DC Step Up Converter Problem

You are on the right path when thinking about start-up in-rush current. 15V at 200mA is 3 watts, or about 1A at 3.3 volts, considering inefficiencies. It's likely your 3.3V supply can provide the extra amp, in the long run, but it apparently has trouble with a fast 1A load step.

The TPS61085 can run at 1.2MHz and apparently is a very fast boost regulator, capable of demanding some serious input current. But I think you have a simple answer in the soft-start function TI provides with the '61085. They show a 0.1uF C_SS capacitor slowing the IC's current-limit to about 2ms, which should be slow enough for your 3.3V supply. Shown at right, figure 7 in the datasheet.

Some switching ICs don't have soft-start pins, and special tricks are required to make them do a soft start. Some designs allow for this on the COMP pin, but apparently not your '61085. But any IC that has a pair of feedback resistors to set the output voltage, must bend to a second trick: slow down the voltage rise with a capacitor across the upper feedback resistor.

Now, I wonder if I've misunderstood your problem. It's not simply during initial startup of the '61085 that your 3.3V supply sags, but whenever you use the enable input EN, and are you going to tell us that the soft-start function doesn't work at this point (although the datasheet text implies that it does)? If so, you have several choices. One is to use the second trick mentioned above.

Another choice would be to force the soft-start to operate by using a small mosfet, etc., to ground the SS pin when the '61085 is disabled. Then the pin's 10uA pullup current would create a ramp programming the maximum input-current limit.

Your 3.3V supply is only good for 5W. The L.E.D.s are consuming 3W. That only leaves 2W for your LAN card. You may be current limiting when the L.E.D.s switch on. How much power does the LAN card take?

Hi Win,

You are very knowledgable. I believe your suggestion is to hook up the SS and EN pins of TPS61085 to the LED PWM driving signal to output the 15V whenever the LEDs are turned on (right now EN is tied to 3.3V and SS is tied to 100nF). I will try that asap. Some more details: the 3.3Vnever sags as long as LEDs are off. I chose to use 1.2MHz because the TPS61085 seemsto fail after sometime (no 15Vout) or after a few power on/off at 650KHz. But everything works fine after I replaced the chip which leads me to suspect the transient response of the 3.3V power is not slow enough and the switched current limit of 2A for the internal mosfet is exceeded. I have no problem with 1.2MHz.

I've calculated the total 15V capacitance of 1.63mF is sufficient to temporary provide about 80mA of current for about 40msec of PWM pulse width (120KHz carrier within 40msec turn on follow by 100msec off). Thgerefore this should reduce the current requirement from the 3.3V to 120mA. This is to reduce the 3.3V current drain. But I've not reckon with the dc converter. This dc converter is switching continously during the 40msec and this is bthe cause of voltage sag. There is no sag at 3.3V rail when LEDs are off. Is there any other way to permanently enable the dc converter (for another project) yet tweak the dc converter into not switching continously during the 40msec? One way is to use mosfet to control the SS pin during 40msec. Another way you've mentioned is to slow the voltage feedback signal with a cap. I think I will try the EN=HIGH with the SS and FB trick.

If all else fails, is there some current limiting circuit that I can place between 3.3V and the inductor to limit the current and remove the voltage sag? Can this method also work? Anyone has some cookbook circuit for this current limiter?

Thanks in advance.

Whoa! You're already using the 0.1uF C_SS cap? This means you already have a slow current risetime that the 3.3V SMPS control loop should be able to handle. So you can toss all my remarks about startup currents out the window.

It appears you are simply drawing too much current, I calculated 1A, from this wimpy supply. I noticed that LG_Dave remarked that the 3.3V supply was only good for 5 watts total. I'm not sure where he got that piece of info, but it's very significant. There's an old saying, that you can't squeeze blood from a turnip. If you get my drift.

Thanks Win,

No. I am using 10nF for Css now. So I wonder if what I mentioned in the previous post can be applied to reduce voltage sag during LED turn on.

Best Regards

Sorry my mistake.

I am at present using 100nF for Css. But in my previous post, I mentioned using mosfet to discharge Css during LED turn on together with the cap across upper resistor in the FB voltage divider to trick the dc converter into current limiting the switching current during LED turn on cause the dc converter is switching continously and I can't made it stop norcontrol the switching frequency/duty cycle.

Do you think this will work to reduce 3.3V sag if I have sufficient 15V buffer capacitance for the LED switching?

Hi All,

If this still fail, can I current limit the inductor current during dc converter switching? Does anyone has any circuit for that?

Thanks and Best Regards

Wish you a Happy New Year

Winfield,

I got the 5W from the first paragraph of Anonymous' post:

The 3.3V SMPS is able to source up to 1.5A

Dave

Hi Dave,

Can you comment if my above post is feasible?

Best Regards

Anonymous,

I think you should determine first if you are at the capacity of your supply. At 15V and 200 ma for the L.E.D.'s, I assume you have 5 in series. If so, put some resistance in series with the L.E.D.'s and get the current down to 100 ma or less and see if the 3.3V rail still droops. If that stops or reduces the droop, that indicates a need for more current capacity on the 3.3V rail, or less current for the L.E.D.'s.

Dave

Excellent. Maybe someday I'll learn to read!

Please explain to a non-engineer: why not just provide a proper independent 15 V supply?

I am surprised that you're using a DC/DC converter rather than a dedicated LED driver.

How are you regulating the LED current?

Where does the figure of 200mA come from:-
10 strings of ordinary LEDs?
1 string of under driven high power LEDs?

Hi All,

Thanks. I think I can solve this problem with using 2 rectifier diode/capacitor branches in parallel from the 3.3V secondary. One to 3.3VDD for digital circuit, the other to 3.3V of dc converter. The diode will prevent voltage sag from affect digital circuit.

P.S. I've to use 3.3V cause this is suppose to be an add-on module and have to tap from existing power.

Thanks!