Professional Scope Purchasing Criteria

[TunerSandwich]

Hameg isn't really mid-range (which starts with the R&S RTM), all the Hameg kit is solidly placed in the entry-level sector. Even their top-of-the-range scope (HMO3000) is still a entry-level scope. 

I was taking into consideration the context of this forum.  I think we have discussed this in the past.  Maybe an edit to that statement of mine is due?

compared to the commonly discussed product tiers, in this forum, I think Hameg represents the mid-range price point.....feature set follows suit.  However I still think they have one of the best UI's and measurement display methodologies on the market...and yes I will agree their screen implementation itself is not fantastic....

I would personally like to see them, or another manufacturer, implement the same measurement markers/tags/cursors into a truly "high end" scope....

One thing I am finding a bit annoying about the LeCroy is the way it wastes screen space for simple measurements.....I mean at least a floating frequency counter would be nice to have (maybe it's there and I don't know it?).....

[Wuerstchenhund]

Hameg isn't really mid-range (which starts with the R&S RTM), all the Hameg kit is solidly placed in the entry-level sector. Even their top-of-the-range scope (HMO3000) is still a entry-level scope. 

I was taking into consideration the context of this forum.

AH, I see. Yes, I think in the context of that the majority of discussions on this forum is around entry level gear it's probably fair to say that Hameg sits quite well in the mid-range.

However I still think they have one of the best UI's and measurement display methodologies on the market...and yes I will agree their screen implementation itself is not fantastic....

It's not that bad (for a truly horrible use of screen area look no further than to the Siglent SDS2000), but yes they could improve it further.

I would personally like to see them, or another manufacturer, implement the same measurement markers/tags/cursors into a truly "high end" scope....

I doubt we will see that anytime soon, as feature-wise most low end manufacturers seem to more or less copy the same feature sets amongst them. Yes, they get cheaper, and the cheaper ones are built better than before, but they still come with the same limited functionality which I think is a shame.

I guess the cheapest new scope with advanced measurement capabilities these days is the LeCroy WaveSurfer 3000 (which starts at roughly $3k) aka the Siglent SDS3000 (which seems to be only sold in China).

One thing I am finding a bit annoying about the LeCroy is the way it wastes screen space for simple measurements.....I mean at least a floating frequency counter would be nice to have (maybe it's there and I don't know it?).....

No, there isn't one. Don't forget that the "Frequency Counter" which is available in some cheaper scopes is exactly that, a hardware frequency counter built into the scope. Such stuff is predominantly found in low end scopes only.

To save screen area you can disable Statistics display in the Maths section and then select "Frequency" for the appropriate channel.

[TunerSandwich]

No, there isn't one. Don't forget that the "Frequency Counter" which is available in some cheaper scopes is exactly that, a hardware frequency counter built into the scope. Such stuff is predominantly found in low end scopes only.

To save screen area you can disable Statistics display in the Maths section and then select "Frequency" for the appropriate channel.

Yes I have been doing that, as well as removing the somewhat useless "histicons"....

I have just been using a cursor to grab a quick view of certain measurements.....I suppose this is where these scopes get good....with all the custom math and measurement tools we can "fake it"

I was going to start playing with the XDEV tool-set.  I was noticing some ability to write UI changes with VB.

One thing I really wish I could do is change that awful purple/blue in the DSO application.....the new black version they have is much better on contrast.....

[Wuerstchenhund]

Yes I have been doing that, as well as removing the somewhat useless "histicons"....

I have just been using a cursor to grab a quick view of certain measurements.....I suppose this is where these scopes get good....with all the custom math and measurement tools we can "fake it"

The X-Stream scopes are great if you need lots of parameters, but I think the display of only one or two parameters was solved a bit better with their older non-Windows scopes. Oh well, I guess we can't have it all.

I was going to start playing with the XDEV tool-set.  I was noticing some ability to write UI changes with VB.

Yes, you can. In fact, you could write your own scope UI, using all the capabilities that are in the scope. It's really great.

One thing I really wish I could do is change that awful purple/blue in the DSO application.....the new black version they have is much better on contrast.....

It's already in the softwar on your scope. What "theme" it uses is decided by the scope model the software installed on.

Also, don't forget that newer models run Windows Vista/7 which support Aero effects, and the nice semi-transparent blocks will look dull on Windows XP.

The black scheme certainly looks more modern, though.

[rx8pilot]

Some more technical insight would be helpful in developing a purchasing criteria.....I.E. how much flexibility do you really want, and what exactly do you envision using the scope for?  What is day to day for me, might be considered "minority" for others...

I can say what I am doing today and what I hope to be doing in the future. At the moment I am designing power related products for industrial/professional imaging systems. This generally means low voltage - medium current from multiple sources being distributed to multiple outputs at various voltages. All of this is managed and monitored with a blend of analog comparators, 16bit A/D, MCU's, i2c, SPI, 1-wire, boost converters, buck converters. I need to see the low mV drop across a resistor compare with an i2c event. Maybe monitor a comparator input to gate timing on routing FETs. Converter design has me looking at shaping the PWM gate pulse while monitoring noise/ripple at various loads. All of these things are normal measurements for most anyone. There are so many things that I am not capable of doing with my limited equipment on my bench right now so I hardly know what to hope for in a new scope. Triggering on digital data would be a great benefit. Right now, trying to temporally tie a Rigol scope with a Saleae logic analyzer is very time consuming. Looking for a specific i2c command and understanding it relationship to an analog event is a pain. Window testing sound really great. As the circuit switches from one power source to another at full load, it can only drop so much before things go bad. That is a simple measurement, but some effort involved in comparing to a previously known good waveform. Some of the advanced math functions I see in the higher-end are outside of my understanding. I would like to learn more about why various math options exist and how they can be used in analysis and understanding of the signals.

For the future, I have been studying FPGA dev for the purpose image transformation. This will bring a whole new crop of measuring challenges that I am not even aware of yet. Hopefully the scope I choose today will be enough to get me through the learning curve and at least to the point where I can release my first FPGA device. The cost of developing a new product will involve way more than just a scope and an FPGA dev environment so I am hoping that I don't have to buy a new scope right at the end of development when cash is super tight. If this happens, I will be presented with very fast serial signals at 3gbps+ and I am thinking of renting a scope for that period of development if possible. There is not a long term need to monitor the serial signals beyond knowing that they are properly making their way into the FPGA and back out. After that is worked out, I would be back into the slower parallel world for oscilloscope measurements and the images will be measured on dedicated image monitoring gear - totally separate from an oscilloscope.  It's really hard to understand what I may be needing for a project like this.

Clearly I am looking for a general purpose scope that will be the least likely to limit my design options in the near-ish future. I would be happy if it was relevant for 3 years but that is a random guess. There is no real way to understand how long it will be before the scope limits my progress. Therefore, I would like to choose something that is significantly more advanced than I need today while not being such a financial burden that it slows down the business. The 'upgradable' scopes seem interesting because I can spread the cost out and choose what I need exactly when I need it. For the moment, I am guessing what the future may bring. I know that I could get everything I would ever need for $30k, but that would mean I would be a slave to the scope for a while. I am also building a small P&P line and spending about $10k+ on other bench gear so I definitely need to balance the budget wisely.

Beyond cost, I am certainly worried about the day-to-day experience and ease-of-use. If two scopes are the same purchase price with nearly identical features, but one is a little slow and cumbersome - it may double or triple the cost of that scope over 3 years in labor and missed opportunities. Intuitive is just as critical as bandwidth IMHO. I have learned that lesson in my CNC shop over and over.

[tautech]

There are a few tricks you can use for the development of high speed stuff.
A client had a memory bus running at 800 MHz that they needed to check timing, signal levels, integrity etc.
Halved the clock speed and it was visible with a 300 MHz DSO.
Answered their questions and they moved on.
Sure a DSO that matches or betters the bandwidths in a circuit is nice but not always necessary.
Its called measurement techniques  and can save lots of money.

Modern DSO's generally have a substantial serial triggering suite and you will have to find one that has the flexibility you need. You may have to option this in for some brands/models.

[AndyC_772]

That sounds like an absolutely classic example of not understanding the difference between clock rate and edge speed. High speed signal integrity is all about rise and fall times, and they're governed by the design of the silicon and the PCB. The number of transitions per second is largely irrelevant, and it's amazing just how many engineers don't really appreciate and understand the difference.

A 300 MHz scope won't show anything useful on a memory bus designed to be able to run at 800 MHz. Very few scopes will; the parasitic effect of the probes dramatically affects the signal shape, regardless of the scope's bandwidth. The only real way to know what signals on a high speed memory interface look like is to simulate them.

[Wuerstchenhund]

There are a few tricks you can use for the development of high speed stuff.
A client had a memory bus running at 800 MHz that they needed to check timing, signal levels, integrity etc.
Halved the clock speed and it was visible with a 300 MHz DSO.
Answered their questions and they moved on.
Sure a DSO that matches or betters the bandwidths in a circuit is nice but not always necessary.
Its called measurement techniques  and can save lots of money.

I have to agree with AndyC_772 here, signal integrity on these busses very much depends on edges, and the major factor are the vias. Aside that just halfing the clock speed doesn't mean you would get the same results even when measured correctly, the 300MHz scope may have shown something, but certainly nothing that was a relevant denominator of signal integrity.

You can properly measure such busses but I'd wouldn't touch it with anything less than a good 2GHz scope and transmission line probes which are designed for such stuff (and even with them the parasitic effects can be noticable impact the measurement).

[rx8pilot]

Verifying integrity of 800Mhz bus sounds expensive and time consuming in addition to needing to be very knowledgeable in how to measure without killing what you are measuring. Even at much lower levels, that has been a challenge for me - don't kill the signal you are measuring or you will be chasing your tail perpetually.

[nctnico]

An 800MHz 'bus' is probably a serial link. A simple TDR setup is all you need to check for impedance mismatches and discontinuities.

[tautech]

Verifying integrity of 800Mhz bus sounds expensive and time consuming in addition to needing to be very knowledgeable in how to measure without killing what you are measuring. Even at much lower levels, that has been a challenge for me - don't kill the signal you are measuring or you will be chasing your tail perpetually.

Yes of course, however the result was a clear pass in this case.
For debugging one would never attempt this, this was more just a test of proof of concept.
Should there have been problems there was funds available for hire of appropriate gear.

[AndyC_772]

Well, OK... but what you'd see on a 300 MHz scope screen looks nothing like the actual shape of signals that exist on a bus capable of 800 MHz. The probe alone will wreck the shape of the signal beyond all recognition, and the low pass filtering effect of the scope ensures you don't even see the wreckage properly.

I'm guessing we're talking about DDR2 or DDR3 SDRAM... right?

[tautech]

Well, OK... but what you'd see on a 300 MHz scope screen looks nothing like the actual shape of signals that exist on a bus capable of 800 MHz. The probe alone will wreck the shape of the signal beyond all recognition, and the low pass filtering effect of the scope ensures you don't even see the wreckage properly.

I'm guessing we're talking about DDR2 or DDR3 SDRAM... right?

IIRC it was DDR2 but not running @ 800 MHz only about half that.

My SDS2304 was on loan to this client at the time for appraisal and this test. Yes there were problems with probe loadings at those frequencies.
The probes that were supplied with my pre-release DSO were rated at 300 MHz and were found to inadequate for that job.

A BK Precision 500 MHz passive probe was then purchased for this test.
We did seriously consider an active probe at the time but chose to to get the BK as intermediate measure as it was only ~ $115.

However the probes supplied with the same DSO today are of better quality.
How do I know, I delivered a new SDS2204 today. 

Though we were well past the -3dB point there was enough info available(rf-loops results SDS2000 thread) to project the DSO bandwidth roll off out to the tested frequency for a rough result.

Most experienced users are aware that any test connection will affect the DUT or give incorrect results, but its about knowing that and taking steps to minimise errors, maybe that includes shaking the moths out of ones wallet. 

[coppice]

Most experienced users are aware that any test connection will affect the DUT or give incorrect results

Is that really true? I find a serious lack of skepticism about the information presented by instruments of various types. We all go chasing through smoke and mirrors on the odd occasion, but some people seem to make their entire career out of that.

[nctnico]

Well, OK... but what you'd see on a 300 MHz scope screen looks nothing like the actual shape of signals that exist on a bus capable of 800 MHz. The probe alone will wreck the shape of the signal beyond all recognition, and the low pass filtering effect of the scope ensures you don't even see the wreckage properly.

A standard probe is the wrong tool for these kind of jobs even if they are rated for 300MHz they simply cannot be used at that frequency in any real circuit. An active or low-Z passive divider probe is the only choice and as a rule of thumb you'll need such a probe for measuring signals over 100MHz.

[coppice]

Well, OK... but what you'd see on a 300 MHz scope screen looks nothing like the actual shape of signals that exist on a bus capable of 800 MHz. The probe alone will wreck the shape of the signal beyond all recognition, and the low pass filtering effect of the scope ensures you don't even see the wreckage properly.

A standard probe is the wrong tool for these kind of jobs even if they are rated for 300MHz they simply cannot be used at that frequency in any real circuit. An active or low-Z passive divider probe is the only choice and as a rule of thumb you'll need such a probe for measuring signals over 100MHz.

Even when you use an active probe you need to take great care about how it is attached to the circuit to get anything close to the true waveform on the screen of a scope. Accurately probing wideband signals is a task for experts, even if they are equiped with the most exotic tools in the world.

[Wuerstchenhund]

IIRC it was DDR2 but not running @ 800 MHz only about half that.

If it was DDR2 then it was running at 200MHz clock. Still, the problems described before remain an issue even at that lower clock rate.

My SDS2304 was on loan to this client at the time for appraisal and this test. Yes there were problems with probe loadings at those frequencies.
The probes that were supplied with my pre-release DSO were rated at 300 MHz and were found to inadequate for that job.

A BK Precision 500 MHz passive probe was then purchased for this test.
We did seriously consider an active probe at the time but chose to to get the BK as intermediate measure as it was only ~ $115.

The thing is that a normal passive probe is completely inadequate for probing a DDR2-800 memory bus, independent who made it or what the bandwidth is.

I'm sure You have seen something on your scope but certainly nothing representative of what's really going on with that line. You really need a low-z passive or active probe for that.

Most experienced users are aware that any test connection will affect the DUT or give incorrect results

Not in my experience. In fact, I've seen many engineers (even experienced ones) absolutely underestimate the impact the test kit can have on the UUT, and then draw conclusions from bogus measurements.

I lost count of how many times I encountered situations as your above example with the memory bus.

its about knowing that and taking steps to minimise errors, maybe that includes shaking the moths out of ones wallet.

The problem is that many engineers aren't even aware of the basics behind the kit thei're probing, let alone the impact introducing that probe into the circuit. The thing is that you really need to know what you're doing, otherwise you'll never realize that the nice measurements you're just taken are worthless.

[coppice]

Most experienced users are aware that any test connection will affect the DUT or give incorrect results

Not in my experience. In fact, I've seen many engineers (even experienced ones) absolutely underestimate the impact the test kit can have on the UUT, and then draw conclusions from bogus measurements.

I lost count of how many times I encountered situations as your above example with the memory bus.

When you help someone out in a situation like that they typically say something like "I learned from that. I won't make that mistake again.", yet the next time you see them there's a high chance they are making exactly the same mistake. People are really bad at learning from their mistakes. Look how many divorces are rapidly followed by another marriage. 

[Wuerstchenhund]

When you help someone out in a situation like that they typically say something like "I learned from that. I won't make that mistake again.", yet the next time you see them there's a high chance they are making exactly the same mistake. People are really bad at learning from their mistakes. Look how many divorces are rapidly followed by another marriage. 

Indeed. Not that institutions (i.e. corporations) are better at that, though. Just have a look at outsourcing.

[Company A]
"We'll save costs by outsourcing IT/HR/whatever"
=> disaster

[Company B]
"We'll save costs by outsourcing IT/HR/whatever. Company A did it and failed, but we're different and will succeed!"
=> disaster

[Company C]
"We'll save costs by outsourcing IT/HR/whatever. Companies A and B did it and it failed both times, but we're different and will succeed!"
=> disaster

[Company D]
"We'll save costs by outsourcing IT/HR/whatever. Companies A, B and C did it and it failed every time, but we're different and will succeed!"
=> disaster

[Company E]
"We'll save costs by outsourcing IT/HR/whatever. Companies A, B, C and D did it and it failed every time, but we're different and will succeed!"
=> disaster

... To Be Continued

[TunerSandwich]

Some more technical insight would be helpful in developing a purchasing criteria.....I.E. how much flexibility do you really want, and what exactly do you envision using the scope for?  What is day to day for me, might be considered "minority" for others...

I can say what I am doing today and what I hope to be doing in the future. At the moment I am designing power related products for industrial/professional imaging systems. This generally means low voltage - medium current from multiple sources being distributed to multiple outputs at various voltages. All of this is managed and monitored with a blend of analog comparators, 16bit A/D, MCU's, i2c, SPI, 1-wire, boost converters, buck converters. I need to see the low mV drop across a resistor compare with an i2c event. Maybe monitor a comparator input to gate timing on routing FETs. Converter design has me looking at shaping the PWM gate pulse while monitoring noise/ripple at various loads. All of these things are normal measurements for most anyone. There are so many things that I am not capable of doing with my limited equipment on my bench right now so I hardly know what to hope for in a new scope. Triggering on digital data would be a great benefit. Right now, trying to temporally tie a Rigol scope with a Saleae logic analyzer is very time consuming. Looking for a specific i2c command and understanding it relationship to an analog event is a pain. Window testing sound really great. As the circuit switches from one power source to another at full load, it can only drop so much before things go bad. That is a simple measurement, but some effort involved in comparing to a previously known good waveform. Some of the advanced math functions I see in the higher-end are outside of my understanding. I would like to learn more about why various math options exist and how they can be used in analysis and understanding of the signals.

For the future, I have been studying FPGA dev for the purpose image transformation. This will bring a whole new crop of measuring challenges that I am not even aware of yet. Hopefully the scope I choose today will be enough to get me through the learning curve and at least to the point where I can release my first FPGA device. The cost of developing a new product will involve way more than just a scope and an FPGA dev environment so I am hoping that I don't have to buy a new scope right at the end of development when cash is super tight. If this happens, I will be presented with very fast serial signals at 3gbps+ and I am thinking of renting a scope for that period of development if possible. There is not a long term need to monitor the serial signals beyond knowing that they are properly making their way into the FPGA and back out. After that is worked out, I would be back into the slower parallel world for oscilloscope measurements and the images will be measured on dedicated image monitoring gear - totally separate from an oscilloscope.  It's really hard to understand what I may be needing for a project like this.

Clearly I am looking for a general purpose scope that will be the least likely to limit my design options in the near-ish future. I would be happy if it was relevant for 3 years but that is a random guess. There is no real way to understand how long it will be before the scope limits my progress. Therefore, I would like to choose something that is significantly more advanced than I need today while not being such a financial burden that it slows down the business. The 'upgradable' scopes seem interesting because I can spread the cost out and choose what I need exactly when I need it. For the moment, I am guessing what the future may bring. I know that I could get everything I would ever need for $30k, but that would mean I would be a slave to the scope for a while. I am also building a small P&P line and spending about $10k+ on other bench gear so I definitely need to balance the budget wisely.

Beyond cost, I am certainly worried about the day-to-day experience and ease-of-use. If two scopes are the same purchase price with nearly identical features, but one is a little slow and cumbersome - it may double or triple the cost of that scope over 3 years in labor and missed opportunities. Intuitive is just as critical as bandwidth IMHO. I have learned that lesson in my CNC shop over and over.

That is VERY similar to the type of work I do.  DO NOT underestimate the need for bandwidth.  I think a lot of engineers falsely assume that "because my SMPS is below 1MHz, I don't need anything more". 

Nothing could be further from the truth.  You need to consider the ability to analyze feedback loop interactions and external resonant tanks.  Do you need to go out into the GHz ranges?  chances are no, BUT it really depends on the interaction between your devices, and the PCB (often overlooked). 

FPGA are very sensitive to step responses, and although most HF artifacts in POL/SMPS are very low in energy, they can cause instability in the sense loops of many POL devices.  These problems can lead to inefficiency and can also lead to thermal issues.  Especially devices that have large current ripple. 

The other consideration (often overlooked) is the rise time of highly critical signals.  Especially in shared system buss architectures.  Remember that bandwidth on the front end also relates to rise time. 

I think you would be selling yourself VERY short if you didn't look at these very real scenarios.  I have been there quite a few times.  Everything will seem fine, until you have an UN-explainable problem.  Then you are really going to wish you had the ability to properly capture fast, erroneous events.

The next MOST critical factor, is your probing solution.  In fact I would be willing to say that your probing solution is MORE important than your acquisition device.  As with most things, the weakest link in the signal chain will dictate your results. 

Another consideration is the use of ext Vref.  Do you have a meter spec'd that is appropriate for calibrating Vref for a 16 bit A/D?  Also do you have a calibration reference for all of these tools? 

The next question would be about clocking solution.  What are your plans for clocking?  The MCU's and other devices are going to need to lock to something external.....and using GPIO pins for PWM on a micro is just not going to work out.  You will be looking at max PWM of around 500ish HZ.  I am assuming you know of that quite well already, and would be using serial i/o for switching, BUT this is generally not a good way of doing this.  You will need to be able to respond to external errors (based on sensor f/b) much faster than the MCU can provision for.  I have tried many experiments on multiplication of MCU PWM, with mostly unacceptable results. 

I could go on and on here about WHY more bandwidth is necessary (beyond 100 MHz) but obviously there is a limit.  I think a 300-400 MHz front end is MUCH more appropriate, and most tools that you are going to consider and NOT going to offer "upgradeable" bandwidth.  Consider this...."upgrading" bandwidth simply means that you ALWAYS had the front end to begin with, and that the manufacturer has simply crippled it, in order to get more $ from you. 

If I was in your situation I would rent or lease some "high end" tools....and do some comparison.  Analyze some of your signals at both 100 MHz and then at 400 MHz or above.  You will gain some serious insight into just what you are missing in the lower BW device.  Also the probing solution is CRITICAL, I can't stress this enough.  If the properties of your probing solution create resonant tanks of their own, you are really flying blind.  Also consider the dynamic range of the probing solution and front end.  You need a solution that can respond to extremely fast state changes and can resolve the step response of your power devices.  FPGAs are tricky that way...

P.S.

one important thing to note is the usefulness of FFT based operators, in characterizing shifting impedance vs frequency.....you obviously need a front end that can acquire pesky harmonics (or fundamentals) but ALSO need the sample rate to makes sense of it.  Since FFT resolution and bandwidth is proportional to half the sample rate of the acquisition device, you aren't going to find many options in large sample rate, low bandwidth scopes.....do you really need to look out to 5GHz+ for oscillation/resonance in your step response?  well no i suppose not, BUT you might miss harmonic interactions in your switching devices, that can cause errors in proper voltage regulation.....

I could post a screenshot of a budget scope FFT (DS2102A-S) vs the WR64MXi....showing the exact same signal (from a POL).  It will clearly show just what you would be missing (in the signal interpolation) between the decreased bandwidth (and crap probing solution) of the Rigol unit, and the increased bandwidth (and proper FET probing solution....). 

Obviously it's an apples to oranges comparison....and that is the point.  The proper way to do that comparison would be with differential probes and a well engineered test, but the differences in captured data are SO SIGNIFICANT, I think the problems will be immediately obvious. 

Harmonic content in your step response, with complex digitally controlled POL devices, is CRITICAL in properly driving dynamic loads.....I can't think of a more picky dynamic load than a modern FPGA/ASIC/SoC

Also these harmonic interactions and large current oscillations (as a byproduct) can cause massive premature damage to multilayer SMD caps....and if your sensors are sensitive to harmonics and those physical resonances/harmonics are existing on the same PCB/plane...well you know where this is going.....you will get huge instability and oscillation down the line....this can even go as far as LF resonances that will easily destroy the inductor of some POL devices.....not to mention the power lost.

[rx8pilot]

Excellent info from someone going down a similar path, but further along than I.

As I have started the educational process of understanding the finer points of measurement, it has become clear that it is a field of specialty all its own. Just like machine shops have a separate metrology department because it is so challenging to measure properly. Probes have REALLY surprised me in range of capability and difficulty of proper usage just to get the signal to whatever instrument I may have without altering it to the point of chasing my tail. At this point I am putting in a significant number on the probe line of the budget - I want my talent to be the limitation, not the probe. I can always learn how to do it better, but it the probe is the wrong choice in the first place, I am doomed regardless of how well I think through the test.

For bandwidth - good points. I realized that the price difference from 100Mhz to 200Mhz is generally pretty small. Makes it easy to get started budgeting for 200Mhz. Stepping up from there starts a steep pricing curve so I need to think pretty hard about where to start. I would certainly prefer to spend the money to get 350Mhz or 500Mhz if I knew that it would save my butt by showing me something that would be invisible on a slower instrument. It is a little hard to guess, however, it is never going to be a problem to have too much bandwidth on the scope from a testing perspective - just a financial one. As others have pointed out, upgrade-ability is not a white knight riding in to save the day.

I could post a screenshot of a budget scope FFT (DS2102A-S) vs the WR64MXi....showing the exact same signal (from a POL).  It will clearly show just what you would be missing (in the signal interpolation) between the decreased bandwidth (and crap probing solution) of the Rigol unit, and the increased bandwidth (and proper FET probing solution....). 

If you have screenshots, it would be helpful for sure!

Thank you to all those that have thrown in various tips, knowledge, experience..... this discussion has been very helpful. The marketing materials and other skin-deep information available from Google searches is not enough to make an informed decision. Actual users making real products offering information is really great.

[LabSpokane]

I realized that the price difference from 100Mhz to 200Mhz is generally pretty small. Makes it easy to get started budgeting for 200Mhz. Stepping up from there starts a steep pricing curve so I need to think pretty hard about where to start.

I was in the same boat as well with the sticker shock on the >200 MHz scopes. I went 200 MHz, 4 channel MSO with the thought that as soon as I outgrew the scope I could sell or trade in. I figure by that time I'll have revenue to justify the equipment upgrade, and the scope I have now will still be good to keep or have value to resell.

[TunerSandwich]

Excellent info from someone going down a similar path, but further along than I.

As I have started the educational process of understanding the finer points of measurement, it has become clear that it is a field of specialty all its own. Just like machine shops have a separate metrology department because it is so challenging to measure properly. Probes have REALLY surprised me in range of capability and difficulty of proper usage just to get the signal to whatever instrument I may have without altering it to the point of chasing my tail. At this point I am putting in a significant number on the probe line of the budget - I want my talent to be the limitation, not the probe. I can always learn how to do it better, but it the probe is the wrong choice in the first place, I am doomed regardless of how well I think through the test.

For bandwidth - good points. I realized that the price difference from 100Mhz to 200Mhz is generally pretty small. Makes it easy to get started budgeting for 200Mhz. Stepping up from there starts a steep pricing curve so I need to think pretty hard about where to start. I would certainly prefer to spend the money to get 350Mhz or 500Mhz if I knew that it would save my butt by showing me something that would be invisible on a slower instrument. It is a little hard to guess, however, it is never going to be a problem to have too much bandwidth on the scope from a testing perspective - just a financial one. As others have pointed out, upgrade-ability is not a white knight riding in to save the day.

I could post a screenshot of a budget scope FFT (DS2102A-S) vs the WR64MXi....showing the exact same signal (from a POL).  It will clearly show just what you would be missing (in the signal interpolation) between the decreased bandwidth (and crap probing solution) of the Rigol unit, and the increased bandwidth (and proper FET probing solution....). 

If you have screenshots, it would be helpful for sure!

Thank you to all those that have thrown in various tips, knowledge, experience..... this discussion has been very helpful. The marketing materials and other skin-deep information available from Google searches is not enough to make an informed decision. Actual users making real products offering information is really great.

The great thing, is that in the used market, bandwidth doesn't fetch quite the premium it does in the new retail market.  The proportional losses of new value, on higher bandwidth front-ends, vs used...greatly benefit the buyer.  I can grab a couple hundred "extra" MHz for virtually no additional money.  Obviously we are talking about upper mid to high end scopes. 

ABSOLUTELY measurement of electronics is an art form and diverse skill set of it's own.  The other factor in a differential probing solution is safety.  A lot of people overlook this.  Remember that the negative signal in an active/differential probe is floated from earth ground.  In fact thinking of it as ground is wholly wrong.  It's a second, isolated signal path.  So in a situation where you have floating grounds in your test device, you won't wind up going down the road of disaster.  This is especially true with HV, and I have to assume that in imaging equipment, you are dealing with HV at some point. 

If I was considering a single scope solution for a "professional" development environment....and I needed versatility and viability for a WIDE range of tasks (eventually you will want to deal with wireless stuff....it's the trend in the market)...I would look at a 3GHz front end, a proper isolated HV differential pre-amplifier, a selection of passive differential probe leads for that PA, a selection of current probes with appropriate BW and rise times...a few FET and passive probes...a calibration/de-skew reference and as many math and analysis options as possible.  I would also look at a proper user grounding solution and a good clean, dedicate, power source.  Also I would look at the devices needed to proper float and decouple various types of grounding scenarios, as well as some isolation chambers to properly characterize field devices. 

Don't under-estimate the value of a good shielded temperature controlled isolation cabinet....it has been my experience that for every 15 minutes of measurement I do, I spend a good couple hours researching and designing the proper test and fixtures to know that the data/metrics I acquire are actually worth something.  The last thing you want to do is go chasing common mode noise or bogus artifacts.  It will save you a lot of resources (money,time etc...) in the long run. 

You can acquire everything I just described, plus a good calibration reference, and a proper DMM (trust me you are going to need this for your ext Vref, and you are going to need 2 units...) and a decent electronic load (another MUST have) for under $30k.  You just have to be frugal and look at what you really need/want.  You are going to have to buy used to fit your budget vs needs/wants.  There is nothing to be scared about, when buying used.  Just make sure the devices are still supported, either direct from manufacturer or by a 3rd party support community. 

Also don't be afraid or get caught up in things like the most possible memory depth or more modern "marketing" ploys.  Yeah a lot of the modern bell and whistles are nice, but wholly un-necessary.  I am chugging along, happily with my 5 year old, used, LeCroy.  Is it the best at everything? NO.  Do I wish it had more memory depth? YES.  In fact my ds2000A still sits on the bench next to the LC "high end" scope, because it offers that finer resolution in the vertical domain, and slightly more mem depth for long captures.  The LeCroy did replace a "only 2 year" old Agilent though.  It just does more of what I need and less of what i don't (the Agilent did a lot of the "don't really need" type stuff). 

I will throw up some pictures of the two scopes, side by side (Agilent is gone already) analyzing the SAME signal, from the same signal test point.  The differences are very obvious.  Again it's apples to oranges, but that is the point.  Let's have a look at the least ideal solution possible, vs the most ideal, with the tools I have available in the home lab.  As soon as the LeCroy earns it's wings (in terms of reliability) it will go down to the real lab, and maybe we can shoot it out against some other "exotic" solutions.  Perhaps a short video would be more beneficial, that way the entire experiment can be described.  The signal chain I can illustrate here, represents well under $20k of total investment, and you will see just how significant the differences are. 

P.S.

THE WORST possible place to install this equipment would be in a "machine shop environment".  The stray EMI and dirty power is NOT going to get you anywhere, other than a constant feedback loop of second guessing your measurements.  I hope part of your budgetary consideration is a dedicated "clean" room for the lab, and some isolated power.  I have seen first hand what PLC controlled devices can inject into a building power system, and it ain't pretty. 

[nctnico]

@TunerSandwich: IMHO you are going over the top. A good engineer knows what should be on the screen of the oscilloscope and for 99.9% of the measurements you don't need to get rid of noise and spikes.

[TunerSandwich]

@TunerSandwich: IMHO you are going over the top. A good engineer knows what should be on the screen of the oscilloscope and for 99.9% of the measurements you don't need to get rid of noise and spikes.

I think you are under-estimating the gentlemans needs......and his very specific applications.  I would re-read his specific requirements.  Also...how can you go "over the top" in scientific measurement? 

If this was, "over the top", then these tools wouldn't be made at all....I think you are under-estimating the art of highly reliable field device measurement and development. 

I suppose I have been going "over the top" for the last few decades ?    It seems to earn me a living

Also where am I mentioning "getting rid of noise" (slapping a 20MHz filter on the inputs of a scope is the method for marketing "low noise"....and isn't a real way to gain insight on potential dynamic load issue).  I am saying quite the opposite....that you NEED to look at all aspects of the signal, and be assured that the injected noise form the test jig itself is removed.....

A "good engineer" is one who takes into account ALL of the available data, and uses the tools at his/her disposal to understand where things can be improved.

A "bad engineer" is one who ignores relevant data and says "well things SHOULD look like XYZ" and I can just ignore anything that doesn't conform to my predictions...

The most relevant and truthful answer to the question of "how much bandwidth do I need?"....is "as much as you can afford"...bit puzzled by your blanket statements about "good engineers"


Let's have a look at what the manufacturer thinks is "over the top" for their PMA software....which is EXACTLY the type of measurement the OP wants to do.....as you can see in their datasheet, their hardware recommendations mirror mine.....

see/read PDF.....is the scope manufacturer "over the top" or wrong in their recommendations to their customers? 
http://cdn.teledynelecroy.com/files/pdf/lecroy_pma2_datasheet.pdf
http://literature.cdn.keysight.com/litweb/pdf/5989-7835EN.pdf
http://www.tek.com/application/power-supply-measurement-and-analysis

also how could you say "99.9% of the measurements you don't need to get rid of noise and spikes"?  The step response of a POL is CRITICAL in power distribution systems, especially with dynamic loads.....are you saying that we shouldn't worry about overshoot, ringing/oscillation in Vsense 99.9% of the time?  If that were true the computer you are typing on right now simply would NOT work....please qualify your statements with some real world, use case, scenarios...like I have. 

The gentleman did say he was looking for "professional criteria".....