Looking for a Probe to Use for 5V DC Power Supply Ripple Measurement

[bendras]

I would like to measure voltage ripple of a 5V DC-DC buck converter with at least 1mV Peak-to-Peak precision. As far as I understand the best way to do it is with some kind of high impedance differential probe (please correct me if I am wrong). Could someone recommend a differential probe suitable for this type of measurement?

[Keysight DanielBogdanoff]

Depending on budget and oscilloscope, a lot of people are using power rail probes for this type of measurement. Something like the our N7020A:
https://www.keysight.com/en/pd-2471132-pn-N7020A/power-rail-probe?cc=US&lc=eng

It was designed to handle offsets higher than 5V, but the power rail probe has been the go-to for a couple years (at least on the enterprise side).

Not sure how helpful it will be for your specific application, but this blog post also talks about making single ended measurements with a differential probe:
https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2018/04/17/use-your-differential-probe-where-you-never-thought-possible

[nctnico]

The easiest way is to solder a piece of coax to the board and use a DC blocker and 50 Ohm termination at the oscilloscope.

[bendras]

Depending on budget and oscilloscope, a lot of people are using power rail probes for this type of measurement. Something like the our N7020A:
https://www.keysight.com/en/pd-2471132-pn-N7020A/power-rail-probe?cc=US&lc=eng

It was designed to handle offsets higher than 5V, but the power rail probe has been the go-to for a couple years (at least on the enterprise side).

Not sure how helpful it will be for your specific application, but this blog post also talks about making single ended measurements with a differential probe:
https://community.keysight.com/community/keysight-blogs/oscilloscopes/blog/2018/04/17/use-your-differential-probe-where-you-never-thought-possible

The suggested probe seems to be suitable for the task but as far as I can tell it is only compatible with Keysight oscilloscopes. Does anybody know of a similar probe compatible with bog standard oscilloscopes or at least the ones with the "TekProbe Level II" interface?

The easiest way is to solder a piece of coax to the board and use a DC blocker and 50 Ohm termination at the oscilloscope.

I have investigated this technique but at 5V it seems to be loading the circuit with a few tens of mA of current draw. That is the reason why I am looking for a way to do differential measurements.

[precaud]

"Ripple" suggests low-freq AC, so why not just use a 1:1 probe into a 1M Ohm AC coupled scope input? The probe's capacitive loading won't be a problem here.

[_Wim_]

You could also go for a preamp like a stanford resaerch SR560 (http://www.thinksrs.com/products/sr560.html) These have a differential input, high input impedance, ac coupling together with very high gain...

These pop-up regurarly on Ebay for about 750€, not cheap, but very versatile in many applications, and also universal usable on any scope...

[Cerebus]

"Ripple" suggests low-freq AC, so why not just use a 1:1 probe into a 1M Ohm AC coupled scope input? The probe's capacitive loading won't be a problem here.

You need a differential probe or all the common mode noise that's generally slopping about will completely swamp the noise that's actually being generated by the power supply and that's all you'll see.

[Cerebus]

Depending on budget and oscilloscope, a lot of people are using power rail probes for this type of measurement. Something like the our N7020A:
https://www.keysight.com/en/pd-2471132-pn-N7020A/power-rail-probe?cc=US&lc=eng

It was designed to handle offsets higher than 5V, but the power rail probe has been the go-to for a couple years (at least on the enterprise side).

The N2792A (200MHz ±20V differential ±60V common mode) is probably a better fit than the 2GHz N7020A (which at a guess is probably $2k a pop). That or the N2791A (25 MHz ±70V differential @ 10:1  or ±700V differential @ 100:1, ±700V common mode).

Trying to get him to spend more money Daniel? 

[bendras]

You could also go for a preamp like a stanford resaerch SR560 (http://www.thinksrs.com/products/sr560.html) These have a differential input, high input impedance, ac coupling together with very high gain...

These pop-up regurarly on Ebay for about 750€, not cheap, but very versatile in many applications, and also universal usable on any scope...

This particular preamp seems to be limited to 3Vpp inputs signals, hence measuring 5V DC rail would be out of spec. (please correct my if I am wrong). Also as far as I understand its 1MHz bandwidth would not allow measurement of high frequency ripple.

"Ripple" suggests low-freq AC, so why not just use a 1:1 probe into a 1M Ohm AC coupled scope input? The probe's capacitive loading won't be a problem here.

You need a differential probe or all the common mode noise that's generally slopping about will completely swamp the noise that's actually being generated by the power supply and that's all you'll see.

This is exactly the reason why I am looking for a way to do a differential measurement.

Depending on budget and oscilloscope, a lot of people are using power rail probes for this type of measurement. Something like the our N7020A:
https://www.keysight.com/en/pd-2471132-pn-N7020A/power-rail-probe?cc=US&lc=eng

It was designed to handle offsets higher than 5V, but the power rail probe has been the go-to for a couple years (at least on the enterprise side).

The N2792A (200MHz ±20V differential ±60V common mode) is probably a better fit than the 2GHz N7020A (which at a guess is probably $2k a pop). That or the N2791A (25 MHz ±70V differential @ 10:1  or ±700V differential @ 100:1, ±700V common mode).

Trying to get him to spend more money Daniel? 

The mentioned N2792A probe has 10:1 attenuation (as far as I understand this would make 1mVpp look like 0.1mVpp on the scope, hence waste scope's vertical resolution). Also this probe seems to have 6mVrms noise floor, which seems far too high for making 1mVpp ripple measurements.

[precaud]

You need a differential probe or all the common mode noise that's generally slopping about will completely swamp the noise that's actually being generated by the power supply and that's all you'll see.

This is exactly the reason why I am looking for a way to do a differential measurement.

OK. How about a Tek AM502 diff preamp? PAR made a similar standalone unit, the 113. Signal Recovery has the 5113, that will be in the same price range used as the SR560, though.

[bendras]

You need a differential probe or all the common mode noise that's generally slopping about will completely swamp the noise that's actually being generated by the power supply and that's all you'll see.

This is exactly the reason why I am looking for a way to do a differential measurement.

OK. How about a Tek AM502 diff preamp? PAR made a similar standalone unit, the 113. Signal Recovery has the 5113, that will be in the same price range used as the SR560, though.

All of these preamplifiers seem to have maximum bandwidth of 1MHz. As far as I know the standard for measuring power supplies is 20MHz. Do you know of any models with 20MHz or higher bandwidth?

[bendras]

The HP Hewlett Packard 1141A seems like a good candidate, but I am a bit confused about the specification. The datasheet https://www.testequipmentdepot.com/agilent/pdf/1141a.pdf on page 54 says that without attenuators (i.e. in the 1:1 mode) it can handle up to +/-20V in "Common-mode Operating Range DC" and up to +/-300mV peak in "Differential Input Range". Does this mean that it is suitable for measuring 20V DC rails with 300mV peak-to-peak ripple on them?

[nctnico]

The easiest way is to solder a piece of coax to the board and use a DC blocker and 50 Ohm termination at the oscilloscope.

I have investigated this technique but at 5V it seems to be loading the circuit with a few tens of mA of current draw. That is the reason why I am looking for a way to do differential measurements.

No it doesn't. The DC block (just a capacitor in series) is placed before the 50 Ohm termination so no DC current flows into the oscilloscope.

[David Hess]

OK. How about a Tek AM502 diff preamp? PAR made a similar standalone unit, the 113. Signal Recovery has the 5113, that will be in the same price range used as the SR560, though.

All of these preamplifiers seem to have maximum bandwidth of 1MHz. As far as I know the standard for measuring power supplies is 20MHz. Do you know of any models with 20MHz or higher bandwidth?

The Tektronix 1MHz AM502 was the standalone implementation of the Tektronix 5A22/7A22 for the 5000/7000 series mainframe oscilloscopes.  It is great for these kinds of measurements but of course only up to 1MHz.

For higher bandwidth in the 7000 series, Tektronix had the 100MHz 7A13 which provides 1mV/div sensitivity over a common mode range of +/-10 volts and can be used as a differential comparator similar to Keysight's power rail probes or as a differential input amplifier.  The problem with the 7A13 (besides being old) is that its switchable bandwidth limit is 5 MHz (1) although this could be modified by changing a capacitor.  My solution is to connect my DSO to the vertical output from the 7000 mainframe and implement the 20MHz bandwidth limit there; so effectively the 7A13 and 7000 mainframe become a 100MHz 1mV/div differential front end for my DSO.

Preamble made a standalone version of the 7A13.  They were bought be LeCroy who still produces their products but that would be an expensive way to go although cheap if you design power supplies for a living.  You might be able to find a used Preamble instrument on Ebay.  Some external differential probes might be suitable also.  The Pintek DP-60HS could be used with external AC coupling capacitors but it only has a 15 MHz bandwidth.

The easiest way and what I would try first, which also happens to be recommended for ATX power supply testing, is using a pair of x1 probes and a 2 channel oscilloscope configured for differential (invert and add) input operation.  The problem is gain matching between the input channels which some DSOs do not handle well and increased quantization noise.  Analog oscilloscopes and some very rare old DSOs can use their analog "variable" vertical controls to match the gain of the channels for pretty good performance when used like this while avoiding an increase in quantization noise.

(1) I think the reason for a 5MHz bandwidth limit instead of 20MHz on the Tektronix 7A13 was that 20MHz had not be standardized for power supply noise measurement when it was designed and the 7A13 is pretty noisy due to its complexity so 5MHz was a good choice for operation at its maximum sensitivity.  The specified noise of the 7A13 is less than 200uVrms and I measured mine at about 100uVrms which is about 10 times what a simple but good single ended 1mV/div input with a common mode range of +/-10mV has.  The difference here is that the 7A13 is differential which doubles the input noise and has a cascode to support operation over +/-10V or 1000 times what a normal oscilloscope input can handle.  When used as a differential comparator, it is like having an oscilloscope position control which operates over 20,000 divisions.

[precaud]

All of these preamplifiers seem to have maximum bandwidth of 1MHz. As far as I know the standard for measuring power supplies is 20MHz. Do you know of any models with 20MHz or higher bandwidth?

Not offhand. I think you're in diff probe territory to get useful CMRR at that bandwidth.

[_Wim_]

This particular preamp seems to be limited to 3Vpp inputs signals, hence measuring 5V DC rail would be out of spec. (please correct my if I am wrong). Also as far as I understand its 1MHz bandwidth would not allow measurement of high frequency ripple.

AC coupled these can withstand 100Vdc according to the user manual. Yes, bandwidth is limited, but high bandwidth AND low noise is not possible. For me "ripple" seemed like a low frequency thing, I was not aware you wanted to measure at least up to 20Mhz.

As suggested, the probes designed for the job are off course ideal, but expensive and typically dedicated to a specific scope brand.

Edit: quote formatting

[NiHaoMike]

Since only AC is important, salvage the coupling transformer from an Ethernet device or DSL modem, connect the primary to the voltage rail with a DC blocking capacitor, and connect the secondary to the scope with a parallel resistor selected to get as flat a frequency response as possible.

[_Wim_]

The N2792A (200MHz ±20V differential ±60V common mode) is probably a better fit than the 2GHz N7020A (which at a guess is probably $2k a pop). That or the N2791A (25 MHz ±70V differential @ 10:1  or ±700V differential @ 100:1, ±700V common mode).

Trying to get him to spend more money Daniel? 

The N2792A and N2791A have much higher noise (6mVrms and 50mVrms) making them not usefull for this kind of application. The N7020A suggested by Daniel achieves 90uVrms combined with an  S-series scope, a top solution if money is not an issue...

[Kosmic]

There's a interesting application note from linear talking about DC DC converter noise and how to measure it.

http://www.analog.com/media/en/technical-documentation/application-notes/an101f.pdf

At page 10 you will find : APPENDIX C Probing Technique for Sub-Millivolt, Wideband Signal
Integrity

There is also a short YouTube video 
https://youtu.be/WxhjLIu-vPg

[David Hess]

All of these preamplifiers seem to have maximum bandwidth of 1MHz. As far as I know the standard for measuring power supplies is 20MHz. Do you know of any models with 20MHz or higher bandwidth?

Not offhand. I think you're in diff probe territory to get useful CMRR at that bandwidth.

In my experience only a moderate amount of common mode rejection is required for these measurements.  The old Tektronix 7A13 achieve a minimum of 300:1 or 50dB at 20MHz from 1mV/div to 20mV/div where despite no input dividers being used, it still allows for a +/-10 volt DC input range.  Since it supports AC coupled inputs, higher voltage measurements are feasible and indeed, switching between AC and DC coupling when measuring the ripple on a 5 volt supply makes no difference unless something is broken.

I have made the same measurement using analog and analog input DSOs in add and invert mode as long as the common mode rejection is adjusted with the variable volts/div control.  The reason I do not do this all the time is simply because none of my oscilloscopes have the native 1mV/div sensitivity that the 7A13 provides although the 7A13 is pretty noisy if its full bandwidth is used.

Ideally what is needed is a low voltage 20MHz differential probe with a wide input common mode range or AC coupling but offhand I do not know of any.  The old Tektronix P6046 sort of meets these requirements since it supports AC coupling.

[precaud]

In my experience only a moderate amount of common mode rejection is required for these measurements.  The old Tektronix 7A13 achieve a minimum of 300:1 or 50dB at 20MHz from 1mV/div to 20mV/div...

I can't recall ever making a realtime noise diff measurement above 1MHz but what you're saying makes sense.

Since only AC is important, salvage the coupling transformer from an Ethernet device or DSL modem, connect the primary to the voltage rail with a DC blocking capacitor, and connect the secondary to the scope with a parallel resistor selected to get as flat a frequency response as possible.

I played around with that once, those xfmrs roll off the low end pretty early, the one I tried was only good down to 1kHz or so.

[MarkL]

Another possibility is the Lecroy DA1855A differential amplifier (which is a distant offspring of the 7A13 that David Hess is eluding to).

It has a switchable gain of x1 and x10, and a BW of 100MHz.  There are selectable filters to reduce the BW to 20MHz, 1MHz, or 100kHz.  CM range is +/-15.5V (x1) or +/-155V (x10).

They can be found on ebay for exorbitant prices, or you can hang out and wait for a deal.  There was one that just sold for USD$330 with a beat up front panel.  More commonly they can be had for around $600 - $800.  It's a good tool to have if your budget allows.

Additional data:

  http://teledynelecroy.com/probes/differential-amplifiers/da1855a

[KrudyZ]

The suggested probe seems to be suitable for the task but as far as I can tell it is only compatible with Keysight oscilloscopes. Does anybody know of a similar probe compatible with bog standard oscilloscopes or at least the ones with the "TekProbe Level II" interface?

You could use a Tektronix P6247 or P6248 if you're looking For a TekProbe interface.
Common mode range of +-7V
To zoom in on the ripple, apply a constant 5V from a reference to the negative input of the probe.
Make sure the ground side of the reference is hooked up to the ground of the cap on the DUT supply that you are measuring your ripple across. Add a local bypass cap for the reference and either twist your leads or use coax for the feed.
In 1:1 mode these probes have a 850 mV range and they are FAST (> 1GHz). This setup will give you DC readings and you can switch the probe to 10:1 to look at the turn-on and turn-off of the regulator.
The P6246 is a slower model at 400 MHz and there is also a P6250 with higher voltage ratings, but it's less common to find and does not have a 1:1 range.

[bendras]

Another possibility is the Lecroy DA1855A differential amplifier (which is a distant offspring of the 7A13 that David Hess is eluding to).

It has a switchable gain of x1 and x10, and a BW of 100MHz.  There are selectable filters to reduce the BW to 20MHz, 1MHz, or 100kHz.  CM range is +/-15.5V (x1) or +/-155V (x10).

They can be found on ebay for exorbitant prices, or you can hang out and wait for a deal.  There was one that just sold for USD$330 with a beat up front panel.  More commonly they can be had for around $600 - $800.  It's a good tool to have if your budget allows.

Additional data:

  http://teledynelecroy.com/probes/differential-amplifiers/da1855a

Having looked through the specs. of this device it seems that it would not only meet my current needs but that it would also be a safe bet for the future 

The suggested probe seems to be suitable for the task but as far as I can tell it is only compatible with Keysight oscilloscopes. Does anybody know of a similar probe compatible with bog standard oscilloscopes or at least the ones with the "TekProbe Level II" interface?

You could use a Tektronix P6247 or P6248 if you're looking For a TekProbe interface.
Common mode range of +-7V
To zoom in on the ripple, apply a constant 5V from a reference to the negative input of the probe.
Make sure the ground side of the reference is hooked up to the ground of the cap on the DUT supply that you are measuring your ripple across. Add a local bypass cap for the reference and either twist your leads or use coax for the feed.
In 1:1 mode these probes have a 850 mV range and they are FAST (> 1GHz). This setup will give you DC readings and you can switch the probe to 10:1 to look at the turn-on and turn-off of the regulator.
The P6246 is a slower model at 400 MHz and there is also a P6250 with higher voltage ratings, but it's less common to find and does not have a 1:1 range.

The Tektronix P6246 definitely seems suitable  . Having said that, since I do not have a compatible scope at the moment it would workout cheaper for me to get the Lecroy DA1855A.

[nctnico]

The suggested probe seems to be suitable for the task but as far as I can tell it is only compatible with Keysight oscilloscopes. Does anybody know of a similar probe compatible with bog standard oscilloscopes or at least the ones with the "TekProbe Level II" interface?

You could use a Tektronix P6247 or P6248 if you're looking For a TekProbe interface.
Common mode range of +-7V
To zoom in on the ripple, apply a constant 5V from a reference to the negative input of the probe.
Make sure the ground side of the reference is hooked up to the ground of the cap on the DUT supply that you are measuring your ripple across. Add a local bypass cap for the reference and either twist your leads or use coax for the feed.

I'm wondering if this isn't an overly obfustigated way of measuring ripple. One of the problems I see is that the 5V offset must be ripple free as well over a large frequency band. Just look at the Jim Williams appnote linked to above which uses a DC blocking capacitor and a coax cable soldered directly to the point of interest.