What is the minimum input and output capacitance if we use only ceramic caps? Ripple is not too important but board space is. The datasheet only gives examples of certain manufacturers which does not help us. I wish there was a "real" datasheet for SimpleSwitcher chips. Those "recipe sheets" aren't so great.
Regards,
Joerg
Hi Joerg,
LM2674 is internally compensated and the sizing of the output cap will depend on the duty cycle and the inductor value chosen.
For the input capacitance, it is important to put enough capacitance to handle the voltage transient in the input which really depends on the duty cycle. The RMS current of the caps should be chosen around 1/2 the DC inout current to avoid saturation.
You can also use our Webench tool for your particular design which provide input and output capacitor values. www.ti.com on the lower left corner.
Ismail
Hello Ismail,
Thanks. However, I cannot use WebBench because my firewall blocks it. Seems it wants to use some sort of flashplayer routine. I know about the input spike, my supply will be very low impedance, just need to know how much is needed for stability. For example, for most comparable Linear Technology converters it is 1uF if you use ceramic caps.
For the output I'd need to know how well it can regulate out a full load swing to 300mA and back to almost zero without under/over-shooting more than 150mV or so. Normally I do this with SPICE but as far as I know there are no models for your converters.
Based on the output current change you mentioned, you can calculate the change in input current based on your input to output voltage ratio. An input cap value, that is big enough to stabilize the input voltage while supplying the current.
For the output Cap, it is part of the transfer function and you need to choose an LC filter that will guarantee a double pole frequency occuring way below the switching frequency and keeping your system stable with the internal compensation provided on the block diagram.
The advantages of ceramic Caps are size and low ESR. So if you want to use Cermaic caps you will remove the ESR componenet from the transfer function equation.
You mentioned that your load step is 200mA and another 300mA with less than 150mV under/overshoot. What is your input and output volatges? your nominal output current?
No chance to put an LC filter there and we don't really need one. Input is 16-32V and output just a few hundred mA. We can keep it under 300mA but that can swing quite a bit, for example from 100mA to 300mA and back. It's probably best to use a chip with an external comp node, and one I can simulate in SPICE.
Hello Joerg,
I was referring to the LC of the buck. Fr=Sqrt(LC) will be the rolleover Freq. You can simulate the circuit on spice if you use the control section from the block diagram using two Gm amplifiers.
Using webench, with your input range and 5Vout, you will need 10uf on Cout and 1uF on the Cin but you will need 220uH inductor to stabilaize your circuit.
Thanks. That is way too big of an inductor in this case, then we will have to use another chip.
As for Webbench, I was never too enthused about that program. Many times it said that the design can't be done (and it's all in production since years ...), other times I really needed a bit more than just a cookbook recipe. For example optimize load change reactions with phase lead caps and so on. A SPICE simulator is much better for that but one must have models with some stuff in behavioral format in there so simulations don't take forever.
This inductor value was chosen conservatively for 5Vout and 0.5mA. This part is old and we dont have a model for it. you can use our LM2267X family that are better parts and we can create a model for you that you can simulate.
Just give me the exact conditions and I will put a request for you.
Thank you for the offer to request a model, that is very kind. But unfortunately I am nearly out of time, the whole circuit design (with a lot of other stuff) has to be done by Friday. So for this design I had to use another IC where there already is a simulator model.
When you discuss things like this with managers, maybe at a meeting, let them know that SPICE model availability is very important these days. Nearly all my designs are born on the PC now, on the simulator. I see others do the same. With a switcher IC this is not so easy and it is often best to have some functions in there described as behavioral (mathematical). Otherwise the simulations take forever. I had that case with a UC-something switcehr chip from TI last year, simulation runs could easily take half an hour.