Why are oscilloscope input impedances so low?

My question is two-fold:

Where does the input impedance come from?

I'm wondering where the input impedance of your average multimeter or oscilloscope comes from? Is it just the input impedance to the device's input stage (such as an amplifier or ADC input stage), or is it the impedance of an actual resistor? If it is the impedance of an actual resistor, then why is there a resistor at all? Why not just the input circuitry?

I measured the input impedance of my oscilloscope with a DMM. When the scope was turned off, the DMM measured about $$\1.2\mathrm{M\Omega}\$$. However, when the scope was turned on, the DMM measured pretty much exactly $$\1\mathrm{M\Omega}\$$ (I could even see the 1V test input applied by the DMM on the oscilloscope screen!). This suggests to me that there is active circuitry involved in the scope's input impedance. If this is true, how can the input impedance be so precisely controlled? Based on my understanding, the input impedance to active circuitry will depend somewhat on the exact transistor characteristics.

Why can't the input impedance be much higher?

Why is the input impedance of an oscilloscope a standard $$\1\mathrm{M\Omega}\$$? Why can't it be higher than that? FET input stages can achieve input impedances on the order of teraohms! Why have such a low input impedance?

I suppose one benefit of a precise standard $$\1\mathrm{M\Omega}\$$ is it allows 10X probes and the like, which would only work if the scope had a precise input impedance that wasn't unreasonably large (like that of a FET input stage). However, even if the scope had a really high input impedance (e.g., teraohms), it seems to me that you could still have 10X probes just by having a 10:1 voltage divider inside the probe itself, with the scope measuring across a $$\1\mathrm{M\Omega}\$$ resistor inside the probe. If it had an input impedance on the order of teraohms, this would seem to be feasible.

Am I misunderstanding the input circuitry of a scope? Is it more complicated than I'm making it out to be? What are your thoughts on this?

The reason I thought of this is that I've recently been trying to measure the common-mode input impedance of an emitter-coupled differential pair, which is much larger than the scope input impedance, so it made me wonder why the input impedance can't be larger.

• The topic is much more complex than you might think. You seem to be considering only the DC response, but in fact, a scope must have a flat response all the way up to its specified bandwidth. This is a huge challenge, and standardizing on 1MΩ/50Ω makes the problem at least somewhat tractable for probe manufacturers. Commented May 4, 2019 at 11:16
• Would you like to use my old scope? It can be configured for 100 ohm input impedance. On the other hand, it was built in 1965, and the standard setup for it is 1MOhm input impedance. 1M seems to have been standard for quite a while.
– JRE
Commented May 4, 2019 at 11:24
• Don't forget that a $\times$10 probe has an input impedance of 10 M$\Omega$ Commented May 4, 2019 at 11:24
• @DaveTweed So it is not feasible to have a FET input stage with high enough bandwidth? What are input stages of scopes actually like? Commented May 4, 2019 at 11:26
• Is it directly into the ADC? No, how would a scope be able to measure 1 mV and 100 V? Usual configuration: BNC - input protection + switchable attenuation - Input stage (often FET based) - ADC. So yes many are FET based. You would not have an active device define the input impedance. There's a 1 M resistor to set it properly. I highly recommend that you study how things are done and ask yourself WHY before assuming: it must be ... it cannot be... Because you will confuse yourself. Commented May 4, 2019 at 12:03

I would say a combination of a few factors.

1. The input stages of an osciloscope are a difficult compromise. They need to be have a wide range of gains/attenutations, they need to be tolerant of user errors, and they need to pass high bandwidths. Adding a requirement for a very high DC resistance would just further complicate matters. In particular attenuators needed to handle the higher end of the scopes input level range would get much more complex/sensitive if they needed to have a very high DC resistance.
2. It's a de-facto standard, changing to something else would lead to incompatibilities with existing probes etc.
3. There wouldn't be much benefit anyway.

To further explain point 3, at moderate frequencies (from a few kilohertz upwards) the 1 megohm DC resistance of the scope input is not the dominant factor in the overall input impedance. The dominant factor is the capacitance, with the cable making probably the largest contribution.

(in fact at UHF/microwave frequencies it's common to reduce the scope input impedance to 50 ohm, so the inductance in the cable can balance out the capacitance and the cable becomes a properly matched transmission line)

What this means is if high input impedances are desirable then it's much better to deal with that at the point of probing than at the scope. The typical compromise of cost/flexibility/input impedance for general use is an x10 passive probe.

If you need a really high DC resistance then the solution is to add a FET based amplifier in front of the scope, preferably as close to the point of measurement as possible.

• Is the input capacitance also specifically engineered like the 1Mohm input impedance, or is it just a parasitic element that is measured? (A non-precise input capacitance wouldn’t be a problem since attenuating probes have variable capacitors.) Would I be correct in saying that: if attenuation circuitry was not needed, and we didn’t worry about impedance matching at higher frequencies (in which case you might have a switchable input to 50ohms), then it would be fine to have input directly into high impedance FET stage? Just trying to get the different reasons for this straight in my head. Commented May 4, 2019 at 23:56
• I guess even then, you’d still have probe/cable capacitance to worry about, but in that case adding 1meg across it is just going to make the impedance even lower. And 10X probes could just have their own 1meg resistor in parallel with the probe output. So basically: ignoring attenuating probes, impedance matching, and attenuation circuitry, I don’t see any other reasons for an input resistance as low as 1meg, since it would just make the input impedance due to capacitance even lower (and the impedance-matching ship would have already sailed at 1meg input impedance anyway). Commented May 5, 2019 at 0:25
• So my understanding so far: 1meg input resistance is preferable due to: (a) required attenuation circuitry, (b) input impedance is dominated by capacitance anyway, (c) it makes attenuating probe design simpler. Impedance matching doesn’t seem to he a reason since you’d go down to 50ohms in such cases anyway. Makes me wonder about multimeter input impedances (normally 10meg), where only (a) seems to apply. Commented May 5, 2019 at 0:31
• Another issue with high impedance inputs it "phantom" voltages when they are not connected to anything. Even at 10 meg this can be noticeable sometimes. Some high end multimeters do actually have the option to switch-out the 10 meg resistor, I have access to such a meter but I don't think I've ever felt the need to use said feature. Commented May 5, 2019 at 0:40
• @PeterGreen see if you can disable the 50/60Hz suppression too, and you have a random number generator instead of a voltmeter while it is not connected to something. Commented May 5, 2019 at 7:28

A lot of things are the way they are because of history, and de facto standardisation.

A general purpose oscilloscope input is a difficult compromise between not loading the circuit, not being damaged by high voltage, having reasonably low noise, and being able to maintain a decent bandwidth.

1 MΩ in parallel with 15 pF to 30 pF satisfies a lot of people for a lot of applications. There's little incentive for manufacturers to build a general purpose oscilloscope with a different input, to address tiny parts of the market.

When you do need better noise, or a differential input, or a higher input impedance, then you use a custom pre-amp. When you need wider bandwidth, you switch to a 50 ohm input impedance.

There are special purpose oscilloscopes made at high prices that do address niche applications.

• Fair enough. So the input impedance (to a scope or meter) does not come from an actual resistor, but from active circuitry instead? (Am I crazy for not being sure about this?) Makes me wonder how they can precisely control it. I wonder if there are any schematics of scope input stages/front ends floating around the internet that I could have a look at. Commented May 4, 2019 at 11:31
• @hddh I still find it surprising that a FET input stage of sufficient bandwidth can't be engineered Says who? There are FET probes with more than 1 GHz BW, for example: keysight.com/main/… Prehaps what you mean is, that you want it inside the scope. That could be made yet it would be unusable that way! You need a cable to connect your testpoint to your scope. That cable has capacitance. The whole point of FET probe is that is has a low capacitance. Commented May 4, 2019 at 11:55
• Pointers: EEVBlog ! Also there are plenty of schematics to be found in service manuals of for example older Tektronix scopes. It clearly can’t be a FET with a 1Mohm input impedance (right?). No wrong, he input impedance is set by a resistor then (often) a FET amplifier is used to amplify the voltage across that resistor. The 1 M is needed to have a properly defined impedance. Here's Dave reverse engineering the popular Rigol DS1054Z scope: youtube.com/watch?v=lJVrTV_BeGg&t=989s Its design is typical of many modern scopes Commented May 4, 2019 at 12:08
• And here's a service manual of a Tektronix 2215 analog scope, it has a block diagram and all the circuits. Yes it is an old design but the input stage will be very similar to modern many scopes: tek.com/manual/2215 for study purposes, this is very useful. Commented May 4, 2019 at 12:17
• ..ADC w/ FET input stage isn’t feasible is because of the attenuation required before it to achieve the desired dynamic range? Yes, dynamic range is indeed the answer. A variable attenuator helps to bring the signal into a range that is appropriate to both the input amplifier and the ADC. Commented May 4, 2019 at 12:18

Actually, it is ridiculously high for a wideband input.

There is no practical connector or cable that actually has an impedance (from a transmission line view. Resistance, but for coaxial cablers, gold platers, and waveguide plumbers. RF dudes.) of 1 megaohms, leaving the input utterly mismatched - even worse, a 15-45pf capacitor across an 1 megaohm (transmission line impedance) input would mismatch it to oblivion.

The reason it is 1 megaohm is for supporting standard 10:1 probes, which you indeed need to not overload the kind of circuit carrying audio frequency signals at high impedance and with high DC offset (think audio vacuum tube circuits, the probe designs are from just that era).

However, once you are dealing with RF or fast digital circuitry, the parallel capacitance of the scope input (which you can't make too small, again because of probes, cables, connectors) will dominate ... and bring the actual input resistance of that input down to 5 to 10 kiloohms once you reach one megahertz, 500 to 1000 ohms once you reach 10 megahertz. Reach VHF (hint: ACMOS or F-TTL circuitry is VHF stuff even if you don't clock it at VHF) and you would be better off with a matched 50 Ohm input, since you could connect a (within reason) long 50 Ohm cable and still have a 50 Ohm input on the circuit end, instead of an even bigger capacitive burden.

With the conventional kind of probe and input, you will overload RF circuitry easily. RF optimized oscilloscopes tend to have inputs that can be switched to 50 Ohm input impedance (any oscilloscope input can, with a parallel/through terminator) - which is, interestingly, BETTER suited, since now you can use probes (eg Z0 probes or active FET probes) that actually can be made to present much higher effective input impedances at the probe point. Or just provide a reliable 50 Ohm connection to your circuit with any old RG58 cable.

• If I understand correctly: So you’re saying that 1megaohm doesn’t help with impedance matching, and you’d be better with 50ohm inputs in those cases. So if the impedance-matching ship has sailed with 1meg, then why is a low input impedance of 1meg necessary? The reason I’ve gathered for this from other answers is that the required input attenuation circuitry makes this infeasible. Are there other reasons? (Also is the scope input capacitance intentional like the 1meg, or is it parasitic? - i.e., could it be easily be reduced?) Commented May 5, 2019 at 0:08
• @hddh it was parasitic once, then it likely became intentional :) Commented May 5, 2019 at 7:24

Most scopes have a compensated input attenuator to set the input signal to a voltage in the range of the input stage which will usually have the highest sensitivity of the scope.

This attenuator is usually designed assuming a 1 megohm input impedance - so the input impedance seen at the input connector usually is a result of physical resistor.

If the measured impedance changes when the scope is powered it probably means that there are relays controlling the input attenuator that are not activated in the unpowered state.

There may be a higher sensitivity selection with reduced bandwidth that is handled by increasing the gain of the amplifier. The gain selection may also be controlled by a combination of varying the gain of the amplifier and the input attenuator.

In the attached schematic section the resistor R108 provides the 1 megohm input resistance when the highest sensitivity is selected. The input JFET Q101 has essentially infinite input resistance. the stray capacitances form the capacitance seen at the input in the highest gain position.

At lower gains the resistors R102, R103 and R104 (together with R105, R106 and R107) that make up the input attenuator determine the input resistance.

Trimmers C107, C108 and C109 are used to adjust the input capacitance in lower sensitivity selections to be the same as in the high sensitivity setting.

There are bench multimeters/voltmeters that have much, much higher DC input resistance (still, not much higher input impedance at RF). Using such a device will actually prove to be extremely confusing to the "average" user (it would certainly to someone working on house wiring, vehicles, machinery rather than component-level electronics projects): When the test leads are connected to nothing or to an open circuit, any capacitance in the test leads, input circuits etc will be charged by whatever electric field is nearby, leading to a display of completely random values rather than zero volts (try working with a 100GOhm input resistance bench multimeter, you will see just these effects in practice...).

Also, a device with an input resistance in the teraohms will need to be extremely rugged against static electricity, since it can NOT just inherently dissipate the charges of potentially thousands of volts easily found in the environment - and ESD protection circuitry that reliably does not introduce leakage resistances or even worse leakage current sinks that would compromise the high input resistance appears to be difficult to make...

Btw, in addition, most hand multimeters (not all bench ones do) use quite sophisticated tricks (eg using an ADC clock that has a relationship to the mains frequency of the region they are sold in) to filter mains hum out of the results, which would otherwise again lead to unstable and random results even at "low" 1MOhm or 10MOhm (compare an open circuit oscilloscope probe...however, on a DMM, there is much more potential for misinterpretation).

• Thanks for the answer. It was informative. How can you filter out the mains hum with a choice of sampling rate? If the sampling rate is high enough, you'll capture the mains frequency anyway. If it's too low, then you could alias 50 Hz down to a lower frequency (e.g., 0 Hz or near it) - this seems worse to me, since now you've got an aliased DC or very low frequency signal that you can't digitally filter out as easily (or at all). Commented Aug 24, 2021 at 0:24
• Mostly, by making sure interference affects two consecutive samples in a mirror-imaged way... Commented Aug 31, 2021 at 11:42
• That works for all the odd-order harmonics (including the fundamental) - since you could use a simple 2-tap moving average filter to completely cancel them. The even order harmonics would get aliased to DC though, which would add error to the reading. I suppose power hum in practice will be mostly odd-order... Commented Sep 2, 2021 at 0:55
• I think the good old ICL7106 multimeter chip datasheet has some hints about it... Commented Sep 2, 2021 at 8:50
• Ah okay, it appears to use an integrating ADC with very long integration times - in which case it tries to reject power noise in the action of sampling itself, rather than digitally. I assume there'd be some analog filtering too. That's cool. Integrating ADCs make a lot of sense for multimeters I suppose. I normally work with ADCs/DACs for audio frequency. Commented Sep 3, 2021 at 0:54

Why is the input impedance of an oscilloscope a standard 1MΩ? Why can't it be higher than that? FET input stages can achieve input impedances on the order of teraohms! Why have such a low input impedance?

The reason for it being 1 MΩ rather than the much higher value a FET can achieve is that oscilloscopes originally used vacuum tubes. The maximum tube grid bias resistance was generally set at ~1 MΩ to minimize the effect of grid leakage current, which is typically in the region of 0.1 μA. By the time FETs became available the 1 MΩ standard was already well established.

Since a 10x probe is usually used (more to reduce probe capacitance than to increase resistance) this 'low' input resistance isn't usually a problem. If higher input resistance without attenuation is required then an 'active' probe can be used.