A quiet SEPIC negative supply

I needed to add a -3.3V to -6V supply to my RIS-796A 250A pulser, analog version. The PCB didn't have much space, so I implemented a simple inverting switcher with an LM2840 buck converter. See circuit and layout on file U7=LM2840_sch+pcb.jpg in the DropBox folder:

formatting link

There's also a sub-folder with relevant datasheets.

The circuit works OK, although it's a bit fragile, with ICs failing instantly when I miss-probe. The L1 ME3215 inductor has 1R5 resistance, and can deliver over 100mA, more than needed. But it has a horrible 200mV output ripple, forcing me to add an LC output filter stage.

Checkout the other file U7=AP3012_sch.jpg It's my idea for a simple SEPIC inverting converter, using a boost- switcher and a coupled inductor. The inductor feeds a continuous current into the output cap, for low ripple, so maybe the LC filter can be eliminated. Now L1 is an LPD3015, with 2R8 ohms, but it can still deliver 100mA. The coupled inductor is only 3x3mm, and I hope the parts will fit into the same space as the original circuit.

I thought about an LMR64010 for the boost IC, but its current-mode sense resistor is only 0.05 ohms, which doesn't bode well for operation at currents like 20mA. I had a hard time finding modest-current switchers in sot-23 packages, until I discovered the AP3012 by BCD, which has a 0.2-ohm sense resistor. Also, they only cost 32-cents! That's 10x less than the $3.32 LT8330 I also considered. BCD was purchased by Diodes, Inc about 7 years ago. and their parts are well stocked.

--
 Thanks, 
    - Win
Reply to
Winfield Hill
Loading thread data ...

OK, after moving parts around, I actually ended up with a Cuk rather than a SEPIC converter. See schematic: U7=AP3012_neg-Vee_CUK.JPG Cuk converters naturally create negative outputs, but one must add an opamp to provide positive FB-pin feedback. My simplification was to move the converter's ground to the -Vee output: U7=AP3012_neg-Vee_CUK_1. It's slightly weird in that the input inductor L1A gets 20% of its energy from the output rail, but hey! Here's a alternate view: U7=AP3012_neg-Vee_CUK_2.JPG Now to fit it onto the PCB. There's a question of the converter's built-in loop stability, with the 4th-order Cuk configuration, but slowing with a larger inductor should handle that.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

OK, I threw in the towel, the Cuk converter simply had too many parts. Realizing my max -Vee load was 30mA, and that -5V, more or less would be fine, a charge-pump converter was an obvious choice. See how simple it is, schematic and PCB layout in this file in the folder: U7=TPS60403_charge-pump.JPG Some parts have always been off to the left. I called the node -4.5 volts.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

The filtering is appropriate. Charge pumps can be noisy.

RCs make good filters, especially with a lot of uF of low-ESR ceramic or polymer cap. RCs don't resonate!

--
John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
 Click to see the full signature
Reply to
jlarkin

Good point. But the 330uH and 10uF resonate frequency is 2.77kHz, for an inductor reactance of 5.74 ohms. But with the LB3218's DC resistance of 7 ohms, the resonate Q is less than 1. The LC should attenuate well at the 250kHz frequency. The 15-ohm Zout makes a 450mV drop at 30mA, hence the -4.5V label. So this is strictly a low-current solution. Guess I'll play with the Cuk negative supply circuit some time later.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

I'm in the process of going to rev B of this (10-layer!) board.

formatting link

Rev A works, but some circuits bounce other circuits on shared power rails. LC filtering fixes switcher noise problems but can make a rail soft for transient loads, which you may not have.

The new plan is to make the switcher voltages a little high and scatter LDOs locally as needed. I'm enthusiastic about using some LTM8078s in the main supply, which are claimed to be inherently quiet. But still no Spice model!

8078 should do positive to negative, maybe positive to two negative rails, maybe even positive to + and - rails.
--
John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
 Click to see the full signature
Reply to
jlarkin

Do you need 40V in? If you could live with 20 max, an LTM8049 will do that.

-- john, KE5FX

Reply to
John Miles, KE5FX

The box will run off a 24 volt wart. The max negative voltage I need to make is -7.

The 8078 will work, and probably make me two negative voltages from

+24, like -2.5 and -7 maybe.
--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Maybe not perfect, but I was supprised how clean these were.

Cheers

Make America Germless Again

Reply to
Martin Riddle

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.