Low ripple low noise small power supply design

[BrianHG]

Here are all the possible MJD44H11 you can use.
https://www.findchips.com/search/MJB44H11
Some places have it as cheap at 37cents for 1.

LM317...
https://www.findchips.com/search/LM317
Or as low as 11cents for 1, but in TO-92.
37cents in SO-8 for 1, but in quantity, it goes down to 10cents.

[VEGETA]

Do you have the Junction to ambient temperature of the 'LM39302'?  If it is in the same package, it is not much different.
Also, because of those 0.5v spikes coming from the AOZ1284, and the regulator dropout of ~500mv at top load, with a little regulation safe zone, you would still be powering the LM39302 with ~+1.3v, 3.75 watts of heat.

I put some caps before the lm39302 to ensure no big spikes happen, 3x 10uf as i remember. plus, 0.5v isn't much to cause such heat. I didn't do any measurements to see actual stuff happening.

The transistor derating you are reading is the transistor not even mounted on a PCB, complete open air.  Even a PCB alone will drain away heat.

you mean junction-to-ambient? I keep reading people say that even putting some copper area for dissipation will not help or won't make a difference.

Just get the cheapest adjustable linear 100ma regulators (LM317 in SMD) which can go to at least 18v input and down to 3.3v out.
You will use the 100ma regulator's output to feed the base of the MJD44H11T4G, multiplying that supplied current by the transistor's current gain curve which could drive ~ 5amps, however, the sweet spot is at the 3amp mark where the transistor's gain is clearly above 100.

you want to use lm317 for its stable output? how will this affect the final 3.3v and 5v without a linear post-regulator?

The trick to preventing your 12V supply's ripple from reaching the 2x 50ma linear regulators is use a 1/2watt 100 ohm resistor from 12v to the regulator's Vin, and at that Vin, have a good 10uf 25v cap to GND (The GND trace by the output connector).  This 1 resistor and cap can power both regulators simultaneously, or, if you want super separation, use 1 resistor and cap for each LM317.  (Remember, if the MJD44H11T4G is driving a full 3amp load, the regulator powering it's base is driving ~15ma + a minimum pulldown resistor = ~20ma total max.)  While each AOZ1284, remember it makes spikes up to 0.5v, plus you want a little headroom, should power the collector with 1.3v more than the output voltage.  ~4.6v for the 3.3v output and ~6.3v for the 5v output.

so from 12v source to 100ohm (0.5w) resistor (+ caps to gnd) then to lm317 input... then lm317 output to gate of npn?

this way yes the resistor can tolerate such low currents.

however, 1.3v dropout still requires heatsink... damn. I don't mind soldering and fixing heatsinks but shipping them will be very expensive and I didn't really check any local Aluminum shop if they can supply the required material cut and drill the hole...

I really hope that the heatsink remains the final issue... then I can be optimistic.

anyway you mentioned that I shouldn't let the ripple reach... what? you mean the resistor + cap will eliminate the ripple? but it is still getting into lm317.

     However, the 2 problems with this circuit is output regulation unless you try something I never had.

yes, this.

with linear post-regulator you can be very safe and precise. with just a multiplier....?

doing this trick is tricky... no guarantees.

The second problem is that there is no true over-current protection.  You will be relying on the maximum current of the AOZ1284 to limit the output power.

no over-current issue will happen since the actual current draw will barely reach 1-1.2 amps at most... even with modded fans.

This wiring configuration relies on the MJD44H11T4G collector absorbing and not passing through all the switching noise to it's emitter.  For this, the output will need a minimum load with a small cap.  Otherwise down at 0ma, some ripple may make it through as the transistor's internal capacitance will transmit some signal through as a slight DC error offset on the output.  This shouldn't be bad as a single transistor like this can operate above 50MHz.

I can send you the full pdf of the schematic to see how i implemented minimum load for each rail.

so lm317 with resistor and caps makes the npn power transistor absorb all noise but never passes it?

It's too bad the collector is the tab and not the emitter.  The collector is where you are getting all the noise from the switching supply source and a hunk of metal heatsink and fat PCB power trace may act as an antenna.

I could make the copper area connected to the emitter pin instead. easy.

As for your illustrated heatsink, for ~5watts of heat, that will be fine, though placing the heatsink under the PCB right under the transistor with stitched vias and a rectangular copper pour on both sides will get rid of the heat more effectively.  Just gluing/taping that heatsink right ontop of the transistor's plastic case may be enough.

I thought of making a hole in the heatsink then screw it tightly to the board where it sits on the copper area (but not soldered). while the transistor will be in the other face of the board with its own copper area too.

[BrianHG]

I put some caps before the lm39302 to ensure no big spikes happen, 3x 10uf as i remember. plus, 0.5v isn't much to cause such heat. I didn't do any measurements to see actual stuff happening.

If you have a +0.5v input to a low dropout regulator, it's output transistor is basically almost nailed 'ON'.  Now in the 'AOZ1284' datasheet, even with it's recommended output caps, it's output dips by 0.4v when the load switches by 2 amps.  To filter this through a lm39302 you would want +1v above your output voltage instead of 0.5v.  If you had something like a 47000uf cap at the regulator input, then you probably could get away with 0.6v above your output voltage.  Just adding a capacitance multiplier inbetween means the drop there will just make more heat for it's theoretical minimum optimum drop of now 0.7v for the transistor + 0.5v drop for filtering 'AOZ1284's 0.4v drop + ripple noise meaning 1.2v drop.  Now you have 1.2v drop on cap multiplier + 0.5v optimum drop on linear regulator.  That's still a total 1.7v drop in heat being injected into your PCB.  We can do better and save money.

The transistor derating you are reading is the transistor not even mounted on a PCB, complete open air.  Even a PCB alone will drain away heat.

you mean junction-to-ambient? I keep reading people say that even putting some copper area for dissipation will not help or won't make a difference.

Yes, if the PCB has no cooling or air flow, the heat up will happen, it will just take a little longer as the PCB itself warms up.  The same is true of a heat sink.  Still, you don't need so much as each transistor will have a ~1.5v drop, 1.2 amps load each meaning a total of 3.6 watts to radiate away from those 2 transistors.  You can shave the 'AOZ1284' down to 1.1v above the desired output voltage, 2.7 watts of heat for both transistors, but, I would test and trim these.

Just get the cheapest adjustable linear 100ma regulators (LM317 in SMD) which can go to at least 18v input and down to 3.3v out.
You will use the 100ma regulator's output to feed the base of the MJD44H11T4G, multiplying that supplied current by the transistor's current gain curve which could drive ~ 5amps, however, the sweet spot is at the 3amp mark where the transistor's gain is clearly above 100.

you want to use lm317 for its stable output? how will this affect the final 3.3v and 5v without a linear post-regulator?

The LM317 + MJD44H11 creates an adjustable 5 amp linear regulator.  You don't need anything else.  And because of the way we are wiring it, you wont even need more than a 10uf cap on the output.

The 100 ohm + 10uf cap feeding the Vin input of the LM317 means no high frequency ripple reaches the regulator's reference or GND pin.
The regulator output pin has a parallel 1uf cap and 1K resistor to GND to prevent output oscillation and guarantee a minimum load.
That output through a series 10 ohm resistor to prevent transistor oscillation feeds the base of the MJD44H11.
The emitter output of the MJD44H11 goes through a feedback resistor to the ADJ pin on the LM317 and that pin has the second feedback resistor going to your GND reference.  Just like in the LM317 data sheet, except the R1 is taken from the transistor emitter, not the regulator's Vout.
On the MJD44H11 emitter, also add a 10uf cap to GND and a 220ohm resistor 1/4watt to GND to make sure the transistor stays on at 0 load.

This is a linear regulator where the output stage is buffered 200 fold at 2 amps, 100 fold at 3 amps according to the transistor datasheet's DC current gain chart.

The reason this setup wont transmit any noise from the 'AOZ1284' to your outputs is that the entire regulator circuit is running at ~12v, coming from a 100 ohm - 10uf RC filter.  Meaning that the regulator circuitry and reference only needs to deal with and filter out  frequencies below 1Khz at the Vin pin, it wont have to deal with 10Khz spikes & 1-2Mhz RF which typically would go right through it.  With 100uf on the LM317 V+ pin instead of 10uf, now the regulator will only have to deal with ripple noise below the ~100Hz range, a frequency range it was well designed for.

The sensitive regulator circuitry never sees all the RF EMI on the MJD44H11 collector pin coming from the 'AOZ1284' output & source switching supply so long as you carefully lay out your PCB while the 85MHz MJD44H11, at low currents like 0.5 amps will still reject much of frequencies above 10Mhz.  I doubt 2Mhz could be seen on the output unless it being picked up and amplified by a looping GND on the PCB.

This eliminates those pesky additional linear regulators and makes less heat and would probably deliver a much cleaner output.

[BrianHG]

You know if you have any old TO-92 LM317 lying around plus a basic NPN transistor, you can try it out on a breadboard.

Even a 2N3904 or 2N2222 will work, except your output will be limited to the capabilities of the transistor.

[VEGETA]

I have made a quick and dirty drawing of the idea... check it out here: https://slow.pics/c/5iWg8qwS

excuse my phone camera, it started to wiggle!

[BrianHG]

I cant see a block of fuzz above the emitter of the transistor, but, that basically correct.
You want 100uf at the LM317 Vin.
You want 1uf at the LM317 Vout.  (SMD ceramic is good here)
You want 10uf or 100uf at the Emitter of the BJT.
The cheapest electrolytic 100uf 25v will do.

You do not need multiple 10uf s anywhere.  As for the 'AOZ1284' switcher, design it to spec in the datasheet.  Use 1MHz like recommended or 2MHz, and the specified recommended caps.  Let the BJT clean up the crap.

Only that feeding the 'base', you have a 100ohm resistor.  This means the output series resistance at the emitter appear to be 100ohm / Hfe200 = 0.5 ohm.  If you truly want to nail that output, using 10 ohm here would mean a series output resistance of 0.05ohm.  This means little as the LM317 will compensate in both cases, just that with 10ohm, the required compensation is a less making the circuit respond a little faster to rapid load swings.

A simpler PCB like this with only 2 switchers and 2 BJT on output with SOT23 LM317 gives you a PCB with maybe a little breathing room if you keep the current size.  You may still want to ferrite bead then cap +12vin when feeding the Dreamcast's +12v so the noise put on that power line from the 2  'AOZ1284's will not propagate to you 12v line in the Dreamcast infecting all the analog lines & GND as well.

[VEGETA]

I will try to re-design the thing from scratch and give you the update... at least the main circuit for one rail.

I cant see a block of fuzz above the emitter of the transistor, but, that basically correct.

it is (current mirror for minimum load circuit), instead of putting a resistor. this is better right?

You want 100uf at the LM317 Vin.
You want 1uf at the LM317 Vout.  (SMD ceramic is good here)
You want 10uf or 100uf at the Emitter of the BJT.
The cheapest electrolytic 100uf 25v will do.

will take these notes.

shouldn't multiple values (100uf elec. cap + 10u ceramic +1n ceramic) together make better result for all frequencies?

parts consolidation is a big thing too, so why using 1uf where I have 10uf used elsewhere and it is better?

As for the 'AOZ1284' switcher, design it to spec in the datasheet.  Use 1MHz like recommended or 2MHz, and the specified recommended caps.  Let the BJT clean up the crap.

I did as datasheet, but with 2MHz option.

Only that feeding the 'base', you have a 100ohm resistor.  This means the output series resistance at the emitter appear to be 100ohm / Hfe200 = 0.5 ohm.  If you truly want to nail that output, using 10 ohm here would mean a series output resistance of 0.05ohm.  This means little as the LM317 will compensate in both cases, just that with 10ohm, the required compensation is a less making the circuit respond a little faster to rapid load swings.

ok, i will make it 10 ohms.

A simpler PCB like this with only 2 switchers and 2 BJT on output with SOT23 LM317 gives you a PCB with maybe a little breathing room if you keep the current size.

in a 50x50 mm board? hmmm maybe if I remove the current mirror transistors and replace them with resistors but still same or more components.

linear regulators are replaced by the power NPN transistors. we have extra components like lm317 and most importantly the heatsinks.

will the heatsinks get hot? I don't want heat inside the thing, this is important.

I don't think one heatsink can be used for both transistors, right?

You may still want to ferrite bead then cap +12vin when feeding the Dreamcast's +12v so the noise put on that power line from the 2  'AOZ1284's will not propagate to you 12v line in the Dreamcast infecting all the analog lines & GND as well.

on the output pin to the dreamcast?

putting all these extra components will make it crowd area.

___

BTW i am using large inductors (10x10mm), is it necessary with 2MHz switching? if I could get away with less size I can then make room for more stuff.

____

to my understanding, the capacitance multiplier here is implemented by the 10uf that is parallel to the 1k resistor right?

[VEGETA]

I have made a very quick kicad schematic. kindly check it

[BrianHG]

You only install R1a, or R1b, not both.
Use the LM317 resistor calculator to
get the values for R1 & R2.  Use 1% resistors.
Choose an R1 value close to desired V output * 100 in Ohms,
Fill in the desired output voltage and the calculator will give you
the best value for R2.

When using R1a, your output voltage target for the calculator will be exact.
When using R1b instead, you need to fill in a voltage 0.7v higher than your
desired output voltage.

LM317 Voltage calculator website:
https://circuitdigest.com/calculators/lm317-resistor-voltage-calculator

Choose the best R1 available in 1% at around 100 X Vout in ohms.
The make sure R2 is easily available at 1%.
You may need to move R1 up or down a little to find an optimum R2.
With both resistor values, the calculator will give you the exact voltage output.
You can then remove R1's value and increase the output voltage by 0.7v and
get a new value for R2b if you want to test.

I'm assuming that you did the feedback resistor calculations correct for the AOZ1284.

On the linear side where you have the 10uf and 100uf, you only need the 100uf.
Cheap electrolytic radial will do.
Use at least 25v for the +Vin on the LM317 since a 16v cap is a little close.

[BrianHG]

You want to really save, make this the 5.0v regulator and on the output of your switcher, place 2 smd 'S5MB R5G' or 'S5KBHR5G' or through-hole 'MUR460RLG' 4-5 amp diodes in series to feed the collector of the transistor of the 3.3v linear regulator section.

Only 1 switcher, though, that switcher needs to deliver enough amps for the 5v and 3.3v.
Mount the diodes at the opposite edge of the PCB so they don't heat the BJTs as much.
Through-hole diodes may send a bit less heat to the PCB.
@1.2 amps on the 3.3v output means each diode will radiate around ~1 watt of heat.

[BrianHG]

I've had good results with these TDK supplies in the past, but they are not PCB mounted:

TDK supplies

This one is also good and cheap:
Triad Magnetics 65watt

Not as well filtered output...
Meanwell board mount, 12v @ 5 amps:
https://www.digikey.com/product-detail/en/mean-well-usa-inc/IRM-60-12/1866-3063-ND/7704688
Meanwell 12v @ 2.5 amps:
https://www.digikey.com/product-detail/en/mean-well-usa-inc/IRM-30-12/1866-3043-ND/7704668

[BrianHG]

If you really want to save on the PSU, use a cheap switcher 3$ 15v, 3amp wallwart, and add a third linear regulator to make a clean 12v from that junky wallwart.

I doubt the 12v on the Dreamcast will take much current as it only power's the CD drive.  And you get a UL/CE approved enclosed switcher with a barrel jack which can plug into your filter regulator PCB.  15v at 3 amps will easily give your 6.5v @ 4amps using your onboard single switcher and since the LM317 & BJT comes in at around 1$ per regulator, now making 3 of them.  This would be a fraction of the 30$ Meanwell PSU which has exposed mains wires meaning no safety approvals for your project as there are exposed mains wiring involved.  The 15v also leaves regulation headroom for the length of cable coming out of the wallwart.

Higher quality 15v bricks (AC power chord wire -> box -> and DC output chord with barrel jack, like laptop power supplies) go for 12$usd.


Cheap example: 15v 3amp Wallwart
Better quality: 15v 5 amp Brick
Better quality chassis mount for 6$: Get 12v 3amp version @6$ and adjust the output to 13.8v with onboard trimpot (less heat on 12v linear reg)
The 6$ chassis mount unit is nowhere near as good as the TDK or Triad Magnetics supplies I listed above.
The TDK and Triad units have proper input power chokes, well isolated from the mains and they are well certified and have fully documented datasheets.  You get what you pay for...

[VEGETA]

If you really want to save on the PSU, use a cheap switcher 3$ 15v, 3amp wallwart, and add a third linear regulator to make a clean 12v from that junky wallwart.

I doubt the 12v on the Dreamcast will take much current as it only power's the CD drive.  And you get a UL/CE approved enclosed switcher with a barrel jack which can plug into your filter regulator PCB.  15v at 3 amps will easily give your 6.5v @ 4amps using your onboard single switcher and since the LM317 & BJT comes in at around 1$ per regulator, now making 3 of them.  This would be a fraction of the 30$ Meanwell PSU which has exposed mains wires meaning no safety approvals for your project as there are exposed mains wiring involved.  The 15v also leaves regulation headroom for the length of cable coming out of the wallwart.

Higher quality 15v bricks (AC power chord wire -> box -> and DC output chord with barrel jack, like laptop power supplies) go for 12$usd.


Cheap example: 15v 3amp Wallwart
Better quality: 15v 5 amp Brick
Better quality chassis mount for 6$: Get 12v 3amp version @6$ and adjust the output to 13.8v with onboard trimpot (less heat on 12v linear reg)
The 6$ chassis mount unit is nowhere near as good as the TDK or Triad Magnetics supplies I listed above.
The TDK and Triad units have proper input power chokes, well isolated from the mains and they are well certified and have fully documented datasheets.  You get what you pay for...

well, the cheap psu is on the users themselves, not mine. so no cost there xD.

I will just provide the small board with nice little psu + a 3d printed small part with a female dc jack on it (my idea is to use a little pcb with a hole in it instead -> 0$ solution per unit).

You can just see this product to see for yourself what I want to make: https://www.indiegogo.com/projects/dreampsu-making-your-dreamcast-cool-again#/



____

I will respond to your previous posts soon.

[BrianHG]

LM317 with available 1% resistor values in ohms:

R1 - R2   = #v.
510-1540 = 5.02v
432-1540 = 5.71v -> may need adjustment depending on Vbe of transistor.

340-560  = 3.31v
249-560  = 4.06v -> may need adjustment depending on Vbe of transistor.

[BrianHG]

well, the cheap psu is on the users themselves, not mine. so no cost there xD.

Ok, fair.  If you are not providing a PSU, then I recommend buying the 2 cheapest pieces of junk in the above link and verify that they deliver the results you will claim on your indiegogo campaign just to be sure.  And I recommend passing the PSU links to your users so you know they will get the same results you are getting.

I also recommend getting 12v adapter versions to see if they provide a good enough 12v to remove that third regulator.  However, you would want good ferrites and caps on the 12v line feeding the Dreamcast.

[VEGETA]

well, the cheap psu is on the users themselves, not mine. so no cost there xD.

Ok, fair.  If you are not providing a PSU, then I recommend buying the 2 cheapest pieces of junk in the above link and verify that they deliver the results you will claim on your indiegogo campaign just to be sure.  And I recommend passing the PSU links to your users so you know they will get the same results you are getting.

I also recommend getting 12v adapter versions to see if they provide a good enough 12v to remove that third regulator.  However, you would want good ferrites and caps on the 12v line feeding the Dreamcast.

That indigogo is not mine. That is a replacement psu which can be found here: https://github.com/PSUThings/PSU

it is pure switching supply using 2 of TPS54525PWPR switching regulators. So I decided to make one using linear stage and much cleaner output.

I have a cheap chinese adjustable psu that can be used, and I recently got another one from amazon. they are good enough to test the thing.

I will try my best to redesign the thing using our approach of lm317 + npn, despite fearing the lm317 behavior since I never took feedback from another point rather than its output.

Also, where exactly in our design that cap. multiplier is implemented?

Will a small smt heatsink be enough for one regulator rail? aside from the idea of using a diy one by cutting an Aluminum sheet (this one is last resort).

I hope that we can use one heatsink for the 2 regulators (if it is hooked on gnd, no shortings needed xD).

Since using 2 MHz switching frequency, do inductors need to be large (10x10 mm)? these 2 take the most space and they are 10uH to eliminate the ripple... datasheet did specify lesser value.

[BrianHG]

I will try my best to redesign the thing using our approach of lm317 + npn, despite fearing the lm317 behavior since I never took feedback from another point rather than its output.

Also, where exactly in our design that cap. multiplier is implemented?

Will a small smt heatsink be enough for one regulator rail? aside from the idea of using a diy one by cutting an Aluminum sheet (this one is last resort).

I hope that we can use one heatsink for the 2 regulators (if it is hooked on gnd, no shortings needed xD).

Since using 2 MHz switching frequency, do inductors need to be large (10x10 mm)? these 2 take the most space and they are 10uH to eliminate the ripple... datasheet did specify lesser value.

The LM317 will just raise the voltage of it's output until the 'ADJUST' input pin reaches 1.25v.  The ADJUST pin is like a negative feedback of an op-amp.  Once the emitter of the BJT gets high enough to make the ADJUST pin 1.25v, (through the resistor divider) the LM317 will stop raising the voltage at the base hence regulating the circuit.

The LM317 feeding the BJT will simulate a battery, not a cap, expect an equivilant 1 farad output with a 20 amp peak capability and an ESR of ~0.1ohm so long as your switcher continues to deliver more than 1v above the output voltage at the collector.  Basically the LM317 is buffering and multiplying it's 100uf at it's Vin (since it is a negative feedback op-amp circuit, this is a huge number) while the BJT takes that multiplied output and further multiplies by another 100.

For heatsinks, if you could press-fit one of these ontop of both BJTs, yes ontop of the plastic case as each will be radiating only 2 watts, it will probably work fine if you even need it.  The space in that Dreamcast is big and your PCB may not build up anywhere near the amount of heat the original linear supply did.

1 Heat Sink for both - mounting permitting
(Too bad it wasn't just a tad longer, you cannot beat 1 unit at 25cents for both.):
(Though 2 of them side-by-side would be perfect for length for 3 BJTs):
https://www.arrow.com/en/products/v5619a/assmann-wsw-components-inc
Individual heat sinks (Yes, each one costs more):
https://www.digikey.com/product-detail/en/assmann-wsw-components/V5618A/AE10819-ND/3511413

I'd say think about using 1 switcher and the 2 diodes.  You will only increase your radiating heat by ~2.5 watts on the opposite side of the PCB.  And the diodes are only  ~16 cents each (S5KBHR5G).  Though your switcher will need to have enough output current for both +5v and +3.3v together.

If you 3.3v supply draws 1.5 amps continuous, each diode will radiate ~1.25 watts + the BJT will radiate ~2.25 watts.
You haven't specified how much current the 5v takes.

[BrianHG]

Hmmm this changes things, Quote: https://bitbuilt.net/forums/index.php?threads/dreamcast-r-d-and-documenting.1368/

Main Voltage lines:
12v
-Not needed to boot
-Used by disc drive laser

5v
-Required to boot
-Uses .42a

3.3v
-Required to boot
-Uses 2.8a
-Powers the 2v and 2.5v lines
-Uses .4a when not feeding onboard linear regulators

It looks like you need the full 3 amps for the 3.3v.  I might tune your switcher to the bare safe minimum of 4.5v making your worst heat output 3.6 watts instead of 4.5 watts.

Now, 0.5amps for 5v, this would be 3.5 watts of heat if you just powered the 5v linear regulator from the 12v supply directly.
Your waste heat with 1 switcher will be ~7 watts.
With 2 switchers, your waste heat will be around 4.5 watts (including the minute added heat of the second switcher).
The difference is 2.5 watts of heat to include a second switcher.
Maybe look for a smaller 1amp switcher for the 5v which may use smaller components & lower power inductor.

The 'AOZ1280CI' looks tiny and dirt cheap giving you 1.2 amps, double what you need for 5v.
You will only need to heat-sink the 3.3v BJT (if at all, I mean I have un-heat-sinked fpgas drawing 5 watts and the PCB spreads out the heat wide enough to radiate it away air convection) as the 5v one will only dissipate 0.7 watts max.

[VEGETA]

Check this one please: https://www.beharbros.com/product-page/dreamcast-power-supply

this one is more modern than dreampsu and it seems to use a heatsink stuck on the regulators... how? how does this press-fit work? I mean can I do it myself? since the assembly service from jlcpcb will not.

I can buy either one heatsink for both transistors or one for each since either solution is gonna be just 1$ total.

this one for example has 48 degrees per watt, assuming say 3 watts from each transistor = 6 watts = 288 degrees? horrible. am I calculating wrong? the other one you posted has 80 degrees per watt.

while this one has only 18 degrees per watt -> 3x18 = 54 degrees as absolute max (real max might be less than half). this one is a bit pricey and i don't know how should i solder it.

check these out: https://lcsc.com/products/Heat-Sinks_441.html

we can drill 2 holes in one of these:

https://lcsc.com/product-detail/Heat-Sinks_XSD-XSD183-097-B_C286209.html
https://lcsc.com/product-detail/Heat-Sinks_XSD-XSD35-014_C286197.html

then screw it to the board to squeeze onto the regulators plastic... dunno if this works.

dirt cheap heatsinks! didn't know they exist there. anyone we could use?

I'd say think about using 1 switcher and the 2 diodes.  You will only increase your radiating heat by ~3 watts on the opposite side of the PCB.  And the diodes are only  ~16 cents each (S5KBHR5G).  Though your switcher will need to have enough output current for both +5v and +3.3v together.

I should search for equivalent on lcsc and jlcpcb.com/parts for this diode.

anyway, putting all items on one side of the board taken into consideration... this forces me to put these 2 very far from the npn transistors. One switcher will be limited to only 4 amps... it is enough though since the ENTIRE dreamcast psu is just 22 watts xD. assuming 6.3v @ 4 amps = 25.2 watts + the 12v sources is gonna be way more than original 22 watts.

but assuming diodes has 2 amps = 2 x 1.1 = 2.2 watts, could be more if used more than 2 amps. however this diode SS52 is better as it seems. It has 0.55 dropout at 5 amps which gives max of 1.65 watts at 3 amps which is about 83 degrees of temperature, so putting 3 of them in series will get the job done, even 4 is ok. price is mere 0.05$ for one.

4 of them = 2.2v drop --> 6.3 - 2.2 = 4.1v for the 3.3v regulator.


___________

The 'AOZ1280CI' looks tiny and dirt cheap giving you 1.2 amps, double what you need for 5v.

this one is about 30 cents which is 10 cents lower than 1284 variant... it is much tiny though.

people do mods for the DC like adding noctua fan (50 ma current @ 5v) and rgb leds... dunno if 1.2a will be good enough. I don't mind spending the extra 10 cents to get a way better switcher.

getting 2 switchers solution is about this:

aoz1284 x2 = 0.8$
npn transistor x2 = 0.5$
lm317 x2 = 0.3 $
10uH or so inductor x2 = 0.5$
caps (all) = ~2$
resistors (all) = ~1$
heatsinks = 2$ max

total = 7.1$

assuming board assembly + shipping + handling + customs + etc = 10$, then total will be 17.1$ assumed to be 20$ per board complete. I think I can sell it with 50$ but I wanted it to be less.

[BrianHG]

Check this one please: https://www.beharbros.com/product-page/dreamcast-power-supply

this one is more modern than dreampsu and it seems to use a heatsink stuck on the regulators... how? how does this press-fit work? I mean can I do it myself? since the assembly service from jlcpcb will not.

Those heatsinks are purchased with a thermal conductive tape/sticker on the bottom where you just peel off the back and press fit it on.
For heatsinks without stickers, suppliers of thermal conductive tape exist.

I did not know the Dreamcast uses 3 amps at 3.3v.  Do not use the diode idea.
As for 5v, even with added LED mods, if they are powered from 5v, ok just use another 4amp switcher again, except dont expect to draw more than an additional amp for LEDs.  5 watts of LEDs means ~50watts of equivilant incandescent light bulbs.  Are you trying to light up a room?  Still, the second BJT with an additional 1amp for LEDs will generate 1.8 watts of heat, half the 3.3v BJT. That's 4.2 watts, 5.4 watts with 1amp for leds, for both BJTs total if you tune each switcher to only supply +1.2v above the final output voltage.

I mean, tuning +1.1v above would be the absolute bottom and might let a occasional power dips, you would need to measure, but with this setup, you are now down to 3.9watts without LEDs, 5 watts with.  This ~0.4 watt savings is becoming silly as you might compromise the quality of what you are trying to achieve.

[BrianHG]

Thermal tape:

https://www.amazon.com/s?k=Heat+Sink+Thermal+Tape

It's double sided tape.  Since you only need a little square on each BJT D2PAK, those 2x25 meter rolls for 10$ will make you something like 5000 units.

Or you can buy heatsinks with the tape already on like here:
Heatsinks With Tape

[BrianHG]

Actually, I don't trust that Amazon tape, go for real 3M brand:
https://www.digikey.com/product-detail/en/3m-tc/1-2-5-8810/3M10315-ND/2649860

[BrianHG]

If you truly want to go massive on the heatsink, just place the connectors on the bottom edges of your PCB, and tape 1 huge rectangular heatsink over the entire bottom of the PCB and mount it with that side up in the Dreamcast.  This means not through-hole components anywhere in the middle of the PCB.  Gonna need one of those really wide rolls of thermal conductive tape.

[BrianHG]

The thermal epoxy (not glue) should do a better job than those Amazon blue tapes:
https://www.amazon.com/s?k=heatsink+thermal+epoxy&ref=nb_sb_noss

LOL, in the past, I used crazy glue in a pinch and for ~5 watts, it was fine for a year until the heatsink came loose.

The some of the authentic high temperature double sided thermal tapes have a super-thin screen wire mesh in them with a gooey white glue, but these tapes cost something like 150$ a roll.

[VEGETA]

kindly check the attached picture. Maybe this is the working main circuit, besides stuff like connectors and so on.

as for heatsinks with these pads, are they gonna stick nicely? I mean moving the device and so on, stop sticking and getting loose with time...etc.

If I get that tape, I will need to cut the pieces myself which may not be 100% the same area size of the plastic of the IC, is it ok?

Here is my little calculations:

3.3v rail has 4.6v inputs and 1.3v dropout -> expected consumption is 3 amps -> 3*1.3 = 3.9 watts.
5v rail has 6.24v input and 1.24v dropout -> absolute max is 3 amps -> 3*1.24 = 3.72 watts.

I think for maximum of 4 watts, these small heatsinks will do the job. I just don't want them to be very hot like frying eggs on them since that will be worse heat than original psu. what do you think?


we are talking about an absolute max of 7-8 watts total... will this be enough: https://lcsc.com/product-detail/Heat-Sinks_XSD-XSD183-097-B_C286209.html

enough for both?

like putting these 2 NPNs next to each other then stick this on top on their plastic.