Switch mode power source (SMPS) the noise that go up can become very bad sometimes. A power source of simple low cost switch (SMPS) the voltage noise that go up, and be in because of these power source almost the fame of noise respect not beautiful and drop. So, how should measure switch power source (SMPS) medium noise?

Switch manostat outputs noise mediumly

With respect to its property character, the output of NSMPS can have noise of a few switch. After all, they are to use pulse width modulation by the design (or pulse frequency modulation) signal from electric current of switch of taller direct current source, use next 2 extremely LC filter has filter wave to its.

The switch movement of MOSFET produces alternant cycle, among them the first electric current flows into inductor, next inductor discharge. This brings about big DI / Dt and big voltage aiguille. We expect this kind of noise. This is a problem, LC filter has in the others field that avoids these big voltage aiguille to transmit circuit how effective.

The typical output voltage of SMPS will be in indication ripple in switch frequency. A main index is when adjuster does not have load, and resistor of load of model of the to load in applying from time to tome how many ripple.

Measure the noise in switch power source

I have a low noise to apply recently, the 3.3 V SMPS; that I think to try to use a very low cost needs the load current of 50 MA only. I have to evaluate board, I receive power source repeatedly with power source of 5 V wall, with simple 10× probe measures output. If the graph is shown 1 times,my measurement is configured.

Graph 1. Use 10 times probe to measure output voltage course.

Dc n is very good when 3.3 V. Depend on ability of 12 resolution of oscillograph of my Teledyne LeCroy HDO 8108 and big deflection, I can offset this voltage, such I can magnify ripple noise and still can search slow fast dc drift. Graph the 2 measurement that showed scale of 10 MV / Div falls voltage noise.

 

Graph 2. Use 10×The measurement on the SMPS output of probe noise, scale is 10 MV / Div.

Switch implement 20 small seconds are periodic – corresponding the switch frequency at 50kHz – be clearly. From inductor electric current charge to anticipate triangular pulse circularly with discharge. But, besides the feature that anticipates besides this, still have the high frequency noise of two kinds of types. Even area is put in noise of peak value of 10 MV peak, aiguille noise can reach peak value of 60 MV peak sometimes.

High frequency noise and acerb noise aiguille are disturb. This is done not have by 2 extremely LC filter filter is divided. If I use this kind of power source, how can I ensure my circuit board the function that can keep enough, although there are these noises?

However, the fact proves, this kind of noise is not the voltage noise on power source output actually. In my exploration, all radio frequency are radio frequency.

Divisional voltage noise and RF collect

Through LC filter the big DI / Dt of medium inductor brings about the big magnetic field that produces around SMPS. Any annulus ways that have low inductance way can produce magnetic induction electricity, produce the tension that we measure with oscillograph thereby.

I made even the 10 times probe that receives SMPS down-lead an annular antenna, can collect these aiguille. The first your idea may be, but of 10× probe most advanced whether to have 9MΩ resistance? Is this one prevents any alternating current to be in annulus road big impedance of the induction?

Most advanced the resistance that has a 9MΩ , but the shunt-wound electric capacity that also has a 10 PF, it is the one part of equalizer circuit, high frequency electric current has flowed through this circuit. When 100 MHz, the impedance of 10 PF capacitance is 160Ω only, very low.

To check these noise whether be the RF collect in probe really the idea that is not the actual noise on power source course, I receive connector solder of a small-sized SMA the output of circuit board to carry, in order to reduce annular antenna area and radiation sensitivity domain. In addition, I am measuring SMPS to output voltage around added another 10 times probe, but use the 2nd probe, most advanced short receive ground wire. This kind of setting allows me to use 10 times probe to measure output course at the same time, measure output course through SMA connector, and noise of this locality RF (probe collect, most advanced short receive ground wire ICfans) . If pursue,3 are shown.

Graph 3. Use two 10× probe and join of a coaxial 1× to measure SMPS to output metrical voltage noise.

Graph 4. The measurement on SMPS output voltage. All passageways are inside range of same 10 MV / Div.

Probe attenuation can affect SNR

Have two important observation results. Above all, the general noise level of 1× coaxial cable is far under 10× probe. Because 10× bougie is not 10× bougie,this is actually, it is 0.1× bougie. It signal attenuation 10 times, its extent reduces 20 DB. When we measure small signal n, for example a few millivolt, measures voltage is very sensitive to the amplifier noise of oscillograph.

Most oscillograph is enough and clever, can identify a passageway to go up with 10× probe. They are met the voltage scale that self-correcting shows, show most advanced tension in order to compensate decuple factor attenuation. Accordingly, when oscillograph shows signal with scale of 10 MV / Div, it uses scale of 1 MV / Div on amplifier actually. What what we see is, the peak value of most advanced noise achieves 10 MV almost to peak value, it is 1 MV about in the noise of peak peak value of oscillograph amplifier actually.

Cable of coaxial of use SMA connective is 1× probe actually. This trace also shows with scale of 10 MV / Div. Below this kind of circumstance, noise of amplifier of peak value of 1 MV peak includes the line in trace more or less wide inside.

This made clear important optimal measure practice: When we observe low extent signal, for example noise of power source course, probe of any 10 times attenuation can make our SNR reduces 20 DB. When every DB computation, do not use attenuation probe please.

Coaxial join and oscillograph probe

The 2nd observation is as a result, in coaxial join nonexistent big and acerb aiguille, however two 10× bougie measures consist in in. As a result of among them a probe is outputted without orbit of bring into contact with even, this shows because RF collect is caused,aiguille aiguille noise is strongly, is not the voltage noise on SMPS output.

This makes clear the 2nd important optimal measure practice: When measuring low extent signal, make exhaust the likelihood is close to coaxial connective to measure a setting, in order to reduce the annulus road area of probe and the effectiveness that its regard aerial as.

If we carry out these two optimal measure practice, we have the noise of ripple of peak peak value of 30 MV in course of 3.3 V voltage. This is the ripple of 1 ％ , suit the SMPS of low cost very much. In addition, high frequency noise is reduced greatly, and short when transient state – regard RF collect as noise actually and not be course voltage noise – serve as switch no longer implement the one part that exports signal shows.

Frequency domain noise

Want me to use the ground level of the power source that stands by me and signal method only, this is a practice of important optimal semiconductor design, from this the signal on the equipment of SMPS power supply and my board will see the harmonic SMPS that arises by 50 KHz only.

Use direct coaxial, low noise joins, I measured the audio spectrum of a confusion of voices on course of SMPS power source. An example is shown 5 times like the graph.

Graph 5. The audio spectrum of a confusion of voices on power source course. Top is when chart of frequency conversion table, exceed 10 seconds, show very stable extent. Here scale falls, 0 DBmV is noise of 1 MV extent.

The peak value in spectrum is the 50 KHz harmonic of switch frequency. The extent of harmonic is 10 DBmV about, namely 3 MV. This is far be less than the voltage of 30mV peak peak value that measures in time domain. Because if this occupies sky comparing lowly,ripple noise is had,this is. In harmonic short when there is too much sine wave in trigonometry pulse. The odd form of form of region medium wave when a large number of Gao Cixie wave express reachs his high frequency content.

All switch noise all under the 10μV extent when making an appointment with 3 MHz. To my application, this is an acceptable noise level, it is very actually low for the SMPS of low to this kind cost.

Conclusion

The article discussed the serious consideration factor of the voltage noise that produces actually about switch power source, introduced two kinds of optimal measurement technique, can help you undertake to the output course of switch manostat accurate oscillograph is measured.