Low ripple low noise small power supply design

[BrianHG]

Your feedback resistors of 10K and 16.5k are really high.
The LM317 ADJUST input has an input current and input capacitance.
The website LM317 calculator website I sent you does not take the input current on that pin into account.
This is why I said for R1, use 500Ohm or less to eliminate that figure.

If I had any LM317 on hand, I would have already tested this circuit with a 2N3904 as an NPN buffer to verify final output voltage.
Studying the way the reference works internally, it looks like you will need to add the 0.75v drop of the NPN transistor into your output voltage calculation.  IE 3.3v = set regulator for 4v.

Also, I would still add an R1a and R1b, and only mount 1 of them.
Placing the resistor at in the R1b position means the regulator feedback sees only 1/100 the current spikes since it needs to travel through the transistor emitter to the base.  This means the circuit is acting like a current multiplier VS more like a regulator.

Note we can switch the LM317 with an op-amp / voltage reverence and make a super precise regulated output, on par with a high quality audio amplifier driving a DC output.  This would guarantee dead exact output voltage no matter the load.  A dual opamp would replace both LM317s.  The voltage reference would be a 78L05 regulator.  This would make your output look like a battery with 0ohm ESR.  In the past, I have actually made this type of circuit, so I know it works really well and you may swap the BJT for a logic level mosfet here as the opamp will compensate for the Vgs and it's drift as the mosfet temperature changes.

As for your heat-sink.  No problem radiating 5 watts, but with 8 watts, it will get warm.

According to ON-Semiconductor's MJB44H11 datasheet, at 5 amps. the Vce saturation voltage is 0.2v. (Actually 0.11v at 3amps)
This means if you want 3.3v out, you can get away with 3.5v in, however, there is a little offset in linearity right at the edge.
Now, your switcher can dip it's output voltage by up to 0.4v during current spikes, so, 0.4v + 0.2 + 0.2 extra margin means minimum = 0.8v above target output voltage.
3 amps *0.8v = 2.4 watts heat *2 = 4.8watts heat total, assuming 2.5amps-12.5 watts of LEDs.
This is as low as I would attempt the circuit with obvious test verification.

Raising the extra margin to 0.4v:
3 amps * 1v = 3 watts heat * 2 = 6 watts heat total, assuming 2.5amps-12.5 watts of LEDs.

Radiating 6 watts will be like having one of those old 120v 5 watt outdoor Christmas light bulbs inside your Dreamcast.  If the old PSU had a linear transformer, that transformer alone would radiate that amount of heat at least.

Using the opamp with a mosfet in a DPAK/D2PAK case, you can go down to ~0.6v, ~0.2v above the switcher's worst voltage dip.  Since the op-amp will not need to drive current with a mosfet, that part of the circuit will run cooler & only needing a 0.6v headroom, the circuit will radiate a total of 3.6 watts, 1.8 watts per mosfet.  You wont need a heatsink here, but, the PCB will still get warm to the touch, but not too hot.  I would still prefer at least 0.8v headroom.

Just so we are clear, (3*3.3+3*5)*1.2(loss in switcher supplies) = 29.9 watts at 12v, or 2.5amps at 12V.  With no room for the CD drive's 12v supply if you are using a 30 watt 12v power supply.

I've attached an image of using an opamp circuit.  Obviously we would use a more modern opamp and something like the mosfet.
Cheap mosfet = IRLR8726TRPBF - (Crss=310pf) https://lcsc.com/product-detail/MOSFET_International-Rectifier_IRLR8726TRPBF_International-Rectifier-IR-IRLR8726TRPBF_C81137.html
Or any 50N03 variant should do, like https://lcsc.com/product-detail/MOSFET_KIA-Semicon-Tech-KIA50N03AD_C112249.html

You can also use a TO-220 Mosfet/BJT with a bolted heatsink mounted down on the PCB:
EG: https://lcsc.com/product-detail/Heat-Sinks_XSD-XSD-heat-sink15-5-10-5-24-Htype_C108928.html
https://lcsc.com/product-detail/Heat-Sinks_Made-in-China-Made-in-China-15-10-20-U_C361579.html (More difficult to accidentally short)
However, they will have active voltage unlike the taped SMD heatsink.

See attached circuit below: ( The + goes to the 78L05 output, for channel 2 on the opamp, 3.3v, well divide the 5v to 3.3v with 2 resistors)  This would be an op amp-corrected/compensated/buffered capacitance multiplier.
https://lcsc.com/product-detail/Dropout-Regulators-LDO_Diodes-Incorporated_AS78L05RTR-E1_Diodes-Incorporated-AS78L05RTR-E1_C90471.html
https://lcsc.com/product-detail/General-Purpose-Amplifiers_STMicroelectronics-MC4558CDT_C435907.html
Input latch-up protection diode not shown in circuit.
https://lcsc.com/product-detail/Switching-Diode_ON-Semiconductor-ON-MMBD914LT1G_C46523.html
Precision CMOS opamp:
https://lcsc.com/product-detail/General-Purpose-Amplifiers_Texas-Instruments_TLC272CDR_Texas-Instruments-Texas-Instruments-TLC272CDR_C9374.html

[VEGETA]

Note we can switch the LM317 with an op-amp / voltage reverence and make a super precise regulated output, on par with a high quality audio amplifier driving a DC output.

I was about to suggest this.

it is not required to be extremely precise like 5.0000v but rather very low noise. If we put 78L05 with 12v input, it will give 5v and we can get 3.33v using 10K//10+10K or so. However, ripple and noise will continue to exist right?

Now all of that will go into the OP-amp and eventually to the mosfet gate. How can this be anti-noise and ripple.

Radiating 6 watts will be like having one of those old 120v 5 watt outdoor Christmas light bulbs inside your Dreamcast.  If the old PSU had a linear transformer, that transformer alone would radiate that amount of heat at least.

The original PSU is here: https://d3nevzfk7ii3be.cloudfront.net/igi/NDTYlWMuiMCcR5nW.full

Using the opamp with a mosfet in a DPAK/D2PAK case, you can go down to ~0.6v, ~0.2v above the switcher's worst voltage dip.

let's assume 1v dropout: 3x1= 3 watts per rail. I don't know if we can get away without heatsink here. even with heatsink i would still get the cheapest ever one xD.

This would be an op amp-corrected/compensated/buffered capacitance multiplier.

it looks like a linear regulator, where is the cap multiplier part?

I mean where to put the R and C? on mosfet gate from the op-amp output? like 100R in series from op-amp to mosfet gate and 10uF ceramic to gnd?

Cheap mosfet = IRLR8726TRPBF

seems fine and available at jlcpcb parts.

I will re-design the circuit based on IRLR8726 + LM358 op-amp (used for both rails).

[BrianHG]

Ok, update your schematic and I will take a look tomorrow.
The 100r series for the gate drive is good.

The opamp VCC and 78L05 VCC will share the same series 100 ohm/100uf-10uf-0.1uf RC filter from the 12v supply.
This is what will separate your switcher's ripple noise from your linear amplifier and voltage reference section.

The mosfet Drain has a capacitance of 310pf to the Gate.
This is where the 2MHz switcher noise is coming from, it will be conducted by that 310pf cap from drain to gate where the op-amp output will be fighting it.  This should be manageable with a strategically chosen cap from gate to GND.  Not too large as to prevent the LM358 from regulating the output fast enough, yet large enough to mute most of the switching noise.  I would say a ~10nf cap would do and you may need to shrink the 100ohm opamp output to gate resistance in half.

With the MJB44H11, the base-collector capacitance was only 130pf, but the opamp would need to drive ~30ma at it's output.

[BrianHG]

I mean where to put the R and C? on mosfet gate from the op-amp output? like 100R in series from op-amp to mosfet gate and 10uF ceramic to gnd?

The 1uf at the + input of the op-amp is the cap which you are multiplying.
Since the op-amp has a 100na input current, this will be amplified a million fold.
Coming from the 5v regulator should be a 330ohm series resistor to the + input of the 5v side.
For the 3.3v side, use another 1k series resistor from the 5v and a 2k pull down to GND for 3.33v. (A little high, maybe add a 27ohm in series with the 1k for 3.30v) Don't forget the 1uf cap from that +input to GND.
These 1uf caps with the series ~330Ohm & ~333Ohm feed are your voltage source filters.

The 100Ohm between the +12v 100uf to GND at the 78L05 & LM358N V+in does the bulk filter removal from the switcher's noise.  It will also have a 1uf cap on the output.  I mean, you could place a 470uf or 1000uf cap in place of that 100uf cap if you want to.  Electrolytic, not ceramic as high uF value ceramic caps lower their capacitance value when run at higher DC voltages.

Don't forget, you still want the right SMD emi ferrite bead & cap between your switchers and 12v in, and again between the 12Vin and the rest of your analog output.  If the 12V feeding you Dreamcast has high frequency ripple on it, the power supply caps for the 12v in the CDRom drive will conduct that ripple back through the GND.

[VEGETA]

Here is the updated schematic. Might be missing some stuff though.

- I didn't find a reason to have 330 ohms in the +5v input of the opamp, and chosen 10k//(10K+10K) to consolidate stuff.

- input side of opamps has 10uF instead of 1uF, for consolidation... unless it is necessary to have 1uf.

- mosfet gate is 100R parallel to 1nF since I happened to use a 1nF.

what do you think?

[BrianHG]

Ok,
Blue arrows, get rid of those 100uf caps, they wont do anything good.
Green arrow at V_5->3.3k, This will make the 5v reference rise up and filter identical to the 3.3v reference.

Dark Green arrows at new 3.3k on the V_FB# - This separates the opamp -input from a harsh supply rail and also allows the 2 other green arrows with the new MMDB914 diodes at the opamp inputs to work.

The 2 new MMBD914 (or equivilant) diodes lifts the V_FB# inputs off the GND rail at powerup and also prevents the + & - inputs from the op-amp from getting too far from each other preventing latch-up on power-up or output inversion. (These are not guaranteed rail-2-rail input opamps.)

Having the 1nf directly on the op-amp output can make the op-amp oscillate.

The red arrows are errors or changes.

The black arrow shows how you need to wire the +12V input to the rest of your circuit.  According to the AOZ128 datasheet, at full load, the input VCC ripple becomes massive and this is what we don't want to transmit to the analog supply section, or, to the Dreamcast.  Choose a FB which has a good low DCR & Current, and cuts/high impedance @ 1-2MHz.

Your only additional improvement here is to use really expensive ultra-low ESR caps, and blindly placing them wont help, or it could make things worse.  Your board layout and trace paths have now become important.

The switchers should be on 1 side of the PCB with their own GND and VCC and power supply input.
The other side should be the analog with a dedicated GND path from the power input and the output of the second ferrite bead with decoupling caps.  That GND path should GND all the analog/linear components.

Additional, You may also use 2 ferrite beads, one for each switcher's VCC.

[BrianHG]

For the ferrite beads, you might be better off just using the same 10uh inductors.

[BrianHG]

Ooops, you are also missing a dummy load on the +3.3v out and the +5v out.
If you are consolidating parts, place a 100ohm on the 3.3v and 2 series 100ohm for the 5v.

[VEGETA]

Kindly check the drawing, I made all the fixes.

Notes:

- I used this ferrite bead: https://lcsc.com/product-detail/Ferrite-Beads-And-Chips_TDK_MPZ1608S600ATAH0_60R-25-at100MHz_C76816.html/?href=jlc-SMT since it is available at JLCPCB SMT service and still provide what we need.

- 100R dummy load for both rails.

- ferrite beads everywhere needed, or so I think.

For the ferrite beads, you might be better off just using the same 10uh inductors.

those 10uH inductors are very big, can't afford adding another one.

[BrianHG]

Ok, here you go.

The purple arrows are where you forgot your bus/wire ties.
The red arrows are the mistakes.
Note that only having a single 100Ohm on 5v will overheat a smd 0805 resistor.  Use 2 in series.

Note that you are already using 100nf for the BST pin on the switchers.  Might as well use them for decoupling the regulator and opamp instead of 1nf.

Green arrow adds 10uf on the +12v to Dreamcast.
If you are using ceramic non-polarized 10uf caps, be careful.  Read the complete data sheet as 12v DC will result in a lower capacitance and higher ESR as ceramics at these capacitance levels at higher voltages de-rate miserably.  So much so that a 1uf cap may do better.

Quote from AOZ datasheet:

For lower output ripple voltage across the entire
operating temperature range, X5R or X7R dielectric type
of ceramic, or other low ESR tantalum capacitor or
aluminum electrolytic capacitor may also be used as
output capacitors.

Finally the ferrite chip bead.  I will look for a better one in the next post.

[BrianHG]

I assume you are using a 10uh inductor like this:
https://lcsc.com/product-detail/Power-Inductors_Wenshan-YT0630-100M_C428419.html
7mmx6.6mm.  69mOhm 4.5amp sat.

Or, if you want a better efficient cooler switcher, you would be using this TDK inductor:
https://lcsc.com/product-detail/Others_TDK-SPM10040T-100M-HZ_C375968.html
11mmx10mm.  26.7mOhm. 5.8amp sat.

Last ugly monster:
https://lcsc.com/product-detail/Power-Inductors_MAGLAYERS-MMD-10DZ-100M-X2_C332199.html
11.5x10mm  27mOhm 7.5amp sat.

This is the largest (impedance increases more at the lower frequencies) affordable ferrite chip bead:
https://lcsc.com/product-detail/Ferrite-Beads_TAI-TECH-HCB2012KF-600T60_C369525.html
I see you placed a bunch all over.  However, the chip beads will only really knock out frequencies in the +10Mhz range.

Just using only 2 strategically placed of the first 7mmx6.6mm inductors would vastly cut out lower frequency transmition down at the +1MHz range.

That's overkill, just read the next message...

[BrianHG]

Just a single strategically placed 3.5mmX3mm 10uh power/choke inductor:

https://lcsc.com/product-detail/Power-Inductors_Cybermax_CMLF0302-100MTT_Cybermax-CMLF0302-100MTT_C405095.html

To deliver 12v to your Dreamcast & Linear regulator section will do much more than all of your 5 ferrite beads combined.

Assuming that the Dreamcast's 12v draws no more than 1amp.

[BrianHG]

Another mistake...
Your 'FSW' resistors are set to 20k.
This is not 2MHz....


  The bloody formula in the datasheet is completely F--KED UP!!! 

You will need to verify that the device is actually running at 2MHz.
Otherwise, there will be ripple going throughout your entire PCB.
The example listed figures don't even come close after fiddling with the decimal point.

[VEGETA]

The bloody formula in the datasheet is completely F--KED UP!!

yes but when you calculate it properly you will result in getting 20k resistors. there is no way we can verify that.

Just a single strategically placed 3.5mmX3mm 10uh power/choke inductor:

https://lcsc.com/product-detail/Power-Inductors_Cybermax_CMLF0302-100MTT_Cybermax-CMLF0302-100MTT_C405095.html

To deliver 12v to your Dreamcast & Linear regulator section will do much more than all of your 5 ferrite beads combined.

Assuming that the Dreamcast's 12v draws no more than 1amp.

well, you mean feeding the 78L05 and 12v to dreamcast pin? then yes, I guess 1A or so is ok since the entire dreamcast power supply is 22 watts.

I think this inductor is perfect: https://lcsc.com/product-detail/Power-Inductors_Sunlord-MWSA0603-100MT_C132141.html/?href=jlc-SMT

I will use it for the rails themselves, I wonder why I didn't see it in the first place and went for a huge 12.5x12.5 mm inductor! 4.5A is more than enough for each rail... heck, small size means we can put more of it on inputs too.

[BrianHG]

I think this inductor is perfect: https://lcsc.com/product-detail/Power-Inductors_Sunlord-MWSA0603-100MT_C132141.html/?href=jlc-SMT

I will use it for the rails themselves, I wonder why I didn't see it in the first place and went for a huge 12.5x12.5 mm inductor! 4.5A is more than enough for each rail... heck, small size means we can put more of it on inputs too.

Don't go crazy on inductors.  Obviously the switchers themselves need them in circuit and your priority is to isolate the Dreamcast's +12v and your linear supply from that nasty side with all the ripple noise.
If it can be done, a good 1000uf to 4700uf cap on the Dreamcast 12v side of the 10uh inductor would really be nice.

A fat flat one like this (10k hour):
https://lcsc.com/product-detail/Aluminum-Electrolytic-Capacitors-Leaded_Ymin-LKMI2001C472MF_C442801.html
Cheaper one (2k hour):
https://lcsc.com/product-detail/Aluminum-Electrolytic-Capacitors-Leaded_CX-Dongguan-Chengxing-Elec-4700uF-16V-16-20_C47896.html

However, due to size price and availability, using 2 of these 2200uf is a better deal...
https://lcsc.com/product-detail/Aluminum-Electrolytic-Capacitors-Leaded_CX-Dongguan-Chengxing-Elec-16V2200uF-10X20_C3328.html

[VEGETA]

Don't go crazy on inductors. 

I will just use the part I linked up instead of the old one, which saves 50% of space. + on more inductor for 12v output to DC. + 1 for isolating the 5v regulator and op-amp circuit. which means a total of 4.

If it can be done, a good 1000uf to 4700uf cap on the Dreamcast 12v side of the 10uh inductor would really be nice.

I think this cap (1000uF) is the way to go due to size and smt placement capability: https://lcsc.com/product-detail/Others_Lelon-VE-102M1C1010-TRO_C249474.html/?href=jlc-SMT

I could put smaller 100uF caps instead but won't be the same amount of capacitance and still takes place.

[BrianHG]

Yup, placing 3/4 of those 1000uf in parallel will do it.

I may be going out of town at the end of the week.
So if you want, it's a good time to finish up and get to the PCB.

Your switcher's 2 frequencies where it generates noise is up at 2Mhz and according to the response time on load transients, those 0.4v dips at 100us, in the 100KHz region.
According to Panasonic's basic 1000uf 16v electrolytic caps, at 100KHz, the cap looks like a 0.06Ohm resistor and can conduct 1.19 amps of current.

This is approximate: Your 1000uf cap
Hz - ~Series Impedance in Ohms
10Hz      - >10 Ohm
100Hz   - >1 Ohm
1Khz     - >0.2 Ohm  result curving due to internal resistance
10Khz   - >0.1 Ohm
100Khz - >0.06 Ohm  lowest resistance
1Mhz     - >0.1 Ohm   resistance is increasing due to lead inductance.
10Mhz   - >1 Ohm      resistance is increasing due to lead inductance.

So for example, if you placed 1 1000uf cap for the 78L05 & LM358, since it is being fed power through a 100 Ohm series resistor, ripple at 100KHz will be attenuated 10000:6.  At 100Hz, the attenuation will be 100:1.  The 78L05 is rated to reject the V+ input by around -60db at 100Hz, but at 1Khz, that's (going off of memory of old datasheets which had charts for this) -20db.  At 10Khz, this was around -10db, and no rejection at 100Khz and up.  Meaning, if you power the 78L05 at 10v with a 100KHz 2v sine wave, the V+ output will also modulate +/-1v at 100KHz.  This only using that RC filter will make those potential +/-0.4v ripples only 0.00024v ripples at the regulator V+ input.  To really rid of the 2MHz ripple, since electrolytics aren't good here, your 10uf/22uf need to be MLCC caps as shown in the table I've attached below.  The op-amp also has a power supply rejection value in it's data sheet as well.  But as you can see, the 100ohm / 1000uf cap + 10uf MLCC cap in parallel filter will leave a final overall ripple noise which you could not measure so long as your +12v input isn't conducting any current to your mains and through the GND pin of your scope.  Unless you scope can see without any noise less than ~0.0005v, or you have an amplified differential probe designed to see these types of signals.  I think the transistor noise alone in the 78L05 output is noisier than this anyways.

Your unknown source supply may also switch as low as 20KHz.

Image #1 has a 560uf Aluminum VS a 22uf MLCC.  Your best choice is a 1000uf plus a 10uf, or 22uf MLCC in parallel to choke out those switching spikes.

Make sure your 10uf caps are the right grade and size, don't go for the smallest.
See image #2 which shows how the capacitance value derates of a 4.7uf MLCC cap.  Both type ans size makes a huge difference.
As you can see, X7R 25v 1210 is your best bet as the 0805 versions at at 12v dropped to down to ~1uf.  And this is a 4.7uf cap.  A 10uf cap would be at least the same if not worse as you try to squeeze more capacitance in the same package.

[BrianHG]

Here is an X7R 25v 1206:
https://lcsc.com/product-detail/Multilayer-Ceramic-Capacitors-MLCC-SMD-SMT_SAMSUNG_CL31B106KAHNNNE_10uF-106-10-25V_C14860.html
Here is an X7R 25v 1210:
https://lcsc.com/product-detail/Multilayer-Ceramic-Capacitors-MLCC-SMD-SMT_SAMSUNG_CL32B106KAJNNNE_10uF-106-10-25V_C39232.html

A bit of a deal, 22uf X7R 16V 1210:
https://lcsc.com/product-detail/New-Arrivals_Taiyo-Yuden-EMK325B7226KM-PR_C412624.html



Because of the price difference, maybe just add 3 in parallel for the 12v and 2 in parallel for the 5v&6v.

[BrianHG]

I just found an EMI filter schematic for feeding the VCC of 2MHz 6 amp buck converters.
They recommend a 1uH 7.5amp power inductor with 2x2.2uf & 2x470nf caps before and after the inductor.
Using a 10uH inductor may dip too much.

This is all you need to separate each switcher's VCC supply feedback ripple from the 12V supply input:
https://lcsc.com/product-detail/New-Arrivals_Sumida-252012CDMCDDS-1R0MC_C492725.html
And they are small at 2.5x2mm.

[VEGETA]

I made a new update taking everything in consideration.

Maybe I went crazy on EMI stuff? well, this is what makes the psu in this application. 

I can re-arrange the schematic into multiple hierarchical sheets to fit the wide chain of stuff... later on when I finish the design. I picked parts suitable with jlcpcb smt service... 1206 of 10uf caps and 100nf caps + 1000uf caps (big enough). + nice 1uH inductor but bigger than yours.

[BrianHG]

Back in a few hours....
Yes, some overkill there in the wrong places.

[langwadt]

Back in a few hours....
Yes, some overkill there in the wrong places.

overkill? nooo, a medium sized tactical nuke is the proper tool for killing a fly

[BrianHG]

Here you go.  I think you are ready for the PCB part placement.

[BrianHG]

1 last addition, I would add 3-4x more 1000uf caps on the Dreamcast +12v side + a single 1000uf cap at the V+12 input.

[VEGETA]

100uF caps on NPN inputs

I really wanted to get rid of them, now is the chance. x3 of 10uF 1206 maybe save space compared with 100uF elec. cap.

_____

suggested small inductor

I know your suggested inductor is better, but it doesn't exist here: https://jlcpcb.com/parts . I really searched for smaller one but couldn't find it.

I know people say PCBWay offers similar service but JLCPCB one seems better at least for now.

1 last addition, I would add 3-4x more 1000uf caps on the Dreamcast +12v side + a single 1000uf cap at the V+12 input.

I wonder if all of those are gonna fit in a 50x50 board with all components on one side. 


____

So you removed the EMI circuit in the bottom for 5v and op-amp? you took them from the 12v output to the DC itself, wouldn't this cause noise and ripple to get into the 5v and op-amp?

____