Frequency of Output Ripple Voltage

[orionsbelt]

Hello,

I am designing a power supply for a highly-sensitive accelerometer. The accelerometer requires a maximum input noise of 100muV/sqrt(Hz). My chosen power supply has an output ripple voltage of 13.75mV, which will affect the measurement of the accelerometer. Now, I have used the TI WEBENCH and the suggested design is to use an LM2698 boost converter, which is suitable for my solution. But in the design report, provided by the TI, the frequency of the output ripple voltage is not mentioned.

Hence, I wonder whether the frequency of the output ripple voltage has the same frequency as the switching frequency of the converter? According to the datasheet of LM2698, the switching frequency is 600 kHz.

I am uploading the design report which might help you further.

Best regards
T.K

 

[KlausST]

Hi,

ripple frequency usually is switching frequency. Take care the switcher is not in "burst mode". This is at light loads and ripple is much higher than in continous mode. Additionally the ripple frequency is much lower than switching freqency.

I recommend one of the folowing: (top: the most easy way ---> bottom the least expectable ripple voltage)
* with the swticher: us an additional load to prevent from burst mode
* and/or use a LCR low pass filter between switcher capacitor an accelerometer.
* use a linear post regulator between switcher and accelerometer.
* use a voltage reference as supply of your accelerometer.

Klaus

 

[FvM]

LM2698 has no burst mode according to the datasheet. But it's essential to check for continuous switching without subharmonics and adjust the loop compensation if necessary.

I think it's generally suggested to design staggered power supply filtering and use a well considered ground layout to achieve the intended low noise operation. A linear post regulator as suggested by KlausST is a good idea.

 

[FlapJack]

Analog Devices ADM7150
Input voltage range: 4.5 V to 16 V
Maximum output current: 800 mA
Low noise
1.0 µV rms total integrated noise from 100 Hz to 100 kHz
1.6 µV rms total integrated noise from 10 Hz to 100 kHz

 

[orionsbelt]

Thank you all for your answers. I find them very helpful.

Now, after looking carefully at the design report and also the datasheet of the accelerometer, where it is stated that the noise from the power supply shall not be greater than 100uV/sqrt(Hz), I did the following calculation to find out whether any filter or LDO is needed:

Vp-p: approx. 13.80mV
Switching freq: aprrox. 600kHz

Hence, the noise density is: <b>13.80mV/sqrt(600kHz) = 18uV/sqrt(Hz) </b>
This is less than 100uV/sqrt(Hz).

So, in such case, can I conclude that no extra filtering is needed?

 

[FvM]

The calculation is flawed. You can't convert a discrete frequency interference level into spectral density.

I think that well considered power supply filtering should be provided in any case.

 

[orionsbelt]

What would be the correct calculation? I cannot find proper information regarding the calculation online. It would a great help if you could provide me some information/derivation!

 

[FvM]

There's no correct calculation, noise density and level of a periodical signal are different quantities.

The correct question is how the device reacts on periodical interferences in the 600 kHz range and which level can be tolerated.