Mosfet Gate Capacitor?

[mneary]

Just a quick correction: Inductors don't come in uF (microfarads). It's microhenries (uH.)

 

[MrAl]

Hi again,

I agree that 5uH is way too small, and so is 15uH really. For 50kHz 100uH is
more typical, but i've used 50uH many times with good results. It has a lot
to do with input to output differential too.

Yeah that must have been a simple oversight...stating uf instead of uH he he.
I think he went with the inductor stated in previous posts because they are
on sale right now for about 2 dollars, while the others seem to be over 4 dollars
each and are rated for less current with the higher inductance.
BTW, two 39uH inductors in series gives 78uH but that would cost 4 dollars
too That would provide for a 78uH 11 amp inductor though...not too shabby.

 

[BrownOut]

Just a quick correction: Inductors don't come in uF (microfarads). It's microhenries (uH.)

I think you were influenced by the OP's drawing, which shows a 33uf inductor.

 

[nickelflippr]

Dunce cap award on inductor values, uH not uf.

Actually the current **broken link removed** output cap is probably doing the job at 100khz. It was picked for a previous small switcher project based on the L4972A chip. It has a low esr 0.054Ω and ripple of 1.26A. For the previously mentioned Rubycon of 470uf/50V and 1000uf/50V the specs are .027Ω/2.05A and .021Ω/3.01A respectively.

As far as amps were concerned, was originally thinking a 10A input from solar. As it is, probably won't be getting any more solar power, so lets stick with the 4.4A for now. Not sure why I would need a current limit, the 12V lead acid battery will just stop accepting current, voltage goes up, and need to prevent over voltage. The blocking diode would protect from hooking up backwards.

The current ripple voltage "appears" to be around 100-150mv, and would seem acceptable. Possibly willing to go for less.

Hall sensor problem solved. AN7 wasn't working, so put it on AN1 of the PIC micro. This was a software problem.

Sounds like the consensus is to leave off the 5uH and 15uH inductors from the shopping list. Have added another 33uH, because they are cheap like me. Don't know where I will find the room on the board.

Thanks again for everyone's suggestions. Still need to update schematic and work on a charging algorithm.

Initial and rough power calculations had this around 85% efficient, does that sound right? Does it make sense to change this into a synchronous buck regulator?

 

[mneary]

If you keep the frequency at 100kHz, the extra inductor shouldn't be needed.

The blocking diode has to prevent the battery from supplying the panel. Draw the circuit, including body diodes, and confirm there is no leakage path. If the loss in the blocking diode concern you, use a MOSFET. If you have the freedom to put the blocking in the (-) line, it can be N-channel.

As you know, synchronous rectification only eliminates the losses in the freewheeling diode. Calculate these losses and judge for yourself whether the added parts are worth the trouble. It would be an N-channel, so it's cheap.

 

[RCinFLA]

From your original schematic it looks like you have a simulation program. You can determine the values from it. Do think about sweeps in conditions that may apply.

The coil charge L*di/dt supports 5 volt drop. Coil discharge supports L* di/dt of 12.5 vdc (added some for negative going diode).

What is not likely in your simulation is satuaration characteristics on your inductor. From your sim get the peak coil current and check it against specs on the coil.

Higher inductance puts the switcher in continuous conduction mode. This is okay as long as coil core can handle it. In continuous conduction mode the coil is not allowed to fully discharge its field through diode before the next MOSFET charge cycle starts. Duty cycle of switcher would be roughly 75%. There is little to no ringing in this mode. Ringing occurs in discontinuous mode when the diode conduction ceases at its diode drop voltage but there is still a small amount of energy left in the coil with nowhere to go.

If you want to better represent the PV panel in your sim replace the 17v voltage source with a variable current source in parallel with 34 diodes in series. You can add a little series resistance between this PV model and your circuit to represent the Rs of the PV panel. Varying the current source represents illumination level on the panel. The main thing is to see how your circuit reacts when illumination drops and the source become a high impedance current source. The input cap has to provide the switcher current peaks.

 

[MrAl]

Dunce cap award on inductor values, uH not uf.

Actually the current **broken link removed** output cap is probably doing the job at 100khz. It was picked for a previous small switcher project based on the L4972A chip. It has a low esr 0.054Ω and ripple of 1.26A. For the previously mentioned Rubycon of 470uf/50V and 1000uf/50V the specs are .027Ω/2.05A and .021Ω/3.01A respectively.

As far as amps were concerned, was originally thinking a 10A input from solar. As it is, probably won't be getting any more solar power, so lets stick with the 4.4A for now. Not sure why I would need a current limit, the 12V lead acid battery will just stop accepting current, voltage goes up, and need to prevent over voltage. The blocking diode would protect from hooking up backwards.

The current ripple voltage "appears" to be around 100-150mv, and would seem acceptable. Possibly willing to go for less.

Hall sensor problem solved. AN7 wasn't working, so put it on AN1 of the PIC micro. This was a software problem.

Sounds like the consensus is to leave off the 5uH and 15uH inductors from the shopping list. Have added another 33uH, because they are cheap like me. Don't know where I will find the room on the board.

Thanks again for everyone's suggestions. Still need to update schematic and work on a charging algorithm.

Initial and rough power calculations had this around 85% efficient, does that sound right? Does it make sense to change this into a synchronous buck regulator?

Hi again,


Your choice of capacitor sounds good at 100kHz. I agree that you probably
wont need an extra inductor at that frequency.

85 percent sounds normal yes.

I'd like to hear who selected your initial inductor choice, because the sat
rating is very good. Because the input can be as high as 4.4 amps and
the input is as high as 21v (rounding off) and output 12v, the output current
(in a buck circuit like this one) could be higher than the input current with
the same ratio of input to output voltage, so that means the output current
can be as high as 7.7 amps average, and the peak of say 1 amp added
to that means the max output current at any time can be close to 9 amps.
Thus, 11 amps is a good choice for the inductor rating so i wonder who
selected that now Did you? If so, good choice

The efficiency without a diode would be slightly greater at least theoretically
but this is more of an issue with very low output voltages. At 12v output
the difference would be the difference between 12v output at say 7 amps
and 12.5v output at 7 amps, times an approximate factor of 0.5 for a
50 percent duty cycle, which comes out to about 2 percent, which isnt
a huge difference for the trouble of going to a synchronous converter
with extra transistor. The diode in the non sync converter of course should
be a Schottky type or similar low voltage drop type.

Let us know how this project works out ok...

 

[nickelflippr]

Hi MrAl et al,

Got around to updating the schematic. Comments welcome. Have started a LTspice schematic, but hung up with some of the details, as this is a first go for me. May have to ask for some help on how to do some of the models and simulation. Have the info plugged in for the inductor, diode and caps. The input and output caps are based on those arriving from the Newark order, so not in circuit yet, later this week hopefully.

The inductor was a lucky guess, as described in earlier post, with the idea that it was the highest readily available value at the rated output. The buck converter idea is from a schematic detailed by the Microchip dsPICDEM SMPS Buck Development Board. Other resources are Microchip SMPS appnotes AN114 and AN1207.

Considering that zero amps are produced at the open circuit voltage of the solar panel (21V), the peak amps you describe seems high. If the rated voltage of 17V is used, then maybe only 5.5 amps peak or so?

Anxious to hook up the battery, but must wait to get feedback loops and failsafes figured out. This will take some time, but will keep progress updates.

 

[MrAl]

Hi again,


Well with 17v input and 4.4 amps from that supply the output can go to approx
4.4*17/12 amps which is about 6.2 amps.

I guess you dont need current limit if your battery can take the full output current.
I wasnt sure what size battery you were using.

I guess the battery will be hard wired into the system too? No possibility of
reverse connection there then.

It sounds like your project is coming along nicely and i cant wait to hear how it
works out after you get all the parts.

 

[mneary]

Make a note that the SFP9540 might not handle your power without a heat sink (5A in 0.2 ohms = 5W). Consider two or three in parallel instead. Each should have its own gate resistor. This sounds extravagant, but in addition to saving power it ends up a lot smaller than a heat sink. (5A in 0.067 ohms is 1.67W total shared by three devices, or 0.56W each.)

With 4.4A from the panel, the input SB560 is approaching its 5A rating. While not a big deal reliability wise, again you might consider that it'll also be at the top end of its voltage drop at 0.7V. This is more important than the freewheeling diode because it's in the circuit at 100% duty cycle. Another diode in parallel would reduce the voltage drop and save a little power. A MOSFET switch it could save even more power. A TO-220 standing up is about the same footprint as the DO-201A diode package.

 

[nickelflippr]

Hi and back again,

The battery will be size 31 deep cycle marine. It is generously rated at 100ah, but wouldn't bet on it. The shape of the battery is questionable, we'll see.

The solar and battery connections will either be by 10A rated pcb screw terminals (current setup) or equally rated polarized pcb header/socket combo. The idea of fool proofing the connections or protection scheme is attractive, because I'm always scratching a loose wire where it shouldn't

Good points about the SFP9540. It was the last type of power PFET in the bin, because I sent two of the IRLR5305's to the graveyard. The Microchip Buck demo board uses the 5305's with an 18volt zener from the gate to the source to clamp any spikes I guess. Did not have a 18V zener, and paid the price.

The SFP9540 is very sturdy with Vds of -100V, Vgs of 30V, Id of -17A and a lousy Rdson of 0.2Ω. The IRLR5305 has o.k Vds of -55V, lousy Vgs of 20V, Id of -31A, and a good Rdson of 0.065Ω. Will make another stab with the IRLR5305, this time with the 18V gate zener, when the parts order comes in. So would rather change chips on the high side switch, than say double or triple up the SFP9540.

Have a new schematic, with the low side power on/off switch suggestion, rather than the schottky diode. Thanks, this will help out a lot. I see running this off a micro or LM339 comparator, with the reference dialed into a low dropout input voltage of say 15-16V? The other half of the LM339 could be run as a low dropout current (say a few hundred milliamps) based off the hall effect sensor reading. The two comparators could be OR'd together with some signal diodes to the on/off power switch driver.

The on/off power switch could be a IRLR7807Z or other low Rdson nfet. The Vds of 30V doesn't look so good. The other specs are Vgs of 20V, Id of 43, and a nice low Rdson of 0.014Ω. The Microchip TC4428A is a dual pfet/nfet driver.

 

[mneary]

The divider for AIN3 needs to be to the right of the IRLR7807Z because yout panel (-) is no longer circuit GND. If the IRLR7807Z is turned on, these points are essentially the same.

I wouldn't worry too much about the 30V IRLR7807Z, it would only have (Vin + Vbatt) if you actually hook it up wrong.

You don't need to be exactly 18V on the zener. It needs to be less than Vgs.max and higher than Vgs where you've achieved adequately low Rds-on. I can't seem to find IRLR5305; who makes it?

 

[nickelflippr]

Huh, you are right about AN3 and the voltage divider. That's a challenge, since I was hoping to use an analog comparator (LM393) as a go-no-go, based on the panel voltage, and independent of the micro. Could blip the low side switch on with the micro, take an a-d reading, then decide. That sort of defeats the purpose of the comparator though. I could be barking up the wrong tree; how to reconcile the two different potentials?

Still working on the schematic. It will show how the comparators are worked in to share the voltage divider and current sensor.

Arrgghhh, my mistake, it's an irfr5305 by International Rectifier. Irlr7807z and Irfr5305 datasheets.

Smaller zeners would do, but the only ones I had were like 5.1V, not gonna work. The 18V zeners should be here any minute.

 

[mneary]

Would the open-circuit voltage of the panel give you useful information?

You could put the divider across the IRLR7807Z. It shouldn't upset anything if the value is high enough, and it gives (Vpanel - Vbat). Your controller should be able to do the math. You'll want a protection diode on AIN3 to block negative inputs when Vpanel < Vbat.

 

[nickelflippr]

Was thinking that the open circuit voltage could be an indicator of when to wake up the circuit in the morning, and shut down during the evening. Sort of like an LDR circuit.

Well with the LM393 comparator (that's to be added), you only get Vdd-2V on the output, so it looks like a LM317 enters the mix to power it, as the mosfet driver needs minimum 2.4V. Could run the LM317 and LM393 combo off the solar panel side of the switch. The logic would be if the panel isn't hooked up, or very low light level then no circuit activation. If circuit activated, and not enough current flowing after micro starts PWM: shutdown micro, wait a bit, then try again.

The AN3 input may not provide any useful info, other than debugging a good startup voltage.