Power & current measurment

Hi,

Spike on the resistor ... should not be, since it is mandatory to put fast capacitors at the VCC pin(s) of the microcontroller. (no resitor between capacitor and IC)

Thus I expect frequencies in the kHz region max. .. with low voltage ripple.

Klaus

KlausST said:

Hi,

Spike on the resistor ... should not be, since it is mandatory to put fast capacitors at the VCC pin(s) of the microcontroller. (no resitor between capacitor and IC)

Thus I expect frequencies in the kHz region max. .. with low voltage ripple.

Klaus

Click to expand...

Thanks for your reply.
Based on dynamic simulation in Innovus the peak of current is about 1.6ampers.
I want to measure this current and report it in a datasheet.

I do not have any capacitor near the IC that can measure the current peak.

Do you have any recommendations for an approach to measuring the spike current?

Thanks & Regards
Saati

Use infinite persistence mode on scope to get to peak values. If you
have color grading (frequency of sample occurrence) maybe turn that
on to get additional information on spike.


Another simple method, but takes longer to evaluate, is simple set scope for
level trigger mode, and keep advancing level until no more triggers occur.
But infinite persistence I think better method.

The value of infinite persistence, set your scope on AC, say 100 mV/box, infinite
persistence, and look at your supply rail. Usually an eye opener, eg. the pk-pk
noise levels on micro supply rails.

Make sure scope acquisition mode is at highest sampling rate, but
thats affected by spike rep rate.

Also use scope FFT to get a better idea of source of spike, eg. more detective
work.


Regards, Dana.

danadakk said:

Use infinite persistence mode on scope to get get to peak values.

Make sure scope acquisition mode is at highest sampling rate, but
thats affected by spike rep rate.


Regards, Dana.

Click to expand...

Thanks danadakk.

The bandwidth of the oscilloscope and probe is not affect the result ?

I need the wideband oscilloscope?

100 MHz bandwidth current measurement with a low-ohmic (e.g. 100 mOhm) shunt needs exact compensation of resistor inductance, calibrated with a VNA or similar tool.

FvM said:

100 MHz bandwidth current measurement with a low-ohmic (e.g. 100 mOhm) shunt needs exact compensation of resistor inductance, calibrated with a VNA or similar tool.

Click to expand...

Thanks for your good guidance.

Can you explain more ?

The famous companies (e.g. STM, TI) how measure the peak current at rising edge of clock?

Thanks & Regards
Saati

I'm not in the IC design business. I imagine that any current measurement device affects VCC current waveform. I'd probably go for miniature current transformer, hall sensor or coreless di/dt sensor (Rogowski coil).

Page 36 covers BW on needed in scope for various applications, attached.

A low C fet probe one possibility. See youtube for some DIY wideband solutions.
Also Jim Williams LTC ap notes on the topic.


Regards, Dana.

Use ground side 50 mV shunt method with FET Diff probe or make a fixture for probe coil spring pins.


R=50mV / 1.6A using R with 2 test vias, spaced for coil probe adapter on each side of R very close to uC between low ESR cap and uC over a ground plane. Then using coil spring to 10:1 probe tip&ring only to test vias.

Or add in-circuit current sense amplifier using high GBW amp.

Either way it needs special tooling fixture to detect device current with high speed, low ESL and without disturbing Vdd to Gnd which is low noise type.

If Ic=CdV/dt your low ESR Cap must be big enough with ESR as low as current sense shunt for excellent performance. Take care between IC capacitance and discrete cap on ground side with current sense resistor.

dV=dI*ESR + Ic*dt/C on supply Hopefully it will perform OK. There are other methods to measure 1ns rise time current to work.

Hi,

I understand now that you want to measure the currents on a VCC pin ... without capacitor attached at this pin.

What USEFUL informartion do you get form this?

The biggest problem that I see is, that .. maybe you are getting some scope pictures ... while the proper function of the chip is not guaranteed.
--> Thus my question: How useful are "current measurements" on a chip that is probably not working like expected?

It´s well known that there will be spikes. And that these spikes may cause problems.
But what problems?
* Problems with the power supply? --> probably not, because this is a non realistic circuit where the "full bandwidth of currents" are fed to the power supply. In a real circuit there ARE decoupling capacitors. For a good reason. The reason is to guarantee proper operation conditions to the chip in first place.
* Problems with the chip? --> yes. These currents may cause problems IF there is NO suitable decoupling installed. That´s exactly what you do: You generate a condition that may cause chip malfunction.

My personal opinion this setup gives no useful information (Again: may personal opinion)

I understand that it´s usueful to get informations about power supply current. Like lowest current, highest current, average working current, .. and also bandwidth of current but WITH cpacitors installed.
The rise time of the signals in the chip ... tells you the bandwidth that is needed FOR THE CAPACITORs (not for the power supply).
And the power supply needs to handle the jumps in current BUT with the CAPACITORs installed.

***
For me this is like testing a receipe for a cake but KNOWINGLY mixed salt with sugar.
--> Even for testing .. or better say "especially" for testing you first need to ensure proper operation. You need to keep on the reciepe, you need to keep on the datassheet. .. and install the capacitors according datasheet.

Saati said:

the power per clock

Click to expand...

Surely is the AVERAGE VCC current ... with properly installed capacitors. Nothing HF at all.

Saati said:

of companies for characterizing the .... current peak in the rising edge of the clock?

Click to expand...

I´ve never seen such a characterisation. (Again: personal. The fact that I´ve not seen it does not mean "it does not exist")

Especially at microcontrollers you don´t know how many gates are switching at a clock edge. It depends on code currently processed. And for sure it also depends if outputs are switching and the load connected to the outputs. ... and maybe on other parameters

Klaus