Ask Hackaday: Help Me Choose A ‘Scope

If there is one instrument that makes an electronic engineer’s bench, it is the oscilloscope. The ability to track voltages in the time domain and measure their period and amplitude is one akin to a light in the darkness, it turns a mere tinkerer with circuits into one in command of them. Straightforward add-on circuits can transform a basic oscilloscope into a curve tracer, frequency response display, and much more, and modern oscilloscopes offer a dizzying array of useful measurement features unimaginable to engineers only a few years ago. And I need your help to pick a new one.

They don't make 'em like they used to! My Cossor portable oscillograph.
They don’t make ’em like they used to! My Cossor portable oscillograph.

The Status Quo

My first oscilloscope came my way in the early 1980s when my school had a lab clear-out. It’s a dual-beam Cossor, probably from the 1950s, and it proudly boasts a 2MHz – or should I say “2 Mc/s”! – bandwidth. The maker’s plate calls it a “Portable Oscillograph”, because it does have a handle on top and if you are a weight-lifter you can probably carry it some distance. Dusting it off from the garage recesses for this article brought back memories of all those hacked-together circuits made from old TV parts, of seeing for myself the mysteries of the PAL colour burst, and of my home-made spectrum analyser.

The Cossor clearly wasn’t going to cut it for an electronic engineering student, so sometime about 1990 I made the trip to the Aladdin’s Cave  of Stewarts of Reading, and bought a bargain Gould dual-scan ‘scope from the mid 1970s. It has a 20MHz bandwidth, and has been my trusty companion ever since. It’s typical of everyday ‘scopes of the era, and since ‘scopes like it can be found for little more than beer money these days I’d have no hesitation recommending one to anybody looking for a basic piece of test equipment.

A quarter century later though, I have a ‘scope problem. As a radio amateur I’ve always wrestled with the Gould’s low bandwidth. It’s also not the smallest of instruments, and the sheer number of things new ‘scopes can do these days are something I just can’t ignore. It’s time I bought a new ‘scope, and this is where you come in.

Choices, Choices

If I held my finger over the badge, would you be able to distinguish it from its competition? Image: SIGLENT TECHNOLOGIES CO.,LTD [PD], via Wikimedia Commons
If I held my finger over the badge, would you be able to distinguish it from its competition? Image: SIGLENT TECHNOLOGIES CO.,LTD [PD], via Wikimedia Commons
To narrow down the selection a little, consider that I won’t be able to spend thousands of pounds on the ‘scope I’d really like. The people who sell ‘scopes at the top of the market will have to wait for my ship to come in. And USB ‘scopes aren’t my thing, I prefer a stand-alone instrument.

Instead I will be looking where I suspect a lot of you will too, at the lower-end Chinese digital ‘scopes from brands like Rigol, Owon, Siglent, Hantek, and others. I’m very familiar with more than one of them from use in contracts, hackspaces, and other people’s benches. They are all compact instruments with fairly similar specifications between brands, in fact many models look similar enough to have been made on the same production lines. They will not perhaps have the spec of the multi-thousand-pound ‘scope when it comes to the edges of the envelope in noise or even sensitivity, but the performance they deliver for the price is more than enough for my purposes.

Why buy a DS1074 when a DS1054 will do. Image: Alex P. Kok (Own work) [CC BY-SA 4.0], via Wikimedia Commons.
Why buy a DS1074 when a DS1054 will do. Image: Alex P. Kok (Own work) [CC BY-SA 4.0], via Wikimedia Commons.
So given an array of outwardly similar ‘scopes which still occupy a range of prices over the budget end of the market, how should I choose? Once I’ve made the decision that I only need 2 channels rather than 4, my basic requirement above all else is for bandwidth, so that seems a good place to start. But even there the picture is muddled, it seems to be the norm for these instruments to have a quoted bandwidth which can be extended with a software hack. Most well-known are the Rigol 1050 series, 50MHz ‘scopes which can achieve 100MHz bandwidth, but they not alone by any means. Perhaps the manufacturers permit such illicit upgrades because they are a valuable sales tool. My gut feeling though is to buy the highest bandwidth ‘scope I can afford and see a later upgrade as a bonus, but not necessarily to do it straight away as I prefer my instruments unbricked and with warranties.

Even homing in on the bandwidth doesn’t give me as clear a picture as it should. For somewhere between £200 and £300 (About $260 to $400) when all taxes are paid, I can buy any of a spread of Chinese 2-channel 100MHz ‘scopes with similar spec. Some of the models even promise a bonus 200MHz upgrade with a software hack, but this is not price-dependent. I’m left looking at differences in the length of the sample memory, and even wondering whether I am sometimes simply being expected  to pay a bit extra to support an emerging brand hierarchy.

The Hackaday readership are a diverse group, among whom alongside the interested readers reside real, not armchair, experts on almost any subject we cover. Some of you will be in my position of looking at new ‘scopes, and many of you will have been through this process yourselves and have the tales both good and bad to tell about your choices. So if you were standing where I am and looking at a budget digital ‘scope, what would inform your choices?

Header images: Binarysequence (Own work) [CC BY-SA 3.0], via Wikimedia Commons

89 thoughts on “Ask Hackaday: Help Me Choose A ‘Scope

  1. You can also start bombarding local tech schools and labs with cheapy chinese scope literature anonymously until they ill advisedly dump one of their solidly built 90s or 00s digital high bandwidth models in the back alley :-D

  2. I have a Rigol DS1054 and would highly recommend it. It’s compact, lightweight, has a large screen, straightforward controls, and four inputs. If it’s lacking any features, I haven’t found them yet.

    The one downside is the built-in cooling fan is a bit loud, but that’s essentially a non-issue.

    1. Seconded. The UI can get rather sluggish, but for £280 it’s a no-brainer. Especially as it’s got LAN and USB connectivity. I’d rate it as better than the crappy £2K Tektronix MSO2012s at work, but nowhere near as nice as a decent (£6k+) Agilent. I didn’t have £6K, but I had £300… :)

    2. I also have a 1054 and I like it. I have used scopes up to 60k$ at work in the past. There is one down side I discovered. I work with big power electronics sometimes and it does seem to not be well shielded for EMI. I was using it next to an older tektronix digital scope which had no problem with the EMI. But I still like it except for that.

    3. I just ordered a new scope yesterday. I went with the new 4000 series from Hantek because it also has an arbitrary function generator. 100mhz scopes go for $300 on amazon and less in Ebay.

      My reservation on Hantek was a review on eevblog that showed ringing on fast transitions near the bandwidth limit. Additionally others complained about buggy firmware that very very occasionally crashes the scope software. Not a pretty picture, yes. But I found complaints about specific rigol owan, and siglent. So it seems like all the Chinese scopes don’t measure up to tektronix or agilents. But I wanted cheap and easy to read (big fonts with well thought out human interface). And to me it looks like hantek wins on price, features, and human interface. I felt the new series which has a redesigned front panel offered the most chance of better firmware so I’m rolling the dice this new one.

      Wish me luck! And post below any specific 5000 series bugs you want me to test on the 4000 series.

      1. I note that all the bugs and claims of ringing were on the older 5000 series. I’ve not seen any reviews of the 4000 series yet. They only recently released the 4000 series. Why they retro graded the series from 5000 to 4000 on a scope that matches all the 5000 series specs I do not understand. Bigger numbers must be better right? Confusing .

        I went with 100 MHz because you need to have about 5x the frequency to see a square wave ( harmonics). And because one suspects the manufacturers lie a bit. So it will be good for 10mhz and at least show activity but not fidelity at 50mhz.

        In demos it had a faster acquisition rate (traces per second) than the long memory owan or riglol. So that looked good. It has 40k memory which is better than say 5k on the comparably priced siglent.

        The Fft is often limited on most cheap scopes to 2048 points so longer memory doesn’t help with that.

        It has a 800 pixel wide screen which many siglent models lack.

        The persistence mode has a limited depth and doesn’t decay like a phosphor emulator so that’s a minus compared to more expensive scopes.

        It’s too new to know if the 5000 series unlock hacks that double the analog bandwidth will work but I don’t want to be hacking test equipment that needs to “just work”.

        You can use the arbitrary function generator at the same time as the scope so it will be a nice tool for pump probe analysis of circuits. For example with a chirp input the synchronized time trace will show the frequency response of your circuit!!! Try that on a regular scope

        1. Okay the Hantek 4000 series scope arrived. Holy cow this thing has some problems. There’s a couple of screenshots over on reviews that show the problems I see

          The Arbitrary function generator is supposed to have a 25Mhz bandwidth and 200 samples per second. However it appears the roll off happens around 100Khz or so. If you create a 1Mhz 4-volt square wave you are a rounded off waveform that is about 160mv peak to peak. It also has about 10% jitter on the waveform period at 5ghz.

          On the DSO portion there’s a few problems, perhaps less severe but still not heart warming. If you use the Horizontal Zoom feature it splits the screen to show a fast trace of the center portion of the slow trace. When you change the number of points collected (4K, 20K or 40K) the seconds per divisions shown on screen are actually wrong. Not just a little wrong. Like factors of 10 wrong. I’ve only seen this is the Horizontal magnification and the error depends on the number of points collected.

          the Pulse trigger allows you to set a minimum or maximum pulse width for a trigger event. Trying to caputure a 100MS negative going pulse I found I actually had to use the positive going pulse setting (so they messed up the software labels). It also would not work at all when triggering in Normal trigger, only using pulse with autotrigger got any traces to scan on the screen. And autotrigger isn’t what you want for reliably caputuring specific pulses. So that feature is crippled.

          Looking at the traces using the built in 1Kz calibration square wave I see that in the 10x setting the square wave looks flat but on the 1x setting there is a noticable (10%) overshoot (no ringing , just damped relaxation) on the leading edge of the pulse. Not too bad but not great. And I can’t tell yet if that’s the probe or the calibration source that is malfunctioning.

          The brightness and contrast ranges are pretty good but not ideal. The problem is that when colors like Yellow and Cyan are about the right bightness (not overly bright), the purple traces are not bright enough.

          Although the initial specs said the FFT would use 2048 points, it seems to use only 1024 and it’s not too swift. So this is a pretty lame spectrum analyzer. THat function is there but it’s not something you are going to be looking at sidebands with.

          The measurement disply is nice in the large fonts it uses and it lets you have a lot of measurements on screen at the same time. However it’s annoying that you have to set each one for a specific channel (and it doesn’t tell you on-screen which channel). Because there are many possible measurements they use a menu to let you set this up, thus doing something simple like flipping all the measurements from one channel to another would be a nightmare of button presses.

          You can’t take a screen shot without having the screen-shot menu in the picture.

          The menus appear to have a software defect in the frequency setting on the Arbitrary function generator. When you press the menu button to set the frequency it often pops up a keypad covering the screen. That would be nice if this were a touchscreen! but it’s not. If you dismiss the keypad the wrong way it resets the frewuency from where it was to 1000Hz. Arg! The frequency set knob is not velocity sensitive so you can’t easily chnage from 1Khz to 10Mhz without many many many twists of the set knob. It needs a course grained range button to atleast get you within an order of magnitude.

          The ARB only has 2000 points so any hope of programming a chirp is sort of useless for any realistic bandwidth.

          The contextual help is pretty useless and to find the sectioin you actually need it’s painful to scroll through zillions of topics to find the one you want, especially if you don’t know what it is called (zoom? magnification? Horizontal? split screen?).

          So the bottom line is this. If they had advertised the Arbitrary function generator as a 50Khz not a 25Mhz device I might even be really pleased with this. At least I do have an audio waveform generator. And now that I know the bugs in the split screen onscreen sec/div i’m less likely to get burned by that. The FFT is sadly aweful but that’s normal on all cheap scopes so no surprises there.

          In the past I’ve use things like Lecroy 9400, 9500, and tek and HP scopes of that same vintage. If I could trade this for one of those I might. Those are a lot more usable and not full of bugs and surprises. They are however pretty darn large. They actually make less noise than the fan on this Hantek but the Hantek is otherwise more freindly on the desktop.

          1. Thank you for the helpful information. For the last one month I’ve been narrowing down my choice to the Hantek DSO4254C, but now I might have to rethink the whole process.

    4. I have an older RIGOL. Sometimes the extra bandwidth and extra channels of the 1054 would come in handy, but I’ll have to figure out how to manage if I don’t want to spend the money to upgrade. If I were buying now, the DS1054 would be on the top of my list too.

  3. Bandwidth is not the only parameter you should be looking at. Arguably more important are captures per second (the newer generation of scope capture 30,000 or more waveforms per second, and superimpose them on screen like an analog scope so you can see glitches) and capture depth (the number of samples that can be stored in a capture. Large buffers let you scroll through a lot detail).
    Take a look at the Rigol Z series as an example of what to look for.

    1. I 2nd this, but with the BIG caveat that thousands of waveforms per second are not likely to be found in a budget scope. Note that if you CAN’T afford a scope with a high update rate, then get what you can, and KEEP the old analog scope. Analog scopes are still wonderful for that purpose.

      Disclaimer: I used to work for Keysight (back when they were called Agilent), and I worked on the main ASIC in Keysight’s low-end scopes, so I may be biased…

      More waveforms per second is VERY valuable when trying to see infrequent events. If you have some sort of digital glitch that happens one out of every 1,000 events, you stand a much greater chance of seeing it if you can see 10,000 waves (capture 10 glitches per seconds) per seconds vs 10 (capture one glitch every 100 seconds).

      A couple of months ago, I was trying to debug the SPI interface on brand-new silicon, and somebody brought over a low-end LeCroy scope, This thing would show about two or three waveform per second, and after 30 minutes I had to get rid of that TOY and stea a real scope (Keysight, in this case) from another desk. Yes, it makes a difference, and you can see it.

      Having a deep memory certainly can be useful too, and I would rank that as a #2 priority behind waves/sec. However, deep memory is easier to implement, and should be easier to find on a cheap scope.

      If you can afford it (or pay to unlock it later), then a mixed-signal scope is nice, as you can see analog and digital (16 channels) at the same time. These channels emulate a logic analyzer is “timing” mode (sorry, no state mode), but can still be VERY useful, if you can afford it.

      Other niceties like FFT would also be useful, especially for radio work, so if that might be an option, consider how much the license is for those features. The good thing is that those features can be unlocked in the future, so no up-front cost.

      Finally, don’t knock the USB-based instruments. The Digilent Analog Discovery 2 seems *VERY* nice for the money ($280), and has a LOT of cool features (from the reviews). A USB instrument cannot REPLACE a real scope, but it can SUPPLEMENT it. A good USB device can give you 80% of real scope functionality for 30% of the cost, and use a LOT less bench space. You still need a real scope for those few cases when the USB instrument is not up to the job, though. I can tell you from experience that the Saleae Logic Pro 16 makes a pretty good logic analyzer (timing mode only), but the analog functionality is not anywhere good enough to even come close to passing for a scope.

      1. Good recommendation to get a built in logic analyzer if you can afford it. Been doing FPGA work for the last year, and being able to watch 16-signals from inside the chip has be a life-saver. (just wire to spare IO and go to town.)

        That said, I would NOT recommend getting a Rigol MSO1074Z at this time. The logic analyzer section has quality problems and quickly develops broken wires/solder-balls. (Which show up as random channels that intermittently only show a short pulse on rising edges) Between my initial purchase and 3 warranty replacements, I’ve had this problem crop up on all 4 MSO1074Z motherboards that have been in my ‘scope.

        1. ….? For FPGAs, you’re almost always better off using one of the built-in internal ‘logic analyzer’ type modules, like Xilinx’s ChipScope or Altera’s SignalTap. Data that comes off the chip has little to do with what’s going on *inside* the chip, and more importantly, tapping a signal to get it *out* of the chip usually changes the routing significantly. The internal debugging tools are usually significantly less intrusive because they can run at much higher clock rates than the I/Os can.

          If you really love physical logic analyzers, then using something like the Agilent Trace Core is a decent idea, but really, the internal tools are usually what you want to do.

          1. I am only looking for logic errors. Sub-clock timing errors mean I messed up the code. Second, JUST learned VHDL for this project. Needed to get stuff done instead of spending 2-3 more months figuring out how to use ISE’s simulator and chip scope. Third my code also needed ALL of block ram, no space for chip-scope’s sample memory. Fourth, I’d hardly expect debugging over JTAG to be as responsive. Fifth, do I need to re-compile if I get my triggers wrong? Sixth, routing 16-pins out to IO messes with routing MORE than adding a big internal state machine connected to the BRAM and JTAG pins?

          2. Learning ChipScope and the simulator vastly increases your productivity in the future, and 2-3 months is really an overestimate. It’s maybe a few days of work. If you used all the BRAM on chip, OK, that’s a separate problem.

            Fourth, I’d hardly expect debugging over JTAG to be as responsive.

            It’s basically instantly responsive.

            Fifth, do I need to re-compile if I get my triggers wrong?


            Sixth, routing 16-pins out to IO messes with routing MORE than adding a big internal state machine connected to the BRAM and JTAG pins?

            Yup. First off, the internal state machine is tiny, logic-wise, and more importantly, it runs at JTAG clock, so the tools know they have all the time in the world. They’ll use any available space they can and in general stay way the heck away. Second, BRAM’s located all over the chip – it’ll find the easiest one to deal with, and use that one. The I/Os are fixed location.

            Routing delays for I/Os aren’t small. Crossing the chip can take a long time (almost ~10 ns), so you can see big skews between signals that aren’t actually there internally. You also have to be careful that the I/Os you use for debugging aren’t included in the timing constraints, otherwise the entire design can be *radically* affected by place & route trying to meet an offset time that you actually don’t care about.

      2. “If you can afford it (or pay to unlock it later), then a mixed-signal scope is nice, as you can see analog and digital (16 channels) at the same time. ” You can also see at the same time if you use the trigger-out of the scope to trigger an external LA. Depending on the LA you can also do the reverse thing. Of course everything on a single screen is more convenient but if you are on a tight budget an external LA is cheaper (especially with these 10€ clones from China).

        Dave Jones talked about this:

    2. In fact, bandwidth and captures per second are just as important. The bandwidth says something about the analog part of the scope being able to follow fast signal changes. The captures per second says something about the digital part being able to actually capture those fast signal changes.

      If your scope has a high cps, but a low bandwidth, all you’ll capture is a very detailed picture of the failure of your scope to be able to follow the input signal changes.

      With a scope with high cps and a low bandwidth, you’ll never be able to see those glitches you’re mentioning. :)

      I know: you said ‘arguably’… ;)

      1. You’re confusing wave form captures with digital samples/second. You want 10x the number of digital samples minimum compared to the bandwidth (Nyquist says 2, but that’s to see anything. If you want any kind of definition aim for 10).
        Waveform captures are a stream of samples. Cheap DSOs only start new streams at best a few hundred times per second and it’s easy to miss glitches and atypical data.

        1. “You want 10x the number of digital samples minimum compared to the bandwidth (Nyquist says 2, but that’s to see anything. If you want any kind of definition aim for 10).”

          This isn’t true. The Nyquist-Shannon theorem says you can reconstruct the signal perfectly, not “to see anything.” Oversampling makes it easy, and can compensate for the low resolution of scopes. But with even decent oversampling (say, 4x bandwidth) you can interpolate pretty easily with limited artifacts. Just need to make sure you’ve got enough of a record length, too.

          1. Nyquist says you need 2x the number of samples to reconstruct a sine wave. The problem is that you’re unlikely to have a pure sine wave. The 10x number is to see undershorts, overshoots, harmonics etc.

          2. No. Absolutely not. If the signal is contained within a bandwidth ‘f’, and you sample at discrete times at a rate of 2f, you get *all* of the information in the signal. Not just if it’s a sine wave. Any signal. So long as it has no frequencies above “f”, you’re fine. The mathematical form can be expressed here, in Theorem 1.1. Sinc interpolation will completely recover the original signal.

            The problem is that in order to meet the theorem’s assumptions, you need an infinite length record, a perfectly band-limited signal, and perfect sampling. Which you basically never have. So oversampling helps you deal with those limitations.

            But it is absolutely *not* just for a sine wave.

          3. According to Nyquist, you only need to sample at just barely over 2x the frequency. According to the math, to reconstruct a signal with up to 50 MHz bandwidth, you only need barely over 100 MHz sampling — according to the math.

            OK. With that out of the way, we have math, and we have real life. For a 50 MHz bandwidth, it is assumed that you will have **NO** frequencies above 50 MHz. None. At all. Have you ever seen a real-life filter that good? Do you have a filter that will pass 50 MHz and block 50.001 MHz? Me neither. This means that you need much more than Nyquist predicts. For a 50 MHz bandwidth, if you sample at 250 MHz, that is 5x the bandwidth. You will start to get aliasing at 125 MHz, but making a filter to pass 50 MHz and effectively block 150 MHz is certainly practical.

            Plus, reconstructing a signal an the hairy edge of Nyquist is difficult. Imagine sampling a 50 MHz signal at 100 MHz. If you got the phase relationship just wrong, you might sample exactly at the zero-crossings, so your scope would report a DC value — very wrong. So, you need to sample at OVER the Nyquist frequency. How much over? In theory, not muc (one Hertz would be enough). In practice? Quite a bit more. Imagine that you were sampling a 49.999 MHz sine wave at 100 MHz. If you just used something simple like sinc interpolation (sin(x)/x), then you would see an AM signal — a 100 kHz signal modulating a 100 MHz signal — not what the signal looks like. Yes, you could reconstruct the signal with enough math, and with enough sampling time, two things that real life scopes will not bother with. Real scopes typically just do a little digital filtering to remove frequencies above the bandwidth, and then do sinc interpolation to make a pretty picture.

            So, because of the impossibility of making a true brick-wall filter, and the impracticality of doing more complex math than sinc interpolation, we need to sample at much greater than Nyquist would suggest. Around 4x and 5x the bandwidth seems to be pretty common.

          4. Yeah, that’s what I said. Although note that this: “imagine that you were sampling a 49.999 MHz sine wave at 100 MHz. If you just used something simple like sinc interpolation (sin(x)/x), then you would see an AM signal — a 100 kHz signal modulating a 100 MHz signal” – is wrong, because sinc interpolation wouldn’t contain a 100 MHz frequency component. It’d resolve back the 49.999 MHz sine wave – but you’d need a long record in order to resolve that.

            The other issue you don’t mention is the sampling fidelity: it takes a more bits per sample to resolve a high-frequency signal than it does a low-frequency signal at equivalent SNR, and ADCs are notoriously low fidelity measuring devices (typically 8 bits).

          5. That’s why Rigol DS1054Z can take: 1GSa/s (Single-channel), 500MSa/s(Dual-channel), 250MSa/s(Fulll-channel). And that’s also why one can hack them to work with higher bandwidth…

    3. I really completely disagree.

      You can hack your way around slow and limited sample depth scopes. Cut down the sampling rate, interpolate in software, or if you’re lucky enough to have a repetitive problem, just stitch together buffers to make a longer sample depth. Is it a pain in the neck? Sure, but it’s *doable*.

      You can’t really hack your way around analog bandwidth. The information just *isn’t there*. To me, a 100 MHz scope is “minimally usable,” but 500 MHz is the edge of what I’d call a “real” scope. Especially with hobby-level development boards getting faster and faster, a 100 MHz scope isn’t going to cut it with a SPI ports able to push 80+ MHz. And at 500 MHz, interesting stuff starts to become viable, like looking at downconverted RF.

      Used Tektronix scopes with 500 MHz analog bandwidth show up on eBay for around $500 pretty often. If it were me, I’d go that route.

      1. I got a lecroy 4-ch 350mhz scope for $100, with two probes.
        The downside is that the shipping price was equal to the purchase price because it’s ginormous and heavy. There’s a lot to be said for scopes from the post-CRT era and the ease in moving them around.

        by the way, about probes. For any decent scope, you’re likely to pay more for probes that can match the scope than you do for the scope itself, so keep that in mind when making your selections. We generally require the probes to have twice the bandwidth of the scope, because probe bandwidth is usually the 6db rolloff point. GHz probes aren’t cheap.

      2. OK, where does this ‘GHz probes cost a fortune’ idea come from?

        You can make a high-impedance, 1 GHz probe out of basically scrap. If you need an active probe, there’s been one featured here that’s got a 3 dB point at 850 MHz. And if you’re buying used stuff on eBay, GHz probes on eBay are under $100 (although you usually need to buy the tip kit as well).

  4. Tektronix and Keysight sell low-end digital scopes. They’re a little bit (but not much) more expensive than the cheap Chinese scopes and have slightly lower specs. But do they offer ANY advantage (besides the cachet of a respected name)?

    1. At least with Tektronix and Keysight you “MAY” have a hope in-heck of getting a bug fixed in the re-branded buggy Chinese ‘scopes they now re-sell. After all, the extra money you paid for in the U.S. brand-name must be worth Something, right?? Or am I wrong to assume this? (Comments from users with experience with this are welcome.)

  5. I have a Rigol DS1054Z and a Siglent SDS1052DL. They’re roughly in the same ball-park price-wise. Where I find the Rigol to shine is with its incredible memory depth – a factor I believe many first time buyers overlook. The Siglent has 32k points, whereas the Rigol has 12Mpts (Std.)! Having the ability to zoom in on far-apart signals can be very handy. The Rigol also feels a little more solid and polished (both in software and hardware).

  6. If you can afford it, get the DS1054z. If you have extra money, then get one of the value added features (signal generator and logic analyzer the only ones that cannot be unlocked in software).

    If you are convinced that you need better than 100MHz of bandwidth, and are willing to buy the active probes and other equipment needed to accurately measured such signals, get the DS1054z for everyday work and a used boat anchor style high speed scope (ex, lecroy lc684 which gives 1.5GHz bandwidth and RIS mode for up to 200Gs/s sample rate for about $1k) or better yet a used sampling scope (ex HP 54121 gives 20GHz bandwidth and an effective 4,000Gs/s sample rate and for about $1k). Then use your savings to buy a network analyzer, high performance signal generator, active probes, etc that are needed for doing real RF work.

    My logic is that while the DS1054z looks mid-range on their spec sheets (bandwidth, sensitivity, noise, sample rate, etc). the software and user interface is as-good-if-not-better than when you will get on a current generation $10,000+ name brand ‘digital phosphor oscilloscope’ style instrument. High resolution screen, 10k+ captures/second, ethernet connectivity, software logic decoders (serial, etc), go/no-go testing, advanced math modes, glitch capture, screen persistence, advanced measurement capability, etc. It is also build very solidly, and we have seen less failures of the Rigol 1xxxZ series then the tektronix DPO2xxx series scopes which cost literally an order of magnitude more (in a student lab environment).

    If you can’t afford the $400 for the DS1054z, then your choices are much more limited, and honestly I would recommend getting a used analog scope and cheap usb scope for sub-$100 and saving up for the rigol in the mean time.

    1. “and are willing to buy the active probes and other equipment ”

      There are plenty of relatively cheap probes that’ll go up to 300-500 MHz, no problem. Capacitance can become an issue depending on what you’re looking at, sure, but it’s still a lot more information than a 100 MHz scope will give you. And that’s only if you *need* a probe.

  7. Decision tree:

    1. Is it free?
    a. Will you wife/sig other flip out if you take it in your house?
    b. Get free scope, no questions asked.
    a. Goto 2
    2. Proceed with trade studies / shopping / etc

  8. You might also look into the last of the “analog” portable scopes. I have and love a 1990s vintage Tektronix 2247A that is rather impressive. 100 MHz, four channel (only two are full featured though), analog signal path and crt, but with an extraordinary amount of digital measurement capabilities. Things like frequency, amplitude, period, rise time, fall time, delays and phase shifts between channels, a real time cycle counter, multiple delayed timebases allowing for zooming in on any part of a signal, various volt meters, cursors, and a whole lot more. The digital measurement stuff has a 200 MHz bandwidth, too. You can find a good one on eBay, etc, for around $400-500.

    1. Likewise the 2467. Plus these are at least marginally repairable. Their newer stuff is great but it’s pretty much signal goes right into unmarked chip one, that connects to unmarked chip two, that connects to a video system, and if something dies you’re hosed.

    1. I Agree 100%. The quality of the FFT tells a LOT about the overall quality of the oscilloscope firmware/software quality and attention to detail. I wish somebody (other than Dave Jones) who really knows how to test this would compare FFT between the current low-end ‘scopes. Kudos to David Jones for his eevBlog content, but he’s really NOT an RF capable Engineer (IMHO). His expertise lies elsewhere.

    2. Not sure about that. Depends on what you intend to use the ‘scope for, I suppose. If the firmware is flawed, you can download the samples to a PC and do FFT there. Keep in mind though that they all use 8 bit ADCs. One notable exception is the Analog Discovery (II) which uses a 12bit ADC (albeit only at 100MS/s and a fairly high noise floor). For low frequency FFT, a PC’s sound card might be the better bet.

  9. I mentioned this above, but really, I’d seriously recommend buying a used Tek scope from the 90s, if you don’t mind the space (and given the vintage hardware, you probably don’t). They’re great scopes. Yes, they’re a little slow (yay 200 MHz pentium CPUs!), but the analog performance is great.

    And, to be honest, you’re talking about increasing the bandwidth to 200 MHz with a hack? How about a TDS744? Used on eBay for ~$500, and you can modify them to 1 GHz bandwidth, and expand them to 1M sample depth.

    OK, of course, I’d rather have *both* scopes, but really, a lot of what people use fancy low-end scope features for can be done with logic analyzers just as good, but you can’t beat high-end analog performance.

    1. I’ve read that the older Tek scopes — TDSxxx series — are plagued by capacitors that leak on to the PCB, etch the traces, and render the scope dead due to PCB failure. You’ll see some Tek scopes on EBay that show the self-test reporting an “acquisition failure” or something like that, which is apparently a symptom of the cap leakage having already done its damage. Anyone have some experience, good or bad, with this issue?

  10. I use old, soviet C1-79 analog scope. It’s build like a tank, works like a tank, and I think it’s bulletproof like a tank. Two-channels, 50MHz with all the features soviet tech could offer. And it weights slightly less than my fridge! With such an indestructible beast one not only can learn quite a lot, but also get in shape by simply hauling it on the desk every day…

    1. What about state of cathode ray tube, it’s still OK? I have DT-516A export version with 25MHz bandwidth and problem with horizontal deflection (straight line is curved upwards). Old analog scopes like that are too cumbersome to use IMHO – too much workspace is used and also you need some time to warm-up the device. It’s better to buy a digital scope unless someone gives you analog for free (I don’t count Tektronix).

      1. Lamp is in a quite good condition. Spares are available too Mostly old tubes made in 80’s that never were used. But I think they still make this scope. I can’t afford a DSO, so I’m using this monster. A real scope for real men…

      2. You know the one great advantage of these beasts…. they stay put… get a little tension or tangle in the probe wires, they don’t fall over on top of your project… turn around to reach for something and knock the bench, they don’t fall over on top of your project…. cat comes in, jumps up on the bench to “help”, they don’t fall over on top of your project.

        ptuk, if I had that one, I might try waving a magnet around it, in case the shielding that side got magnetised, with patience, might be able to undo it, straighten the line out…. unless you put a sine or square wave in it and the peaks and troughs aren’t parallel, then it’s probably something in the yoke or drive.

  11. I don’t do radio work, so a 50MHz bandwidth is more than enough for me—I’m trying to decide between the Rigol DS1054Z and the Digilent Analog Discovery 2. The software looks better on the Digilent USB scope and it is certainly more portable, but how good is it really for small signal work?

    Has anyone got both the Rigol and the Digilent and compared features as implemented, rather than as advertised?

      1. I happen to have and like them both and consider them good deals (but then, I’m financially more at ease than I was twenty years ago). They are different beasts though: the Rigol 1054z is simply a well done entry level DSO, the Analog Discovery (II) is an educational tool.

        The Rigol, as a ‘proper’ stand-alone oscilloscope with (fairly) traditional user interface and robust triggering is much better suited for trouble-shooting, the Discovery, with it’s 14bit (not 12bit as I erroneously wrote in a different comment earlier) ADC *and* DAC and 16bit LA is so much more than just an oscilloscope and if you’re content with building circuits on a breadboard, than the 100MS/s sampling rate isn’t a limitation.

        I was pleasantly surprised to find that I was able to install the software (WaveForms 2015) on an old laptop running Windows XP SP2 and use it battery powered (floating). It doesn’t seem to trigger very reliably though (done in software on the PC?) — forget about hunting runts using it. Otoh, It’s quite fun to build a simple circuit on a breadboard, connect the AD as network analyzer and compare the result with the simulation (and ponder why there is such a difference).

        I’d think beginners should consider the AD, more advanced hobbyist and professionals will have a clearer understanding of their needs.

  12. I’d recommend the Rigol DS1xxxZ series because of the sample depth. I have DS1102E which I’ve been quite pleased with except for the slow transfer times writing data to USB. Supposedly there is a firmware fix, but there were no release notes for the fix at the time I inquired. I was told there would be release notes “soon” but that was over a year ago and still no word. Lately I’ve been studying the mathematics of compressive sensing which has absorbed all my mental cycles so I haven’t followed up on it.

    The notion that you can hack a 50 MHz scope into a 100 MHz scope by changing the firmware strikes me as very naive. I strongly suspect that all the models are built on the same line and then get labeled based on what specs the unit meets at final test. Scope front ends are seriously nontrivial to implement. When the Z series scopes came out the top end model was quite pricey, but a few months later had dropped quite a bit. I suspect that was entirely due to yields on the production line.

    I would love to see test results on the front ends on some early DS1054Es. I’d bet they don’t meet the specs for the DS1102E. The yields may have risen to the point that the hack will work on many units today, but I wouldn’t want to count on it without having test gear to check it. Try building a <1 ns rise time pulse generator and you'll get an object lesson in just how hard meeting the rise time requirement for a 100 MHz scope is. I was introduced to the difficulty while repairing a Tek 465. After several iterations I found an affordable Tek 105 and gave up my pulse generator project. I did get the 465 working, but it was a lot of work. Corroded transistor sockets and bad solder joints are tedious to find and fix.

    1. No it’s well established that the scopes all have the same analog band with circuits but they degrade the signal in the software. Siglent Hantek and rigol all do this trick. Most people hearing about this have the same initial doubt you do which is a good one but in fact it’s actually true. One aspect of this however is calibration and certification. It could be you hacked scope is neither calibrated or traceable so just because you unlock more analysts g bandwidth it does not mean the software properly corrects the amplitude and phase response needed for waveform shape fidelity. But in practice it seems to work . For me it’s better to use test equipment as test equipment not an experiment so I don’t plan to do this

  13. If you have more time than money, you can get an “oscilloscope kit” on Amazon for under $25. It’s not going to replace a good traditional scope, but it puts one in the reach of more people.

  14. I’ve gone the route of old tekronics scopes. My most recent find is a 500Mhz 7934 storage mainframe oscilloscope from 1992. It has amplifier modules for single channel 1Ghz, dual channel 400Mhz and dual channel 200Mhz. It has two slots for amplifier modules so I can display 4 signals at once. It also has two time base modules so I can trigger on two different signals. The storage doesn’t seem to work quite right so it go nothing on a more modern digital storage scope. I only paid $100 for it at a hamfest. Last year I picked up a much older 453A that was only dual channel 60Mhz for the same price.

    The old scopes still work just fine. The high bandwidth of my 7934 scope is great as it lets me make measurements on 2M.

    I’d still like to find a digital storage scope from the late 90’s. I can recall our benches in college have scopes with really nice measurement features built right in.

  15. I love my Siglent SDS1102CML, but I’m hardly a power user. I will echo some of the concerns others have here:

    The UI is not instantaneous…it triggers, samples, then proceses for a moment before displaying anything. It’s not a problem for me but it is absolutely not real time in the sense that an analog scope is. The information on the screen represents really an infinitesimal moment in recent history, and especially if you’re using auto-trigger, that can be kind of misleading.

    Also, while it has a fairly decent-sized sample buffer, it is not “huge” by any stretch. I can’t generally zoom in much in the time domain after the sample is acquired. I find myself re-triggering with a different time-step to try to get enough detail if the interesting part isn’t right at the trigger point. I kind of wish it had a much more complicated triggering mode, like “wait 1ms after the trigger before sampling”, but I don’t wish it enough to flash an experimental firmware on it, or even to read up on any obscure triggering options it might have. :)

    I think my satisfaction is largely based on how crappy my previous scope was. For best flavor, use low expectations.

    1. I discovered the exact same thing! With a very slow sweep rate, it’s obvious to see that the screen does NOT respond to the input signal. It acquires a record of data internally, and then dumps the data points onto the screen all at once. Lather, rinse, repeat. It is NOT real-time.

  16. To measure rise time or bandwidth for scope or probes, I use a 35Mhz – 4.4Ghz rh signal source (on your favorite auction site), the output is a 1 Vpp square wave, most likely with rise time < 100ps (I measured 500ps but this was the limit of my scope ). It is also interesting to see how much signal one can see way above the advertised bandwidth for instance on a 35Mhz analog scope I could still see a 1ghz sine wave from this generated (albeit heavily attenuated)

  17. I bought an el-cheapo Owon SDS7102V a while back. And in all honesty, it’s fulfilled its purpose and I’m more than pleased. I do plan on moving to something better, but as an upgrade from an old Tek that weighs as much as a school bus, it’s great. Is it better than a Rigol? No idea. It does what I want for a price I can’t argue with.

  18. I bought a Rigol DS-1102E a few years ago and I could not be happier with it. because of that, I highly recommend going with a DS-1054Z, which has very a very similar analog front end and AD converters, but has 4 channels, a larger screen, and a bunch more features.
    And if you’re concerned about bandwidth, check out this project I did stretching the scope’s frequency limits:

  19. I’m using a Rigol DS2072 and DS1054 regularly, and played around with several other oscilloscopes, young and old, from different companies, so I guess I’m at least not terribly unqualified in this discussion.
    What’s important, yet often overlooked, is ease-of-use. Modern scopes have tons of features and far fewer buttons than features. You have to rely on menus. Are the menus structured in a sensible way? Do you usually find what you’re looking for in little time? Do the rotary encoders work in a sensible fashion or do you constantly have to alternate between different encoders for no reason?
    One would think getting it done correctly shouldn’t be such a problem, it’s mostly software after all. However, sadly, many chinese manufacturers either don’t get it or don’t seem to care. On paper, their scopes are quite nice, but they are just no fun to use at all. If you do the same thing over and over and over again, or have a configuration preset that you can load over the network, this might not matter. But for multi-purpose instruments, you want to be able to quickly configure them to your liking.
    Usability is where the expensive traditional manufacturers excel. Using a Tektronix, or a LeCroy, or an HP/Agilent/Keysight/howeverthey’recalledwhenI’mdonewriting scope is just very nice. Is it nice enough to warrant the extreme premium over Chinese manufacturers? Depends, but I don’t think so. However, go for a nice Chinese manufacturer. I tend to prefer Rigol, they seem to put a bit of thought into their software, so usually you can find what you’re looking for with only minor amounts of frustration. The DS2072 is far nicer than the DS1054; the DS1054 is very cramped and a bit fiddly. However, still usable.

    The DS1054 is a steal. If you need a scope and you’re not quite sure what you want, get the DS1054. Period. It’s built to a price and has its flaws, but all in all, it’s an absolutely great piece of test equipment that will serve you very well.
    If you don’t need 4 channels, but would like the larger memory, higher bandwidth, better recording features (they are available on the DS1054, but getting there involves many levels of nested menus), and overall better/nicer ease-of-use, get the DS2072. It’s about twice as expensive as the DS1054, so it’s a tough decision.

  20. I’m not sure what upholstery has to do with oscilloscopes anyway but, the cheapest “real” scope I can find is under USD $250 shipped from HK, the UNI-T UTD2052CL. Does anyone have any experience with these or the handheld combined multi-meter and scope (UT81B) that they also make?

    I have to set up a lab/maker space for a group of kids so I’m after the best gear for them to learn the concepts on and don’t require the fancier gear often mentioned in online dick-waving competitions, if you know what I mean…

    I also have a pile of surplus ex corporate HP and IBM desktop PCs runing Linux that I can use with the Hantek 6022BL then pipe the data into whatever software I need, even scilab etc. so FFT in the actual scope may not be relevant.

    Does anyone have any practical experience that they feel is relevant and would like to share?

  21. The OWON – I got mine at Saelig. – The series has all the bandwidth but the big thing is 10 Megasamples per channel!
    This means you can rewind and zoom in on the signal if the triggering event is a while after or before. You could be looking at 10mS and zoom into microseconds.
    It also has an optional battery which helps if you need to take it to the field or to your vehicle (I do a lot of automotive).

  22. I recently traded my Tektronix 7000 series mainframe for other hardware, the seller was willing to trade several thousand dollars’ hardware for it so away it went…

    I liked it fine, it was a 100% analog 150 MHz ‘scope, and you could have four channels, dual timebases, a plugin logic analyzer, et c. But, I bought a Tektronix 2432A to replace it — 300 MHz, 250 megasamples/second, GPIB, two channels. I paid less than $250 shipped for it, from a major test equipment calibration company. It came with a 90-day warranty, the manual, and a recent calibration. It’s a fine little ‘scope, my first personal non-analog unit, but I now understand why no one in college seemed to be able to figure out my old 7000-series: there’s an “AUTO” button that basically sets the thing up for you!

  23. I basically checked all local second hand web sites, and ebay for oscilloscopes, kept looking up their manuals, and compared them. Slowly this builds a kind of ordering of different models of scopes, according to your needs, and sometimes you see a broken or AS IS scope, I kept interpreting the pictures of the symptoms, and read the repair manuals if available, to estimate the probability of success. I got sick of trying to sort them all out a couple of times, and shied away from the task for half a year or so each time. Then the last time I was going trough them all again, I found a Tektronix TDS5104 1Ghz 4ch oscilloscope for one US kilobuck, with a “boot device not found” BSOD pictured on the scope on ebay. I found forum posts of others copying their almost failing HDDs to SSDs, so I decided do take the risk of there being more than just the HDD problem. It arrives, and I pay about $300 transport and $300 customs, and then sub $100 for a new SSD. I successfully backed up the drive contents to SSD, the scope even boots faster now! So it cost me $1700 and I now have a working scope the same model is being advertised by others for around $4000-5000 on ebay…

  24. One option is also that you made one yourself for free. Open source, cross platform and completely free! Software, hardware and firmware included. Windows, MAC and Linux. You can get free chips samples from manufacturers, so the biggest expense would be the PCB board.

    Take a look at or .

    Some basic specification:
    – 100 MHz Real-Time Sampling rate / 3.2 GHz Equivalent-Time Sampling rate
    – 2 analog oscilloscope channels (10-bit)
    – 2 analog generators (12-bit)
    – 16 digital channels (logic analyzer / pattern generator)
    – USB 2.0 interface

    And you can easily hack-it :-)

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.