Best method "Measuring Voltage Ripple"

[Tomsik]

Hello, I'm looking for the best method for measuring voltage ripple.
I've read document from Analog Devices AN-1144.
I intend to use 50 ohm terminated line of oscilloscope, but my oscilloscope doesn't have it. Therefore I design dc blocker and 50 ohm terminator.

And when I use my design for two channels of oscilloscope by Dave's recommendations https://www.youtube.com/watch?v=Edel3eduRj4&t=1951s, I get best results.
Will it work?

Next question is:
In the AN-1144 document is described this method by adding external blocking capacitor and using an internal 50 ohm terminating in oscilloscope. Why?
Every oscilloscope that it has 50 ohm terminator also has blocking capacitor inside, therefore it just AC mode, isn't it?.

Why do they suggest add next external blocking capacitor?




Thank you

[bob91343]

I don't think a 50 Ohm termination is proper.  I think high impedance is what you want.  If you still insist on 50 Ohms you certainly need a blocking capacitor or you will damage the terminator in the oscilloscope.

Ripple, especially if small, isn't measured casually.  You will need to be precise in your connection of the oscilloscope, especially the ground lead.

[tautech]

For scope 50 ohm inputs there is a typical 5V max spec, exceed it and you damage the input !

[thm_w]

I don't think a 50 Ohm termination is proper.  I think high impedance is what you want.  If you still insist on 50 Ohms you certainly need a blocking capacitor or you will damage the terminator in the oscilloscope.

Ripple, especially if small, isn't measured casually.  You will need to be precise in your connection of the oscilloscope, especially the ground lead.

High impedance won't work unless you are using a scope probe instead of the 50R cable, see Daves recent video on measuring the RD6006.
1x scope probe works, but that limits bandwidth. 10x probe works, but then noise level is increased.


edit: to actually address OPs question

Next question is:
In the AN-1144 document is described this method by adding external blocking capacitor and using an internal 50 ohm terminating in oscilloscope. Why?
Every oscilloscope that it has 50 ohm terminator also has blocking capacitor inside, therefore it just AC mode, isn't it?.

Why do they suggest add next external blocking capacitor?

When you use AC coupling + 50R, some oscilloscopes will have a very high low frequency cutoff. For the TDS744A for example, its telling me 200kHz.
With the external AC coupling capacitor, you can make it much larger. I assume that is the reason in AN1144.

Your idea will work but the bandwidth will depend on how well it is implemented.

[DimitriP]

Show of hands, who read AN1144  before responding ?   

[DimitriP]

Every oscilloscope that it has 50 ohm terminator also has blocking capacitor inside, therefore it just AC mode, isn't it?.

I don't know about "every" and neither do they.
They are also including Spectrum Analyzers not just oscilloscopes.
When the difference between "hoping" DC will get blocked and "knowing" you are blocking it , is one capacitor, use the cap! 

[Kosmic]

Every oscilloscope that it has 50 ohm terminator also has blocking capacitor inside, therefore it just AC mode, isn't it?.

I don't know about "every" and neither do they.
They are also including Spectrum Analyzers not just oscilloscopes.
When the difference between "hoping" DC will get blocked and "knowing" you are blocking it , is one capacitor, use the cap! 

My Lecroy oscilloscope has 50Ohm termination but only DC (no cap). 1Mohm termination has AC/DC.

[srb1954]

Generally 50 inputs on scopes don't have a DC blocking capacitor and are DC coupled, which usually limits the maximum DC input voltage to around 5V to avoid excessive power dissipation in the 50 termination resistor. If you did have a DC blocking capacitor in the scope you would need a really large one (some 10's of uF) to get any sort of useful low frequency (<1kHz) response . Such large capacitors have a lot of undesirable parasitic elements eg. inductance that are not conducive to accurate measurements at higher frequencies where the 50 input termination is normally used.

Measuring PSU noise with a scope using a quality 10M probe would be usually sufficient unless you are interested in measuring noise components >100MHz where the 50 termination will give better signal fidelity and more accurate measurements. With a 10M probe selecting AC coupling on the scope input has much less effect on the low frequency response and you can still get good measurements down to low frequencies (<10Hz). However, obtaining good signal fidelity and accurate noise measurements with a 10M probe requires careful attention to the grounding of the probe: use the shortest possible ground lead on the probe and connect it as close as possible to the circuit point where the signal is being measured.

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes. This is why the connection diagram shows the blocking capacitor as this is essential for the spectrum analyser. You could go even further and add a power limiter to spec analyser input to protect against the voltage spike that occurs when the DC blocking capacitor charges up. Such a power limiter would be recommended if the PSU under test has an output >10V. Otherwise set the input attenuation on the spec analyser to its maximum level to protect the mixer diodes.

[Kosmic]

Generally 50 inputs on scopes don't have a DC blocking capacitor and are DC coupled, which usually limits the maximum DC input voltage to around 5V to avoid excessive power dissipation in the 50 termination resistor. If you did have a DC blocking capacitor in the scope you would need a really large one (some 10's of uF) to get any sort of useful low frequency (<1kHz) response . Such large capacitors have a lot of undesirable parasitic elements eg. inductance that are not conducive to accurate measurements at higher frequencies where the 50 input termination is normally used.

Measuring PSU noise with a scope using a quality 10M probe would be usually sufficient unless you are interested in measuring noise components >100MHz where the 50 termination will give better signal fidelity and more accurate measurements. With a 10M probe selecting AC coupling on the scope input has much less effect on the low frequency response and you can still get good measurements down to low frequencies (<10Hz). However, obtaining good signal fidelity and accurate noise measurements with a 10M probe requires careful attention to the grounding of the probe: use the shortest possible ground lead on the probe and connect it as close as possible to the circuit point where the signal is being measured.

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes. This is why the connection diagram shows the blocking capacitor as this is essential for the spectrum analyser. You could go even further and add a power limiter to spec analyser input to protect against the voltage spike that occurs when the DC blocking capacitor charges up. Such a power limiter would be recommended if the PSU under test has an output >10V. Otherwise set the input attenuation on the spec analyser to its maximum level to protect the mixer diodes.

Good explanation. 

Some spectrum analysers have internal dc blocking cap and can tolerate DC voltage on the input. But normally their range is limited. Ex, the Agilent E4406A I have here can tolerate up to 26VDC.

[tautech]

Measuring PSU noise with a scope using a quality 10M probe would be usually sufficient unless you are interested in measuring noise components >100MHz where the 50 termination will give better signal fidelity and more accurate measurements.

It's commonly overlooked that a PSU rail is a low impedance source and as such ripple is adequately measured with 1x probe which in turn displays any ripple better as it's un-attenuated.

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes.

This may be the case with some SA's or older boat anchor models but it is not the situation with many recent offerings to the marketplace as they often have 50V max input ratings due to DC blocking caps on their inputs which also serves to provide some quite reasonable electrostatic protection.

[bob91343]

  Remember that the oscilloscope measures the difference between the probe tip and the ground lead.  So if you want a true picture you need to locate the ground lead to a reference at the DUT with a short ground lead.  Preferably the ground and probe tip should be very close to one another at the measured device.  Even better is one of those probe tips with the two pins to connect to both sides of the source.

[Kosmic]

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes.

This may be the case with some SA's or older boat anchor models but it is not the situation with many recent offerings to the marketplace as they often have 50V max input ratings due to DC blocking caps on their inputs which also serves to provide some quite reasonable electrostatic protection.

I think as soon as the frequency range is too large, they just drop the cap. Implementing DC protection from 5Khz to 26Ghz would be tedious.

For example the Keysight N9040B 3Hz - 50Ghz

[tautech]

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes.

This may be the case with some SA's or older boat anchor models but it is not the situation with many recent offerings to the marketplace as they often have 50V max input ratings due to DC blocking caps on their inputs which also serves to provide some quite reasonable electrostatic protection.

I think as soon as the frequency range is too large, they just drop the cap. Implementing DC protection from 5Khz to 26Ghz would be tedious.

For example the Keysight N9040B 3Hz - 50Ghz

Yes well few of us have even seen a 50 GHz SA let alone it's inputs. 

This is a typical input setup of one of Siglent's SA products in the 1.5 - 7.5 GHz range.

[srb1954]

Measuring PSU noise with a scope using a quality 10M probe would be usually sufficient unless you are interested in measuring noise components >100MHz where the 50 termination will give better signal fidelity and more accurate measurements.

It's commonly overlooked that a PSU rail is a low impedance source and as such ripple is adequately measured with 1x probe which in turn displays any ripple better as it's un-attenuated. 

A x1 probe can be used but you need to be aware of the limited bandwidth (typically <20MHz) of x1 probes. Dave Jones did a video on this subject. A x1 probe would be perfectly adequate for measuring the low-frequency ripple on a linear supply but with a switching supply you would potentially not see the higher frequency transients due to the limited bandwidth of a x1 probe. The reason for using a X10 probe is not for minimising loading effects but to preserve the maximum bandwith from probe tip through to the scope.

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes.

This may be the case with some SA's or older boat anchor models but it is not the situation with many recent offerings to the marketplace as they often have 50V max input ratings due to DC blocking caps on their inputs which also serves to provide some quite reasonable electrostatic protection.

A DC blocking capacitor by itself will not provide any protection against electrostatic discharge. The fast rise time of these transients will be coupled straight through the capacitor. To achieve some ESD protection you need some fairly severe low-pass filtering to eliminate the fast transients (not a useful technique on a SA!) or some transient suppressor diodes or diode clamping circuitry to bypass the transient away from sensitive inputs.

[Tomsik]

Your idea will work but the bandwidth will depend on how well it is implemented.

Yeah, it depends on performance. I've made a simulation with LTspice:
It looks that it will work from 1kHz - it's great for me


PCB design is here:

[Kosmic]

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes.

This may be the case with some SA's or older boat anchor models but it is not the situation with many recent offerings to the marketplace as they often have 50V max input ratings due to DC blocking caps on their inputs which also serves to provide some quite reasonable electrostatic protection.

I think as soon as the frequency range is too large, they just drop the cap. Implementing DC protection from 5Khz to 26Ghz would be tedious.
...

Yes well few of us have even seen a 50 GHz SA let alone it's inputs. 

That's an example of a modern instrument. Just took the first wide frequency range spectrum analyzer I found on keysight website.

This is a typical input setup of one of Siglent's SA products in the 1.5 - 7.5 GHz range.

This is the limited range I was talking about. Anything higher than 6-7 Ghz will probably not have any internal dc block.

[Wuerstchenhund]

The situation with spectrum analysers is different; these almost always have a fixed 50 input with no internal blocking capacitor. The input to a spectrum analyzer can't tolerate any DC  voltages as even a small DC voltage (<1V) can destroy the input mixer diodes.

This may be the case with some SA's or older boat anchor models but it is not the situation with many recent offerings to the marketplace as they often have 50V max input ratings due to DC blocking caps on their inputs which also serves to provide some quite reasonable electrostatic protection.

This has nothing to do with the age of the SA but with frequency. In general, integrated DC blocks can only be found in SAs with very low frequency ranges, i.e. up to around 8GHz. Pretty much anything above that is DC sensitive.

A DC blocking capacitor by itself will not provide any protection against electrostatic discharge. The fast rise time of these transients will be coupled straight through the capacitor. To achieve some ESD protection you need some fairly severe low-pass filtering to eliminate the fast transients (not a useful technique on a SA!) or some transient suppressor diodes or diode clamping circuitry to bypass the transient away from sensitive inputs.

This is correct. The purpose of a DC block is not ESD protection, it's purpose is the removal of parasitic or unwanted DC compnents which mostl likely would damage the input mixer, for example the DC supply of an active antenna or amplifier. Which is common in the telecoms field, which is why many SAs aimed at that market come with a built-in DC block.

[little.tesla]

For reference, the blocking capacitor shown in the AN-1144 is a Hewlett & Packard 10240B. Google should reveal a datasheet for it. The internal capacitor is a 0.18uF with a voltage rating of +/- 200V.

[Tomsik]

For reference, the blocking capacitor shown in the AN-1144 is a Hewlett & Packard 10240B. Google should reveal a datasheet for it. The internal capacitor is a 0.18uF with a voltage rating of +/- 200V.

I've compared 0.18uF vs my 10uF. In the case with 0.18uF is cut-off frequency bigger than with 10uF capacitor.
In standard -3dB it is:
0.18uF - ~17kHz
10uF - ~320Hz

The mentioned 0.18uF capacitor is good for me if I use 1Meg terminating of an oscilloscope, but it is useless.
The real cutoff frequency will be other, it depends on performance and used components.

[vk6zgo]

Show of hands, who read AN1144  before responding ?   

Link?

[vk6zgo]

Hello, I'm looking for the best method for measuring voltage ripple.
I've read document from Analog Devices AN-1144.
I intend to use 50 ohm terminated line of oscilloscope, but my oscilloscope doesn't have it. Therefore I design dc blocker and 50 ohm terminator.
(Attachment Link)
And when I use my design for two channels of oscilloscope by Dave's recommendations https://www.youtube.com/watch?v=Edel3eduRj4&t=1951s, I get best results.
Will it work?

Next question is:
In the AN-1144 document is described this method by adding external blocking capacitor and using an internal 50 ohm terminating in oscilloscope. Why?
Every oscilloscope that it has 50 ohm terminator also has blocking capacitor inside, therefore it just AC mode, isn't it?.

Why do they suggest add next external blocking capacitor?

(Attachment Link)


Thank you

Some oscilloscopes put the termination before the ac coupling capacitor.
Of course, by definition, an external termination will always be before the ac coupling cap.
To the DC, it is just a 50 ohm resistor, & depending upon the DC voltage setting of the supply, without an external blocking capacitor, you may overheat your termination badly.

[David Hess]

In the AN-1144 document is described this method by adding external blocking capacitor and using an internal 50 ohm terminating in oscilloscope. Why?
Every oscilloscope that it has 50 ohm terminator also has blocking capacitor inside, therefore it just AC mode, isn't it?.

Why do they suggest add next external blocking capacitor?

As pointed out, an external blocking capacitor is required if you want a lower cutoff frequency.

However not all oscilloscopes or other instruments place the AC coupling capacitor before the 50 ohm termination.  Some sources require a DC path to ground so the AC coupling capacitor is placed *after* the 50 ohm termination.  Examples of this include tunnel diode pulsers used for oscilloscope calibration and time domain reflectometry and some TTL outputs intended to drive 50 ohms.

[srb1954]

You should be aware that a simulation in SPICE does not always accurately represent the actual results that will obtained in real world measurements. This is because the component models often don't exactly simulate some of the real behaviour of physical components.

From your simulation graph I gather that you have used a 10uF coupling capacitor and from the PCB design it appears to be a ceramic cap in a 2220 package. If this cap is rated at 50VDC then it would have to be an X7R dielectric or similar to fit in this package size.

X7R dielectric is great for bypassing applications but is less suitable for signal coupling applications such as this. The main problem is that the capacitance of X7R types varies quite considerably when DC voltage is applied across them; at their full rated voltage they typically lose 50% of their capacitance. This has the effect, as the DC voltage across capacitor increases, of increasing the LF cut-off frequency of your coupling circuit.

Another effect of the voltage dependence of X7R is that it introduces harmonic distortion to a signal, particularly at lower frequencies where a greater proportion of the signal voltage appears across the capacitor.

Ideally, you would use a ceramic capacitor with a C0G dielectric in this application as this doesn't exhibit the voltage dependency of X7R and also has better temperature stability and lower losses. However,  a 10uF 50V C0G ceramic capacitor will be quite large and eye-wateringly expensive.

[trobbins]

Tomsik, can you describe the converter you are probing, and its likely ripple frequency/waveform and magnitude?

Do you have a defined part such as an SMT cap or equipment/pcb terminals/link pads that you want to use to define the measurement location - imho that is preferably across a bypass cap and can be important as measuring in other locations can give different levels so best to identify a benchmark.

Have you at least tried a simple 10:1 probe with the slip-on ground tip applied to the probe tip  (rather than a ground wire and crocodile clip), as shown in Fig 10 of that link - that is likely to meet most needs for pcb based equipment.

[Tomsik]

You should be aware that a simulation in SPICE does not always accurately represent the actual results that will obtained in real world measurements. This is because the component models often don't exactly simulate some of the real behaviour of physical components.

From your simulation graph I gather that you have used a 10uF coupling capacitor and from the PCB design it appears to be a ceramic cap in a 2220 package. If this cap is rated at 50VDC then it would have to be an X7R dielectric or similar to fit in this package size.

X7R dielectric is great for bypassing applications but is less suitable for signal coupling applications such as this. The main problem is that the capacitance of X7R types varies quite considerably when DC voltage is applied across them; at their full rated voltage they typically lose 50% of their capacitance. This has the effect, as the DC voltage across capacitor increases, of increasing the LF cut-off frequency of your coupling circuit.

Another effect of the voltage dependence of X7R is that it introduces harmonic distortion to a signal, particularly at lower frequencies where a greater proportion of the signal voltage appears across the capacitor.

Ideally, you would use a ceramic capacitor with a C0G dielectric in this application as this doesn't exhibit the voltage dependency of X7R and also has better temperature stability and lower losses. However,  a 10uF 50V C0G ceramic capacitor will be quite large and eye-wateringly expensive.

Yeah, you are right. But choosing a proper capacitor is difficult. I think it is a good to measure from 1kHz (not from 17kHZ), therefore I need a bigger capacitor.
I see C0G is a best ceramic capacitor but it has just hundreds of nF.
I've chosen 10uF and X7R on 50V(Kemet C1210C106K5RACTU) because I want to measure my power supply designs with a maximum voltage around 5V(voltages around MCU and FPFGA core) - I expect that voltage ripple won't be more than hundreds of millivolts.

I know that real capacity decreases due to applied voltage, but I think that 5V on 50V capacitor X5R will be okay, isn't it?