Hi,

I tried to modify a +15V/-15V buck/boost module to produce around +10V/-10V output voltage. The operation succeeded but the patient died. I want to try again with a +12V/-12V module that I have. Before I do I'd like to know if there's anything obviously wrong with the method or what might have cause one of the chips to burn.

The modules have a CN3903 for the positive output and an MP4560 for the negative output. Each has a resistor divider network that goes to the feedback pin to determine the output. I got the MP4560 working at around -10V, then modified the CN3903 network. Proceeding to test the CN3903 output, the MP4560 started to burn, see "15V bottom.png" of the modified module. Also see the partial schematic of the unmodified module. It's missing the diode and inductor that must be there in each output and some other stuff. It's difficult to get a complete picture.

I think it should in principle be possible with reasonably sized resistors between 10k - 100k. As you can see the module sat on unsoldered pins on a breadboard receiving 12V in. Now I suspect that the third inductor is part of a pi-filter to clean up the input voltage. Could it be that intermittent connection of the input supply voltage may have caused inductive spikes in the pi-filter that burnt the MP4560?

 

It is tempting to assume you can do things like this but switching regulators do not operate like their linear cousins. They are inherently non-linear. Depending on the input voltage and the conversion ratios which adjust the duty cycle of an oscillator you cannot be certain that the adjustments you make are achievable without additional modifications.

If you want modules for specific outputs, then get ones designed for that purpose. It is just not worth your time screwing around and potentially letting out the magic smoke.

 

Going from 15V to 12V is not a big step. I would make the modification.

 

Going from 15V to 12V is not a big step. I would make the modification.

It is a buck-boost topology. That is NOT +15 to +12, rather is +15 to -12V. The nominal implied duty cycle is 12/27 ≈ 0.44. This is close to the 0.5 level where the output can go less than -15V which would be a bad thing. I would not be comfortable making these kinds of mods on something this sensitive.

The formula for this topology is:

\( V_o\;=\;-V_s\left(\cfrac{D}{1-D}\right) \)

Change the duty cycle from 0.44 to 0.52 and the 15V supply gives -16.25V on the output. Similarly, if we drop the duty cycle D to 0.36 the output drops to 8.44 Volts. That's quite a range for just a bit more than ±18% change

 

Going from 15V to 12V is not a big step. I would make the modification.

I was going from 15V to 10.5V-ish, now I'll try 12V to 10.5V-ish. I'll just try the mod again.

One thought I had is that after poking at it with a hot air tool I probably should wait longer before I test the circuit. Not waiting long enough may have killed it.

 

I think it should in principle be possible with reasonably sized resistors between 10k - 100k.

For change V_out not only R2 should be changed, but at least change L1 value and optimize C_ff and R3:

 

MP4560 does not say "current mode" but it does say the PWM cycle ends when the current reached a level set by the error amp. So, I think it is current mode, which is forgiving of changes in the line and load changes.

The CN3903 is a constant on time PWM so the frequency will vary with input and output voltage.

I am certain I can find PWM designs with output voltage of 1 to 25V set by a knob.

When I first designed offline power supplies, we made 120V and 240V supplies. Later we added a switch 120/240. When we moved over to "current mode" PWMs I made 120-240 supplies. The problem is Japan had 100V 50hz lines East/ 60hz west. The slower 50hz caused the PWM to see large ripple voltage so the PWM needed to work down to 80V. On the other side the 240 could easily be 230V. Remember there is probably 20 or 30V of ripple on the voltage the PWM sees. In India the voltage could be almost anything. We wanted the PWM to work from 80 to 250V.