TPS55288: Description of the OCP_DELAY field in Register 3 of the I2C registers. Just how does current limiting work with this part ?

Hello Peter,

For your three questions:

1. TPS55288 has two current limitations. One is the output current limitation which is sensed by the ISP, ISN pins. The output current limitation value could be set by register 2. The delay time could be set by register 3. The delay time means when a heavy load is applied to TPS55288 Vout, the internal current limit circuit won't work for a certain delay time. After that, TPS55288 works in constant output current mode.

Another one is inductor average/peak current limit which is set by RILIM pin resistors. With a 20kohm RILIM resistance, the inductor average current limit is set to 16A typically. You could set this resistance to proper value to avoid inductor is saturated.

2. You could take a look at figure 8-12 in datasheet. After the delay time (response time), the device starts to regulate the output current not higher than output current limitation value set by register 2.

3. Right. 

Thank you for that response, it all makes sense.

This is why I like this part- the ability to fine tune how it works with the I2C interface.

As you ca see in the scope pictures, when at 12V out a 4A load is applied there is a very brief (sub mS) drop of about .25 Volts. This was also setting the overcurrent flag in the status register. Now that I know what this looks like, I'll try the 3 mS response time and see if the over current protect status bit is set. Note also when the load is removed, which takes the part from 4A load to no load, there is a .25 voltage spike up that ramps back down in about 10 mS. You can see the "stair-step" that corresponds to the frequency of the switching. None of these are a problem, it's a 2% change and just about everything out there that wants 12V has a +- 5% tolerance at a minimum. More typically changes as processors change their current draw are far less that 4A and as such in these cases, regulation should be better than 1%.

w/r/t/ inductor current, I had the room on the PCB for a big inductor, the ones you see are Irms=24A, Isat=35A. The big win with this part is the 6.8 mOhm DCR. Tradeoffs of size and efficiency, and I had a fixed size for the PCB.

For other forum members, with both supplies at 12V and outputting 4 A, at room temperature no part got hotter than 6 degrees C (that was the 10 mOhm current sense resistor). The TPS55288 chip and inductor got about 4 degrees C above ambient. The board passed the "At room temperature, can I put my finger on any part and leave it there for a long time" without any issues. This was a 4 layer PCB, with OSHpark stack-up. The top and bottom layers were 2 oz copper, which performed better than the proto type which was done with OSHpark and as such had only 1 oz copper. The electrolytics are 20 mOhm aluminum organic polymer. 4 input caps close to the FETs are 10 uF 50v, the 7 caps on the output are 22 uF 25V. There are TVS diodes on the battery input, the 3 volt supply (created on board), on both outputs and on all other connectors to the board. Current sense resistors are 3 mOhm with the TI "gain of 100" current sense amps. All analog signals into the uP are RC filtered, typically 1K into .01 uF with the cap up by the uP running a radial ground to the analog ground. There is an external 2.5V reference. This results in a steady measurement that is not affected by the ground noise which is pervasive in any switcher like this. 1% resistors are used for the voltage measurement into the uP, and I'm seeing about a .4 % total measurement accuracy there.

The battery input is heavily filtered first by ferrite and then by the big 600uH 15A/5.3 mOhm common mode choke. Enclosed pix is a sweep of just the filter section on a PCB. This is conducted noise- the spectrum analyzer was connected with the tracking generator on one side, and the input on the other side of the filter section. Not terribly scientific or accurate, but, it shows the impact of the big choke which was not an inexpensive part. Note also the fusing on both the hot and ground, necessary for any vehicle scenario. You don't want your supply and server mother board to be used as the ground path for some high current motor if the vehicle ground was compromised. A measurement of conducted noise at the battery input with the supply off and then on at a 96W load showed that the background noise when up by about 10 dB worst case. Local radio signals made that another "ballpark" measurement. The entire system will undergo a "real" emission test at a lab, I don't foresee any problems. Enclosed is the pix of that test.

I purchased the evaluation module for the TPS55288  (and the accompanying USB interface) and highly recommend that approach for anyone designing with this part. It will let you see if the part can do what you want. You get a PCB placement that will work. You can try different inductors and frequencies. You can change the I2C registers on the fly and see what changes. My first protoboard worked, and changes made were just tweaking for the specific layout. I'm testing the first run of 25 boards now, it all looks good. I suspect a 4 layer board is necessary. the 2nd layer was just a big ground plane. Layer 3 had interconnect and some large traces for ground and power to the switchers. Layer 4 was any signals that could not fit on layer 3 and more and and power to the switchers. I also saw an improvement in performance by using both small vias (10 mil) to connect power and larger hole (38 mil) when bringing power a long distance on the board. I suspect the 38 mil holes had lower resistance and inductance and allowed me to take advantage of the power runs on layers 1, 3 and 4. Note also that a 4 layers board with all the large copper pours does a very good job of distributing heat.

That's everything I can think of that was specific to this part and this particular application (server/PC supply for mobile use). If there is interest, in a few days I can post a pix of the entire server and power supply.