We have been working on a boost regulator circuit using the LTC3115-1 part from Linear

http://cds.linear.com/docs/en/datasheet/31151fb.pdf


We input +12V and output +38V. It outputs around 500mA peak current with each pulse
lasting around 1ms.

As we increase the current from 0A to 500mA, we find we are getting noise on our input.
It is a repeating wavefom (almost sinusoidal) of 1V pk pk at a frequency of 1.7KHz. Our
output voltage does not show the noise and is perfectly stable. We have 47uF ceramic
cap on our input and 400uF of electrolytic cap on output with a 1uF ceramic as well.

The regulator is powered from another PCB via a header and supplied. That board is
supplied from a bench top PSU. We have gone in to great detail to make sure the compensation
network is calculated as per the LTC3115 simulation model. It is stable with a phase margin of 60deg
and has a bandwidth of 5KHz.The switching frequency is 754KHz. We used a type 3 comp circuit.

Assuming our layout is correct(8 layer board solid 0V planes etc…), which it appears to be, can anyone think of a reason
why the input of the regulator is going bananas like this? I had thought that the spikey current draw was exciting
a resonance in my input supply but reducing or increasing the input cap didn’t make any difference. I put 2200uF
on the input and it only helped a tiny bit.


Do you think a snubber circuit might help?

 

It sounds like there's a problem with your +12V supply.
With a proper supply the input voltage should not vary significantly.
Do you see this voltage variation at the bench supply terminals?

So even with 2200uF on the input you are still seeing 1Vpp at 1kHz?
That's about 12A peak current!!
Is there any inductance anywhere in the input circuit?