Low ripple low noise small power supply design

[BrianHG]

 

[VEGETA]

at last we can go for it?

gonna send you the panel just in case xD 

when I start selling them, I will surely give you a cut. I insist and you deserve it.

[BrianHG]

at last we can go for it?

gonna send you the panel just in case xD 

when I start selling them, I will surely give you a cut. I insist and you deserve it.

Yes, go for it.
Relax, my help is free...

[VEGETA]

at last we can go for it?

gonna send you the panel just in case xD 

when I start selling them, I will surely give you a cut. I insist and you deserve it.

Yes, go for it.
Relax, my help is free...

Now it is in your PM.

I know you helped for free but I also like to give rights to those who deserve it. We'll wait and see what happens.

[VEGETA]

The KIA mosfet doesn't have any charts showing the Vgs curve.

Try this one : https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KND3403A_C382143.html

Or you can trust the one you picked or find a data sheet with the Vgs curve table shown.

Why did you pick this particular mosfet? I need to know what is in the Vgs curve we need to look for.

gate voltage is 3.3 and 5v, so I assume it should be in the linear region at that voltage in order to regulate?

I have ordered the PCBs but still going to need to order the top board components... this is why I need to know the MOSFETs.

[VEGETA]

I realize I made a mistake in picking the inductor!

the inductor footprint is 7.3x7.3 but the one in the BOM (and paid for) is 12.5x12.5! it is the 10uH for both switchers and one in the top side.

I am searching for replacements but none are with enough current!


Edit:

I found these replacements:

3.3uH @ 4.2A current: https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YPRH0704-3R3M_C516316.html
2.2uH @ 6A current   : https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YPRH0704-2R2M_C514805.html
1.0uH @ 9A current   : https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YPRH0704-1R0M_C338840.html >> I presume this is too low of inductance.

I sent an email to their support telling them to change it to this:

https://lcsc.com/product-detail/Power-Inductors_Sunltech-Tech-SLH0704S2R2MTT_C216193.html
which is 2.2uH @ 6.2A since it was the only thing available.

this one is slightly bigger at 7.4x7.4mm:

https://lcsc.com/product-detail/Power-Inductors_Gotrend-Tech-GSDRK74P-4R7M_C142420.html

but it is 4.7uH @ 5A.


____

so picking the 2.2uH inductor here is some datasheet calculations:

ripple current = [V_out * (1-V_out/V_in)]/(frequency*inductance)

which equals to 0.680A if you plug 2MHz frequency and 2.2uH inductance with 6.24v and 12v.

inductor chosen can easily handle the maximum current but I am not really sure about the ripple in final output... both voltage and current ripple?!

notice that nothing we can do now since the boards are already paid for. Keep changing the thing made us not aware of the change in footprints.

[BrianHG]

The KIA mosfet doesn't have any charts showing the Vgs curve.

Try this one : https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KND3403A_C382143.html

Or you can trust the one you picked or find a data sheet with the Vgs curve table shown.

Why did you pick this particular mosfet? I need to know what is in the Vgs curve we need to look for.

gate voltage is 3.3 and 5v, so I assume it should be in the linear region at that voltage in order to regulate?

I have ordered the PCBs but still going to need to order the top board components... this is why I need to know the MOSFETs.

It's the avalanche curve at ~2v, almost completely just at 3v at 100amps for that mosfet.  It is a low voltage logic level mosfet.  The cheaper mosfet you chose has no spec in the data sheet and I assume it is a cheap clone of:
IPD50N03 https://lcsc.com/product-detail/MOSFET_Infineon-Technologies-IPD50N03S2-07_C536767.html
or STD50N03.

Compare the voltage difference between gate and source when drawing ~5-10amps.

This tells you how many more volts the opamp needs to send to the gate compared to your desired output voltage when drawing ~5amps.  Note it may be a little higher since we are operating so close to the voltage at the Drain.

[BrianHG]

I realize I made a mistake in picking the inductor!

the inductor footprint is 7.3x7.3 but the one in the BOM (and paid for) is 12.5x12.5! it is the 10uH for both switchers and one in the top side.

I am searching for replacements but none are with enough current!


Edit:

I found these replacements:

3.3uH @ 4.2A current: https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YPRH0704-3R3M_C516316.html
2.2uH @ 6A current   : https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YPRH0704-2R2M_C514805.html
1.0uH @ 9A current   : https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YPRH0704-1R0M_C338840.html >> I presume this is too low of inductance.

I sent an email to their support telling them to change it to this:

https://lcsc.com/product-detail/Power-Inductors_Sunltech-Tech-SLH0704S2R2MTT_C216193.html
which is 2.2uH @ 6.2A since it was the only thing available.

this one is slightly bigger at 7.4x7.4mm:

https://lcsc.com/product-detail/Power-Inductors_Gotrend-Tech-GSDRK74P-4R7M_C142420.html

but it is 4.7uH @ 5A.


____

so picking the 2.2uH inductor here is some datasheet calculations:

ripple current = [V_out * (1-V_out/V_in)]/(frequency*inductance)

which equals to 0.680A if you plug 2MHz frequency and 2.2uH inductance with 6.24v and 12v.

inductor chosen can easily handle the maximum current but I am not really sure about the ripple in final output... both voltage and current ripple?!

notice that nothing we can do now since the boards are already paid for. Keep changing the thing made us not aware of the change in footprints.

Check these out.  Lower resistance, higher current, the better:
https://lcsc.com/product-detail/Power-Inductors_TAI-TECH-TMPA0605SV-100MN-D_C305195.html
https://lcsc.com/product-detail/_Chilisin-Elec-HPPC06030-100M-Q8_C329517.html
https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YSPI0750-100M_C718402.html
https://lcsc.com/product-detail/Power-Inductors_SXN-Shun-Xiang-Nuo-Elec-SMMS0650-100M_C149584.html

[VEGETA]

Check these out.  Lower resistance, higher current, the better:
https://lcsc.com/product-detail/Power-Inductors_TAI-TECH-TMPA0605SV-100MN-D_C305195.html
https://lcsc.com/product-detail/_Chilisin-Elec-HPPC06030-100M-Q8_C329517.html
https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YSPI0750-100M_C718402.html
https://lcsc.com/product-detail/Power-Inductors_SXN-Shun-Xiang-Nuo-Elec-SMMS0650-100M_C149584.html

these are 7.1*6.6mm but our footprint is 7.3x7.3mm.. will these be ok? even in smt assembly machine? I can solder them by hand though.

however, as a worst case scenario... is using that 2.2uH will cause harm to our final output?

mosfet choice

well, we will have a Vgs of 3.3v and 5v, this mosfet (https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KND3403A_C382143.html) has high current at these VGS values to be fully on since it is a very low voltage logic level mosfet.

with our drain-to-source (Vds) voltage being about 2.5v~3v, it shows +60A on Vds vs I curve.

can you explain a bit more since I find it kinda acting like a fully on switch at gate voltage of 3.3v and 5v, not in Linear region which what regulation is all about.

[BrianHG]

Check these out.  Lower resistance, higher current, the better:
https://lcsc.com/product-detail/Power-Inductors_TAI-TECH-TMPA0605SV-100MN-D_C305195.html
https://lcsc.com/product-detail/_Chilisin-Elec-HPPC06030-100M-Q8_C329517.html
https://lcsc.com/product-detail/Power-Inductors_YJYCOIN-YSPI0750-100M_C718402.html
https://lcsc.com/product-detail/Power-Inductors_SXN-Shun-Xiang-Nuo-Elec-SMMS0650-100M_C149584.html

these are 7.1*6.6mm but our footprint is 7.3x7.3mm.. will these be ok? even in smt assembly machine? I can solder them by hand though.

Check if the pads will solder on from the cad drawings.
If not, enlarge your solder pads so you support more inductors in the same space.
I'd say make sure you can place all 4 of the ones I listed so you have plenty of stock to choose from.

however, as a worst case scenario... is using that 2.2uH will cause harm to our final output?

mosfet choice

well, we will have a Vgs of 3.3v and 5v, this mosfet (https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KND3403A_C382143.html) has high current at these VGS values to be fully on since it is a very low voltage logic level mosfet.

with our drain-to-source (Vds) voltage being about 2.5v~3v, it shows +60A on Vds vs I curve.

can you explain a bit more since I find it kinda acting like a fully on switch at gate voltage of 3.3v and 5v, not in Linear region which what regulation is all about.

See example mosfet charts below...
Now, with the 5v out mosfet drawing 3 amps, can the op-amp give enough voltage to the gate of the second mosfet?

[VEGETA]

Check if the pads will solder on from the cad drawings.

yes it should fit as seen in attachment. the 7.3 full length of the inductor should cover the pads easily... thus, it is actually better to have 6.6mm width to save a bit of room.

I will make a slight modification in the next batch to make the solder pads a bit longer inside so it can fit smaller footprints too. this batch is already sent and paid for,  no surrender!!

See example mosfet charts below...
Now, with the 5v out mosfet drawing 3 amps, can the op-amp give enough voltage to the gate of the second mosfet?

here is what I think:

our op-amp takes the clean filtered 12v as power. so it can drive output voltage up to say 10v since it is not rail to rail or so good. that is why first mosfet is totally ok since all we need is 2 Vgs to achieve 5v output @ 3 amps maximum. that means 5v + 2v = 7 volts from op-amp... no worries.

as for 2nd mosfet... we need 4.5v Vgs, means 4.5v+5v = 9.5v from opamp... I say yes we can but kinda tight tolerance.

so first mosfet is the best choice.. and since I will buy components myself, then I will get other mosfets too to try them and see.

[BrianHG]

here is what I think:

our op-amp takes the clean filtered 12v as power. so it can drive output voltage up to say 10v since it is not rail to rail or so good. that is why first mosfet is totally ok since all we need is 2 Vgs to achieve 5v output @ 3 amps maximum. that means 5v + 2v = 7 volts from op-amp... no worries.

Our opamp is being powerd through a 100ohm series RC resistor cap to filter power.  Only expect it's VCC to have ~10v.

so first mosfet is the best choice.. and since I will buy components myself, then I will get other mosfets too to try them and see.

Just look at the mosfet's data sheet's Vgs transfer characteristics.  You will see the approximate voltage difference between Gate and Source at a specific current draw at usually 3 temperatures.

Your snapshot is that of a capacitor, not inductor.  Make clean snapshots of 1 layer for easier reading...

[VEGETA]

Our opamp is being powerd through a 100ohm series RC resistor cap to filter power.  Only expect it's VCC to have ~10v.

now 2nd mosfet won't work since 10v is the peak and this opamp is not rail to rail. we stick to first one and I will try to search for other alternatives although this one is very cheap in +100 quantity.

Just look at the mosfet's data sheet's Vgs transfer characteristics.  You will see the approximate voltage difference between Gate and Source at a specific current draw at usually 3 temperatures.

so my analysis above with little calculations are good? aside from the rail being 10v not 12v. we stick to our first mosfet since it works and cheap.

Your snapshot is that of a capacitor, not inductor.  Make clean snapshots of 1 layer for easier reading...

now attached for the inductor, it does fit.

actually, no need for any snapshot since the length is 7.3mm exactly as the footprint. the difference is only the width which is not important at all!

[BrianHG]

That all looks good...

[BrianHG]

So, does it work?

[VEGETA]

So, does it work?

I have problems with JLCPCB since they keep cancelling the order. One time cuz the wrong inductor part and the other one for small mouse bites, they want them to be 5mm. I did that and will send the offer again today xD.

I will keep you posted always.

[VEGETA]

and now I am up to another freaking problem, the new visa credit card classified my payment move to JLCPCB as "possible fraud" when paying via paypal or directly via the card! I am trying to see the solution but it will take until Sunday if I am lucky, but most likely won't be good.

the laughable thing is that the previous card is mastercard credit and it accepted the payment a week ago. from the same bank!! same central bank!!

I messaged pcbway to see what can they offer as a turnkey solution.

my hope for jlcpcb is if i can transfer money from my credit card to my paypal then use paypal's own money to pay directly to them. this worked well for me on a previous situation with hetzner. however i used to pay my friend then he refunds them to paypal... this will take losses which I cannot afford for +300$ payments.

I just cannot believe this is happening

[VEGETA]

Hello,

boards arrived and they are with DHL to be delivered to me within days.

Our chosen Mosfet (https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KND3403A_C382143.html) is out of stock thus I searched and found these alternatives:

https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD3006_C479048.html
https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD15N10_C479050.html
https://lcsc.com/product-detail/MOSFET_SINO-IC-SE4020B_C396085.html

what do you think?

I will also get the 1000uF cap: https://lcsc.com/product-detail/Aluminum-Electrolytic-Capacitors-SMD_ROQANG-RVT1C102M1010_C72470.html

Let's pick a MOSFET soon, so I can order fast.

_______

The remaining items are the connectors and heatsink. Connectors from Mouser, I got all P\N needed. I will try to see what to do about heatsinks locally.

However, if you can, please search for a good thermal tape product on Mouser to be used to press-fit the heatsinks. I can't get stuff from digikey due to some problems, and since I am getting a big package from Mouser I will get all stuff from them. If you can't find cheap stuff on Mouser, then maybe I can get this one single item via ebay or amazon but I prefer mouser now.


_________

Meanwhile, I will test the thing with installing IRLIZ44N mosfets (through-hole). It should work since it requires 2v for maximum of 3 amps. I only need it to show that it works well, I have heatsinks I can bolt it on it if I really need to fully operate the thing.

However, I will still need to get the top-side inductor too... so I could just short the inductor terminals just to see if the whole thing works... regardless of output quality. good move right?


Also, what noise and ripple is expected to show on output? single digit uVs or more?

I really hope it works as intended. Do you honestly think it wouldn't work? this could be a nightmare for me really! Don't like to even think about it. However, I really don't think there are any reasons that it wouldn't work... what about you?

[BrianHG]

Hello,

boards arrived and they are with DHL to be delivered to me within days.

Our chosen Mosfet (https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KND3403A_C382143.html) is out of stock thus I searched and found these alternatives:

https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD3006_C479048.html
https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD15N10_C479050.html
https://lcsc.com/product-detail/MOSFET_SINO-IC-SE4020B_C396085.html

what do you think?

I will also get the 1000uF cap: https://lcsc.com/product-detail/Aluminum-Electrolytic-Capacitors-SMD_ROQANG-RVT1C102M1010_C72470.html

Let's pick a MOSFET soon, so I can order fast.

_______

The remaining items are the connectors and heatsink. Connectors from Mouser, I got all P\N needed. I will try to see what to do about heatsinks locally.

However, if you can, please search for a good thermal tape product on Mouser to be used to press-fit the heatsinks. I can't get stuff from digikey due to some problems, and since I am getting a big package from Mouser I will get all stuff from them. If you can't find cheap stuff on Mouser, then maybe I can get this one single item via ebay or amazon but I prefer mouser now.


_________

Meanwhile, I will test the thing with installing IRLIZ44N mosfets (through-hole). It should work since it requires 2v for maximum of 3 amps. I only need it to show that it works well, I have heatsinks I can bolt it on it if I really need to fully operate the thing.

However, I will still need to get the top-side inductor too... so I could just short the inductor terminals just to see if the whole thing works... regardless of output quality. good move right?


Also, what noise and ripple is expected to show on output? single digit uVs or more?

I really hope it works as intended. Do you honestly think it wouldn't work? this could be a nightmare for me really! Don't like to even think about it. However, I really don't think there are any reasons that it wouldn't work... what about you?

I have no clue about the brand's reputation, but, based on the data sheets, only this mosfet has a very low Vgs.
https://lcsc.com/product-detail/MOSFET_SINO-IC-SE4020B_C396085.html

Meanwhile, I will test the thing with installing IRLIZ44N

What's the Vgs at 3-4amps?

I really hope it works as intended. Do you honestly think it wouldn't work? this could be a nightmare for me really! Don't like to even think about it. However, I really don't think there are any reasons that it wouldn't work... what about you?

It should.  I've seen fake LM358 oscillate on the output, but, without a scope, you would never know.  The average output voltage would still look fine on the volt-meter.  A temp solution is usually a pull-down resistor on the output or 1000pf to 10000pf cap on the output of the op-amp.
Assuming the switchers work fine, the problem I listed above is the worst case scenario unless we did not catch something in the schematic.

[VEGETA]

I have no clue about the brand's reputation, but, based on the data sheets, only this mosfet has a very low Vgs.
https://lcsc.com/product-detail/MOSFET_SINO-IC-SE4020B_C396085.html

This has the lowest Vgs on 3 amps, but the others are also the in the good range. At least similar to our original pick. what is their problem in your opinion?

What's the Vgs at 3-4amps?

Here is the datasheet: https://www.infineon.com/dgdl/Infineon-IRLIZ44N-DS-v01_02-EN.pdf?fileId=5546d462533600a401535664286325cb

it is less than 3v max so I guess it is ok.

It should.  I've seen fake LM358 oscillate on the output, but, without a scope, you would never know.

This is what I think too, it should work. However, I don't have a scope now. I am thinking of assembling one unit then send it to you via EMS to check and take photos of the signals and especially the noise. I can test some analog scopes of my company but it is gonna take a lot of time to get one out...

A temp solution is usually a pull-down resistor on the output or 1000pf to 10000pf cap on the output of the op-amp.
Assuming the switchers work fine, the problem I listed above is the worst case scenario unless we did not catch something in the schematic.

I don't think we missed anything in the schematic, no way.

the output of the opamp already has 100R resistor and a 1nF capacitor on its output, we did this to make sure it won't oscillate if you remember.

now the funny thing is... JLCPCB for some unknown reason, didn't solder the switchers!!!!!!!!!!!!!!! now I cannot test anything at all 

I will order the switchers with the mosfets and other components, solder them manually...

_______

On another matter, I got quotes from PCBway to fully turnkey manufacture the panels and solder all components on all sides... it was way more expensive than this approach. the components alone were about 750$ and their assembly service is 500$ or a little bit less while PCB manufacture is just 135$. Do find 135$ + 500$ reasonable to this situation? assuming soldering all smt and through hole components on both sides. I could get the components myself and deliver them directly to them to save a bit of that huge 750$ but the main service is the same regardless.

while on JLCPCB costs about 360$ so far without extra components... gonna collect all data once we finish.

Got a simple conversation of a certain Aluminum local supplier, he said he can do the required heatsink very easily and cheaply... about less than 70$ for the entire 125 pieces. It won't be finned heatsink of course, just a rectangular cube (1.5 cm thickness) cut to suit the board... maybe 2.5x2cm of area. so heatsink is solved.

[VEGETA]

any news from you?

[BrianHG]

any news from you?

I'm stuck in a quarantine lock-down...
Bored to death...

(Everything else you selected looks fine...)

[VEGETA]

I'm stuck in a quarantine lock-down...
Bored to death...

you guys still on that!? we have normal life but after 12:00 midnight no one is allowed.

(Everything else you selected looks fine...)

here are summery of my selections:

Mosfet: gonna get all these 3 but mainly your suggestion for the default position.

https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD3006_C479048.html
https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD15N10_C479050.html
https://lcsc.com/product-detail/MOSFET_SINO-IC-SE4020B_C396085.html

1000uF cap:

https://lcsc.com/product-detail/Aluminum-Electrolytic-Capacitors-SMD_ROQANG-RVT1C102M1010_C72470.html

Inductor:

https://lcsc.com/product-detail/Power-Inductors_Sunltech-Tech-SLO0630H100MTT_C207842.html

Output connectors: these 2 are the ones to go, one is an official replacement of the other.

https://www2.mouser.com/ProductDetail/538-09-48-4069/
https://www2.mouser.com/ProductDetail/538-09-48-3064/

3-position connector (2 are used):

 https://www2.mouser.com/ProductDetail/571-1-1123724-2/

2-position connectors:

https://www2.mouser.com/ProductDetail/5716404452/

need 375 for 125 boards. Now under restocking from Mouser, so I got these alternatives:

https://www2.mouser.com/ProductDetail/TE-Connectivity/640388-2/?qs=%2Fha2pyFadugQXFuu0XZrQDI83R0hWQJW3xfg4GqlfBI%3D

Total price for extra components:

Switchers: 144$
Mosfets: 23$ (I will get 5 of the other 2, not calculated here)
Capacitor: 35$
Inductor: 11$
Connectors: 46$ + 14$ + 126$
Shipping: 66$ + 0$

TOTAL: 465$

What I paid so far: 358$ (PCB) + 271$ (customs)

Expected expenditure: ~50$ for 3D printed small part + ~30$ for DC connector (panel mount on the 3D printed part) + 150$ for extra customs and fees.

GRAND TOTAL: 1325$
cost per part: ~10-11$

[BrianHG]

I'm stuck in a quarantine lock-down...
Bored to death...

you guys still on that!? we have normal life but after 12:00 midnight no one is allowed.

(Everything else you selected looks fine...)

here are summery of my selections:

Mosfet: gonna get all these 3 but mainly your suggestion for the default position.

https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD3006_C479048.html

No...
https://lcsc.com/product-detail/MOSFET_PUOLOP-PTD3006_C479048.html

[gnuarm]

Update:

I am using AOZ1284 with 20KOhm frequency set resistor... that means 2 MHz switching frequency.

I take the 12v into the AOZ1284 then output 3.8v for the 3.3v LDO and 5.5v for the 5v LDO (input of LDOs have 3x 10 uF ceramic caps). The LDOs are LM39302 (Chinese clones, cheap) which are followed by 1000uf elec. cap (10mm x 10mm one) + 5x 10 uF ceramic caps. The inductors are 10uH big size.

I was reading through this thread waiting to find someone to explain that a capacitance multiplier won't do anything a linear regulator won't do.  Then I see that you are misusing your linear regulator which is why the output is not very well regulated.  You don't provide any spec you are trying to meet, so I can't offer a real solution, but I can point out that giving a 3.3 volt output linear regulator 3.8 volts input is not enough to allow it to regulate fully.  If you read the spec on the linear it will specify a minimum drop out voltage, but at some specified drop in output voltage that will likely not be acceptable if it appears as ripple. 

Boost the input voltage to your linear regulator to at least 4.5 volts and you will see a big improvement in the regulation.

You guys mentioned 2MHz devices, well, here is one used... also followed by linear LDO with good amount of caps.

How much ripple do you think we get?

If you mean a low frequency (<10 kHz) you should see ripple below 1 mV.  However, keep in mind many linear regulators have specific requirements on the capacitors and the effective series resistance (ESR) to prevent oscillations. 

Higher frequencies on the power rail are referred to as noise and a linear regulator can't do much about them as it is out of their pass band.  That requires small ceramic capacitors and potentially power planes.

I imagine someone has pointed out to you all this, but I'm not going to read 10 pages of this tread to find out.  Good luck.