Ripple voltage measuring: confusion with using of capacitor along with oscilloscope probe

Hi,

1) I´ve never seen installing an additional capacitor just for scope measurements. But you are free to do so.
When I install such a capacitor I do this to reduce the rippl at the load. (also to reduce EMI noise)
BUT you need to be carful with wiring.

Don´t do star wiring: Don`t do: power_supply --> capacitor AND power_supply --> load

Better do linear wiring: Power_supply --> capacitor --> load. For both output and GND.

2) The power supply datasheet tells.

3) Again (like 1) I don´t see why one wants to suppress noise on the scope. My thinking is rather the opposite: I want to see everything what really happens at this node. Wanted and unwanted signal. Jitter, noise, ringing ....
Thus I personally don´t suppress these "errors". They are real, thus I want to see it. Even if truth hurts ;-)

***
For a detailed discussion we need to see photos showing all of the circuit: Power supply, capacitor, load, scope probe connection, and all the related wiring..
Photos: 100kBytes should be sufficient.

Klaus

KlausST said:

Hi,

1) I´ve never seen installing an additional capacitor just for scope measurements. But you are free to do so.
When I install such a capacitor I do this to reduce the rippl at the load. (also to reduce EMI noise)
BUT you need to be carful with wiring.

Don´t do star wiring: Don`t do: power_supply --> capacitor AND power_supply --> load

Better do linear wiring: Power_supply --> capacitor --> load. For both output and GND.

2) The power supply datasheet tells.

3) Again (like 1) I don´t see why one wants to suppress noise on the scope. My thinking is rather the opposite: I want to see everything what really happens at this node. Wanted and unwanted signal. Jitter, noise, ringing ....
Thus I personally don´t suppress these "errors". They are real, thus I want to see it. Even if truth hurts ;-)

***
For a detailed discussion we need to see photos showing all of the circuit: Power supply, capacitor, load, scope probe connection, and all the related wiring..
Photos: 100kBytes should be sufficient.

Klaus

Click to expand...

Thank you for reply. Here is picture of circuit I had assembled and tested (and, nfortunately, it looks like it was star wiring)

Those prototype boards are TERRIBLE for doing any kind of measurements of noise or ripple, etc. And you’ve got the actual device you’re trying to measure connected by long wires to your terrible proto board? This is a useless setup.

barry said:

Those prototype boards are TERRIBLE for doing any kind of measurements of noise or ripple, etc. And you’ve got the actual device you’re trying to measure connected by long wires to your terrible proto board? This is a useless setup.

Click to expand...

Thank you for comment.
Yes, I'm aware that protoboards are noisy and long wires also adding noise (anyway this particular wallplug designed with 1.2 meter wire).
And I agree that this circuit will be useless if i want to know exact values of actual noise/ripple voltage.

But I'm much more interested in role of connected capacitor (Should I think of it as a part of oscilloscope probe?)
regardles of circuit/device (will it be noisy or "quiet") to be tested.

Also, the Noise is generally measured as RMS, not peak-to-Peak.
Your set-up is completely inappropriate, you're using protoboard, your GND connection is too long and measurement node is different than actual output.
In additional to these, the noise is measured juts at the output of the supply, not 1.2meter long and EMI open set-up.

Hi,

I agree with the others. I can´t see any useful in this setup.

Regarding my recommendations:
* I asked for showing the complete circuit. Now I see two photos .. the power supply is not conected at all. But the wiring is important.
* I recommended linear connection... I can´t see this in your photos

***
Additionally: Maybe the GNDs of the two probe channels are connected inside the scope. If so you generate a new path for current flow. Indeed it´s a loop, an antenna, with it´s own resonance. Ringing expected.

Klaus

BigBoss said:

Also, the Noise is generally measured as RMS, not peak-to-Peak.
Your set-up is completely inappropriate, you're using protoboard, your GND connection is too long and measurement node is different than actual output.
In additional to these, the noise is measured juts at the output of the supply, not 1.2meter long and EMI open set-up.

Click to expand...

Thank you for reply

I have some questions regarding your comments:
1) GND connection: do you mean GND connection of the oscilloscope probe?
2) Actual output: Can I think about protoboard "red" and "blue" rails as it is output of one assembled "Wallplug+wire+protoboard circuity + protoboard strips" device (of-course noisy one device) which I'm trying to test using method described in abovementioned article?
3) Actual output: Anyway this wallplug will be used with it's 1.2m wire. Load will be connected via barrel jack at the end of this wire. Why hould we test it's noise before wire?

Sorry if my questions too naive

KlausST said:

Hi,

I agree with the others. I can´t see any useful in this setup.

Regarding my recommendations:
* I asked for showing the complete circuit. Now I see two photos .. the power supply is not conected at all. But the wiring is important.
* I recommended linear connection... I can´t see this in your photos

***
Additionally: Maybe the GNDs of the two probe channels are connected inside the scope. If so you generate a new path for current flow. Indeed it´s a loop, an antenna, with it´s own resonance. Ringing expected.

Klaus

Click to expand...

Thank you for reply.
I'll re-assemble, re-test and share result for circuit with linear connection tomorrow, please wait.
(Pictures attached in my previous post I have made before startingthis thread and before your posted recommendations, i just shared what i had already done)

The article "how to reduce voltage ripple" is misleading in my view. It's rather about reducing switching noise contributing to apparent Vpp value measured by an oscilloscope. The decision what you consider as real voltage ripple is however arbitrary, depending on which parameters of the power supply you want to characterize. Switching noise can be very real when it e.g. exceeds EMI limit values.

If the articel is just about hiding noise from the measurement, you can apply a low-pass filter.

An important point is completely ignored in the article, it's common mode noise dropping along the probe cable and this way injected into the measurement.

If you want to measure ripple voltage at smps output, look up "DIY Coaxial probe"...there have been many postings on that in this forum. Measure it near the terminals of an output cap........i dont see the point in putting a cap on the scope tip...but yes, power integrations always show an app note where that is done.

If you cant make a DIY coax probe, i just cut the end off a cheap scope probe and use that...ie, using it as coax....solder it to cct and measure......scope on x1 obviously....put a 47R or so resistor to the tip too if you want to be really snazzy.

The 10:1 probe coil is not suitable for remote sensing with dual probes far apart > 3cm as this creates ~8 nH / cm inductive ground path resonating with coax capacitance 80 ~ pF/m ( if I recall)

A common 0V point must be used. which we call ground. with no current path thru it.

This will eliminate the high Q series RLC ground loop. One can also by adding 470 ohm to the probe ground to suppress > 20MHz probe resonance or use the scope 20 MHz filter.

But for SMPS ripple , I prefer to send only AC to on 50 ohm coax terminated on a T-BNC with an axial resistor ( 50 to 500 Ohms) or 50 inside DSO. I depends on the bandwidth of noise you need to measure.

This shows what was in your link and how I prefer it. This may the cap on board, or in your case hopefully simulating the cap (s) in target load. Obviously, this > 1V ripple needs much more bulk C to reduce ripple.



My guess is your little black brick suffered a meltdown with high ESR caps inside showing on the corner and that's why you are measuring ripple.

The waveforms do not look like switcher ripple to me. Or you've
scrubbed off all the harmonics by the components and "board".
Or you're BW-limiting the scope channel.

DC-DCs need more than "noise filtering". The output filter cap
also has to take up the "quick time" current of any load-step,
as the control loop is too slow and plays catch-up.

ESR and ESL of the filter cap (which includes routing to / from)
can ruin the HF suppression. But here it seems everything is a
sine wave.

The cap-blocked coax can keep HV DC off the 'scope channel
which might save your equipment (unless the cap fails). A
probe's attenuation (50/1M or 50/10M) will often be OK but
a naked coax could put hundreds of volts (if that's your output)
to expensive gear that probably says "Don't!" somewhere in
the manual, if not in print on the panel.

Thanks Dick. that's why I prefer my method for cleaner textbook waveforms.

Considering the battery as an ultra capacitor, it will have an ESR < 100 mohms and C>5kF so you could use anything >> 100 uF for your measurement but the thermal issue I mentioned is a hint of problem.

Then the single 10:1 probe with tip/ring coil method will be adequate.

Hello all!

Taking into account recommendations from cupofftea and D.A.(Tony)Stewart, I have ordered some BNC coaxial cable, and I will post an update to this thread in few days (after receive it and asemble coaxial probe).

Would it be better to accept a 50% attenuation, and put 50 ohms
series to better match the line? The 'scope photos up top do seem
to have a lot of ringing at the transitions, but whether this is line
match or just the result of wireball Ls and Cs, would have to be
figured out on the bench.

Every inductor will have some interwinding C that will let edge
dV/dt blow through to the filter.

Every filter cap will have some L which degrades its ability to deal.

And then there's the wireball.

Hello All!

At last I’m able to share results of re-testing of this noname wallplug adapter. I used seven different setups and different self-made probes, so hope it can be interesting to somebody (at least it was interesting for me).

Taking into account advices from cupoftea (reply #10), D.A.(Tony)Stewart (reply #11, #13), dick_freebird (reply #15) and information from these articles: Match impedances when making measurements, Build Your Own Power Rail Probe , I decided to make two custom probes and use one standard 10:1 probe with short grounding and compare its performance while testing that wallplug adapter.
One more point: I also include test of breadboard mounted setup to evaluate “breadboard noise”.

PROBES

- Probe 1: 47nF Capacitor series probe with 50Ω termination;

- Probe 2: 51Ω series probe with 50Ω termination. 51Ω series resistor will work as wallplug adapter impedance. so wallplug adapter, probe cable and oscilloscope impedances will be matched. But 51Ω series and 50Ω termination will form voltage divider and make ~50% attenuation for oscilloscope input);

- Probe 3: Standard oscilloscope 10:1 probe with short grounding tip mounted on it.

TEST SETUPS AND RESULTS PART 1 (1,2,3 of 7, Part 2 in next reply)


TEST SETUPS AND RESULTS PART 2 (4,5 of 7, Part 3 in next reply)

TEST SETUPS AND RESULTS PART 3 (6,7 of 7)


Other details:
-I calculated attenuation factor by comparing wallplug Vdc output, and DC voltage measured by oscilloscope before each test.
-During performing of tests, I tried to move all available wires in my setup: tried to entangle or align wires in straight lines, to move it +/-1m in different directions, make coils, tangles etc). The only effect that I saw is about +/-10% changes in amplitude of afterspike ringing.


My conlusions/quesstions
1)
As for main question of this thread – role of filter cap: is it added just to hide actual DUT noise (for example switching noise) to make ripple waveform more clear on oscilloscope screen, or is it added to hide some not related to DUT noise (picked-up noise or some oscilloscope/probe resonance related noise).

I’m still not 100% sure in but I’m inclined to the former. My reasoning is: In my tests, when filter cap is present, spike and afterspike ringing vanishes, but ripple Vpp rising and became longer, so I believe that it is because energy of actual spike absorbed by filter cap and added to ripple waveform. And filter cap worked as low-pass filter. Can it be correct answer?

2)
As for self-made probe performance, it looks like best probe is PROBE 1 (47nF Capacitor series probe with 50Ω termination). Maybe because it is wrapped by aluminum foil tape. I’ll try to upgrade PROBE 2 with aluminum foil next time and compare.

3)
I'd love to hear If anyone have comments and suggestions for abovementioned testing setups

You misunderstood me. I meant only cap on board and 50 ohm at scope.. This prevents reflections and attenuation and block DC into for more gain on scope not it is not a ripple filter, rather a ripple prevention method with a coax matched load to a very low impedance Voltage source.

Ripple from a long 10:1 probe clip >2cm is induced from ground wire L ( 10nH/cm and coax capacitance) .

If 10:1 probe is used, it MUST be without tip and gnd clip and ONLY use coil spring adapter and tip when impulses or step waves exist with BW > 20 MHz.,

Thus both methods will remove impulse resonance error and reveal true DC ripple. But the methods are different.,

It is just a high pass filter and eliminates probe ground inductance .

Reducing scope to 20 Mhz is an option but may be rjecting some true noise > 20 MHz.

When I mentioned a R/R divider , I meant 9R/R divider high impedance AC coupled as a another alternative. But the high Q LC ground path is the main issue I think even adding a 1k resistor to long ground clip (not tried) ought to dampen the probe resonance. FYI Sorry for lack of clarity.

D.A.(Tony)Stewart said:

You misunderstood me. I meant only cap on board and 50 ohm at scope.. This prevents reflections and attenuation and block DC into for more gain on scope not it is not a ripple filter, rather a ripple prevention method with a coax matched load to a very low impedance Voltage source.

Ripple from a long 10:1 probe clip >2cm is induced from ground wire L ( 10nH/cm and coax capacitance) .

If 10:1 probe is used, it MUST be without tip and gnd clip and ONLY use coil spring adapter and tip when impulses or step waves exist with BW > 20 MHz.,

Thus both methods will remove impulse resonance error and reveal true DC ripple. But the methods are different.,

It is just a high pass filter and eliminates probe ground inductance .

Reducing scope to 20 Mhz is an option but may be rjecting some true noise > 20 MHz.

When I mentioned a R/R divider , I meant 9R/R divider high impedance AC coupled as a another alternative. But the high Q LC ground path is the main issue I think even adding a 1k resistor to long ground clip (not tried) ought to dampen the probe resonance. FYI Sorry for lack of clarity.

Click to expand...

Thank you for reply.
Sorry for bothering and sorry if asking you to tell me same thing but in other words, but can you please look at next few pictures and confirm is this wat you meant?

1)
As for 10:1 probe, here is picture of what i have used (reffered as "Probe 3" in my previous post). Is this what you meant?


2) As for "only cap on board and 50 ohm at scope" probe. Here is schematic of probe I made. (reffered as "Probe 1" in my previous post). Did you mean something like this? (I decided to add barrel jack socket because it is easier to connect it with wallplug adapter)

Yes this works with ground trace close to the signal with proper decoupling cap across the IC or output of power supply. How much ripple do you get on your output now in peak to peak?
This should eliminate all the ringing above 20 MHz

Hello
Thank you for clarifying! Now Im getting this output:

1) 10:1 probe (Probe 3)
- Without capacitor across output of power supply: Spike 2.67Vpp. Ripple 0.122Vpp;
- With 47nF capacitor actoss output of power supply: No spike. Ripple 0.260Vpp


2) Capacitor probe (Probe 1)
- Without capacitor across output of power supply: Spike 0.662Vpp, Ripple 0.118Vpp;
- With 47nF capacitor actoss output of power supply: No spike, Riple 0.262Vpp;