What is the derating of a 1Mohm oscilloscope input?

[ballsystemlord]

Hello,
I have an MSO5000 scope and I wanted to measure op-amp slew-rate and settling time. It occurred to me that I might consider a direct connection to the scope, but I can't find a derating curve in the datasheet or user manual telling me at X frequency I can put Y amount of voltage into the inputs.

It's a 1Mohm input w/17pF capacitance +/-3pF, but if you measure it with an LCR meter (East tester ET3501), you get somewhat different results than just a simple 1Mohm. At 100Khz You get 9Kohms of series resistance, 80Mohms of parallel resistance, and -1.7pF series and parallel input capacitance. At 1Khz you get 6Mohms of series resistance, 900Mohms of parallel resistance, and -2pF of series and parallel capacitance. It should be noted that if you reverse the VCC and GND leads of the LCR meter, you get very different results.

What is the derating of a 1Mohm oscilloscope input?

Thanks!

[Sensorcat]

I have an MSO5000 scope and I wanted to measure op-amp slew-rate and settling time. It occurred to me that I might consider a direct connection to the scope, but I can't find a derating curve in the datasheet or user manual telling me at X frequency I can put Y amount of voltage into the inputs.

In general, is is preferable to connect the scope to the DUT with the probes, especially because the MSO5000 has no 50Ohm inputs and slew-rate or settling time measurement may involve fast signals. For fast measurements, you also need the 10x probe setting to get the full bandwidth, which is not available at 1x (only about 13MHz). Apart from that, 10x means less capacitive loading of your circuit. The capacitive load can make your circuit behave much different.

Not to specify the input derating was an omission from Rigol, but it shouldn't matter in your case, unless you have an extreme and rare high-voltage op-amp.

It's a 1Mohm input w/17pF capacitance +/-3pF, but if you measure it with an LCR meter (East tester ET3501), you get somewhat different results than just a simple 1Mohm. At 100Khz You get 9Kohms of series resistance, 80Mohms of parallel resistance, and -1.7pF series and parallel input capacitance. At 1Khz you get 6Mohms of series resistance, 900Mohms of parallel resistance, and -2pF of series and parallel capacitance. It should be noted that if you reverse the VCC and GND leads of the LCR meter, you get very different results.

Try to measure the input resistance with your DMM first. The scope must be turned on for this. Should be almost exact 1MOhm in both polarities.

The impedance measurement requires a model: For a scope input, the appropriate model is resistor in parallel to capacitor, usually shown as Rp and Cp on LCR meters. If you select a different model on the meter, it won't match the input. Also, the meter must be calibrated open and close prior to the measurement, with the same wires in the same arrangement as in the measurement. There should be no negative values if you do this! Some frequency dependence has to be expected, because the resistor is more difficult to measure at higher frequencies (gets less current compared to the capacitor), but it should be possible to measure stable 17pF at 1kHz to 100kHz. I do not observe a dependency to polarity, but there could be dependency to stimulus voltage level of the impedance meter.

[ballsystemlord]

In general, is is preferable to connect the scope to the DUT with the probes, especially because the MSO5000 has no 50Ohm inputs and slew-rate or settling time measurement may involve fast signals. For fast measurements, you also need the 10x probe setting to get the full bandwidth, which is not available at 1x (only about 13MHz).

Wait, I thought that 13MHz limit was for the probes at 1x. The oscilloscope, because the BW was measured as a full system BW, could probably go beyond it's normal scope+probe BW if you connect it directly to the DUT. Or did I miss something?

It's a 1Mohm input w/17pF capacitance +/-3pF, but if you measure it with an LCR meter (East tester ET3501), you get somewhat different results than just a simple 1Mohm. At 100Khz You get 9Kohms of series resistance, 80Mohms of parallel resistance, and -1.7pF series and parallel input capacitance. At 1Khz you get 6Mohms of series resistance, 900Mohms of parallel resistance, and -2pF of series and parallel capacitance. It should be noted that if you reverse the VCC and GND leads of the LCR meter, you get very different results.

Try to measure the input resistance with your DMM first. The scope must be turned on for this. Should be almost exact 1MOhm in both polarities.

I'll have to get back to you on that one.

The impedance measurement requires a model: For a scope input, the appropriate model is resistor in parallel to capacitor, usually shown as Rp and Cp on LCR meters. If you select a different model on the meter, it won't match the input.

I tried both CP-RP and CS-RS with the LCR meter connected to the oscilloscope. Grated, I used the alligator leads, so the reading might not be 100% perfect. The values I got are stated above. The test voltage was 1v RMS without any DC bias applied. I cannot say that the LCR meter is calibrated, only that almost every 1% part I apply to it it reads within that 1% tolerance at the specified frequency. E.G. C0G parts take 1Mhz, although I can only do 100khz with my LCR meter. 1% resistors from questionable sources are where it says things are not within tolerance. But even then I think if I connected a bit closed/farther out on the leads I'd get that 1% resistor to be within spec.

Also, the meter must be calibrated open and close prior to the measurement, with the same wires in the same arrangement as in the measurement.

The alligator leads were attached and the meter was calibrated open/closed prior to taking a reading. I didn't let the oscilloscope and LCR meter warm up for 30min prior to taking the reading.

There should be no negative values if you do this!

Negative readings indicate an inductive, not capacitive, value. I found this out through trial and error.

Some frequency dependence has to be expected, because the resistor is more difficult to measure at higher frequencies (gets less current compared to the capacitor), but it should be possible to measure stable 17pF at 1kHz to 100kHz. I do not observe a dependency to polarity, but there could be dependency to stimulus voltage level of the impedance meter.

I'm happy to try whatever you suggest to get accurate readings. The only thing I can think up that might be causing error is that the oscilloscope and LCR meter both use active components. Is it possible that they're interacting in an unpredictable way?

Thanks!

[_Wim_]

Wait, I thought that 13MHz limit was for the probes at 1x. The oscilloscope, because the BW was measured as a full system BW, could probably go beyond it's normal scope+probe BW if you connect it directly to the DUT. Or did I miss something?

Yes, you are correct, that is only applicable with a 1x probe (is related due to the "special" coax used in the probe with high resistance). Dave has a good explanation here:

I'm happy to try whatever you suggest to get accurate readings. The only thing I can think up that might be causing error is that the oscilloscope and LCR meter both use active components. Is it possible that they're interacting in an unpredictable way?

An auto-balancing bridge LCR (like your east tester) is not well suited for measuring grounded devices.
See page 36 of this Agilent document: https://wiki.epfl.ch/carplat/documents/rcl_agilent.pdf

[David Hess]

The oscilloscope input is modeled as a 1 megohm shunt resistance in parallel with the input capacitance of about 20 picofarads, however there is a lot more going on.  The input protection requires a high impedance series element between the 1 megohm shunt resistance and the high impedance buffer.  Typically this is a bypassed high resistance, so at low frequencies this series resistance hides a major part of the input capacitance, and at high frequencies, all of the input capacitance is exposed.  Between 470k and 1 megohm, and 0.001 microfarads and 0.0056 microfarads is typical for these parts.  Some examples shown below.

The bandwidth of a x1 probe assumes a 25 ohm source impedance is driving the tip.  If the driving impedance is lower, then the bandwidth will be higher.

Driving the oscilloscope input directly with a short cable or no cable works where lower noise, higher bandwidth, and higher sensitivity are required, as shown below.

[ballsystemlord]

Which brings us full circle back to the original question: What is the derating of a 1Mohm oscilloscope input?

I'd be applying no more than 36v p-to-p.

Maybe if you don't know the answer you have experience with similar situations and so have a reasonable safe value you can inject into an oscilloscope's 1Mohm input?*

Thanks

[David Hess]

You can calculate it based on that parallel RC element in series with the protection network.  I have done this before and gotten results pretty close to the safe overload specifications.  The protection diodes can handle up to a given amount of average current which defines the derating versus frequency.

In practice the input will withstand way more than 36 volts peak-to-peak at *any* input attenuator setting. 

[ballsystemlord]

Try to measure the input resistance with your DMM first. The scope must be turned on for this. Should be almost exact 1MOhm in both polarities.

As promised, I've used my DMMs to try to examine the oscilloscope. Note: It's in "positive to positive" / "positive to negative" format.

DMM	Calibrated	Mohms	pF*	AC coupled pF*
GW Instek GDM-9061	Yes	1.000634-6 / same	224 / 476	19 / 179
FLIR DM93 (original)	Yes	0.9606-1.0421 / same	Cannot stabilize.	34310 / 35000
TA801C (Chinese)	Are you insane?	0.996-9 / same	46-41 / same	51 / 51

* The GW Instek 9061 and the TA801C measure capacitors using a saw tooth wave. I'm uncertain what exactly the DM93 is doing to measure capacitance. The Instek is a benchtop DMM, the others are hand held.

It's kinda funny to me that the cheap Chinese DMM got the closest to what the value of the scope's input capacitance should be.

[ballsystemlord]

You can calculate it based on that parallel RC element in series with the protection network.  I have done this before and gotten results pretty close to the safe overload specifications.  The protection diodes can handle up to a given amount of average current which defines the derating versus frequency.

In practice the input will withstand way more than 36 volts peak-to-peak at *any* input attenuator setting.

I know how to do RC power calculations, but how do I know how much power the scope's inputs and protection network can handle?
Thanks!

[Someone]

You can calculate it based on that parallel RC element in series with the protection network.  I have done this before and gotten results pretty close to the safe overload specifications.  The protection diodes can handle up to a given amount of average current which defines the derating versus frequency.

In practice the input will withstand way more than 36 volts peak-to-peak at *any* input attenuator setting.

Having looked rather carefully at such matters I'd want to see the schematic (and part sizes) of the specific front end before waving away "any input attenuator setting". For the older 1054 it was an attenuator part that was the first limiting factor in frequency derating:
https://www.eevblog.com/forum/testgear/rigol-ds1202z-e-input-voltage-vs-frequency-derating/

[ballsystemlord]

You can calculate it based on that parallel RC element in series with the protection network.  I have done this before and gotten results pretty close to the safe overload specifications.  The protection diodes can handle up to a given amount of average current which defines the derating versus frequency.

In practice the input will withstand way more than 36 volts peak-to-peak at *any* input attenuator setting.

Having looked rather carefully at such matters I'd want to see the schematic (and part sizes) of the specific front end before waving away "any input attenuator setting". For the older 1054 it was an attenuator part that was the first limiting factor in frequency derating:
https://www.eevblog.com/forum/testgear/rigol-ds1202z-e-input-voltage-vs-frequency-derating/

Dave has already done a teardown. Here is the album and I think I found the frontend picture for you.
https://www.flickr.com/photos/eevblog/albums/72157701904033821/with/45805998381
https://www.flickr.com/photos/eevblog/45805998381/in/album-72157701904033821

Thanks

[Someone]

You can calculate it based on that parallel RC element in series with the protection network.  I have done this before and gotten results pretty close to the safe overload specifications.  The protection diodes can handle up to a given amount of average current which defines the derating versus frequency.

In practice the input will withstand way more than 36 volts peak-to-peak at *any* input attenuator setting.

Having looked rather carefully at such matters I'd want to see the schematic (and part sizes) of the specific front end before waving away "any input attenuator setting". For the older 1054 it was an attenuator part that was the first limiting factor in frequency derating:
https://www.eevblog.com/forum/testgear/rigol-ds1202z-e-input-voltage-vs-frequency-derating/

Dave has already done a teardown. Here is the album and I think I found the frontend picture for you.
https://www.flickr.com/photos/eevblog/albums/72157701904033821/with/45805998381
https://www.flickr.com/photos/eevblog/45805998381/in/album-72157701904033821

Why for me?

[das_strobel]

Which brings us full circle back to the original question: What is the derating of a 1Mohm oscilloscope input?

I'd be applying no more than 36v p-to-p.

Maybe if you don't know the answer you have experience with similar situations and so have a reasonable safe value you can inject into an oscilloscope's 1Mohm input?*

Thanks

To answer your questoin you should look at the datasheer of the MS5000. On page 18 it states it.

[Fungus]

What do you hope to gain? The measurement bandwidth of the oscilloscope is maximized by a x10 probe, even though it causes other problems. A direct connection may also load down your op-amp and alter the reading.

What signal will you be applying? The derating curve of an oscilloscope mostly applies to continuous sine wave inputs, not sparse pulses.

The best solution is probably active probes but even without investing in those you could look into improved probing techniques, eg. Get rid of the ground clip and use probe BNC adapters.

[ballsystemlord]

You can calculate it based on that parallel RC element in series with the protection network.  I have done this before and gotten results pretty close to the safe overload specifications.  The protection diodes can handle up to a given amount of average current which defines the derating versus frequency.

In practice the input will withstand way more than 36 volts peak-to-peak at *any* input attenuator setting.

Having looked rather carefully at such matters I'd want to see the schematic (and part sizes) of the specific front end before waving away "any input attenuator setting". For the older 1054 it was an attenuator part that was the first limiting factor in frequency derating:
https://www.eevblog.com/forum/testgear/rigol-ds1202z-e-input-voltage-vs-frequency-derating/

Dave has already done a teardown. Here is the album and I think I found the frontend picture for you.
https://www.flickr.com/photos/eevblog/albums/72157701904033821/with/45805998381
https://www.flickr.com/photos/eevblog/45805998381/in/album-72157701904033821

Why for me?

Because you literally just said you wanted to see, "the specific front end..." so now you can see it.

[ballsystemlord]

Which brings us full circle back to the original question: What is the derating of a 1Mohm oscilloscope input?

I'd be applying no more than 36v p-to-p.

Maybe if you don't know the answer you have experience with similar situations and so have a reasonable safe value you can inject into an oscilloscope's 1Mohm input?*

Thanks

To answer your questoin you should look at the datasheer of the MS5000. On page 18 it states it.

(Attachment Link)

Erm, that's for a transient condition, right? For measuring settling time my op-amps would be injecting a continuous input power.

[G0HZU]

If you allow continuous waveforms (like a sine wave) up at HF through VHF frequencies, the 1 Meg input of a modern DSO can typically be derated to something like 10 Vrms up at 30 MHz or so.

Go even higher in frequency and the derating of the 1Meg input can dip to 5 Vrms (or less) in some cases. There are often some small and vulnerable components at the front end of a typical modern DSO and RF can easily cause damage to them.

Often the voltage derating begins within the frequency range of 10 kHz to 100 kHz and it derates at about x10 per decade of frequency down to something like 10 Vrms. The manufacturer should ideally provide you with this info but sometimes they don't include this on the datasheet.

[das_strobel]

Which brings us full circle back to the original question: What is the derating of a 1Mohm oscilloscope input?

I'd be applying no more than 36v p-to-p.

Maybe if you don't know the answer you have experience with similar situations and so have a reasonable safe value you can inject into an oscilloscope's 1Mohm input?*

Thanks

To answer your questoin you should look at the datasheer of the MS5000. On page 18 it states it.

(Attachment Link)

Erm, that's for a transient condition, right? For measuring settling time my op-amps would be injecting a continuous input power.

As I read it, transients are rated like that: "Transient Overvoltage 1600 Vpk"

Following what this datasheet says you can put 300Vrms/400Vpk on the analog inputs without damaging them. Should be ok for your use case  as far as I can see.

[David Hess]

I know how to do RC power calculations, but how do I know how much power the scope's inputs and protection network can handle?
Thanks!

Here is how I did it.

I assumed that the input protection diodes could handle 20 milliamps average forward current, although I could look up the actual values in my case.  These diodes are low leakage and have low capacitance, so this is a reasonable number.

Then the series RC network made up of the 470 kilohm resistor in parallel with 1000 picofarads limits the current into the diodes.

For DC, 400 volts across 470 kilohms is about 1 milliamp, so well below the 20 millamp rating of the diodes.  Other considerations like high voltage breakdown now limit the maximum DC voltage.

For AC, 20 volts/microsecond through 1000 picofarads produces 20 milliamps, but the diode rating is for *average* current, so in pulse applications the slew rate may be much higher until the point where the diodes overheat.

Erm, that's for a transient condition, right? For measuring settling time my op-amps would be injecting a continuous input power.

Unless it is an operational amplifier producing more than 400 volts, it will never be fast enough to cause a problem.

Settling time measurements are usually made with a false summing node for maximum sensitivity.  Oscilloscopes are not good at measuring settling time directly because of limited front end performance and susceptibility to overload, and higher resolution oscilloscopes do not solve these problems.  There is an exception here for sampling oscilloscopes which are excellent at measuring settling time, but they are outside this discussion.

[ballsystemlord]

What do you hope to gain? The measurement bandwidth of the oscilloscope is maximized by a x10 probe, even though it causes other problems. A direct connection may also load down your op-amp and alter the reading.

The 10x probes I have are the default PVP2350 variety. They have an input capacitance of 50pF +/-20pF in 1x mode and 10pF +/-5pF in 10x mode plus the compensation capacitance which is 10-25pF. So I think I'd be loading the op-amps down regardless.

What signal will you be applying? The derating curve of an oscilloscope mostly applies to continuous sine wave inputs, not sparse pulses.

I'm imagining applying a pulsed DC input (from my scope's AWG?), to the op-amp and then feeding the op-amp's output into the BNC input of the oscilloscope. The resultant waveform will show me the slew rate.
Then I'd feed the pulsed DC, input into one of the logic probes (in addition to the op-amp), and have the scope trigger off of the digital channel with the analog channel in AC coupled mode and the scope set to high-res mode. This would reveal the settling time.
UPDATE: As pointed out, it seems my settling time test idea won't work. I'll have to modify my plans accordingly.

So what I'd get is a very sudden high frequency input to the oscilloscope with a lessening amount of HV power input to it over time.

The best solution is probably active probes but even without investing in those you could look into improved probing techniques, eg. Get rid of the ground clip and use probe BNC adapters.

Umm, I'm connecting to an op-amp, which doesn't have a BNC attached to it -- yet. I normally use the ground spring, not the clip anyway.

I hope that clarifies my intended test setup.

Thanks!

[David Hess]

UPDATE: As pointed out, it seems my settling time test idea won't work. I'll have to modify my plans accordingly.

Measuring operational amplifier settling time is a well explored topic.  Two applications notes from Jim Williams are attached discussing it.  It is not as straightforward as attaching the output to an oscilloscope, but it is not difficult either with some simple circuits.

Nothing you do with an operational amplifier is going to damage an oscilloscope input whether connected directly or through a probe.

For a moderate performance settling time test circuit, I would use a couple of transistors and a reference to make an accurate pulse, and then buffer the false summing point to drive the oscilloscope.  The problem with this by itself is that the oscilloscope will suffer from overload as discussed in the application notes.

[Fungus]

Following what this datasheet says you can put 300Vrms/400Vpk on the analog inputs without damaging them.

That's correct, but it's unrelated to this question. High frequency inputs can burn things up even when they're only a few volts.

Watch this video at about 13 minutes. Joe melts a multimeter just by raising the input frequency:

[Conrad Hoffman]

FWIW, I was looking at the inexpensive Tektronix TBS1000C series and they say the following:

Input impedance
1 MΩ ±2 % in parallel with 14 pF ±2 pF

Maximum Input Voltage
300 VRMS, Installation Category II; derate above 4 MHz at 20 dB per decade to 200 MHz

[G0HZU]

Here's the spec for one of Rigol's newer scopes, the HDO1000 series:

 Maximum Input Voltage
CAT I 300 Vrms, 400 Vpk (DC + Vpeak)
 
Frequency derating (assumes sine wave input):
400 Vpk until 40 kHz. Then derates at 20 dB/dec until 6 Vpk.

 Remarks
 No transient overvoltage allowed whether the probe
 is used or not.

 Use this instrument only for measurements within its
 specified measurement category (not rated for CAT II,
 III, IV).
 

[ballsystemlord]

After researching the subject, I've determined that measuring settling time of an op-amp accurately would take several weeks of work (from a student like me), to get a circuit with repeatable characteristics.
I reasoned that because op-amps were used in so many applications, and that many of those applications, by virtue of using op-amps, measured their settling time by way of circuit behavior, that doing the measurements with an oscilloscope would be easy. But it's anything but easy!

So now I get to look into what it would take to measure slew rate accurately...