Design AC-DC power supply using Innoswitch and custom transformer

[xvr]

Which polifuse you plan to use? Proper selection of them could be challenging 
Good quality once can costs more than all other stuff (including HT-Link module). Cheaper one could trip in short circuit case after all PSU will burned out 
Plain fuses here could be safer.

[mariush]

They're probably used in apple chargers because iPhones can negotiate voltages higher than 5v, so they needed a 25v capacitor I guess.. you don't.

On 5v and 3.3v you could use 6.3v rated or 10v rated solid capacitors...
These NXH 3300/25v are 13mOh esr, 3.6A ripple current in 16x25 package...

Former Samxon, now X-Con / Man Yue Tech 1500uF 10v ULR series 8x16 solid capacitors are 10 mOhm esr, 5.9A ripple current and they're 10 cents each at lcsc : https://www.lcsc.com/product-detail/Solid-Capacitors_Man-Yue-Tech-ULR158M1AF16RR_C840181.html

Aishi SPZ 1500 / 6.3v  1500 6.3v has 10mOhm / 5.4A current and it's 15 cents each : https://www.lcsc.com/product-detail/Solid-Capacitors_AISHI-Aihua-Group-SPZ0JM152F11O00RAXXX_C160193.html

Apaq 1500/6.3v 7mOhm 6.1A 10x11.5mm 3000h @ 105c are around 25 cents each : https://www.lcsc.com/product-detail/Solid-Capacitors_APAQ-Tech-6R3AREP152M10A2T_C2681208.html

They're 2000h but the lifetime rating of solid capacitors is calculated differently than electrolytics... you double the lifetime with every 10c drop on electrolytics, while with solid capacitors the formula is different, something like  L = L orig x 10^ (Tm-Ta)/20

So for example 10k@105 electrolytic at 65c ambient is 20k@95, 40k@85, 80k@75, 160k@65c  and a 2000h solid capacitor  is 2000 x 10^ (105-65)/20 = 2000x10^2 = 2000 x 100 = 200,000 hours

[VEGETA]

Which polifuse you plan to use? Proper selection of them could be challenging 
Good quality once can costs more than all other stuff (including HT-Link module). Cheaper one could trip in short circuit case after all PSU will burned out 
Plain fuses here could be safer.

chosen components so far:

CMC: RC212-0.5-10M
fuse: BK250-400-SZ a bit higher than required?
resistor: YAGEO PNP300JR-73-6R8
x cap:  PANASONIC ECQUAAF224K
y cap:  TDK CD45-E2GA102M-VKA
varistor:  Brightking 561KD14

still not final though.

They're probably used in apple chargers because iPhones can negotiate voltages higher than 5v, so they needed a 25v capacitor I guess.. you don't.

On 5v and 3.3v you could use 6.3v rated or 10v rated solid capacitors...
These NXH 3300/25v are 13mOh esr, 3.6A ripple current in 16x25 package...

Former Samxon, now X-Con / Man Yue Tech 1500uF 10v ULR series 8x16 solid capacitors are 10 mOhm esr, 5.9A ripple current and they're 10 cents each at lcsc : https://www.lcsc.com/product-detail/Solid-Capacitors_Man-Yue-Tech-ULR158M1AF16RR_C840181.html

Aishi SPZ 1500 / 6.3v  1500 6.3v has 10mOhm / 5.4A current and it's 15 cents each : https://www.lcsc.com/product-detail/Solid-Capacitors_AISHI-Aihua-Group-SPZ0JM152F11O00RAXXX_C160193.html

Apaq 1500/6.3v 7mOhm 6.1A 10x11.5mm 3000h @ 105c are around 25 cents each : https://www.lcsc.com/product-detail/Solid-Capacitors_APAQ-Tech-6R3AREP152M10A2T_C2681208.html

They're 2000h but the lifetime rating of solid capacitors is calculated differently than electrolytics... you double the lifetime with every 10c drop on electrolytics, while with solid capacitors the formula is different, something like  L = L orig x 10^ (Tm-Ta)/20

So for example 10k@105 electrolytic at 65c ambient is 20k@95, 40k@85, 80k@75, 160k@65c  and a 2000h solid capacitor  is 2000 x 10^ (105-65)/20 = 2000x10^2 = 2000 x 100 = 200,000 hours

X-Con, SamXon, Man Yue Tech, Aishi... do you recommend these brands?

I will definitely use solid caps in next revision... actually in my current project (saturn psu) i am using ceramics + solid polymers at output (+ one small 10u elec for damping), for switchers input i use ceramics + one sold cap (270u) and the main output from the module is CLC filter and the caps are elec. caps

for saturn, 5v is main rail with 2 amps.. right now i put 5 22u 1210 ceramics + 3x470u polymers + one 10u elec to help with damping a bit.

 

[mariush]

The solid capacitor division of Samxon was branded / named X-Con ... both brands are owed by Man Yue : https://www.manyue.com/02about06.html  - so all three are same thing.

Samxon has a decent reputation, the capacitors are used in Corsair power supplies and other brands, the solid capacitors are used often on video cards ... for example here they are on a nvidia gtx 970 card : https://www.reddit.com/r/AskElectronics/comments/122b95q/anyone_know_the_exact_capacitor_on_this_msi_gtx/

Apaq capacitors are used a lot on Asus motherboards on VRMs, I think also on Gigabyte boards and Asrock motherboards... for example you can see the logo all over this Asrock board : https://www.techpowerup.com/review/asrock-b550-pg-riptide/3.html  or on this Gigabyte board https://www.gigabyte.com/Motherboard/B550M-AORUS-PRO-P-rev-12#kf 

Aishi are less common, it's more  popular inside led lightning, they're less known for solid capacitors

[VEGETA]

The solid capacitor division of Samxon was branded / named X-Con ... both brands are owed by Man Yue : https://www.manyue.com/02about06.html  - so all three are same thing.

Samxon has a decent reputation, the capacitors are used in Corsair power supplies and other brands, the solid capacitors are used often on video cards ... for example here they are on a nvidia gtx 970 card : https://www.reddit.com/r/AskElectronics/comments/122b95q/anyone_know_the_exact_capacitor_on_this_msi_gtx/

Apaq capacitors are used a lot on Asus motherboards on VRMs, I think also on Gigabyte boards and Asrock motherboards... for example you can see the logo all over this Asrock board : https://www.techpowerup.com/review/asrock-b550-pg-riptide/3.html  or on this Gigabyte board https://www.gigabyte.com/Motherboard/B550M-AORUS-PRO-P-rev-12#kf 

Aishi are less common, it's more  popular inside led lightning, they're less known for solid capacitors

I will add them to my list of accepted brands then, first time i know they are used in such very high end stuff!

for saturn as i said i used 5x 22u 1210 x7r caps + 3x 470u polymer 7mOhm Lelon. do you think i am over-doing it? notice that saturn psu will have better routing and layout as seen in attached images... still not complete though. I did a forced ground routing to caps negative terminals and the bottom ground pour (hidden in images) not connected to ground connections at output to prevent current from going away from caps negatives... but rather connected up near the main ac-dc module itself.

i will see after Eid how stuff is gonna be and will get a small prototype

if it is fine, then i will do a revised edition of DC one to be like this with fewer caps and more solid caps in general + better layout. I could also try hi-link ac dc module.

[mariush]

for saturn as i said i used 5x 22u 1210 x7r caps + 3x 470u polymer 7mOhm Lelon. do you think i am over-doing it? notice that saturn psu will have better routing and layout as seen in attached images... still not complete though. I did a forced ground routing to caps negative terminals and the bottom ground pour (hidden in images) not connected to ground connections at output to prevent current from going away from caps negatives... but rather connected up near the main ac-dc module itself.

i will see after Eid how stuff is gonna be and will get a small prototype

if it is fine, then i will do a revised edition of DC one to be like this with fewer caps and more solid caps in general + better layout. I could also try hi-link ac dc module.

Without knowing the maximum output current on each voltage and what switching regulators you plan to use, I really can't tell if the layout is good.

But most likely you don't need X7R capacitors, especially in that big capacitance value, as they would be expensive.

Just paralleling capacitors to reduce ripple is ok, but you'd get more results with some filters like capacitor - inductor - capacitor , or you could get less noise if you go with a 2 phase dc-dc converter

Also, considering the small size of the board and the low volume of boards you may sell and high markup, it may be worth going with a 4 layer board as it wouldn't cost more than maybe a couple dollars extra per board. Then you'd be able to have a whole inner layer and most of the bottom as ground and use top and some of the bottom for traces.

Some random google searches point to being several revisions , some with 9v instead of 12v, some with 3.3v instead of 5v... you'll have to account for that. seems like max 21w in total (they use a top202yai which is advertised as 15-30w in the original psu shown here https://www.pixelrefresh.com/resaturn-psu-review-when-your-sega-saturn-power-unit-dies/ ) or something in that range...

The replacement psu in the link above used jumpers for selection, I'd consider using some elegant surface mount slide switches or dip switches instead of jumpers

[VEGETA]

for saturn as i said i used 5x 22u 1210 x7r caps + 3x 470u polymer 7mOhm Lelon. do you think i am over-doing it? notice that saturn psu will have better routing and layout as seen in attached images... still not complete though. I did a forced ground routing to caps negative terminals and the bottom ground pour (hidden in images) not connected to ground connections at output to prevent current from going away from caps negatives... but rather connected up near the main ac-dc module itself.

i will see after Eid how stuff is gonna be and will get a small prototype

if it is fine, then i will do a revised edition of DC one to be like this with fewer caps and more solid caps in general + better layout. I could also try hi-link ac dc module.

Without knowing the maximum output current on each voltage and what switching regulators you plan to use, I really can't tell if the layout is good.

But most likely you don't need X7R capacitors, especially in that big capacitance value, as they would be expensive.

Just paralleling capacitors to reduce ripple is ok, but you'd get more results with some filters like capacitor - inductor - capacitor , or you could get less noise if you go with a 2 phase dc-dc converter

Also, considering the small size of the board and the low volume of boards you may sell and high markup, it may be worth going with a 4 layer board as it wouldn't cost more than maybe a couple dollars extra per board. Then you'd be able to have a whole inner layer and most of the bottom as ground and use top and some of the bottom for traces.

Some random google searches point to being several revisions , some with 9v instead of 12v, some with 3.3v instead of 5v... you'll have to account for that. seems like max 21w in total (they use a top202yai which is advertised as 15-30w in the original psu shown here https://www.pixelrefresh.com/resaturn-psu-review-when-your-sega-saturn-power-unit-dies/ ) or something in that range...

The replacement psu in the link above used jumpers for selection, I'd consider using some elegant surface mount slide switches or dip switches instead of jumpers

switchers are AP62300, switching frequency is fixed 750khz. 5v is at 2 amps while 3.3v is so much lower, most likely just half an amp or so. i already have circuit for 9v or 12v selection but it is not shown in the images, using jumpers for now since people here advised me previously that sliding switches do not handle high current. the current need to pass the switch\jumper in this design.

i think this these parallel low esr caps + good layout shown i can get low noise performance. going 4 layers is a viable option but first i will test normal 2 layer board to see first prototype result and then i judge what is going to happen.

looks like i got convinced on using Hi-Link power module suggested here instead of mornsun one.... due to price most likely. can't beat 2.5$ for quantity of 100! especially that in both cases i need to do the input protection myself which i did.


the replacement psu you linked requires external adapter as you know. they always suggest high powered adapter but really the saturn is rated at about 12w so a 15w power module is good enough for it. testing will make sure this is correct too.


EDIT:

a had a stupid idea for the DC power supply to use 2 parallel ac-dc modules (mornsun ls15). 2 rated at 15W to be 30W. both modules + input protection would be less than high quality recom one. is this solution reasonable or as i said, stupid?

[jkostb]

CMC: RC212-0.5-10M
fuse: BK250-400-SZ a bit higher than required?
resistor: YAGEO PNP300JR-73-6R8
x cap:  PANASONIC ECQUAAF224K
y cap:  TDK CD45-E2GA102M-VKA
varistor:  Brightking 561KD14

Regarding Y1 capacitors. You can also pick 2 SMD Y2 capacitors and place in series (which is equivalent to Y1 cap). Have a look at safety capacitors from Johanson.  You need to verify with the applicable safety standards but I think you need a bleeder resistor  with X-cap.  You need to investigate in further detail the inrush current of Mornsun module. I think the module you selected has no soft start. I see inrush currents of 20A in datasheet. You need to ensure that resistor/fuse can handle this. Since you have no way to perform conducted emission and other EMC tests, you should follow the recommendations in datasheet. Datasheet recommends input choke.

[VEGETA]

CMC: RC212-0.5-10M
fuse: BK250-400-SZ a bit higher than required?
resistor: YAGEO PNP300JR-73-6R8
x cap:  PANASONIC ECQUAAF224K
y cap:  TDK CD45-E2GA102M-VKA
varistor:  Brightking 561KD14

Regarding Y1 capacitors. You can also pick 2 SMD Y2 capacitors and place in series (which is equivalent to Y1 cap). Have a look at safety capacitors from Johanson.  You need to verify with the applicable safety standards but I think you need a bleeder resistor  with X-cap.  You need to investigate in further detail the inrush current of Mornsun module. I think the module you selected has no soft start. I see inrush currents of 20A in datasheet. You need to ensure that resistor/fuse can handle this. Since you have no way to perform conducted emission and other EMC tests, you should follow the recommendations in datasheet. Datasheet recommends input choke.

i do have common mode choke, what other choke you mentioned? you mean "LDM"?

i will look into having 2 y caps in series for sure.

i tried applying "Application circuit 3——Universal system recommended circuits for indoor industrial environment".

note 2 says:

Note 2: According to the certification requirements, the X capacitor needs to be connected in parallel with the bleeder resistance, the recommended
resistance value is less than 2.5MΩ, and the actual need to be selected according to the certification standard.

so it is just a resistor, say 1 mega ohm 1 watt or so, in parallel with x cap? seems easy and cheap.

[jkostb]

Please note that there the bleeder resistor selection must meet requirements in safety standards. Resistor value selection is base d on discharge time in safety standard.

common mode chokes are for damping common mode currents. The Mornsun power supply draws impulse input currents. For this reason you often see a choke (LDM)  in series with input. This way the impulse currents are filtered by input capacitor. You need to pick the application note section for indoor class B emission. Please also note that the Mornsun application section for indoor has no surge protection. My advice is add surge protection! The mornsun power supply does not specify switching frequency. You need to select common chokes matching with mornsun power supply. Hope this helps

[VEGETA]

Please note that there the bleeder resistor selection must meet requirements in safety standards. Resistor value selection is base d on discharge time in safety standard.

common mode chokes are for damping common mode currents. The Mornsun power supply draws impulse input currents. For this reason you often see a choke (LDM)  in series with input. This way the impulse currents are filtered by input capacitor. You need to pick the application note section for indoor class B emission. Please also note that the Mornsun application section for indoor has no surge protection. My advice is add surge protection! The mornsun power supply does not specify switching frequency. You need to select common chokes matching with mornsun power supply. Hope this helps

i was hoping to design input protection circuit which suits not just mornsun module but others like hi-link ones. more of a universal input protection which can work for most options.

adding a series inductor for low current of 0.5A is not a big issue, they specify 1.2mH meaning about 1200uH? many of them on LCSC like these 2:

https://www.lcsc.com/product-detail/Power-Inductors_COILMX-MACM5045-102_C19076710.html
https://www.lcsc.com/product-detail/Power-Inductors_PROD-Tech-PSPHAQ127-102M_C2983085.html

what surge protection is to be added in your opinion? I mean what more am I missing to make the circuit very safe and robust?

I could look for mornsun\meanwell recommendation and follow it with using other brands like hi-link. since most of it is the same, even switching frequency is similar which is on the lower end.

[mariush]

switchers are AP62300, switching frequency is fixed 750khz. 5v is at 2 amps while 3.3v is so much lower, most likely just half an amp or so. i already have circuit for 9v or 12v selection but it is not shown in the images, using jumpers for now since people here advised me previously that sliding switches do not handle high current. the current need to pass the switch\jumper in this design.

i think this these parallel low esr caps + good layout shown i can get low noise performance. going 4 layers is a viable option but first i will test normal 2 layer board to see first prototype result and then i judge what is going to happen.

looks like i got convinced on using Hi-Link power module suggested here instead of mornsun one.... due to price most likely. can't beat 2.5$ for quantity of 100! especially that in both cases i need to do the input protection myself which i did.


the replacement psu you linked requires external adapter as you know. they always suggest high powered adapter but really the saturn is rated at about 12w so a 15w power module is good enough for it. testing will make sure this is correct too.


EDIT:

a had a stupid idea for the DC power supply to use 2 parallel ac-dc modules (mornsun ls15). 2 rated at 15W to be 30W. both modules + input protection would be less than high quality recom one. is this solution reasonable or as i said, stupid?

You're not switching the voltage, you're switching between two resistors on the feedback of the switching regulator.

For example, if you stay with AP63200, the datasheet recommends 31.6k and 10k for 3.3v out,  and 52.3k and 10k for 5v out.

So you could have 31.6k in series with a 20.7k ohm resistor (or a 20.5k if you want to stick to E48 resistors and get  ~4.97v out) and if you want 3.3v out, the jumper/ slide switch simply shorts out the 20.7k resistor creating a 0 ohm path, so the regulator will only see 31.6k and 10k  - With a jumper there's a small downside as in if the jumper is missing, the regulator defaults to 5v instead of 3.3v, there's less risk of damaging a console if you default to outputting the lower voltage.

With a slide switch, you could use a double slide switch and have a status led for each voltage light up, depending on the position of the slide switch only one of the leds would light up... so user could set the slide switches before connecting the cable to the unit, see the voltage leds light up and be sure he got it right, then turn off power and connect cable to console.   


As for 12v/9v out obviously won't get 12v out with 12v in, but you could just use a 15v DC out converter (your regulators are good up to 18v in) or  most good AC-DC converters will have a TRIM / ADJ pin that you could take advantage of to make the AC/DC converter output 12.5v ... 13.5v instead of 12v

You can get 15v 40w open frame power supplies for 8.5$ a piece if you order at least 5 : https://www.digikey.com/en/products/detail/mornsun-america-llc/LO45-10B15/13168116

or sealed ones for 8-11$  : https://www.digikey.com/short/vzn9v3ct

AP63200 is fine, relatively cheap ... if you want to optimize some more you could use buck regulators with 2 or more outputs.

For example, Richtek RT2723 is around 1$ on lcsc, around 2.12 EUR if you get 10 at Mouser , around 2.4$ at Digikey if you buy 10, but has 3 outputs, supporting 3A/2A/2A continuously , or maximum 3.5A/2.5A/2.5A

Digikey : https://www.digikey.com/en/products/detail/richtek-usa-inc/RT7273GQW/3986811

Mouser : https://eu.mouser.com/ProductDetail/Richtek/RT7273GQW?qs=amGC7iS6iy9%2Fyt7eO0%252B6kw%3D%3D

LCSC : https://www.lcsc.com/product-detail/DC-DC-Converters_Richtek-Tech-RT7273GQW_C3028966.html

See example at page 9 in datasheet... it's a more pain in the ass footprint, and you'd have a bunch more passives (resistors and ceramic capacitors), but you would get 3.3v , 5v and 9v with a single chip and could have 12v straight from the ac - dc converter.   

You could also get some inspiration from the evaluation board : https://www.richtek.com/m/~/media/EVB%20PDF%20File/EVB_RT7273GQW/EVB_RT7273GQW-00_EN.pdf

[VEGETA]

You're not switching the voltage, you're switching between two resistors on the feedback of the switching regulator.

i know about this but i was referring to the solution where i have 12v input, this way i need 15v input.


I can use this module: https://www.lcsc.com/product-detail/Power-Modules_HI-span-style-background-color-ff0-LINK-span-HLK-15M15C_C5120968.html

which outputs 15v to deliver it to your suggested triple buck IC, which I really found it a month ago when I was searching. I don't really know if this total price will be lower than my own using AP62300.

meanwell irm-20-15 is +8.3$ for 100 quantity while above module is about 2.5$. meanwell is better name brand + does not require external protection components while hi-link does. even mornsun module LS15-23B15DR3 requires external components.

mornsun LD15-23B15R2-M seems cheaper and does not require much external components.

so now i have 4 options:

1- use very cheap Hi-link module and build all input protection myself.
2- use Mornsun LS15-23B15DR3 module and build all input protection myself, similar to option 1.
3- use Mornsun LD15-23B15R2-M module, just put a mov and input series resistor as everything else is inside.
4- use Meanwell module which has everything built in.

money-wise, full input circuit protection is going to be around 1$ or so.

rough estimate for options in 100 quantity:
1 = 3.5-4$.
2 = 6$
3 = 8$
4 = 9-10$

For example, Richtek RT2723 is around 1$ on lcsc, around 2.12 EUR if you get 10 at Mouser , around 2.4$ at Digikey if you buy 10, but has 3 outputs, supporting 3A/2A/2A continuously , or maximum 3.5A/2.5A/2.5A

Using Richtech one, Vout1 = 5v, Vout2 = 3.3v, Vout3 = 9 or 12v by switching feedback resistor. certainly can be done but the issue is that the power supply area is a rectangle with about 50mm width or less. no big room but will see if i can make it work.

evaluation board used relatively cheap inductor, 4.7uH with about 23mOhm resistance. I prefer using highest switching frequency but lower resistance inductors will be big in size. achieving low ripple is important but also low heat. looks like for 2.2 mhz i need 600kohm resistor. what do you use in such situations to achieve smallest ripple available with good thermals?

for filtering output caps, i could still use the 22uF 1210 X7R ones I used before... plus some polymers in parallel. having lots of ceramics in parallel can be achieved in small spaces.

there is the compensation which i am always afraid of. they suggest 4.7nf cap in series with 20kohm resistor for the output caps of 44uF ceramic. they suggest doing a ratio to reduce it for 22uf output caps it should be 10k resistor. this way if i want 4x22u = 88u caps + 470u poly it would be very small compensation cap... more like 1.58k ohm or something. i could not find an equation to do it.

it is fair to say that I need more filtering caps for the big amps rail of 5v 2a, the others 3.3v at 0.6a or 12v\9v at very low amps probably won't require more than 2x22u ceramics.

__________

I am kinda confused now since I wanted this PSU to sell for max of 60$. if i go with best options:

meanwell irm-20-15: 8.5$
richtek switcher: 0.8 (assume all from lcsc, which now is just 24 units in stock)
ceramic caps: ~0.3
other caps: 0.2
resistors: 0.1
connectors: 2 or so
pcb: 7.5
inductors: 0.6

total = 20$ for 100 quantity. more for less quantity.

it can be sold for 60$ and even for 50$ but if i have the other options i could save more.... if you were in my place, what would you choose?

[mariush]

The HiLink doesn't have input filtering  (common mode choke, ptc/ntc for inrush current limiting, fuse, x2 capacitors ... you can find all these on those open frame power supplies I mentioned) - that's how you get that cheaper.

You have to consider the old consoles output analogue signals, which may pick up the noise the switch mode psu may output back into the grid

You're not that space constrained that you couldn't just put 4 AP62300, or the fixed voltage versions of it for 3.3v and 5v (ap63205, ap63203) and a couple adjustable versions ... if you go with AP63200 / AP63201 then all your regulators would support up to 32v input and output 2A each ... if you want 3A there's AP63300 that can do 3A and it's adjustable.


and as they require much fewer passives you could afford to even buy some RF shields and put them over the regulators to further reduce switching noises ...see https://www.digikey.com/short/zrfqr7m7 for examples

Personally I'd put a barrel jack connector (or two, the most common two sizes 2.1 ID and 2.5 ID) and sell the kit with a mass produced certified 15v-16.5v-19.5v or even 20v / 24v laptop adapter, then all the AC switching can sit down on the floor near the socket, you just do the dc-dc conversion in the psu.

Alternatively you could have a usb type c input, and use a usb trigger ic to attempt to negotiate to 12v or 15v or 20v and you could bundle with a charger that supports usb power delivery or offer the board at a discount (ex 10-15$ off if you don't bundle the power supply)  ... works also for you as shipping costs will be less if you don't bundle a heavy power supply.

with usb type c you don't have the problem of power supplies barrel connectors being different sizes (ex 2.1 ID / 5.5 OD versus 2.5 ID / 5.5 OD) but not all chargers out there can go above 12v so it's not a perfect solution, could get people buying the wrong chargers.

Without the AC-DC converter on the board, you could make the whole board smaller and panelize it to get more boards for the price.

[VEGETA]

The HiLink doesn't have input filtering  (common mode choke, ptc/ntc for inrush current limiting, fuse, x2 capacitors ... you can find all these on those open frame power supplies I mentioned) - that's how you get that cheaper.

i know it does not, which is why i told you i need to do all input filtering myself. open frame stuff are not pcb solderable which is a requirement to me.

You have to consider the old consoles output analogue signals, which may pick up the noise the switch mode psu may output back into the grid

my psu was very low noise and ripple, image was very clear. no problem there and i want the same for saturn version.

You're not that space constrained that you couldn't just put 4 AP62300, or the fixed voltage versions of it for 3.3v and 5v (ap63205, ap63203) and a couple adjustable versions ... if you go with AP63200 / AP63201 then all your regulators would support up to 32v input and output 2A each ... if you want 3A there's AP63300 that can do 3A and it's adjustable.

AP62300 seems better choice than these. 2 of it in the saturn version fits as i send the pictures previously. 5v is maximum of 2 amps but i got the 3 amps just in case.

actually AP62200TWU here: https://www.lcsc.com/product-detail/DC-DC-Converters_Diodes-Incorporated-AP62200TWU-7_C2895288.html

is very cheap, < 16.38$ for 200 units (enough for 100 boards). while 3 amps equivalent is just 17$ so i think 3 amps is better choice. this is the T version, i will have to get it instead of regular one.

and as they require much fewer passives you could afford to even buy some RF shields and put them over the regulators to further reduce switching noises ...see https://www.digikey.com/short/zrfqr7m7 for examples

i really never thought of this. but it is safe to say i need to test without it first as a simple prototype then see what happens.

Personally I'd put a barrel jack connector (or two, the most common two sizes 2.1 ID and 2.5 ID) and sell the kit with a mass produced certified 15v-16.5v-19.5v or even 20v / 24v laptop adapter, then all the AC switching can sit down on the floor near the socket, you just do the dc-dc conversion in the psu.

such solutions exist now, people do not like them at all. everyone wants a complete high quality universal plug-and-play solution. my DC one got the hype and people are anticipating it eagerly.

preserving same look and feel is very respected in this hobby.

so for getting the ac -dc power module built in, what do you recommend? notice that safety and heat are important but not overly obsessive... like, no certification needed just ensured performance.

in DC, i opted for recom medically rated module. i wanted meanwell irm-30-12 but it was big and i could not fit it inside, especially that i used lots of big caps and 2nd stage...etc. (would revise that for sure). but for saturn i want something cheaper

i make them with pcbway and i have no complains but i feel board price and assembly are kinda not cheap. they offered to panelize the last batch of 40 units, they did a panel of just 2 boards. i wonder if i could panelize myself and make it cheaper by a significant amount.


_____

EDIT:

one trick i could use to lower cost and save lots of space is to use a small lower power inductor for 3.3v (and for 9 and 12v if used a regulator for them):

this one i found: https://www.lcsc.com/product-detail/Power-Inductors_cjiang-Changjiang-Microelectronics-Tech-FTC252012S3R3MBCA_C5832373.html

it has 80m Ohm DCR, not very good but switching frequency is not high and current is not high at all. it has very small footprint too while being 3.3uH.

here is another one with 34mOhm: https://www.lcsc.com/product-detail/Power-Inductors_DMBJ-CD54-3R3M-3-3UH_C2826659.html

but this one has exposed un-shielded wires...

while the one i currently use and will be used for 5v rail is: https://www.lcsc.com/product-detail/Power-Inductors_APV-APH1040T3R3M_C5349719.html

11.8mOhm but massive 11mm one.

as for alternative,
this one: https://www.lcsc.com/product-detail/Power-Inductors_Ceaiya-CR5040-3R3N_C520346.html

and this one from sunlord: https://www.lcsc.com/product-detail/Power-Inductors_Sunlord-SWPA5040S3R3NT_C305173.html

both are 24mohm and very small footprint. i don't know why i hate this package!

current solution for 9v output is a 7809. don't know if i can use the same solution for 12v-9v switching with resistor switching via slide switch... lm317 is in mind but would want it to be as cheap... found this lm317: https://www.lcsc.com/product-detail/Linear-Voltage-Regulators-LDO_UMW-Youtai-Semiconductor-Co-Ltd-LM317LIPK_C3029045.html

it should be ok since 9v and 12v are just for PAL scart auto switching... just a signal. since for supported tvs when they read voltages of 9 or 12 they switch to scart automatically

[NiHaoMike]

it should be ok since 9v and 12v are just for PAL scart auto switching... just a signal. since for supported tvs when they read voltages of 9 or 12 they switch to scart automatically

If that's all it's used for, wouldn't just 9V work for all revisions? For that matter, maybe it would make more sense to have the main supply output 5V or 3.3V and use a small boost converter or charge pump for the 9V.

[VEGETA]

it should be ok since 9v and 12v are just for PAL scart auto switching... just a signal. since for supported tvs when they read voltages of 9 or 12 they switch to scart automatically

If that's all it's used for, wouldn't just 9V work for all revisions? For that matter, maybe it would make more sense to have the main supply output 5V or 3.3V and use a small boost converter or charge pump for the 9V.

not quite sure it works for all, but i read about 9 and 12 being different... aspect ratio thing.

having 5v is mail output will save me a regulator but the ripple will be high since these modules are not optimized for low ripple. they produce 150-200mv ripple but i intend to get 20mv max... i achieved that with dreamcast one, more or less.

since the module for 5v and 15v kinda same price then it will be ok for having small extra for good switcher... while cheap out for 9\12v rail.

[VEGETA]

as for the new saturn PSU, I decided to use Meanwell IRM-20-12 or IRM-20-15 which give 12v and 15v output respectively.

the 15v one is 2.8$ cheaper. however, it would require a regulator for it to output 12v or 9v.

I wanted to use slide switch to switch between the feedback resistors to get 9v or 12v but i remembered another jumper switch option is required at one of the output pins, to switch between 5v or 3.3v. there i guess i cannot use slide switch since it will get higher current if 5v was the choice even if there is another 5v pin near it...

if so, then i would revert back to using a jumper for 12v\9v to keep it the same...

having 15v module seems better since 2.8$ price difference is too big... i mean i am currently using a cheap chinese lm317 chip with different feedback resistor swapped by a jumper. switching back to 12v module and 7809 won't save me 2$ difference. what do you think?

besides those big stuff, the rest is just cleanup and misc. stuff.

_______

For the revised version of Dreamcast PSU, since it uses an expensive and big lead time Recom module with high quality medical rating.. i found a Meanwell equivalent which is MFM series, namely MFM-30-5 which outputs 5v. it is the cheapest of the series tied with 24v one but if i use 24v module i would need an extra buck for 5v + higher voltage caps which will be noticeably more expensive... so 5v it is... as long as my filtering caps can reduce its noise.


 i use it with buck regulator (ap63357) for getting the main 3.3v rail, 5v rail is directly gotten from the module but with many polymer + ceramic caps to reduce ripple, and the 12v rail is via a boost regulator (tlv61048 + small inductor + ceramic output caps).

this saves some money, especially if i ditch the expensive input common-mode choke since i think the one inside module is enough + get a cheaper output connector alternative from alibaba (i got a sample and it works fine).

maybe i can get its price down to 75$ which will make it significantly more appealing.

[VEGETA]

I actually found some slide switches which can tolerate high current. 5v has 2 amps in total, so if 5v was chosen it will have 2 amps on 2 pins, roughly 1 amp per pin.

I got this:

https://www.lcsc.com/product-detail/Slide-Switches_SHOU-HAN-SS12D10G5_C7431054.html

2 amps rated, there are 3 amps ones and up to 6 amps...

while the slide switch required for switching between feedback resistors for 12 and 9v output is significantly cheaper and available.

i think i will go with these choices.

[mariush]

If you actually trust a tiny lubrified steel ball sitting between two contacts with amps of current... and on a 7 cent part... 

Still think it's best not to switch the actual voltages, switch the adjustment resistors or something like that.

There's making things cheap and then there's being ridiculous. You're planning to charge 50$ on something that costs you under 10$ to make and you're still looking where to cut costs.

[VEGETA]

If you actually trust a tiny lubrified steel ball sitting between two contacts with amps of current... and on a 7 cent part... 

Still think it's best not to switch the actual voltages, switch the adjustment resistors or something like that.

There's making things cheap and then there's being ridiculous. You're planning to charge 50$ on something that costs you under 10$ to make and you're still looking where to cut costs.

it will cost about 20$ according to my estimates, but still good.

I don't really trust these slide switches but they are one of my choices. the LM317 circuit for 9v/12v can use a cheap one to switch feedback resistors... that is done.

the remaining is to switch between 5v and 3.3v... 5v always on, but some variations require 3.3v alongside 5v. here i need to switch voltages not feedback resistors... since 5v regulator already exist and it is on. the 3.3v regulator works but it is either switched on to the load or not.

jumper seemed best choice overall as long as slide switches are outside the picture.

another possible robust solution is a toggle switch like this one: https://www.lcsc.com/product-detail/Toggle-Switches_YUEN-FUNG-ST-0-103-A01-T000-RS_C1788488.html

but the jumper is way cheaper and does the same thing... just that the switch is easier and more elegant.

what would you do?

[mariush]

You can use power switches with an EN pin , or p-channel mosfets, or ideal diodes with EN pin, and use the jumper / slide switch to pull the EN pins high or down, or to connect voltage to the gate of a mosfet or not (by default gate being pulled up or down with a high ohm resistor)

There's also solid state relays you could use...

Examples of power switches

TPS2121 , good for up to 4.5A of current : https://www.digikey.com/en/products/detail/texas-instruments/TPS2121RUXR/9859001  or https://www.digikey.com/en/products/detail/texas-instruments/TPS2121RUXT/9859003


from datasheet

If both inputs are valid, CP2 is low, and PR1 is pulled high, (higher than VREF, 1.06-V typical), then IN1 is used.
If both inputs are valid, CP2 is low, and PR1 is pulled low, then the highest voltage input is used.

So put 3.3v on IN1 , 5.0v on IN2 ... pull PR1 high and you get 3.3v out , pull PR2 low and you get highest voltage of the two, which is IN2 (5.0v)

For lower currents

LM66200  dual ideal diode , 2.5A  (auto switches between highest voltage) : https://www.digikey.com/en/products/detail/texas-instruments/LM66200DRLR/15856663?s=N4IgTCBcDaIDIFkBsSwAY0gLoF8g

The auto switch won't helpful, you ignore that, connect both inputs to same voltage.

It has an EN pin that disables both inputs, so you can have two of these, one that has both inputs on 3.3v , one that has both inputs on 5.0v, and connect the outputs... and  you turn off the ideal diode you don't want. 

For less current, LM66100 is the single ideal diode version that does up to 1.5A : https://www.digikey.com/en/products/filter/power-management-pmic/or-controllers-ideal-diodes/

Same deal, has an enable pin, pull high to turn it off (it uses a p-channel mosfet, so it's on by default)


Solid state relays ... TLP241A does 2A in a optocoupler package : https://www.digikey.com/en/products/detail/toshiba-semiconductor-and-storage/TLP241A-TP1-F/6574627

[VEGETA]

ideal diode seems good idea, found these ICs cheap on LCSC.

However, the on-resistance will have its effect. at about 0.7 amps for 3.3v rail it will be 0.7*0.7*0.091 (LM66100) = ~0.045v drop -> 3.3v rail will be 3.255v so compensation must be done... that is if it does not exceed 0.7 amps. for 5v at 2 amps, assuming 1 amp per pin with LM66200 will be 1*1*0.04 = 0.04v drop -> 4.96v total.

for 5v, if i made it 5.05v then it will be ok but will have 2 pins of 5v rail being of different voltage.. on is 5.05v and the other is 5.0v. don't know what to do with this.

by ideal diode i think they mean a mosfet switcher circuit right? so can we do it in a cheaper way + using very low resistance mosfet?

[mariush]

Unless I have a brain fart, the voltage drop is V = Current x Resistance =  0.091 x 0.7 = 0.06v  ... Power is I^2 x R

I suppose you could configure the regulator to output 3.4v but probably not needed ... you're gonna have voltage drop on the wires between your psu and the console anyway, so the whole console won't be that sensitive about exact voltages...

[VEGETA]

Unless I have a brain fart, the voltage drop is V = Current x Resistance =  0.091 x 0.7 = 0.06v  ... Power is I^2 x R

I suppose you could configure the regulator to output 3.4v but probably not needed ... you're gonna have voltage drop on the wires between your psu and the console anyway, so the whole console won't be that sensitive about exact voltages...

it was my mistake, yes drop will be 0.06v for that. still needs compensation.

however, i found this p-mosfet: https://www.lcsc.com/product-detail/MOSFETs_Diodes-Incorporated-DMP1005UFDF-7_C461046.html

i can get 2 of it, hook some resistance from its drain to gate to make it off by default. then get the slide switch to connect ground to desired mosfet gate to make it active.

wouldn't that work as well? still cheaper and this mosfet has very small resistance so it would not affect anything.

it was inspired by this: https://www.ti.com/lit/an/slva716a/slva716a.pdf?ts=1713171404977&ref_url=https%253A%252F%252Fwww.google.com%252F


I also found these:

https://www.lcsc.com/product-detail/Power-Distribution-Switches_Diodes-Incorporated-AP22966DC8-7_C441830.html
https://www.lcsc.com/product-detail/Power-Distribution-Switches_Texas-Instruments-TPS22976NDPUR_C106796.html

first is very cheap... both has 2 outputs so i can hook my 2 sources at inputs and connect both outputs together... make the slide switch activate only one output at a time.


EDIT: I looked but could not find Rload of 200-300 ohms as you mentioned. some tables has 10 ohms. i guess i will be ok.