Would you like to be notified by email when Jason Sachs publishes a new blog?
My coworkers and I recently needed a new oscilloscope. I thought I would share some of the features I look for when purchasing one.
When I was in college in the early 1990's, our oscilloscopes looked like this:
Now the cathode ray tubes have almost all been replaced by digital storage scopes with color LCD screens, and they look like these:
Oscilloscopes are basically just fancy expensive boxes for graphing voltage vs. time. They span a wide range of features and prices: you can buy a used scope on eBay for less than $100, or a top-of-the-line ultrafast scope for hundreds of thousands of dollars: Agilent's DSAX93204A oscilloscope is a 4-channel 33GHz scope that lists for $286,000.
The first thing you need to know when buying an oscilloscope is that you need to understand the features you need.
The biggest 3 features of oscilloscopes are the channel count, sample rate, and bandwidth.
Almost all oscilloscopes are either 2-channel or 4-channel (obviously you pay more for a 4-channel scope).
Sample rate and bandwidth are key features: you will pay much more for high-bandwidth scopes; most low-end scopes are in the 40MHz-200MHz range. Bandwidth is usually tested by measuring 10-90% rise time of a high-speed pulse: the standard formula relating rise time and bandwidth is that the product of the two is 0.35 -- a 100MHz scope will have a rise time of 3.5 nanoseconds. It is also important to note that this bandwidth applies to the oscilloscope itself; the usable bandwidth depends on the oscilloscope probe, and high-speed scopes either have active probes (input is buffered by a high-speed transistor amplifier) or require coaxial cable inputs.
Many oscilloscopes these days also have one or more of the following:
I was going to call this section "Bells and whistles" because there is almost an infinite variety of software features, many of which will sit unused on the oscilloscope except in rare instances, but there are some fairly common features used in legitimate instances:
Some oscilloscopes are Windows-based computers that are essentially a special-purpose PC that can run arbitrary software and can connect to a keyboard, mouse, and external monitor.
Other features include eye diagrams, mask testing, harmonic analysis, histogram display, etc.
The kind of scopes I look for are low-to-midrange new oscilloscopes for someone who works as an engineer on embedded systems with analog and digital features. If you are a hobbyist with a very limited budget, the following advice may seem extravagant. On the other hand, if you are an RF engineer or someone who works on high-speed digital buses, this may seem cheap.
First of all, bandwidth: 99% of scope measurements I take are on signals that need less than 100MHz bandwidth; these commonly include things like:
A 100MHz scope is going to handle all of these just fine -- I have a 1GHz scope on my bench, but I never use the high-speed features, and I plan to downgrade soon with a 200MHz bandwidth scope, because people keep pestering me to borrow the 1GHz scope. I figure maybe once or twice a year I need an oscilloscope that has at least 200-300MHz bandwidth. If you absolutely have to save money on oscilloscope by accepting reduced performance one feature in an oscilloscope, save money by using lower bandwidth. Scope costs are very closely tied to bandwidth more than any other feature. Again, chances are, you won't need more than 100MHz bandwidth -- and if you do, and you can't afford it, you have a problem anyway. If I were working in a small engineering shop with limited budget and I were buying more than one scope, I'd make sure that one oscilloscope has enough bandwidth to cover my highest-frequency signals, and pick the other oscilloscopes for my budget as long as they handle most signals I'll need to look at.
Note that if you are doing work on RF circuits or high-speed switching power supplies, you may need more than 100MHz and will have to bite the bullet and get a higher-bandwidth scope that meets your needs.
I always try to get a mixed-signal scope with 4 analog channels and 16 digital channels. Two analog channels is not enough -- I often have to view 3 or 4 channels at once. (maybe once every few years I wish there were a 5-channel scope!) The mixed-signal scopes are invaluable for displaying digital signals where you don't care about the analog aspects of the waveform -- if you have a questionable digital signal with ringing, overshoot, or poor output impedance, then you probably need to view it on an analog channel, but otherwise a digital-only display is fine. You can view an amazing amount of information at once on these mixed-signal scopes.
Modern digitizing oscilloscopes have several acquisition modes (usually found under some "Acquire" menu) that usually include these options:
To understand these modes, you have to consider what the oscilloscope is doing. The primary goal of an oscilloscope is to use its input analog-to-digital converter to sample the input channels and obtain waveform samples. The ADC samples are not the same thing as the waveform samples: the ADC samples are the raw input values, and the waveform samples are the data points that are actually kept in memory and plotted on the oscilloscope screen.
Let's take an example. Suppose you have a 500Msamples/second 100MHz oscilloscope, and you've got it on a 100usec/division display with 20,000 samples per waveform. The scope has 10 horizontal divisions (almost all do) for a total time of 1msec. Each waveform sample therefore corresponds to a 50nsec interval of time. But the ADCs sample every 2nsec. So each waveform sample can take into account 25 ADC samples.
Sampling mode takes one ADC sample for each waveform sample.
Peak-detect mode finds the minimum and maximum ADC samples during the interval covered for each waveform sample.
Averaging mode is used with periodic signals: the waveforms from multiple triggers are averaged together. For nonperiodic or slowly-changing signals, averaging mode smears together multiple traces.
Most manufacturers have scopes which also have another acquisition mode called "hi-res" or "extended resolution", which takes all the ADC samples in the interval covering the waveform sample, and averages them together. In our example above, this would take 25 ADC samples and take the average of them for each waveform sample. Most signals have a small level of noise, and the rules of statistics can be used to show that if you take the mean value of N samples, the variation in the mean is 1/sqrt(N) as large as the variation in the raw samples. This effectively gives you higher resolution by oversampling the input signals.
Hi-res mode is a vital feature for oscilloscopes. I will not even consider purchasing a scope that doesn't have it, and I recommend you do the same.
There are often instances where you need to use the oscilloscope waveform data in further analysis, and a printout of the scope waveform isn't enough.
In the 1990s, the standard methods for waveform data transfer were a 3 1/2" floppy disk drive to save data, and/or an RS-232 connection to transfer data to a PC via the serial port.
Now I'd look for a USB port for a "thumb drive", and an Ethernet RJ-45 port. The Ethernet port is usually more convenient, as most scopes nowadays have a built-in embedded web server: you just connect the scope up to the local network, open up a web browser on your PC, and type in the hostname or address of the oscilloscope, and you can save the waveform data to a file easily through your PC's web browser. The USB port is important for those instances where you're out in the field away from your PC and network and you need to save data.
I have a rant against Tektronix for their waveform data saving, which I hope they have fixed by now: Tektronix oscilloscopes have historically saved one waveform per file. This makes it cumbersome to archive multi-waveform data from the same oscilloscope waveform capture: if you have 4 scope channels, you get 4 files. It takes one step to save each one to a thumb drive (4 steps total), and you then have to remember to keep together on disk, and you have to read these files and merge the data together. That's a big stumbling block to getting things done.
It's important to me that the oscilloscope is not a general-purpose PC with built-in oscilloscope features. (Regardless of the fact that I dislike Microsoft Windows, I would feel the same repulsion to a scope with MacOS or a Linux GUI.) It's wrong for several reasons:
Some of the oscilloscopes are based on Windows CE, and I'm fine with that, as long as I don't notice the scope's operating system as kludgey or crash-prone or hard to navigate.
Some oscilloscopes come with DPO mode, which emulates the old analog CRT-based scopes that had phosphor on the CRT to effectively display multiple traces at once. The new DPO scopes act like they are painting with light, and each trace is displayed, so that when multiple traces overlap, they add brightness.
It is essential, in my opinion, to be able to turn DPO mode off. DPO mode is great for seeing jitter. It's a mess for displaying serial data when each trace is different.
Furthermore, if you can turn DPO mode on for "Run" mode (oscilloscope continuously acquiring new waveforms), you should be able to turn DPO mode off for "Stop" mode. I like DPO mode in some instances, but when I hit "Stop", all I want is to see the last waveform that was acquired.
The persistence of DPO mode should be configurable: maybe you want to see 1 second's worth of data, but maybe you want to only see 1 millisecond's worth of data, or maybe -- and this is important -- you just want to see a single trace. The brightness of the traces should be well-designed by the manufacturer to reflect the number of traces displayed at once. I've seen scopes that have DPO mode where if you want to see only 1 trace, all you get is a very dim display to match the brightness of the 1 trace when it's part of 1000 displayed traces.
DPO mode is great, but it needs to be implemented in an intuitive way.
Those are the major issues for me. But there are a number of other things to consider as well:
A new oscilloscope costs as much as a used or even a new car. Would you buy a car before you tried driving it? Don't buy an oscilloscope unless you have used the model in question. Most major test equipment distributors will lend you an oscilloscope to try out for a few days. Use the opportunity to check out all the features you would consider using.
I don't have a really good definition for this one. There are some subtle features of the way an oscilloscope behaves, which are important for being able to quickly navigate and visualize the right waveform data.
In the US, the three major brands of oscilloscopes are Tektronix, Agilent (formerly HP), and LeCroy.
Skip the low-budget Tek oscilloscopes (TDS1000, TDS2000, TDS3000 series) unless you're strapped for cash. They tend to be a bit noisy, and unless they've added something recently, they don't have hi-res mode, so you can't get around that noisy analog input.
The Tek MSO/DPO3000 and 4000 series are pretty decent (MSO = mixed signal, DPO = "digital phosphor" = analog channels only). The MSO/DPO2000 series does not have hi-res mode -- BOO HISS! I looked at this series and at first glance, it seemed to be just right for price/performance, but it lacks this one important feature. Sigh.
Tek's higher-priced scopes (MSO/DPO 5000 and 7000 series) look intriguing but above my budget.
Again, skip the low-budget Agilent scopes (the new 1000 series) -- it doesn't have hi-res mode.
The Agilent InfiniiVision 2000 and 3000 series were released in either 2010 or 2011 and are a pretty good introductory scope. They have variants with and without mixed-signal (16 digital channels). Both have hi-res mode.
Agilent has a bunch of other higher-bandwidth scopes as well. ($$$$$$)
I'm not as familiar with LeCroy as I am with Agilent and Tek. They include the same general features as Agilent and Tek do, with a wide range of price/feature sets, and options like mixed-signal, digital phosphor mode, etc.
I would avoid other brands for benchtop oscilloscopes, such as Rigol or B&K Precision -- there are just too many subtleties in making oscilloscopes. (But I'm open to hearing input from any of you who have used these scopes and are happy with them.) B&K makes low-budget digital multimeters and signal generators that are just fine, but it's really hard to make a low-budget scope that doesn't just suck.
There's a growing series of "handheld" oscilloscopes that are sort of halfway between a digital multimeter (DMM) and oscilloscope -- they are small, battery powered, and have rudimentary oscilloscope features. These have two advantages over benchtop oscilloscopes: you don't need to have an electric outlet nearby, and the oscilloscope inputs don't need to be ground-referenced. These are made by Fluke, Agilent, Tektronix, and AEMC. (All reputable brands -- I'd avoid others.)
Also new in recent years is a series of digitizers that connects to a PC via USB. These have no displays or front-panel controls, but turn your PC into a low-end oscilloscope. There are a lot of independent companies producing these things; I have no experience with them, so shop carefully. (Jack Ganssle wrote a 2005 article in EE Times reviewing a few of these)
A related concept that was recently introduced is the Oscium iMSO-104 that plugs into an iPad and turns it into a handheld oscilloscope, with a single channel 5MHz analog input and 4 digital channels.
(The "me" in this section refers to me, not you. You should form your own opinion; I'm just sharing mine. :-)
I test-drove a half-dozen oscilloscopes in the last few years. For me, it comes down to Tektronix or Agilent. The LeCroy scopes are similar and seem to have more software features, but I find them more cumbersome and less intuitive to use.
I've used the Tek MSO3000 series and the Agilent 3000 series; both are very similar, but Tektronix seems to be better at handling digital inputs, and Agilent at analog inputs. Tektronix's digital decoding and display was easier for me to manipulate. But the analog signals on the Agilent scope were cleaner-looking.
There are two other things I like about Agilent's oscilloscopes.
One is their vertical offset knob, which instead of merely controlling a digital offset, appears to be an analog offset: if you look at a small AC signal with a large DC offset on an Agilent oscope, you can tweak the offset until the AC signal comes into view. We've looked at the terminal voltage of batteries under oscillating loads, and it's impressive how easy it is to set the vertical range from, say, 47V - 49V without losing resolution. (This is different from AC-coupling, which most scopes have, and puts the input through a high-pass filter that loses the DC offset -- with Agilent's scopes, you can still measure DC quantities such as mean/min/max) This has been a feature of many of Agilent's scopes for over 10 years (I first noticed it on our 54624 scope about a decade ago).
The other nice thing is that Agilent's newer series of oscilloscopes (including the 2000 and 3000 series) is that they have "upgradeable bandwidth". Basically, Agilent has made the decision to make all of their scopes in a series based on a single design, and benefit from economies of scale; the lower-bandwidth scopes are nearly identical but are "crippled" through software and/or hardware. You can buy the MSOX3014A 100MHz scope, and later upgrade it to higher bandwidth (to 200MHz through software licensing; to 350 or 500MHz by sending it back to the factory) if or when you need that extra bandwidth.
So my oscilloscope of choice is the Agilent MSOX3024A 4-channel 16-digital-input 200MHz scope; we plan on ordering one soon.
I'll leave you with a few things to look out for when using oscilloscopes.
When you're trying to measure DC voltages accurately, use a multimeter. Oscilloscopes do not have the voltage gain accuracy that a DMM has. Most oscilloscopes are specified as 1% accurate in voltage reading. Furthermore, most digitizing scopes have 8-bit input resolution. They're meant to give you visual indicators of waveforms. If you want higher-resolution input, you can use hi-res mode -- you won't get any better accuracy, but you might get 10 or 12 bits of effective resolution. Or get a data acquisition system -- it won't have as high a bandwidth or sampling rate as an oscilloscope, but you can get better accuracy and resolution.
Oscilloscope probes are tied to the oscilloscope through a coaxial cable with the outer conductor tied through the oscilloscope to chassis ground, which connects to earth ground through the oscilloscope's power cord.
If you're measuring something that's floating, you need to connect the oscilloscope ground to somewhere in your circuit. If the thing you're measuring has any connection to a low-impedance voltage source relative to ground, you have to be really careful where you connect your oscilloscope ground, or you could cause a short-circuit and damage components. This includes circuit boards with USB connections to a PC (If the PC plugs into the wall, then the USB signals are relative to ground), or anything that connects without galvanic isolation to either terminal of a battery, or anything that connects to a motor terminal or transformer or AC plug.
Be especially careful when connecting or disconnecting the ground clip -- if you're at all clumsy, make sure your circuit is unpowered when you make or remove the connection. If you drop the clip onto other circuitry, you'll connect it to earth ground. (BZZT!)
If you need to make a differential measurement, use a differential probe. These are expensive, and you won't get the full oscilloscope bandwidth, but they allow you to make measurements between voltage terminals that are different from earth ground. Or you can use a battery powered scope, or a special-purpose oscilloscope like the Tektronix TPS2000 which has isolated input channels.
You could also try using channel math on the oscilloscope to subtract one channel from another, but this is not a true differential measurement, and the resultant signal quality can suffer from noise, reduced resolution, or gain mismatch.
The standard oscilloscope probe is an ancient thing that is way behind the development of oscilloscopes themselves:
The ground lead on a standard oscilloscope probe is a cheap alligator clip connected via 3-4" of wire to the probe cable, and the probe tip is a 2" "witches-hat". These probes made sense 25 years ago, when circuit boards with through-hole components were still common. Now that most circuit boards use small surface-mount components, these probes are huge and unwieldy -- attach them to something on the circuit board and you have a 6" lever that's likely to pull off something from the circuit board accidentally.
On top of that, the long connection between probe ground and cable adds inductance and has a large loop area, both of which will pick up noise from nearby circuitry. The easiest way to tell if you have a noisy circuit is to connect scope probe tip to the ground clip, and see what signal you see, both near to and touching the circuit in question. If you see noise near the circuit, you've got radiated electromagnetic emissions; if you see noise only when you touch the circuit, you've got conducted emissions.
The next time you see noisy signals looking at an oscilloscope trace, perform this basic test (tie scope tip to ground, near and then touching your circuit) to make sure it's not general noise in your circuit rather than noise on the specific circuit node you're trying to measure.
Add a Comment