Oscilloscope Buying Advice
So you've decided that you need an oscilloscope. Oscilloscopes have been around in varying forms for a very long time, and there are a lot of them on both the new and used markets. They have varying capabilities, varying feature sets, and varying degrees of applicability to a particular set of tasks.
I've been asked from time to time for advice on purchasing oscilloscopes. Every time I'd sit and think about it, and eventually type up a considered reply, which usually ended up being pretty much what I sent to the previous query. Eventually I decided to tidy everything up and put it all together here.
Features To Look For
All oscilloscopes perform the same basic function: They present an input signal as a graph of voltage in the Y axis against time in the X axis. Wrapped around that basic function, however, is a highly variable feature set. As with everything else, some of those features are things you'll need most all the time, while others are less important.
Dual trace is pretty important for many applications. Having multiple traces just means that the oscilloscope can display multiple signals simultaneously. Generally this is done by either displaying the signals one after the other on alternating sweeps, or by chopping rapidly between them, drawing a short segment of one signal, then a short segment of the next, etc from left to right on the screen. The former is most often used for higher sweep speeds while the latter is used at slower sweep speeds. There are some quad-trace oscilloscopes out there, but dual trace has been the de-facto standard for many years.
Dual-beam oscilloscopes aren't seen much anymore, but they do have one very impressive attribute: They have the abililty display two signals simultaneously, but with different sweep speeds for each. Examples of dual-beam oscilloscopes are the Tektronix 555 and 7844.
Some oscilloscopes will have trigger threshold presets, like "TTL" for example, which will set the trigger threshold to a value that is appropriate for TTL logic circuits. Such features can be convenient, but are generally not all that important. It doesn't take much effort to set the trigger threshold with a knob or a keypad.
The "megahertz myth" from the Intel PC world works in favor of knowledgeable people in the used oscilloscope market. Nontechnical folks have been taught by salespeople for many years that more megahertz means more powerful and better...and, to the well-trained consumer, not new means old, and old means bad. Since a 150MHz PC would obviously be useless for most applications, then a 150MHz oscilloscope must be useless too, right? Wrong. There's no correlation whatsoever. But I hear this flawed argument all the time. More vertical amplifier bandwidth does not mean a better oscilloscope, it just means you can look at faster signals in more detail. If you're not looking at faster signals, that additional capability is wasted, and sometimes it can actually be a hindrance by showing too much high-frequency information when that's not what you're looking for. This is why some oscilloscopes have a "bandwidth limit" setting which limits the vertical amplifier bandwidth, usually to around 20MHz, to address this problem.
For example, I frequently look at digital logic signals with my Tektronix 2465A, which is a pretty fast analog oscilloscope with a bandwidth specification of 350MHz. When I want to pay attention to signal transition times or patterns, all the power supply hash, switching noise, and other garbage picked up by the probe just gets in the way. It all goes away when I hit the "20MHz bandwidth limit" button, and I see only the signal I was looking for.
The minimum bandwidth you'll want for general work (say, audio and low-speed digital troubleshooting or exploring) is 50MHz or so. Nearly all oscilloscopes will handle this.
Old vs. New, which sorta means Analog vs. Digital
A common attitude today is "digital anything is always better, no matter what, and anything analog sucks, and is automatically old and obsolete". This attitude is misguided, short-sighted, and naive. More importantly, it just doesn't apply where oscilloscopes are concerned, for the most part. My first bit of advice: Get an analog oscilloscope. I will explain in detail.
Fundamentally Different Instruments
Now don't get me wrong; I have several digitizing oscilloscopes and I love 'em. I have a Tek TDS3012, two HP 54111Ds, an HP 54120B, and a few others. They are all very good instruments. The problem here is that a "digital oscilloscope" isn't just "a better digital version of something we used to do with analog", as is the usual thing we expect when something "goes digital". Actually, digital and analog oscilloscopes are fundamentally different instruments.
Digitizing oscilloscopes are discrete-time systems, while a analog oscilloscopes are continuous-time systems. The quantization in both X and Y is often a big source of limitations and ambiguity. There be dragons here, and these dragons are easily missed unless you spend a lot of time in front of an oscilloscope.
Right now you might be saying "Yes, yes...I know how to avoid aliasing!" Any engineer worth his salt is on a first-name basis with Nyquist (Harry) and knows how to deal with aliasing. Of course that presupposes some foreknowledge of the nature of the signal to be analyzed, but let us take that as a given for now. Aliasing isn't the only difference. Let's look at some of the other limitations.
The Display System
First, the display system. Since the display is the only mechanism of information transfer between the oscilloscope and the user, it must do its job well, or whatever goodness going on inside the instrument is for naught. Digitizing oscilloscopes are discrete-time instruments, as stated above, but they're also discrete in terms of spatial information presentation on the display. Analog oscilloscope displays do not suffer from pixelation and its associated loss of information, because analog oscilloscope displays have no pixels. There are practical limitations such as minimum dot size, beam focus, and phosphor grain size, but in practice, the display system resolution of an analog oscilloscope is often many orders of magnitude greater than that of a digitizing oscilloscope.
If you're looking at a signal from, say, an operational amplifier on an analog oscilloscope and see just a hint of fuzz around the base waveform, an experienced engineer would recognize that as a potential sign of high-frequency oscillation. Information like that will rarely even be picked up by a digitizing oscilloscope, much less displayed. That point highlights another issue...picking it up in the first place.
The Input Stage
Modern digitizing oscilloscopes tend to be relatively insensitive. My TDS3012 tops (bottoms?) out at 1mV/div. Most high-end analog scopes go down into the double-digit microvolt range, and I have a plugin for a scope made way back in the 1960s goes clear down to 10uV/div. Does one always need that kind of sensitivity? No, but when you do, you do...and without it, the fine structure present in many waveforms just goes unnoticed.
In Linear Technology app note #90, my hero Jim Williams said: "Diehard curmudgeons still using high quality analog oscilloscopes routinely discern noise presence due to trace thickening. Those stuck with modern instruments routinely view thick, noisy traces." He really hit the nail on the head; I've seen this difference side-by-side with my own eyes. Many young engineers today will never know what they're missing, because in their quest for "new!", they'll never have seen a sharp, detailed oscilloscope trace.
But digital oscilloscopes have all these neat features!
The convenience features of digitizing oscilloscopes, however, are undeniable. My advice is to get both. After awhile, though, when getting a circuit working becomes a higher priority than shiny whiz-bang features, you may find yourself going to the analog scope more frequently.
But that means buying a used instrument!
It's a simple fact of life that it's cheaper to make digital instrumentation than analog instrumentation. This is one of the main driving factors for the disappearance of analog oscilloscopes. Another major factor is decrease in demand for analog instruments, in turn largely due to the flawed "anything digital is automatically better, and anything analog is automatically old and/or bad" mentality.
This means, to find an analog oscilloscope, you'll usually be relegated to purchasing a used instrument. Fortunately, there are a lot of them out there.
Why would an "old" scope be acceptable? As you know, all an oscilloscope does is display a graph of voltage in the time domain. Oscilloscope technology has essentially been "perfected" for a very long time. There's absolutely nothing that newer technology can bring to the table on the inside of a device that graphs voltage against time, aside from features that are separate from the basic functionality, power consumption, and physical size. Power consumption is largely irrelevant unless you run it all the time. And physical size, well, even "big" scopes just don't take up all that much space. And a top-quality, spare-no-expense instrument from 30 years ago is still a well-designed instrument today. Their ability to graph voltage against time does not diminish with the passage of time.
Yes, it's true that, as of mid-2013, some companies are making impressive oscilloscopes very high bandwidths. Agilent is shipping a 33GHz oscilloscope, and LeCroy just demonstrated a prototype of an oscilloscope with 50GHz real-time bandwidth. The Agilent unit costs more than a quarter of a million dollars. And when was the last time you needed to see a 33GHz signal in the time domain? I work in microwave RF. Most of the work in that world is done in the frequency domain, with spectrum analyzers and vector network analyzers...not oscilloscopes. For those few applications in which such unbelievable speed in an oscilloscope is actually useful, you have to have the budget to order one!
Don't listen to the salesman and the college kids, and default to the "if it's old, it's bad, and if it's not new, it's old" mentality. Use the right tool for the job, whether it's this year's model or if it's decades old.
Getting Good Equipment Cheaply
I'm very lucky to have the lab that I have. I have pretty much the best of everything, but I'm not a rich man...I don't pay list price for anything. The secret? I buy a lot of stuff broken. I research a particular make and model of an instrument, see if I can find schematics and service info, and go find one, broken, for pennies on the dollar. I just got a new function generator (HP 3324A) that way, in the spring of 2013, for less than $100. An afternoon of troubleshooting led me to a dead bottom-side transistor in the output amplifier's differential driver stage. One 2N3866A later, and I had a nice upgrade to replace my old function generator, an HP 3325A. I'll be putting that up on eBay soon, and will probably get $250 for it. Yes, it's possible to get a nice upgrade and make a profit at the same time.
But not everything comes that way. I work primarily as a consultant, so I have no qualms about re-investing a portion of my income in test equipment, because that equipment is how I keep myself and my family fed. I still mostly shop on eBay though, because good deals can be had if you know what to look for, and if you aren't afraid to make some repairs to get a better deal. And you'll learn something in the process!
The late Jim Williams, one of the most respected designers in the history of electronics, said this:
"The inside of a broken, but well-designed piece of test equipment is an extraordinarily effective classroom... The clever, elegant, and often interdisciplinary approaches found in many instruments are eye-opening, and frequently directly applicable to your own design work. More importantly, they force self-examination, hopefully preventing rote approaches to problem solving... The specific circuit tricks you see are certainly adaptable and useful, but not nearly as valuable as studying the thought process that produced them."
I've taken that to heart, and I've learned a great deal on the inside of broken top-end HP and Tektronix equipment. And then I get a working instrument at the end, almost as a bonus!
Don't Forget About Probes
An oscilloscope probe is much, much more than just a piece of coaxial cable. It's the point at which a signal enters the oscilloscope system, and as such, will set some very basic limits on what that oscilloscope system can do. Basic characteristics such as impedance, divider ratio, and capacitance all come into play to very dramatically change what you see on the screen if you don't pay attention. Don't skimp on probes!
The issue to consider here is the system rise time. Each component in an oscilloscope system has its own rise time: the oscilloscope's input amplifier circuitry, the probe, everything. The system rise time is defined as the square root of the sum of the squares of the rise times of each individual component in the system.
Very basically, one goal is to have the rise time of the probe be fast enough such that it does not become a major contributing factor to the system rise time, so it will not degrade the basic specifications of the oscilloscope.
A square wave of frequency F consists of many sine waves of different frequencies, the sum of odd harmonics from F, to 3F, to 5F, etc to infinity, in decreasing amplitude. This is a Fourier series. If you are trying to display a square wave, and your oscilloscope's frequency response (as limited by the system rise time) attenuates everything above, say, 3F, that square wave won't look very square at all.
(actually ALL waveforms, regardless of their shape, can be decomposed into sine waves of various frequencies, magnitudes, and phases)
The relation of rise time (Tr) to frequency response is: Tr in seconds should be at least 0.35X the desired system bandwidth in Hz.
A classic intro to probes and probing is Tektronix' ABCs of Probes Primer.
Digital Oscilloscopes: What To Avoid
Cheap USB devices
About the new crop of USB "dongle" oscilloscopes, I'm sorry to say that I'm not at all impressed by these devices. They're handy to have around as an "at your desk" (as opposed to "on your test bench") scope, but they're not Real Test Equipment. There are some that are pretty good, the PicoScope for example, but they're expensive. With those, you may end up with something approaching a respectable instrument, but if you're going to be spending that much money anyway, it may be better to have a standalone instrument than something that is useless without being tethered to a PC running just the right release of just the right OS with just the right software installed. Then you're stuck in a situation where you need to dig through menus to find functions, rather than just reaching over and turning a knob. On top of all that, there's all the electromagnetic interference that computers generate, and its potential to corrupt the measurements you're trying to make.
These toy oscilloscopes also have all the disadvantages of a discrete-time system, potentially magnified by the fact that the firmware may have been written by either a hobbyist with little or no real-world signal processing experience or some guy in China who may or may not know what he's doing.
Further, and potentially much worse, is the analog front-end. Oscilloscopes display analog signals...at some point, there must be analog circuitry involved, even on a "digital" oscilloscope. This is unavoidable, of course, if the goal is to measure analog signals. Something has to come before the ADC input, and a great deal of analog expertise is required there. No cheap USB oscilloscope I've seen so far has anything reasonable in the analog front-end. The electrical characteristics of an oscilloscope's input must be precisely known if reliable, repeatable measurements are to be made. At a minimum, one must know the input impedance and capacitance to be able to know what effect the oscilloscope probe is going to have on the signal being measured. Knowing the vertical amplifier's rise time is critical to understanding exactly how the waveform being displayed on the screen differs from the waveform present at the probe tip...because it always does! An analog front-end requires good wide-bandwidth amplifiers and attenuators, good high-frequency switching, precision resistors, fast rise-time circuit layout, just for starters, to make a viable analog subsystem whose characteristics are precisely known. Too many people seem to be under the impression that they can just plug a signal into a microcontroller's A/D converter input and be done with it. In the real world, it just doesn't work that way.
All in all, I wouldn't use them for anything beyond "yes, there's a square wave there" sort of signal analysis.
Many people say "Oh, it'll do, I'm just a hobbyist, I won't use it often, and I'm just messing around!" This is the worst possible situation. Someone who doesn't know a lot about electronics needs a better instrument, not a lesser one! A lesser instrument will require more skill, knowledge, and experience to be able to understand its limitations, work within them, and interpret the results. A better instrument will provide more predictable results that are easier to understand, both of which are critical for less-experienced users.
Rigol, a shining star in a universe of crap
As an interesting aside about modern digitizing oscilloscopes...If we ignore the cheap USB-connected "oscilloscopes" flooding the market these days, there are some instruments that at least, on the surface, appear to be Real Test Equipment, but from manufacturers that are not one of the heavy hitters of the oscilloscope world. Generally speaking, if it's Tektronix, HP/Agilent, LeCroy, you'll do ok. You'll also spend anywhere from $5,000 up into the $50,000 range. That would be a "divorce class" oscilloscope for most of us...but still, we don't want crap equipment! At the top end of the range, Agilent is shipping a model now that lists for over a quarter of a million dollars.
What's a guy to do? What about these other brands that have popped up in recent years? Distinctly oriental names give the discerning, experienced engineer the "cheap Chinese crap" vibe, so we avoid them like the plague. This is a good reaction...because most of them actually are cheap Chinese crap. Will the displayed waveform bear any resemblance to what's really going on at the probe tip? Who knows! Don't make your life difficult by taking the chance.
There's one brand, however, that's got a lot going for it: Rigol. My experience with Rigol oscilloscopes is limited, but I did have the opportunity to use one once, and I was very impressed. I was so surprised about how nice that oscilloscope was that I went and did a little bit of research about them. I found this:
Yes, it appears that Agilent's lower-end line is made for them by Rigol. Even before using one, this was enough of a vote in Rigol's favor for me to just buy one right off the bat, but I really don't need yet another oscilloscope, so I haven't. Since then, I've had a chance to put one through its paces, and it is every bit as impressive as one might expect from a piece of "Agilent" equipment. And they're very inexpensive.
No, they're not even in the same class as a high-end Agilent or Tektronix instrument, even a decades-old one. But they do seem to be pretty darn good.
Analog Oscilloscopes: Finding the Sweet Spot
There's a "sweet spot" where age, quality, usefulness, and usability meet. The idea here is to look for an instrument that is old enough to be affordable, high-end enough to give you great performance, reliability, and maintainability, not so old as to run up your power bill (like the Tektronix 500 series), and not so large as to take up the whole bench (ditto on the Tek 500 series).
All of the specific instruments that I will recommend are practically indestructible and very easy to use, and are (or were) top-end, lab-grade models, not toy "hobbyist" scopes. If you're going to do something, do it right, or don't do it at all. "Home" or "hobby" should never mean "garbage", and "good equipment" doesn't have to mean "expensive".
Tektronix 453 and 454
If you want the best bang for the buck, look for a Tek 453 or 454. Expect to pay $50-125 for one of those. Here's some info on these models:
My friend Larry Snyder, a Tektronix employee from 1975 to 1986, points out that CRTs and HV transformers for the 453/454 are difficult to find. I agree, but this shouldn't discourage you from obtaining one, because the failure rate is very low, and getting an entire spare oscilloscope is easy and cheap.
The 465 (not the "TAS 465", that's a completely different instrument) came after that, and was the "gold standard" of oscilloscopes for many, many years, with many still in service in labs all over the world even today...they do everything an oscilloscope needs to do, and they just don't die. The 475 is a higher-bandwidth follow-on model. Expect to pay $100-250 for a 465 or 475 in good condition.
More notes from Larry Snyder: "465 -- well over a million of them out there. That in itself says support won't be an issue. The only common problem was the contacts on the trigger view switch tended to get crapped up from the fan, due to the normal airflow in the box. The 465B addressed that and several other potential issues. I have a B. It's great. 475 is essentially a 465 on steroids. Later serial numbers are better. ... For almost anyone, at least today, I'd say 465B."
Tektronix 22xx series
I recommend avoiding the 22xx series; they were Tek's attempt at "low cost" scopes and are generally not up to their usual standard. That they're still much better than many of the cheap import scopes that are flooding the market today, but one can find better instruments.
Tektronix 24xx series
I strongly recommend these instruments.
The big drawback of the 24xx series is the middle-of-the-road CRTs (except for the 2467), and their use of custom chips in the front-end that are expensive to replace. Otherwise, these are amazingly good oscilloscopes by nearly every metric.
The 2445 is a solid, top-notch instrument with a very complete feature set. It's a much more modern instrument than the 4xx series, with neat and useful features like on-screen cursors to aid in measurement. They are also a bit lighter than their predecessors, so they're easily portable.
Around the time these were built, Tektronix made the transition from pure analog instruments to analog instruments with digital user interfaces. All the controls are digital; tap-up/tap-down buttons, shaft encoders instead of plain rotary switches and variable resistors. This facilitates configuration storage and presets, as well as on-screen textual display of scale parameters, trigger threshold, etc. This also opens the door for remote control of those settings, and some of these later models have GPIB (IEEE-488) interfaces to do just that. I have a soft spot for the 2445 in particular, because I used one for a long time when I worked at Princeton. Despite the excessive number of oscilloscopes in my lab (sometimes I worry that I might have a "problem") I would still like to have a 2445 if I can find a good deal on one. Expect to pay $125-300 for a 2445. This is a tremendous amount of functionality for the money, and they're solid, well-built instruments for which support, spare parts, and service information are readily available.
These are great instruments. They're basically 2445s with some technology updates and bandwidth/sweep speed improvements.
I have a 2465A; I bought it used in 1999 for about $1200. It sees pretty heavy use here, and it has never needed repair of any kind. These days, expect to pay $250-$600 for a 2465.
The 2467 uses a microchannel plate CRT. The microchannel plates is an electron-multiplying technique that results in more brightness at very fast sweep speeds. The problem here is that they're expensive; they typically fetch from $700-1200 on the used market.
Tektronix 7000 series
I also recommend these instruments, if you can afford the bench space.
The 7000 series has essentially no serious drawbacks, in my opinion. They're extremely well-built; recall that these are "spare no expense" instruments from the 1970s, built for the labs that had to have the very best of everything, and had big budgets to match. Once you use one, you'll never forget that level of quality. They rarely break, but when they do, all the schematics and service manuals are available as PDFs for free or on paper at minimal expense. Those manuals also include "theory of operation" chapters that guide you through the circuitry and tell you how and why everything works. I've learned a lot of tricks in those manuals! The 7000 family makes use of a few custom chips, but only very sparingly, and they tend not to be big failure points. In practice, they're not really a cause for concern.
These oscilloscopes are plugin-based, and one can readily see the flexibility this brings. They were also produced in VAST numbers, so you can get pretty much any plugin you want, and most of them are very cheap. There are plugins for different oscilloscope-class features, like low-impedance vs. high-impedance inputs, but there are also plugins for very different types of things. In particular, one can readily find Tektronix 7000-series plugins that, when you plug it into your 7000-series mainframe, turn it into a:
- RF spectrum analyzer
- Analog sampling oscilloscope for ultra-high-speed applications
- Digital logic analyzer
- Digital multimeter (with on-screen display)
- Semiconductor curve tracer
- Time-domain reflectometer
This type of flexibility is tough to beat.
Specifically I would recommend a 7854 equipped with a 7B92A timebase and a 7A26 vertical amplifier. This is a good general-purpose, high-end configuration. If you want a faster frame (the 7854 tops out at 400MHz), go with the 7904A, which is good to 500MHz. As you should know, though, the occasions where 400MHz vs. 500MHz bandwidth is a significant difference are rare. The 7904A is much easier to find, but the 7854 has a wonderful digital front-end that combines the best of both the analog and digital oscilloscope worlds. It's an analog mainframe with digital storage capability.
It's a full analog scope in every way, but the mainframe contains a pair of digitizers that capture the analog outputs of the plugins. The CRT can display those analog signals in real-time, or the digital subsystem can display stored traces, or both can be displayed simultaneously. You can also retrieve the digitized waveform via GPIB.
When running in standard analog mode, it's simply not subject to the effects of aliasing, quantization (in either axis) or pixelated displays. But at any time, you can digitize the displayed trace, freeze it, store it, put cursors on it, and measure/calculate/etc all sorts of parameters digitally with a built-in programming language of sorts, download it to a computer for further processing or publication, etc., long after the trigger event that produced the trace is gone.
It really is the best of both worlds. It is my "go to" scope for nearly all work in my lab, whether I'm repairing antique electronics or debugging modern circuitry for work.