Thinking about it: You would never find a "Grounded" multimeter as robust and useful if a path to ground through the multimeter were introduced, modifying the circuit's behaviour and possibly damaging the multimeter with currents.

Why are so many oscilloscopes earth referenced? Upon reading some educational material, a majority of the "common mistakes made by students" are placing the grounding clip incorrectly and causing poor results - when the o-scope is just being used as a fancy voltmeter!

I've heard of a Tek scope having an isolation transformer within.. however ignoring that, and taking in to account that newer DSOs may have plastic cases (isolated from you most importantly I would assume) could I just remove the earthing pin, and install a 1:1 AC transformer inbetween the o-scope and outlet and be on my merry way probing various hot/neutral/earthed sources with no worries about a path to ground any longer through it?

  • 9
    \$\begingroup\$ The common mistake made by students when they place the grounding clip incorrectly is usually not related to earthing. The mistake is that they don't realise that voltage must be measured between two points, so they do odd things. I've seen students leave the clip off, connect it to the same thing they're probing, short out multiple pins with it, try to connect it to mains earth (by pulling the plug halfway out...!), all sorts of strange and bizarre behaviour. For "grounding clip" read "black lead of multimeter" in the educational material you found. \$\endgroup\$ Feb 23, 2012 at 7:52
  • 15
    \$\begingroup\$ @OptimalCynic, many many O-scopes hard tie that to earth. I have seen projects go up in smoke from connecting it to something driven and it pulling it to ground which happened to be what the power supply they were using referenced for neutral. \$\endgroup\$
    – Kortuk
    Feb 23, 2012 at 8:05
  • 2
    \$\begingroup\$ Oh sure, that's true, and I think the point of the original question (which is why I posted a comment, not an answer). I just wanted to point out that most student errors, in my experience, relate to not understanding voltage measurement rather than not understanding earthing. But then I taught mostly electronics engineering students, so we're lucky the lab didn't burn down more often. \$\endgroup\$ Feb 23, 2012 at 8:08
  • 14
    \$\begingroup\$ @optimal at least all of the students I have taught, the common mistake was connecting the ground pin to something that wasn't ground. This is because they did understand voltage to be relative and thought it was just like a multimeter. \$\endgroup\$
    – Kellenjb
    Feb 23, 2012 at 12:43
  • 3
    \$\begingroup\$ electronics.stackexchange.com/questions/12023/… \$\endgroup\$
    – tyblu
    Feb 23, 2012 at 13:11

10 Answers 10


Oscilloscopes usually require significant power and are physically big. Having a chassis that size, which would include exposed ground on the BNC connectors and the probe ground clips, floating would be dangerous.

If you have to look at waveforms in wall-powered equipment, it is generally much better to put the isolation transformer on that equipment instead of on the scope. Once the scope is connected, it provides a ground reference to that part of the circuit so other parts could then be at high ground-referenced voltages, which could be dangerous. However, you'll likely be more careful not to touch parts of the unit under test than the scope.

Scopes can also have other paths to ground that are easy to forget. For example, the scope on my bench usually has a permanent RS-232 connection to my computer. It would be easy to float the scope but forget about such things. The scope would actually not be floating. At best a fuse would pop when it is first connected to a wall powered unit under test in the wrong place.

Manufacturers could isolate the scope easily enough, but that probably opens them to liability problems. In general, bench equipment is not isolated but hand-held equipment is. If you really need to make isolated measurements often, you can get battery operated handheld scopes.

  • 5
    \$\begingroup\$ Yes, but why not simply make the scope inputs differential? \$\endgroup\$
    – nibot
    Feb 23, 2012 at 14:42
  • 6
    \$\begingroup\$ Differential like this? tek.com/differential-probe (Note the price tags.) \$\endgroup\$
    – markrages
    Feb 23, 2012 at 16:11
  • \$\begingroup\$ USB oscilloscopes (or DAQ cards, they're almost the same thing) can be powered bu PCI or USB, which are continuous supplies; they often provide fully differential inputs (and you can find them for 300+ euros/dollars). \$\endgroup\$
    – clabacchio
    Feb 23, 2012 at 16:34
  • 7
    \$\begingroup\$ This answer answers the question, "why isn't the scope chassis ground isolated from earth ground", but it doesn't answer the question, "why don't scopes usually have differential inputs." \$\endgroup\$
    – nibot
    Feb 23, 2012 at 20:36
  • 7
    \$\begingroup\$ @nibot: Right. As you said, I answered the question that was asked, not the one you made up. \$\endgroup\$ Feb 23, 2012 at 21:04

WARNING: Means of being less likely to die or to kill others is discussed below.
Reading may be considered useful.

Consider this "informed opinion" rather than holy writ.

In an oscilloscope you want the best result you can get for the money and it's far easier to build a single ended amplifier rather than a differential one when you want ruler flat gain for as much of the bandwidth as can be had and constant phase delay. An oscilloscope more often than not is required to measure a non sinusoidal signal, so it's ability to faithfully deal with the frequency components "across the range" is vital to it providing you an accurate picture of what is really there. In many cases single ended is good enough so they can give you dual earth referenced channels for more than a single differential channel but less to noticeably less than due differential channels.

Many better oscilloscopes have a dual channel subtract mode which gives you an approximation to a differential amplifier. The approximation is readily obvious in the ability to independently adjust channel gains so as to null common mode pedestal or to add in some component of it if desired.

People who "must have" true floating differential mode can buy the optional add on sure-to-be-dear differential module.

"Floating" an oscilloscope so that the local ground is not at true ground is a time honoured and usually somewhat frowned upon practice. In many cases it works well enough. The common practice in my (long ago) university days was to have a power plug with a pick-a-back socket. The plug had the earth pin cut off and said plug was usually coloured bright red in an attempt to forestall the various inevitable consequent events. For this to work the amplifiers must have been built single ended but floating with the scope ground wire the only connection to local (real) ground. This places this method in the same class as the one described in the question - the "1:1 transformer is substituted for by an isolated power supply. Where a SMPS is used for the mains supply (as will almost invariably be the case with equipment made at any time since Noah retired) then one may have to think about possible effects of X & Y capacitors connected between scope P&N and scope ground. These are isolated from true PNE by the 1:1 transformer but may still do interesting things to an attempt at emulating a proper balanced amplifier.

A BIG problem occurs (or can) when you try to be on your merry way probing various hot/neutral/earthed sources with no worries about a path to ground any longer using two channels at a time. Or trying to. The newfound freedom to use your probe tip + ground as Vin1-Vin2 does NOT extend to two channels where Vin2 is not the same for both channels. This is blindingly obvious either before you read this or as you do BUT the fact that you did not include it or note a caveat in your question shows how easy it is to overlook. Even if YOU are aware of this not so subtle limitation, it can still kill someone who uses your equipment or who works with you or who just happens to be physically in contact with the earth clip on channel 2 when you decide to connect channel 1 earth to phase. Obviously , a bit of common sense makes this practice safe [tm] for a rather more limited range of values of safe than most people are used to. Your local occupational safety inspector (names vary by country) or boss or small daughter may find the practice wanting.

All that said, most of us do things regularly as part of our workaday lives that are able to kill or maim us or others in a moment if we violate social contract or step outside acceptable agreed bounds of behaviour. Driving an automobile down a 2 lane road with oncoming traffic on the other side is a good example. Your car and your scope can kill you and others if used wrongly. Making really really really sure that the scope doesn't is a minimum requirement for playing the games that you describe, A floating isolation amplifier MAY be a better choice.

  • 3
    \$\begingroup\$ +1 for "can kill you", some inner workings, and other options. I will have to read this a few times over when I am ready. Thank you Russell. \$\endgroup\$
    – Transient
    Feb 23, 2012 at 23:52

A big reason that scopes are earth-grounded is that they generally have two or more channels which share a common ground reference. If a scope isn't earth grounded and one of the probe's ground inputs is connected to an ungrounded exposed metal chassis while another ground input is connected to AC120, the first probe could electrify aforementioned chassis with AC120, thus creating an extremely hazardous condition. The design philosophy of the scope is that if one probe's ground might be connected to an exposed metal object and another probe's ground clip touches a hazardous voltage, it would be better to have sparks fly at the latter point (or worse) than to create a hazard at the former.


Its to stop the 'scope reading the EMI flying around.

Earthing the 'scope probe ground is connecting the long cable screen of the probe to earth. The screen stops radiated interference disturbing the voltage on the measurement wire at the centre of the probe cable.

Otherwise your measurement could contain radiated noise from the equipment under test, mains lighting and whatever else is kicking around (usually lots in a test lab').


I've never been in a situation where I would need to isolate the o-scope via an isolation transformer. I use the isolation transformer on the equipment I'm measuring, like a TV, or a SMPS. Their first stages of the power supply are usually non-isolated. My experience comes from working at a TV repair shop a few decades back, and we always assume the unit is not isolated or its isolation is compromised until tested for leakage currents of exposed metal on the exterior.

I've never seen the need for those diff probes because if you select the correct o-scope, it will handle that input correctly.

Also to answer the tittle to your question, it is because its a requirement that exposed metal should be grounded according to appliance regulators like UL and CE.


To save cost. That is the only reason now.

You can buy fully isolated scopes now - especially with the advent of fully digital scopes, they can literally have an isolated ADC unit for each channel which then transfers the waveform data digitally over a digital isolator with zero loss in quality/linearity.

Fully isolated scopes are far superior in all ways (except cost.) You don't have to worry about ground path loops between the shields of neighboring probes, or between earth ground and your circuit under test. You can simultaneously measure the voltage drop across multiple points when none of them are ground referenced on either side.

But why is non-isolated so prevalent if isolated is so much better?

Back in the olden golden days, scopes started out being single channel, and big, and using primitive components. It was quite a feat of engineering just to get the scope to respond linearly from 0 to 100Mhz.

Isolating an AC signal is easy - just use a transformer. Linearly isolating a DC signal is much more difficult, although possible using opto-isolators and feedback loop, or a voltage to frequency converter and an isolation transformer -- but that doesn't work for 100Mhz because you'd need Ghz frequencies.

Remember, the oscilloscope was invented about 120 years ago. What is easy - or at least plausible now - was not back then. Isolating linearly from DC to even 1Mhz using tubes and such was basically out of the question - and since most scopes were single channel, or maybe 2, technicians had to just deal with the fact that both probe ground leads were connected to each other as well as to ground, and the scope was still extremely useful with this constraint.

That of course set quite a tradition of "We've been doing this for a hundred years, we'll keep doing it that way."

But nowdays non-isolated scope design is just stupid. Scopes are often 4 channels, and we often need to measure voltage differences in circuits which are not ground referenced - and the technology is totally here now make fully isolated multi-channel scopes.

Modern scopes just convert the incoming voltage to digital anyway. At that point, why not pass it through isolators and have an isolated scope?

But people are lazy. Rather than put the effort in to update their designs, they just apply minor adjustments from last year's model and call it good.

At some point a leading manufacturer will just go all-isolated (for the same price) then the game will be over, all the manufacturers will have to go full isolated or go away.

  • \$\begingroup\$ consider how expensive a digital isolator that can work at the kinds of speeds an oscilloscope needs would be... Not to mention the need to power the isolated side, the effort needed to isolate everything properly, and you get something far more expensive than is really worth it. Just get a differential probe and a regular oscilloscope; it'll be far cheaper. \$\endgroup\$
    – Hearth
    Oct 14, 2020 at 1:52
  • 1
    \$\begingroup\$ @Hearth, It's not as bad as you think. Gigabit ethernet is isolated. Look how common and affordable that is. High speed digital isolation is nothing new right now. And isolating power is nothing hard either - use the same technology that is in every laptop CCFL backlight circuit. Now, I just went and looked at a Keysight 100Mhz differential probe - $1700. Get 4 of those and a 4 chan DSO, and you're out $8000. Compare that to a Tektronix TPS2024B which is 200Mhz, 4 channel, fully isolated, 8 hour battery, only $6000. And that's Tek, and that's with very little competition of isolated scopes. \$\endgroup\$ Oct 14, 2020 at 2:19
  • \$\begingroup\$ Remember that your digital isolator needs to work at the sample rate of the scope, not the analog bandwidth of it. The TPS2024B is a 2GS/s scope, and since it's 8 bits per sample, even assuming that's 2GS/s total rather than per channel (it probably is total, anyway) that's two gigabytes per second of data, or 16 gigabits. A fair bit more than gigabit ethernet! Though you are correct that a Tek TPS2024B is cheaper than four diff probes plus a scope. I wasn't aware of the TPS series, and I'll concede the point that isolation evidently can be done cheaply enough to be worthwhile. \$\endgroup\$
    – Hearth
    Oct 14, 2020 at 15:30
  • \$\begingroup\$ However, note that you can also get a TBS2204B 200 MHz 4-channel scope without the isolation for just $3000. And if you want to go even cheaper, you can get a Rigol DS1104Z for less than $700, though that's only 100 MHz and from a less well-respected brand. Isolation adds a lot of cost to a scope, and if you don't need it to be isolated, you probably want to save yourself the $3000 and get a regular scope--or spend that $3k on some other feature you need more. Then you can just get diff probes for one or two channels as needed, and it's still cheaper. \$\endgroup\$
    – Hearth
    Oct 14, 2020 at 15:34
  • \$\begingroup\$ @Hearth wtf!? 16gbps is not a lot of data for optical isolation. 10G SFP+ modules will isolate any CML bitstream from the ADC for $20. 25G increases the cost to $40. These are all retail prices and includes components you don’t need such as the laser and fiber interface so it’ll be even cheaper in volume. Isolating the power supply essentially means you gotta buy a separate psu for each channel, but I can’t imagine that adding over $100 to a typical scope’s BoM. \$\endgroup\$
    – Navin
    Apr 11 at 7:07

If an earth scope is not earth referenced must have all its channeld isolated from each other, both positive and negative probes (to prevent shortcutting one channel's negative input with the other counterparts).

Portable or handheld oscilloscopes with a single channel work this way and you can use them as a multimeter (example Velleman HPS50, Velleman HPS140i). About dual or more channel oscilloscopes, no idea if channels are isolated one from each other.

  • 1
    \$\begingroup\$ Assume that channels on a multi-channel oscilloscope are not isolated, until proven otherwise. Multi-channel oscilloscopes with galvanically isolated channels do exist, but those are a specialty. \$\endgroup\$ Jul 19, 2015 at 17:50

Safety earthing conductive touchable parts prevent people killing themselves while measuring high voltage potentials - otherwise the scope and everything connected to it would be at dangerous potential.

Isolating the channels from eachother and from the frame of the scope would increase the price level of the scope. At same time need to prevent any conductive parts to be touchable. Some TEK scopes implemented this 10-15years ago.

If you know what you are doing getting the scope floating is not a problem but then the person must understand fully the safety issues, preventing to be able to touch conductive parts/devices connected to the scope and/or isolate the DUT also.

  • \$\begingroup\$ why isn't a bench DMM earth referenced then? the earth referencing is not related to safety \$\endgroup\$ Apr 12, 2020 at 12:09

The external case must be grounded, but inputs not at all. Changes might have to be made making the BNC exposed ground obsolete. Easy to do and should be. Regards CB


Try thinking about what would happen if the scope was floating. It would be reasonable to connect the probe clip to one side of a differential (eg RS422) line, and the probe tip to the other leg. The tip side would typically see 10M/10pf loading, but the clip side would see an ill defined capacitive load back through the mains transformer to mains and neutral. Anything measured above the audio spectrum would be highly suspect.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.