LTC3129 DC1922A Feedforward Network

On the DC1922A board, a 2.2k resistor was added to the feedforward network (which was not discussed in the LTC3129 datasheet). How do I decide if I need to add the resistor, and what value should be chosen? Is looking at transient step response in a LTSpice simulation a good way, or should I do an AC simulation to look at the gain and phase margins (seems difficult)?

I'm needing to convert li-ion voltages to 3.5V, at up to 75 mA. I plan to set PWM mode during sensitive data acquisitions to keep the output voltage more steady (though I'm not sure if this is 100% necessary, will test). Based on the datasheet, I should use the feedforward cap to help stability at low output powers.

We are ordering the DC1922A board, and will use it to power our circuit, but I would be grateful for any insights you can provide.

Schematic with feedforward resistor


  • +1
    •  Analog Employees 
    on Aug 14, 2019 10:09 PM


    A feedforward network as mentioned in the datasheet is recommended to reduce the output voltage ripple in burst mode. However it is optional and you do not have to add if your application doesn’t require it. The part would still be stable and work fine as expected. A resistor was also included in the demo board for the feedforward network to reduce the high frequency noise. So the resistor value should be chosen such that it reduces the gain at very higher frequency thus reducing the high freq noise.  This feedforward network also improves the phase margin and is might be required for light load applications in PWM mode.

    As the load for your application is 75mA you should probably not require the feedforward Cap and Res. I would recommend you to use the demo-board as the reference design for your application. You can also keep the option of feedforward network in your circuit and place the components only when needed. 


  • Your reply makes sense. I did see a reduction of the HF noise in a LTSpice simulation (with the extra resistor), but wasn't sure if it would be a simulation artifact or not (due to spice not modeling board parasitics). The RC combination gives a cutoff frequency of 2.2 MHz, which is about double the switching frequency (1.2 MHz). In my design, I'll plan to keep the RC product equal to that of the evaluation board.

    I failed to mention that 75 mA is the max load. Generally the product will be sleeping, where the converter will supply a few microamps. Based on this, I'm expecting that the FF capacitor will be necessary for stability as mentioned in the datasheet.

  • +1
    •  Analog Employees 
    on Aug 16, 2019 3:10 PM in reply to nconrad

    Hi , 

    If the application will have very light load most of the time then I would suggest you to run in BURST mode. As you can compare the efficiency in the datasheet(Page4 and 5), BURST mode will have way better efficiency than in PWM unless you do not burst ripples in the output voltage.

    It is good to keep option of feed forward Cap and Res in your design and place the components only when you see it is required during the test.