What's the input capacitance of an oscilloscope's 50 ohm input?

[bezzada]

Typical specs look something like this:

Input Impedance 50 Ω ±2.0%, 1 MΩ ±2.0% || 16 pF

I understand the 1Meg input has an 1Meg resistor shunted by 16pF of capacitance.  Where is that capacitance coming from? I thought it may have something to do with the physical shape of the input connector, i.e. two pieces of metal close together cause a certain capacitance.  Or does it also come from the 1Meg circuitry?  Otherwise it is not clear what happens to the 16pf when the input is switched to 50 ohm.  If the same capacitance is still there, wouldn't it still shunt 50 ohm at very high frequencies?   

[TimFox]

I believe the 50-ohm input setting comes from switching in a 50 ohm resistor in parallel with the 1 megohm/16 pF input impedance.
Some of the 16 pF is due to the actual input capacitance of the input amplifier.
Of course, 16 pF is important compared to 50 ohms above, say, 100 MHz.
The specification of 16 pF is important when using a x10 probe with the high-Z input setting, each probe model can be compensated (screw adjust) for a given range of load capacitance at the mating BNC connector.

[tggzzz]

You ask sensible questions.

Many scopes do indeed just have a 50ohm resistor in parallel with 1Mohm//15pF. The specification of the input will typically include the VSWR <=1.3. That gives a big hint that there will be frequency dependent results affected by the length of the cable.

Others have a true 50ohm input, e.g.the Tektronix 485 which has both 1Mohm and 50ohm attenuators.

Your second picture shows a classic Z0 resistive divider probe, best used with a true 50ohm input.

So-called high impedance *10 probes are indeed lower impedance at high frequencies. Their cable is a special lossy coax cable, which avoids the consequences of poor VSWR.

FFI see the 1969(!) Tektronix Oscilloscope Probe Circuits  reference mentioned in https://entertaininghacks.wordpress.com/library-2/scope-probe-reference-material/

[radiolistener]

Where is that capacitance coming from?

This is parasitic capacitance of RF frontend, it includes all things together, include amplifier, relay switch, PCB traces and connector wires. Note that this capacitance can vary at different frequency and at some frequency it turns into inductance...

what happens to the 16pf when the input is switched to 50 ohm.

50 Ω mode just switch to another input circuit. It is designed in such way to keep impedance close to 50 Ω for entire working bandwidth.

Also note that 50 Ω mode doesn't means that there is no reactance. This mode has some parasitic reactance which is increased at higher frequency and can switch into inductance.

[G0HZU]

It's worth noting that there is some advantage in deliberately adding a tiny amount of inductance in series with the 50R termination resistor for some higher frequency scopes.

Otherwise, this resistor would ideally need to be a pure 50 ohms in parallel with a negative capacitor (to cancel the input capacitance and improve the VSWR in 50 ohm mode) but a negative capacitor obviously isn't a realistic option. The next best thing is to add a few nH of inductance in series with the 50R terminating resistor that gets selected in 50 ohm mode.

If done correctly, this should improve the input VSWR up at VHF and UHF as the series RL will (kind of) mimic a resistor in parallel with a negative capacitance here.

[tggzzz]

what happens to the 16pf when the input is switched to 50 ohm.

50 Ω mode just switch to another input circuit. It is designed in such way to keep impedance close to 50 Ω for entire working bandwidth.

No, it depends on the scope. Many scopes just add a 50ohm resistor in parallel with the 1Mohm//16pF.

Check the scope's VSWR specification.

[radiolistener]

No, it depends on the scope. Many scopes just add a 50ohm resistor in parallel with the 1Mohm//16pF.

that resistor has some reactance which has inverted frequency curve in order to compensate reactance of RF frontend input. So it allows better matching than using external 50 Ω pass-through dummy load.

[bdunham7]

Check the scope's VSWR specification.

What would be good?  I think Siglent is one of those that pretty much just adds a resistor.  My Siglent SDS2354X+, measured with an imperfectly calibrated NanoVNA (I couldn't find a BNC short so I approximated...) gives me 1.43 @ 500MHz an mostly below 1.2 below that.  My Tek 485 is specced for <1.25 @ 350MHz and <1.15 in the higher ranges.  Tek 2465B is <1.3 to 300MHz, <1.5 to 400MHz.  I didn't try different ranges on the Siglent.  So they seem close?

[bezzada]

Others have a true 50ohm input, e.g.the Tektronix 485 which has both 1Mohm and 50ohm attenuators.

Your second picture shows a classic Z0 resistive divider probe, best used with a true 50ohm input.

Is the circuitry of a true 50 ohm input inherently less capacitive, perhaps fewer parts, shorter traces, etc?   I am curious why the specs just say "50 ohm" not "50 ohm || 2pf"  or something like that.  It is almost as if it's implied that the capacitance is close to zero / negligible or is somehow compensated to be zero. 

Your second picture shows a classic Z0 resistive divider probe, best used with a true 50ohm input.

I ran some simulations and of course the resistor probe would have an almost "unlimited" BW without the parasitics. 
But to make it more realistic, I first added 3pf of input capacitance and that resulted in BW=2Ghz.

 
With a predicted rise time of 170ps which agrees with the empirical rule of Bandwidth x risetime = 0.35.


Here are the results after I added a more realistic tip capacitance and inductance:
 
 
The BW is about 1Ghz but the Bode plot predicts severe ringing at >500-700Mhz even though the pulse shape shows no ringing in the simulation above.

[bezzada]

that resistor has some reactance which has inverted frequency curve in order to compensate reactance of RF frontend input. So it allows better matching than using external 50 Ω pass-through dummy load.

Is that some special kind of a resistor with specific reactance baked in, and it is not parasitic?

[bezzada]

This is parasitic capacitance of RF frontend, it includes all things together, include amplifier, relay switch, PCB traces and connector wires. Note that this capacitance can vary at different frequency and at some frequency it turns into inductance...

So you are saying the 50 ohm resistor has some reactance baked in to cancel these parasitics?
Why couldn't the 1M input have some kind of a resistor like that to try and cancel the parasitic capacitance as well?
Is there anything special in the design of the 50 ohm input that let the vendors spec it as a pure resistive 50 ohm input?

[tggzzz]

that resistor has some reactance which has inverted frequency curve in order to compensate reactance of RF frontend input. So it allows better matching than using external 50 Ω pass-through dummy load.

Is that some special kind of a resistor with specific reactance baked in, and it is not parasitic?

I would like to see the reasons "radiolistener" thinks that.

Parasitics are a possibility: a rule of thumb is that 1mm of wire is ~0.8nH, and old Tektronix scope schematics show capacitors that an annotation notes are part of the circuit board. Parasitic compensation will, of course be better at some frequencies than others.

[tggzzz]

Read that old Tektronix reference about the different types of scope probe. A little research to understand what others have learned will save you from repeating old mistakes and misunderstandings. Let's make new mistakes

Others have a true 50ohm input, e.g.the Tektronix 485 which has both 1Mohm and 50ohm attenuators.

Your second picture shows a classic Z0 resistive divider probe, best used with a true 50ohm input.

Is the circuitry of a true 50 ohm input inherently less capacitive, perhaps fewer parts, shorter traces, etc?   I am curious why the specs just say "50 ohm" not "50 ohm || 2pf"  or something like that.  It is almost as if it's implied that the capacitance is close to zero / negligible or is somehow compensated to be zero. 

Nothing is perfect, but a competently implemented 50ohm attenuator will have a lot lower capacitance than 15pF.

But to make it more realistic, I first added 3pf of input capacitance and that resulted in BW=2Ghz.

Why is 3pF realistic, and more realistic than what?

In your simulations I will assume that your LTRA is 50ohm. Since a Z0 probe lead is only 1m long, I'd use a conventional lossless transmission line model. (You can't do that for the "high" impedance *10 probe, of course)

There is no reason to have the source R3 50ohm resistor in the simulation. In the special case where the source is 50ohms, you would connect it directly to the scope input. In most cases the source impedance won't be 50ohms; a typical logic gate (e.g. 74lvc1g*) has an output impedance in the 7ohms region.

C2 of 2pF is not reasonable for a Z0 probe. The 50yo HP10020 probes are "<0.7pF".
L1 of 15nH can be significantly reduced by the way the probe is connected to the UUT.

[bezzada]

Why is 3pF realistic, and more realistic than what?

More realistic than zero, which goes back to my original question of why 50 ohm inputs are spec'ed as a purely resistive "50 ohm" not "50 ohm || X pf".  I've seen 3pf on the diagrams in some Agilent videos I believe but it is an arbitrary number aka "some low parasitic capacitance of a "true" 50 ohm input". I have no idea how close it is to reality and it probably differs from scope to scope.

In your simulations I will assume that your LTRA is 50ohm. Since a Z0 probe lead is only 1m long, I'd use a conventional lossless transmission line model. (You can't do that for the "high" impedance *10 probe, of course)

Yeah, sqrt(200n/80p) = 50 ohm.  It is actually a lossless line since I let R=0. I find it more convenient to use ltline and set C L and the length rather than the propagation delay and Zo that tline requires.  It is the same thing.

There is no reason to have the source R3 50ohm resistor in the simulation. In the special case where the source is 50ohms, you would connect it directly to the scope input. In most cases the source impedance won't be 50ohms; a typical logic gate (e.g. 74lvc1g*) has an output impedance in the 7ohms region.

C2 of 2pF is not reasonable for a Z0 probe. The 50yo HP10020 probes are "<0.7pF".
L1 of 15nH can be significantly reduced by the way the probe is connected to the UUT.

Good to know, thank you.

[G0HZU]

Check the scope's VSWR specification.

What would be good?  I think Siglent is one of those that pretty much just adds a resistor.  My Siglent SDS2354X+, measured with an imperfectly calibrated NanoVNA (I couldn't find a BNC short so I approximated...) gives me 1.43 @ 500MHz an mostly below 1.2 below that.  My Tek 485 is specced for <1.25 @ 350MHz and <1.15 in the higher ranges.  Tek 2465B is <1.3 to 300MHz, <1.5 to 400MHz.  I didn't try different ranges on the Siglent.  So they seem close?

I suspect that some (low cost) modern scopes may actually use a resistive divider to make up the 50R input termination.

Most scopes have something like 20 ohms ESR deliberately added directly at the input port anyway. If this was done then the scope could exploit this to achieve a fairly low VSWR over a large bandwidth just by switching in a shunt 30R resistor after the series 20R. This would give the equivalent of a 50R termination and it would mask some of the high capacitance seen with the cheaper scopes (eg Siglent). This would give a low VSWR out through VHF and into UHF.

The penalty would be a slightly higher noise level because the (20R series + 30R shunt) divider will also act as an attenuator.

[tggzzz]

Check the scope's VSWR specification.

What would be good?  I think Siglent is one of those that pretty much just adds a resistor.  My Siglent SDS2354X+, measured with an imperfectly calibrated NanoVNA (I couldn't find a BNC short so I approximated...) gives me 1.43 @ 500MHz an mostly below 1.2 below that.  My Tek 485 is specced for <1.25 @ 350MHz and <1.15 in the higher ranges.  Tek 2465B is <1.3 to 300MHz, <1.5 to 400MHz.  I didn't try different ranges on the Siglent.  So they seem close?

"good" is an adjective. I think I'm going to start a "Stamp Out Adjectives" society, with the motto "numbers not adjectives"

More seriously, you previously measured your scopes with a NanoVNA. https://www.eevblog.com/forum/testgear/need-50-ohm-input-on-an-(old)-oscilloscope-that-has-only-1m-ohm-inputs-bnc-tee/msg4484371/#msg4484371

I added my 485+NanoVNA https://www.eevblog.com/forum/testgear/need-50-ohm-input-on-an-(old)-oscilloscope-that-has-only-1m-ohm-inputs-bnc-tee/msg4485067/#msg4485067



and you can see the S11 is in the range -35dB to -31dB, corresponding to a VSWR of 1.035 to 1.06. That's much better than the spec of 1.25!

Here's the TDR of a 485 and 2465 input. The TDR has a ~100ps risetime, i.e. better than ~10* faster than the scopes' response.
Vertical sensitivity is 50mρ/div (+100mρ=>60ohms, -100mρ=>40ohms).
Horizontal sensitivity is approx 10cm/div.
Division 1 marked ^ corresponds to the end of the cable.

485:



2465:



Interpretation:

• the 485 shows brief equal-and-opposite inductive/capacitive spikes presumably going between the BNC connector and through the RF relay selecting the two different attenuators. These spikes will probably only be visible in the frequency domain well outside the scope's bandwidth
• the 485 attenuator is a pretty clean 53ohms
• the 2565 input is much messier: ripples are all positive and last much longer, probably into the scope's bandwidth (I'm too tired to workk it out!)
• the 2465 input is a little closer to 50ohms, eventually

[radiolistener]

I am curious why the specs just say "50 ohm" not "50 ohm || 2pf"  or something like that.

There is no sense to show reactance, it's just assumed that it is small enough for good match with 50 Ω at working bandwidth. Reactance can be inductive, capacitive and can be missing at all depends on the frequency.

If you want to know reactance, it's better to show a curve - reactance vs frequency, because if you write "2pf" it doesn't represent actual value... It can be 2 pF at 1 GHz and at the same time 2 nH at 2 GHz...

[bdunham7]

There is no sense to show reactance, it's just assumed that it is small enough for good match with 50 Ω at working bandwidth. Reactance can be inductive, capacitive and can be missing at all depends on the frequency.

If you want to know reactance, it's better to show a curve - reactance vs frequency, because if you write "2pf" it doesn't represent actual value... It can be 2 pF at 1 GHz and at the same time 2 nH at 2 GHz...

I'm going to respond to tggzzz in some detail later, but looking at the Smith charts of the scope inputs, it seems pretty clear that there isn't simply a capacitance in parallel with the input.  I got the bright idea of connecting an LCR meter to the scope input in the 1M mode, and while I'm doubtful that my calibration is very good (I'm still missing one or two bits) I ended seeing 18pF with 9900R ESR.  This was at 100kHz test frequency, 10kHz resulted in an OL reading and I don't have anything over 100kHz available.  That result may not be accurate, but it is something to consider if you're trying to model the front end of these scopes as the OP is.

[G0HZU]

In terms of how it loads a circuit, a typical 100-200MHz scope input can be modelled fairly well up into VHF with the basic circuit below.
The series resistance R1 will often vary a bit depending on the range setting of the scope and also it varies from model to model. The circuit below represents typical values for the series resistance and the shunt capacitance.

You can see that the 1Meg parallel input resistance (Rp) declines rapidly to a much lower Rp above about 1MHz. This is because of the series resistance R1 at the input of the scope.

[EPAIII]

I think the best answer to this question is that you need to check the specs. of each individual scope. And none of them are perfect.

A 50 Ohm input is intended to be used with a 50 Ohm transmission line connected to it. It is not intended for use with a probe. You terminate the transmission line in the scope. And you must be careful with power levels. Depending on the power involved, an inline attenuator may be needed.

[EPAIII]

There is nothing magic about the resistors.

With the same input circuit, with and without a 50 Ohm resistor in parallel, the capacitive reactance is the same for any given frequency. The "magic" is that a capacitive reactance value of perhaps 10 MOhms has a much larger effect on a 1 MOhm input (10%) than it would have on a 50 Ohm input (0.005%). So the scope makers simply do not see any reason to worry about it with the 50 Ohm input configuration. It will be far below the specified, overall accuracy of the scope.

And likewise for any other value of capacitive reactance.

I am sure some of the purists here will argue, but that's basically it.

This is parasitic capacitance of RF frontend, it includes all things together, include amplifier, relay switch, PCB traces and connector wires. Note that this capacitance can vary at different frequency and at some frequency it turns into inductance...

So you are saying the 50 ohm resistor has some reactance baked in to cancel these parasitics?
Why couldn't the 1M input have some kind of a resistor like that to try and cancel the parasitic capacitance as well?
Is there anything special in the design of the 50 ohm input that let the vendors spec it as a pure resistive 50 ohm input?

[tggzzz]

I think the best answer to this question is that you need to check the specs. of each individual scope. And none of them are perfect.

A 50 Ohm input is intended to be used with a 50 Ohm transmission line connected to it. It is not intended for use with a probe. You terminate the transmission line in the scope. And you must be careful with power levels. Depending on the power involved, an inline attenuator may be needed.

True, but misleading

"Resistive divider" "Z0" probes have a 50ohm transmission line and must be connected directly to a 50ohm input.

The HP10020A probe has a 1.5GHz bandwidth. Ditto the Keysight 10442B probe.
The Tektronix P6150A probe has, depending on option, up to a 9GHz bandwidth.
The Tektronix P6156A probe has a 3.5GHz bandwidth.
There are others.

Since they all have low tip capacitance (some 0.15pF, some <0.7pF), they all have a higher impedance than so-called high impedance *10 probes. That's rather important when probing high frequency signals.

[tggzzz]

There is nothing magic about the resistors.

Probably.

But 1mm of wire (e.g. series PCB trace or simply the length of the resistor) has a rule-of-thumb inductance of ~0.8nH, depending on precise geometries.

With the same input circuit, with and without a 50 Ohm resistor in parallel, the capacitive reactance is the same for any given frequency. The "magic" is that a capacitive reactance value of perhaps 10 MOhms has a much larger effect on a 1 MOhm input (10%) than it would have on a 50 Ohm input (0.005%). So the scope makers simply do not see any reason to worry about it with the 50 Ohm input configuration. It will be far below the specified, overall accuracy of the scope.

Now use some realistic values for either a high impedance scope input or a "high" impedance *10 probe, and you will see your 0.005% is not a useful figure. 15pF is typical for either.

15pF at 100MHz is, of course, ~100ohms, or 22ohms at 500MHz.

High frequency probes go to considerable lengths to minimise tip capacitance. Values from 1/20 to 1/100 of 15pF are normal.

[Siwastaja]

If you want to know reactance, it's better to show a curve - reactance vs frequency, because if you write "2pf" it doesn't represent actual value... It can be 2 pF at 1 GHz and at the same time 2 nH at 2 GHz...

Get your basics straight. Reactance has unit of Ohm, not F or H. Capacitive or inductive reactance is still expressed in Ohms. Capacitance and inductance are circuit elements in equivalent circuit and are always present there, not turning on and off magically when crossing SRF. At high frequencies, parasitic capacitance does not turn into parasitic inductance, but both these parasitic elements are always there, it's just that the ratio of their impedances turns in favor of another.

[dietert1]

For a 50 Ohm input the internal line from the scopes input connector to its amplifier input can be a 50 Ohm terminated line, with the 50 Ohm resistor sitting very close to the input amplifier. The amplifier input capacitance is small, like 1..3 pF.
In high impedance mode, this isn't possible. Instead the connection effectively behaves as a capacitive load that even gets an additional capacitor to make a high bandwidth compensated divider with an adjustable capacitor in the scope probe.
A 1 MOhm || 20 pF input with direct connection (no divider) is the worst of all. How bad it is depends on the type of cable. In my opinion those probe cables with a 1:1 - 1:10 switch are a trap. Scope probe cables are very, very special with center conductors the thickness of a hair. Still their characteristic impedance cannot reach more than some hundred ohms.

Regards, Dieter