LT8642S- spread spectrum rejection

I have been testing a new part, the LT8642S. All looks good but I have seen a low frequency signal content on the output ripple.

This is the test circuit:

Looking before and after the output ripple filter I can see the ripple due to switching at 1MHz is 200mVpp at J4 (red) and it is reduced to 1mVpp at J5 (blue). This is good and what I expected.

But if I look on a longer time scale there is a bigger ripple content on the output at about 6mVpp:

On a spectrum it shows up at 3.4kHz, peak 417uVrms.

From LT8642S data sheet the spread spectrum modulation frequency is 3kHz, so I suspect that this is what I am seeing. There are also other harmonics at 10kHz, etc (OK, data sheet says spread spectrum modulation is a triangle wave).

But I would have thought that at such a low frequency the control loop should remove any disturbances on the output introduced by the spread spectrum.

Looking at the controller gain/phase I get 41dB gain at 3.4kHz, so with an output of 417uVrms, that implies the "disturbance" due to the spread spectrum is 47mV.

Turning off the spread spectrum the low frequency signals drop, and get a peak at the switching frequency, 1.033MHz:


So it does look like what I am seeing on the output is due to the spread spectrum, and I need to use the SS to keep the HF content low. We tested LT8640 demo board and got similar issues.

Hence the question is; if there is any way to remove the LF content at 3kHz. e.g. perhaps there is a way to increase the controller gain at this frequency? Is there anything that can be done the reduce the low frequency ripple with the LT8642S? Thanks.

  • 0
    •  Analog Employees 
    on Nov 20, 2018 7:48 PM


    In this schematics, the LT8642S output capacitors has 1Ohm added ESR (R29A and R29). Those large ESR cause a couple hundred mV output voltage ripple at J4, since the inductor current ripple is more than 2A. With only low ESR ceramic capacitors, the ripple at 1.1MHz switching frequency should significantly decrease.

    The low frequency ripple is caused by the spread spectrum modulation frequency, which is 3kHz. Apparently, 1.1MHz ripple is easier to filter than 3kHz ripple. So, I’ll suggest to use forced continuous mode, with low ESR high value output capacitance, and adding output ripple filter.

    Another option is to add a LDO at the output side, because LDO can offer “quiet” power. With LDO, the spread spectrum mode is OK since LDO usually has high PSRR at low frequency.