authorization, or else that object/service won't be available to you.

IMHO, this is a real threat to electronics at all, basically, the future will be just buy, use, dispose of.

In the past, I was really considered buying the "121GW DMM", just because I thought it was an open-source project. (https://www.eevblog.com/forum/testgear/121gw-why-no-open-source/)

Is there a project for an open-source scope (https://www.eevblog.com/forum/testgear/a-high-performance-open-source-oscilloscope-development-log-future-ideas/?all)

I'll really consider buying it instead of the best brands, just to support this project and see it improve even more.

[AaronLee]

With the current trend, I see everything will be closed and locked-down and encrypted, first to a certain brand and environment, and very soon (as in a few decades away from now) we will see everything (everything as in objects, services, etc.), even your faithful DMM or your faithful handgun, everything will require a personally encrypted key and a valid personally authorization, or else that object/service won't be available to you.

IMHO, this is a real threat to electronics at all, basically, the future will be just buy, use, dispose of.

In the past, I was really considered buying the "121GW DMM", just because I thought it was an open-source project. (https://www.eevblog.com/forum/testgear/121gw-why-no-open-source/)

Is there a project for an open-source scope (https://www.eevblog.com/forum/testgear/a-high-performance-open-source-oscilloscope-development-log-future-ideas/?all)

I'll really consider buying it instead of the best brands, just to support this project and see it improve even more.

After reading the post just before yours (RoGeorge's post that you were responding to), I thought to myself how open-source hardware would affect things. Then reading your post, was surprised that already it's being seriously considered.

Most large companies / high-end professionals likely won't be interested in open-source hardware, but certainly hobbyists and semi-professionals would if it meant a big savings and/or equipment that would last much longer due to the public-source nature and being able to fix, maintain, and perhaps even upgrade it easily.

As time goes by, open-source (not just test equipment hardware, but all sorts of open-source projects) become more and more advanced, and the difference between A-level mass produced stuff and the open-source alternatives becomes less and less. Take scopes, for instance. I doubt many engineers would have the expertise to design a scope at the same level as one of the A-level companies could. Even though an engineer may be expert in one particular area that applies to designing a scope, most likely they aren't in all areas. The engineers that are expert in all areas are likely working for a scope manufacturer. But in time that information leaks out, and becomes more general knowledge, and it starts to make open-source practical and good enough to be in the same ball-park as professional lower-end gear. And even if you don't have one engineer doing the whole things, you get enough well-qualified engineers willing to donate their time to an open-source project, and you still have a shot at getting it right.

Add on things like 3-D printing, where in some cases you don't need to mass produce the housing, being producing molds for the housing alone can cost a very large amount. Or some hobbyists might not care, and be able to just take some off-the-shelf case to make a scope. Especially if the scope was in two parts, whereby a PC is used as the screen, and a generic box with knobs and such. (Unlike a typical USB scope which lacks the physical controls of a bench scope). Anyways, that's just a thought off the top of my head, and maybe there's some big issues that would make it impractical. The point is, I think open-source is only going to grow in the future, and grow at a much faster rate, and it's inevitable in my mind that really practical and good open-source test equipment will be on option someday.

[james_s]

There was at least one impressive attempt at an open source DSO, but the problem is that it doesn't result in savings. IIRC it was projected to cost as much or more than a comparable commercial offering, and that's before you factor in the time and effort to build the thing and figure out some sort of enclosure for it. In the end it only appeals to people who are either charitable and idealistic enough to want to support an open source project for the sake of being open source, or the very tiny minority who have the skills and motivation to get deep into the code and customize it to suit their needs. Personally I thought it was interesting from the standpoint of seeing how a modern DSO works, but I'm not going to pay more for something I have to build myself and then rely on community support for when I can go out and buy a perfectly usable scope for under $500 that comes in a nice molded housing with a carrying handle and professional support if I have a problem with it.

[Trader]

After reading the post just before yours (RoGeorge's post that you were responding to), I thought to myself how open-source hardware would affect things. Then reading your post, was surprised that already it's being seriously considered.

There was at least one impressive attempt at an open source DSO, but the problem is that it doesn't result in savings.

I want to highlight 2 successful open-source projects that still improving more and more:

1) NanoVNA (Vector Network Analyzer) (https://nanorfe.com/nanovna-v2.html). A few years ago, a similar VNA could cost a Lot, now it's ~$50. Alan (w2aew) and Joe Smith did excellent videos about its awesome capabilities.

2) DSP Receiving Radios - the Si4732/35, and PU2CLR library (https://github.com/pu2clr/SI4735) allows you to build an excellent DSP radio. (https://www.facebook.com/groups/532613604253401/)

I also know about some people building an open-source Spectrum Analyzer, very promising.

I think this success is because the Chinese companies got the NanoVNA and PU2CRL libraries, created the hardware, and released new Mass Products for a cheap price.

When this open-source DSO achieve very good capability, I'm sure some Asian company will manufacture and commercialize it for a competitive price (like the NanoVNA).

The DSOs will be like PCs, you can buy good hardware (and upgrade CPU, Memory, Disk, etc), and install/update the new OS distribution version (like Linux flavors) but with All Options opened (Mpts, MSO, FFT, PWR, AWG, Serial Decode, etc), no need to hack it, and people could contribute developing more plug-ins/add-ons, like THD measure, statistical graphics, etc.

Maybe the best the big brands can do right now to keep the big market share is, at least, allowing people to develop apps to be installed in their DSOs.

[AaronLee]

The DSOs will be like PCs, you can buy good hardware (and upgrade CPU, Memory, Disk, etc), and install/update the new OS distribution version (like Linux flavors) but with All Options opened (Mpts, MSO, FFT, PWR, AWG, Serial Decode, etc), no need to hack it, and people could contribute developing more plug-ins/add-ons, like THD measure, statistical graphics, etc.

Maybe the best the big brands can do right now to keep the big market share is, at least, allowing people to develop apps to be installed in their DSOs.

Yes, and for some people, having the software being open-source as well will allow them to modify it for their particular needs. There's probably lots of new ways the equipment can be used or used much more efficiently in existing ways, if the users are just allowed access to the software. But the big brands don't have the resources to tailor their products to each and every need. Only they tailor it to mainstream needs where they can sell a lot of units to pay for their development costs. As the number of niche applications are expanded via open source, the number of protentional users also expands, making open source able to take a bigger piece of the total pie.

[TimFox]

Another related question:  Before I retired, my employer purchased some expensive test equipment that ran Windows XP internally.  Did Agilent/Keysight upgrade the OS thereafter?

[David Hess]

It is not just the lack of service manuals that make the repair of newer test instruments difficult. The tend to use FPGAs / CPLDs with fine pitch. So soldering and measurements can be difficult and without the programmit does not help to replace the chip.
The modern electronics got quite reliable, but if it fails a repair is difficult.

This change happened a bit earlier, IMHO, when CRT oscilloscopes started using cutom ICs and hybrids. Older CRT scopes are more repairable, but more likely to be broken. At the low end, the Hameg CRT oscilloscopes are made with standard parts - but even things like dual JFETs are difficult to find these days.

I disagree for two reasons:

With some very specific exceptions like the improperly mounted hybrids in the Tektronix 2400 series analog oscilloscopes and Mostek ROMs, the custom ICs and hybrids, and this extends to programmable logic now, are more reliable than other parts and tend to not be what fails.

The most common parts to fail tend to be commonly available ones.  Most of the repairs that I do are for bad or worn out capacitors, mechanical failure including those edge wipe sockets from Texas Instruments, and failure in jelly bean semiconductors.

[james_s]

Another related question:  Before I retired, my employer purchased some expensive test equipment that ran Windows XP internally.  Did Agilent/Keysight upgrade the OS thereafter?

I seriously doubt it. I've got two medical vital signs monitors on the bench right now, both are embedded PCs, one of which runs XP embedded and the other Win7. They are treated as appliances by the manufacture, they do not get upgraded. Once the OS is out of support the device is out of support, and these are machines that cost $10k to replace.