Hello

I've built a circuit around an LTC3111 Buck-Boost Converter chip following Linear's design. I keep burning them out (Vcc and GND become internally shorted) in this configuration and I am not knowledgeable enough to correct my error. My guess is that the voltage divider, the only difference between my implementation and LT's design, is the problem. Can anyone offer insight into this problem?

Thanks


Fig 1: The Buck-Boost Circuit I've constructed (PWM connected to Vcc)



Fig 2: LT's design from their website

 

Hello,

You have connected the PWM/SYNC input wrong.
The maximum voltage at that input is 6 Volts.
Connect it to the VCC pin, see the datasheet for more info:
http://www.linear.com/docs/44189

Bertus

 

Bertus,

Sorry my SPICE schematic is misleading in that way. I've connected PWM to Vcc on my breadboard. I've made a small note of this on Fig. 1. Thank you for your reply.

 

I've added 1M resistors in series with the Vout voltage divider and got the circuit to work. Then I started playing with the pot within that divider and I blew another IC. It seems it is very sensitive to this ratio. I suppose I'll talk to LT tech support on Monday. Thanks again.

 

I've added 1M resistors in series with the Vout voltage divider and got the circuit to work. Then I started playing with the pot within that divider and I blew another IC. It seems it is very sensitive to this ratio. I suppose I'll talk to LT tech support on Monday. Thanks again.

How did you connect the pot? Connecting the pot wrong can ask the circuit to put out voltages that are too high...

 

I believe the wiper is grounded and one of the other terminals is attached to FB. Before the last failure, when it was finally outputting a regulated voltage, it was at around 4v.

 

Nothing jumps out at me, but when you find out the root cause, I'm keenly interested in seeing the solution.

I'm assuming it worked in simulation, and worked at least for a little while on the breadboard? If it failed after being under load for a while, I would worry that the chip was getting hot, and burned out. If it failed quickly under no load, I'd look for voltage transients. I'd make sure that those 0.1 uF capacitors were actually connected, and that the 10 uH inductor on the chip was actually 10 uH. If I had a scope, I'd connect it to the inductor on the breadboard to see what's going on.

 

Placing a pot from FB directly to ground is risky; accidentally adjusting the pot too low could cause the output to exceed maximum ratings.

 

What is the target output voltage?

With the values shown in your posted schematic, the programmed output voltage will be set to 10.267v
0.8*(1+(1M/84.5k))=10.267

 

Mr. Young; Yes simulation results were confirmed before building the circuit. The chip worked for a short while before I started messing with the pot. At that point the chip did start getting hot. I have a sense that transients are not the issue as I have been slowly ramping Vin up with a nice power supply. The inductor is probably not exactly 10uH (I believe it is +/- 30%) but I chose the one which LT specified in the data sheet. The load I'm using is an oscilloscope and my understanding is that it should draw very little current due to high impedance.

SgtWookie; I agree that the ratio changed beyond some limit that I am not well aware of. I've since added 1M in series to closely mimic LT's design but again it failed upon messing with the ratios. My next idea is to add more resistance in series with the FB to GND resistance.

eetech00; Output target is 10.4V which is what the simulation produces. After using that formula to obtain a general value we tweaked said values to obtain our result. Hence the pot. But at this point I'd be happy with any regulated voltage to move forward.

Thanks for all of the replies, they are highly appreciated.

 

Can you run in burst mode until you find out why they parts are frying?

The only thing I can think of is that the change in feedback is causing the loop to not be compensated correctly and producing gain... have you been able to look at the output with a scope to see if there's any oscillations?

You are using a power inductor, right? Not all 10uH inductors are created equal! What is the input voltage, output voltage, and load current? What current is your inductor rated at... and what is the resonant frequency?

 

Hmm...the output side of the device has current limiting. Seems like something on the input side is causing the failure.

 

tindel; I believe it was in burst mode the last time it fried. Before adding the extra 1M resistors into the divider network there were what looked like spikes, not oscillations, on the output. After adding those resistors the output was regulated DC.

The inductor is a power inductor and the particular part number was selected by LT in their design.

(chrome-extension://oemmndcbldboiebfnladdacbdfmadadm/http://product.tdk.com/en/catalog/datasheets/inductor_commercial_power_slf_en.pdf)

At last failure the input voltage was around 10v and the output was around 4v. The current drawn from the PSU was about .2A (but that is just a guess). The inductor is rated up to 1A and I have no idea what the resonant frequency is.

I agree that that the compensation dynamics are affected by my implementation and I do not yet know enough to handle the issue. The LTC3111 data sheet mentions that the values of the resistor network are critical for this very reason, but doesn't say much more than that.

eetech; That is a very interesting observation. What could it be?

 

I agree that that the compensation dynamics are affected by my implementation and I do not yet know enough to handle the issue. The LTC3111 data sheet mentions that the values of the resistor network are critical for this very reason, but doesn't say much more than that.

There are 5 pages in the datasheet dedicated to compensating the feedback loop.

 

Yep, thats pretty embarrassing. But never the less I'm having trouble making sense of it. I don't know very much about power circuitry and loop compensation is a completely new topic. One thing that it mentions is a Thevenin equivalent resistance over 100k. Does that mean the resistor network responsible for the output voltage must have an equivalent resistance above 100k?

It seems that in the most simple of cases, type 1 compensation, I must ensure that the cross over frequency (when gain is unity?)of my resistor network is setup such that it stays away from the resonant frequency. Also it mentions that type 1 compensations has poor transient performance which shouldn't be an issue for me as I am just trying to regulate a steady state DC to power a load cell (is this correct?).

 

It does sound like the parallel combination of your feedback resistors should be above 100kohms.

They give you everything you need in the datasheet to compensate the loop properly. But basic understanding of control loop theory is required. Making this an adjustable output power supply will be difficult. The trim pot is a bad idea at first anyway. Get it working reliably in one condition. Then add functionality.

I can help compensate the loop... But won't be able to help until next week. LTspice is a fantastic tool for doing this.

 

Tindel,

I agree with everything that you've said. At this point I'd be happy for it to output any regulated voltage above 2.5v (as it should) based on LT's design. Also, I'm working on understanding the fundamentals of switching power supplies and this will take me some time. Unfortunately, working with SMD components without a printed board leads to some very crude arrangements which I'm sure are interfering with performance.

I think that the pot, which I've wired as a rheostat, is important for fine tuning. I made the mistake of selecting a 100k pot and that, I've learned, is unsuitable. I appreciate the words of wisdom.

Sd510