Voltage regulation circuit 12V to 5V in 5A range

[Demon]

EDIT: Figured I'd concentrate my power supply circuit here instead of my MSFS thread.

-----------------------------------------------------------------------------

Figures that I would hit a roadblock after wiring an ammeter to be able to see current and voltage in real-time. That DC-DC converter under the meters has too much noise on VDD and VSS for stable ADC readings.

9806


I made a 7805 circuit for testing, but it's only rated for 1.5A.

9807


I wish I had seen the LM1084 before; it can handle 5V up to 5A, more than enough for my design. At least it's a piece of cake wiring a through-hole LM1084 with some caps instead of the DC-DC converter.
:)

[tumbleweed]

I wish I had seen the LM1084 before; it can handle 5V up to 5A
Just be aware that if you're starting with 12V VIN then the regulator is going to dissipate roughly 5W with 700mA out (from your other thread).
Pd = (12V-5V) x 0.7 = 4.9W

That's going to get quite hot... over 100degC rise depending on the package type.

[Ioannis]

Well, the SMPS you used are crap and although they say it is a LM25xx in fact they are just bad copies. Also the caps and inductors are crap too.

I would design my own SMPS and make sure to follow the manufacturer advices (either TI or Microchips). Having a goal of 20mV p-p is totally acceptable and doable.

Linear regulators will get too hot and while DC output, if well designed, may be clear enough, it is wasting too much power. You will need a big (and expensive) heat sink also, with ventilation holes on the final product.

Ioannis

[Demon]

I had 2 options originally:

- stay under 500mA and use USB power
- use a wall adapter and add a voltage regulator component.


USB power is no longer an option with the 4 LCDs. I don't have any "simple" design for a SMPS that would have noisy power.

The LM1084 "should" be as clean as the 7805 (I hope).

I can use the TO-220 package with cooling fins, and even add a tiny cooling fan. Tests will confirm how much heat I will generate with everything.

[Demon]

I bought a bunch of these a while back; aluminium heatsink with double-sided tape.

It fits just right on the back of a IRFZ44N (TO-220). I don't seem to have to bother with a screw/nut (unless you highly recommend a model with a fastener).

9809


Aliexpress:

8.8 x 8.8 x 5, I tried removing the fin once attached, and it doesn't come off unless I usetools.

https://www.aliexpress.com/item/1005004079109702.html


My alternative was this screw/nut model with thermal paste for maximum efficiency:

https://www.aliexpress.com/item/32882672963.html

I'd just prefer in bare aluminium; I don't like paint on a cooling surface, seems counter productive to me.

[Ioannis]

Paint on heatsinks is not for cosmetic reasons. Black heatsinks radiate more IR energy than bare aluminum ones.

As for the power loss on the regulator (either 7805 or 1084 or 317, it does not matter which), it depends on the Vin and the current you will need on the load.

Example:
Vin 12 volts
Vout 5 volts
Max current 1A

The power loss on the regulator will be (12-5)*1A=7 Watts.

According to datasheet Power dissipation diagram, at about 35 degrees celsius ambient temp, you will need better than 10°C/W heatsink. So if you opt for 50°C ambient temp, then around 5°C/W maybe just enough.

How this 5°C/W translates to a physical heatsink part? something like this https://www.mouser.ca/ProductDetail/Aavid/529801B02500G?qs=7jyBjEprRBgptH2UqBRD6Q%3D%3D&_gl=1*mpx3ow*_ga*MTY3NTMxNjY5NS4xNzI5NjgwMDk0*_ga_ 15W4STQT4T*MTczMDczMDczMC41LjEuMTczMDczMjU4NS4xNi4 wLjA.

Just tested a 7805 with 1A load and 12 volt input, roughly 7 watts of heat and the 7°C/W I used, is over 80°C with just 24°C ambient !!!

So, you will need a large and expensive piece of aluminum heat sink...

Better use a SMPS, I insist on this.

Ioannis

[Demon]

Option 1:

Do you have a schematic for a reliable SMPS that can do 12V to 5V somewhere under 5A?

I don't have a final current consumption cause I haven't built a complete unit.

And most importantly, will it generate a noise-free VDD/VSS?

[Demon]

Option 2:

How about these with a fan?

25mm x 34mm x 12mm
https://www.amazon.ca/Easycargo-Insulator-Rubberized-Regulator-Transistor/dp/B07H5GQVTD/?th=1


25mm x 7mm 12V brushless fan

https://www.amazon.ca/GDSTIME-25x7mm-Radiator-Brushless-Cooling/dp/B08CMMB6VJ/?th=1

[Demon]

Option 3:

Would it be easier to have 3 LM1084 with that regular black heatsink:

- one to power all PICs and ICs.
- one to power the 3 LED strips.
- one to power the 4 LCDs.

That way I'd get minimal noise on VDD/VSS, and limit heat loss to more manageable levels.

A sort of distributed power management system.

[amgen]

2 cents....... the LCD and LED's stuff aren't that badly affected by slight power supply fluctuations ... slight. The PIC's and IC's need good stable power 7805 or similar on their own circuit.

[Ioannis]

Are you sure you need 5Amps at 5 volts? That is 25 Watts of power.

And about 35 Watts of power loss on the Linear regulator. This means that you will need a monster heatsink like intel CPU with a fast spinning fan. Not anything close of what you posted.

Better use seperate linear regulator as amgen said for the PIC and a switcher for the power hungry LED and Displays.

Or check these:

Vin 4.5-24, Iout 3A: https://www.ti.com/product/TPS56339

Vin 4.5-17, Iout 5A: https://www.ti.com/product/TPS565208

Following the design examples you may have less or equal to 20mV noise. For me good enough even for PIC using ADC.

Ioannis

[Demon]

2 cents....... the LCD and LED's stuff aren't that badly affected by slight power supply fluctuations ... slight. The PIC's and IC's need good stable power 7805 or similar on their own circuit.

Yup, I'm thinking of running a dual system.



Are you sure you need 5Amps at 5 volts? That is 25 Watts of power. ...

Yeah, I'm waiting for a 9V 2A wall adapter to help reduce the waste heat. I'm getting 7V 2A wall adapter tomorrow to reduce that even more.


I'll probably need 2A, but I hate running stuff at peak power, hence why I said 5A.

A component running near max is generating more heat than another running at 50% power; that's not good for long life.

[Demon]

I did encounter a problem with the LM1084-5. Voltage drops when I add load.

I expected a specialized IC like the LM1084 to remain at a stable 5V throughout its workable current range.

I used the bottom of figure 7.2.8 with 2 electrolytic capacitors:

9810

[Ioannis]

That 50 ohm resitor will worsen the regulating output. It forces the upper IC to have higher than 5V output but regulation depends now on the resistor.

Have you had a look at TPS565208? It is a chip with low noise and ripple, easy to use SOT-563 package, very small footprint and cheap. I would not mess with any Linear, heatsinks, wall adapters etc.

Only downside (maybe) is that it accepts only up to 17 volts DC. But if your source is upt to 12volts your are more than good to go.

Ioannis

[Demon]

I only used the bottom regulator on that schematic.

I have 9v 2a wall adapter coming in today, and a 7v 2a coming in tomorrow.

That should reduce wasted heat a lot.

[amgen]

I know you already ordered some things, I have used few of these things..... no other regulator needed. The output voltage is pretty stable.9811

[Ioannis]

Camparing the various chinese power supplies I ended up with Sunny brand, that is represented in Europe by a reputable firm. https://en.sunny-group.com

Opening one of them reveals that the design is in pretty good level.

Please be very carefull with all these cheap SMPS that are sold at low prices. They can have really catastrophic failures (I was almost hurt by this), with very bad "protection" circuits that fail after the big-bang...!

The idea of having a 5V wall adapter may seem good but there is a risk that the user may by mistake connect a 12V adapter at the 5V entry...

Ioannis

[amgen]

.

The idea of having a 5V wall adapter may seem good but there is a risk that the user may by mistake connect a 12V adapter at the 5V entry...
Very true. I wouldn't use that for product for resale, only for building and testing some circuits or maybe a simple small product. But I always hardwire power supply in to avoid exactly what you said....... That was interesting what you said about some dangerous failures, those SMPS's 'should' have the input separated from the output with some kind of high frequency transformer for 'galvanic isolation' meaning no actual electrical connection from in to out.

[Ioannis]

Yes, all of these SMPS do have that high frequency transformer and do isolate the secondary. Few have poor isolation.

Many have bad fuses in the line input and they blow after the explosion of low quality capacitor/bridge/MosFet. And had cases of pcb traces that were melted also.

These two images show WellMax LED panel drivers that drop my three phase protection fuse on the wall distribution box! The pcb fuse was the last to blow after burning pcb traces!

Had other similar case of 12V/5A power supplies that were poor made. I do understand that inrush currents are high enough and the fuse should be relatively slow, but they should add a thermistor then to lower that current and have a faster fuse.

Ioannis

[Demon]

I just noticed the LM1084-ADJ is claimed to "provide 5 A at high efficiencies and very low drop-out." I just ordered a few from Digikey.

Ioannis, will the TPS565208 or TPS56339 voltage drop when I add load?

Current consumption so far:
- 800mA, ICs and 4 LCDs
- 510mA LED strip 1
- 510mA LED strip 2
- 510mA LED strip 3

I'm at 2.330A and I have a few other ICs to add, hence why I target a power supply that can manage 5A.

The thing with the LED strips is that they can be dimmed from 0 to 100%, that's a respectable load change.

(I'll be limiting the LED strips to 350mA or so, but still, I want to be safe if they get stuck "full ON").

[Demon]

Piece of crap 7V wall adapter worked 10 minutes...

https://www.amazon.ca/dp/B091YHX8RY

[amgen]

Yes, all of these SMPS do have that high frequency transformer and do isolate the secondary. Few have poor isolation.
....
....
wow, looks like you were trying to take 100 amps from that little power supply..... (just kidding)
.
adding..... should probably never try to use a power supply at it's full rating, maybe 1/2 of rating is safer.







9

[richard]

chinese amps are quite a bit smaller than normal ones. 0.75 seems close

[tumbleweed]

I just noticed the LM1084-ADJ is claimed to "provide 5 A at high efficiencies and very low drop-out." I just ordered a few from Digikey.

The only difference between the LM1084-ADJ and the fixed output LM1084-5 is the latter already has the resistors to set the output voltage built inside it. You will only get 5A out of it if you have the input > the Vdo drop out voltage spec, and stay within the max power dissipation. With no heatsink you're not going to get anywhere near 5A out, not with 12V in.

You said the output voltage of the LM1084-5 dropped when you added a load. That shouldn't happen as long as you're within the datasheet specs.

If you're trying to get 2.5A out, with your 12V adapter that gives a Pd = (12v-5v) x 2.5 = 17.5W !!


Yeah, I'm waiting for a 9V 2A wall adapter to help reduce the waste heat. I'm getting 7V 2A wall adapter tomorrow to reduce that even more.

You'll need an adapter with more than 2A out if you want 2.5 or 5A out of the linear regulator.

[Ioannis]

Not so tumbelweed.

With Linear regulators, your Input current is the same with the output current! Beware of that! Plus the current of the regulator of course, though this is small enough.

The opposite happens with SMPS. In that case wattage is the same plus the losses in the regulator itself.

The voltage drop with increased load is called Load Regulation and is there in the specs. And of course it is different for every regulator, either linear or SMPS. Besides that spec, which is typically small enough (less that 1% for the TPS565208), the voltage will drop for many other reasons, like design of PCB, part selection, thickness of wires and the point where output voltage sample is taken for the regulator to compensate for. the drop.

More output current, more copper on PCB and thicker wires are needed. After all, the regulator will have the nominal voltage on the output but if your wire is 10 meters long, you will have voltage drop on the cable, right?

I think TPS565208 is one of the easiest to handle and very good in terms of cost, pcb estate, output noise and efficiency.

Ioannis

[Ioannis]

Piece of crap 7V wall adapter worked 10 minutes...

https://www.amazon.ca/dp/B091YHX8RY

Noname chinese rubbish... They don't even have the proper distances or creapage on the PCB...

Prefer good brands like Mean Well or Sunny.

Ioannis

[tumbleweed]

That's what I said in the second part about getting more than 2A out.

You can't get 5A out if the adapter can only supply 2A

[Demon]

That's what I said in the second part about getting more than 2A out.

You can't get 5A out if the adapter can only supply 2A

The 2A wall adapter was a quick fix to be able to continue testing.

The 9V 2A is holding up for now.

[Demon]

Received the LM1084-ADJ, still waiting on the recommended caps (to properly test ripples), but it's already running nicely with what I have on hand.

Also got a thermal imager to help identify hotspots. Looks like I fried 2 PICs with those 12V and 7V bursts (lesson learned).

LM1084-5V on right, supplies PICs, ICs and LCDs, essentially everything that is "always ON", with a slight tweak on ADJ pin to get it slightly above 5V to compensate for voltage drop.

LM1084-ADJ on left, supplies LED strips, cooling fans (to be determined) and anything else that doesn't care about ripples.

(HIKmicro E02)
9814
9815

No thermal compound yet, this was a temporary setup to help prevent forest fires.

9816

I'm sure there's better ways to skin a cat, but this was the simplest technique I could come up with. It remains to be seen if the ripples are "manageable" for the ADC on the PIC circuit.

[Demon]

Wasted so much time trying to figure out why I STILL had a lot of ripples with the LM1084.

I checked directly on the wall adapter:

9818

GIGO; can't expect the LM1084 to do miracles with that.


So I pulled out a 12V-5A Fqcmogu AL-1250 adapter (no idea what it's for):

9819

Definite improvement.


This is the LM1084-5V running at 5.0V (LED strips and whatnot):

9820


This is the LM1084-ADJ running at 5.2V (PIC, LCDs and other ICs), note that is had less ripples as stated in the specs:

9821

[Aussie Barry]

chinese amps are quite a bit smaller than normal ones. 0.75 seems close
ROFL...
Good one, Richard!

[amgen]

if you didn't already........ put your probe lead on 'ground/negative' at a distance from the negative/ground probe connection..... and probe to various ground/neg locations. Sometimes noise is 'introduced' into circuits from other sources called common mode noise.....
--

Key Takeaways

Noise appears in two forms in an electrical interconnect: as differential mode and common mode noise.
Differential mode noise is measured between two sections of an interconnect with equal and opposite polarity, while common mode noise applies to interconnects with the same phase and polarity.
Both forms of noise are induced in an interconnect via Faraday’s law from external radiation.

[Demon]

if you didn't already........ put your probe lead on 'ground/negative' at a distance from the negative/ground probe connection..... and probe to various ground/neg locations. Sometimes noise is 'introduced' into circuits from other sources called common mode noise.....
--

Nope, haven't done that. Good idea.

I'm also redoing my prototype board to have a somewhat "star pattern" for VDD and VSS. Right now the 2 rails run all over the place.

Right now I'm just basking in the glory of having found the root cause of most of that noise on the lines, and started looking for a source of affordable "good" 9V 5A power adapters.

[amgen]

Right now I'm just basking in the glory of having found the root cause of most of that noise on the lines, and started looking for a source of affordable "good" 9V 5A power adapters.
-
not trying to 'rain on your parade'..... it is possible that isn't where your noise is coming from...... if you haven't already, put a 10 or 100 microfarad electrolytical and a couple of .1 or .01 microfarad caps at the output of the power supply..... and I didn't look to see but you would want to put the same capacitor duo on each separate circuit board..... that SHOULD squash the noise. Plus..... a bunch of interconnected grounds running here and there are often culprits for noise. And keep in mind that often serious engineers and engineering is involved with noise and interference

[Ioannis]

Also breadboards are the worse for low noise circuits!

Ioannis

[Demon]

VDD with all ADC pots OFF; VDD is super stable.

9822


I control the contrast on all 4 LCDs with PWM on a 2N2907:

9823


Guess what happens to VDD when I turn up the contrast:

9824

[richard]

that seems a strange way to control the contrast.
how does it go if you set the contrast the normal way ?
9825

anecdotally contrast by pwm needs a substantial LPF otherwise all sorts of nasty artefacts are introduced

perhaps a schematic would enlighten things

[Demon]

that seems a strange way to control the contrast. ...

Sorry, contrast is hardwired to prevent the user from burning out the LCD. It's the backlight that's PWMed.

9826


Backlight is controlled by PWM from a 2N2907A because it controls the 4 physical LCDs simultaneously as well as the radio in-game via USB:

9827

[Demon]

I'm waiting on some IRF5305P hexfets to replace the 2N2907A. Hopefully they're less noisy.

[richard]

does that drive all four backlights ?
where does led cathode go ?

if the cathode goes to vss then no wonder its noisy
the fet should make it worse

[Demon]

does that drive all four backlights ? ...

Yes



... where does led cathode go ?...

VSS

[richard]

i would start with a 100R resister between each cathode and vss , if the backlight is not brite enough
drop the R incrementally .

[richard]

you should have specs on max current for the backlight, the actual R used should limit to that value or less at 100% pwm modulation at 5v. imo

[Demon]

you should have specs on max current for the backlight, the actual R used should limit to that value or less at 100% pwm modulation at 5v. imo

Roger, Richard.

Specs say 90mA. Gonna have to test to see if that's into the "overbright" range with pixels bleeding through.

I remember setting this "by eye" at an intensity that was "aesthetically pleasing". I was also trying to save on AMPS too since I was initially thinking of running off USB power.

9828

[Ioannis]

The 2N2907 is a PNP transistor while IRF530P is a N-Channel FET. So the circuit should be different.

In the case of 2N2907 you source current to the LCD but with IRF you should sink current to the ground.

And of course the polarity of the control signal should be inverted for the IRF. Hope you are aware of that.

As for the noise, your circuit does not have any decoupling capacitors very close to your switching transistor (either 2N or IRF) and to the ground. Put a 100nF and a 10uF in parallel and I am sure things will be better.

Also ground loops etc will have massive effect on this. Do not trust breadbording for this. Noise will be much more on such circuits even with capacitors.

Ioannis

[Demon]

IRF5305P is P channel.

https://www.mouser.ca/ProductDetail/Infineon-Technologies/IRF5305PBF?qs=9%252BKlkBgLFf3WAOOEkAEJ1w%3D%3D&srsltid=AfmBOorOoA9Hf8u9uxnUIulSp-2hwL5m7Ex_TbELALuFut1Niv57Vgi1

[amgen]

I could be wrong but the way I read your instruction sheet I see this..... think the module has the dimmer mechanism internal and is controlled by 0-5 volt on the Vo pin
-
-
9829

[Demon]

I could be wrong but the way I read your instruction sheet I see this..... think the module has the dimmer mechanism internal and is controlled by 0-5 volt on the Vo pin



Yeah, as I corrected myself in post #38,
https://www.picbasic.co.uk/forum/showthread.php/26823-Voltage-regulation-circuit-12V-to-5V-in-5A-range?p=156583#post156583
it's the backlight that I'm controlling.

This is white-on-black LCDs. I've set V0 to a "visually acceptable level" (still have to test Richard's suggestion), and I control the brightness of the LCDs only with the backlight.

https://youtu.be/G87bqw30w-M?si=_7N4VQmrWjQ2aYpJ

[Ioannis]

Obviously I mistyped the part number. Yes it is a P channel.

From the datasheet the Vth seems to be 4Volts min. Have you tested yet this MosFet?

Please check with an Osciloscope that the MosFet is fully closed at PIC low state, so that there is no voltage drop on S-D pins.

Was there a specific reason that you selected a P channel instead of a N channel control of the PWM signal? Usually the N channel have better characteristics in terms of Rds.

Ioannis

[Demon]

Obviously I mistyped the part number. Yes it is a P channel.

From the datasheet the Vth seems to be 4Volts min. Have you tested yet this MosFet? ...

Nope. But I'm using a IRFZ44N n-channel to PWM the LED strip with has similar specs.



...Please check with an Osciloscope that the MosFet is fully closed at PIC low state, so that there is no voltage drop on S-D pins....

Good idea. Choosing the proper resistors to drive the MOSFET is still a black art to me. :D



...Was there a specific reason that you selected a P channel instead of a N channel control of the PWM signal? Usually the N channel have better characteristics in terms of Rds.

Ioannis

Because when I first started testing blue-on-white LCDs, I was PWMing both the Contrast (V0) and Backlight (LedA). I didn't even consider MOSFETs and went straight to transistors since we're not talking about a lot of current. Why I chose PNP transistors back then is a mystery; maybe I saw a schematic on google and copied that.

I've since switched to white-on-black LCDs, I don't need to have adjustable Contrast (V0). So controlling the Backlight with a PNP transistor is just "the old way of doing things". But since noise has become an issue, getting a P-channel MOSFET seemed like the logical way to replace the PNP.

But if I had thought about it for 2 seconds, I would have realized I have access to both the Anode and Cathode on the Backlight. Since I already have an IRFZ44N on the circuit, it would cost less at JLCPCB to use another IRFZ44N instead of a second type of component.

So that's the long-winded way of me saying that now I'm going to swap the 2N2907A PNP to a IRFZ44N n-channel.

Funny story, I finally realized the letter at the end of a MOSFET seems to indicate the channel. I don't know if it holds true for all, but it's working for the first two that I researched.
:)

[Ioannis]

Switching to MosFet from bipolar, as Richard noted, it will introduce MORE noise because they are faster.

So, do add capacitors (100nF plus 10uF) as close as possible to the Mosfet power supply, taking care of ground loops at the same time.

Ioannis

[richard]

So, do add capacitors

a waste of time until he address' the correct current limiting resistors for each display

if its really 90mA per display then a 1000uF might help a little.
The peak current as it stands could be very high depending on led vf , could be amps , who knows

[Ioannis]

OK, but having a 4.000uF on the 5V rail is too much, isn't it?

Ioannis

[richard]

You seem to be moving the goal posts. One 1000uF at the pnp tranny emitter since all four back lights are parallel driven from that source

[Ioannis]

By "if its really 90mA per display then a 1000uF might help a little" seemed to me that you meant 1000uF per display.

In any case though, the 1084 or even the humble 7805, do not have any problem with large capacitance.

I bet a reverse biased diode, across output/input pin, would be needed to protect the regulator, in case the Reg. input is grounded for whatever reason while the outuput cap is still charged.

Ioannis

[Demon]

...
I bet a reverse biased diode, across output/input pin, would be needed to protect the regulator, in case the Reg. input is grounded for whatever reason while the outuput cap is still charged.

Ioannis


Yup, I saw that in the datasheet.

Haven't done any of the suggestions for the LCDs yet. I'm doing voltage regulator heat tests for now; looking for the best heatsink setup.


This is how I'm running the LM1084-ADJ now:

9831


Black is from the datasheet, red is my adjustments.

[richard]

Haven't done any of the suggestions for the LCDs yet. I'm doing voltage regulator heat tests for now; looking for the best heatsink setup.

if you solve the backlight issue then you will most likely find the wall warts simple smps are not the problem and the regs and heatsinks are unnecessary

[Ioannis]

The LCD datasheets note that the backlight is 5v at 90mA, isn't it?

So I understand that no current liminting resistor is necessary, as it is included on the LCD board.

Ioannis

[richard]

So I understand that no current liminting resistor is necessary, as it is included on the LCD board

you may be correct but that does not explain why the backlights cause so much noise.
it would be a 2 minute job to apply 5v to the backlight and measure the current .[i would used a cc regulated ps just in case]
i have never come across an lcd of that type with anything onboard to limit led current.

[Ioannis]

i have never come across an lcd of that type with anything onboard to limit led current.

Well, there are not many but they exist. I use them all the time to avoid external resistors.

If the datasheet marks it as 5V then the LCD includes the limiting resistor. Otherwise they note 4 or 4.2 volts at specific current and specific resistor for 5V supply.

Ioannis

[Demon]

I've put on hold debugging/testing LCDs, ADC and whatnot until I get a handle on power regulation.

Thanks to Ioannis, the datasheet for the TPS voltage regulators led me to TI's WEBBENCH; that tool is just remarkable for the newb like me. I've spent the last 2 days playing with various ICs, checking their availability on JLCPCB and such.

The 2 best features has to be that Alternate and Simulate features; you can chose equivalent parts that are no available at JLCPCB, and see exactly what the output ripple will look like.

I've narrowed down my choice to a TPS56637. I can run 5V @ 5A using 7V to 12V at 11.4mV Vout Peak-Peak; it can do more, but the output ripple degrades as you push it.

This is awesome because it gives me a wider range of power supply adapters. I was initially fixated on running on 9V to limit heat from conventional voltage regulators.

The datasheet even gives a very detailed PCB layout:

9838

Thanks mucho Ioannis!

[Ioannis]

You are most welcome! Thats what friends are for, right?

The selected TPS is difficult to handle by hand. Needs heatsinking on the PCB by special design. I do understand why you selected this one over the TPS565208, that is super easy to handle.

Amazing chips TI has...!

Ioannis

[Demon]

Y...The selected TPS is difficult to handle by hand. Needs heatsinking on the PCB by special design. I do understand why you selected this one over the TPS565208, that is super easy to handle.

Amazing chips TI has...!

Ioannis

Definitely, that's why I did my best to mimic the "suggested" layout I posted above. I figure there's countless hours or even days put into that design.

I came up with this 4-layer PCB, top layer:

9839


Both inner layers are identical:

9840


And the bottom layer:

9841


I did my best to use the same outlines, but I didn't have identical sized components, so I had to make due.
- Rpg goes to VCC on the datasheet drawing, but goes straight to 5Vout on the generated schematic.
- they have a single Cin cap, but mine is split into 2 using alternate parts.
- my inductor seems to be twice as large as theirs.

Also, the Webench schematic for 7-12Vin to 5A-5Vout didn't have these resistors:
- Rmode
- Renb
- Rent
So I don't have that trace from PGND going up to VIN.

The last thing I have to adjust is the copper layers; I remember reading something about needing 2 ounces top and bottom, and 1 ounce inners (or something like that - got reading to do).

Looks like this in 3D:

9842

9843

[Demon]

ARGH! Of course it's only now that I notice Webench generated a layout specifically for that schematic.

9844


I couldn't find any mention of copper thickness in the datasheet. No mention in the Evaluation Board for the TPS56637, but it does have a much nicer layout with A LOT MORE cooling holes and a wider area.

9845


I must have read about the copper thickness in the datasheet for another TPS. I just can't remember which cause I looked at so many.


EDIT: No mention of copper in their Quick Reference Guide:

https://www.ti.com/lit/an/slva958b/slva958b.pdf

[Demon]

Bingo, I must have seen this while researching the LM1084. Digikey PDF for the LM76002 on page 43:
https://www.digikey.ca/en/htmldatasheets/production/2903149/0/0/1/lm76003rnpt


TI recommends using a four-layer board with the copper thickness, for the four layers, starting from the top one, 2 oz / 1 oz / 1 oz / 2 oz.

[Demon]

TI's Eval board for the TPS56637 is 51mm x 64mm. I figure it would be wise for me to use the same dimensions until I have an actual board to check under the infra-red camera.

Increased the amount of cooling VIAs like on the Eval board.

Used a DIP header for the leads for a cleaner look once on the prototype.

Increased thicknesses to 2-1-1-2 ounces and tweaked the trace width/space according to JLCPCB minimum requirements for a 2oz board.

9846


I still have to figure out how I can make my VIAs have large copper rings like the Eval board. I probably have to replace them with actual holes (like I have on the 4 corners - nice visible copper ring).

[Demon]

Had to make the hole a Part, not included in Schematic, BOM or To Be Populated.

Discovered the Create From Selection / Create Array feature; real time saver to make grids of parts.

9847


Starting to look 1/2 decent. :)

[Ioannis]

Wow! You did a lot of work!

I am pretty sure that the extra layers TI suggest are for less noise ad better heat dissipation. Have you checked how much extra cost is added by JLCPCB?

In general if you want/need to make the board smaller, you can move all componets to the left, connector closer to the inductor and the right board edge also moved to the left. I don't see why this pcb needs to be that large. Saves cost also I guess.

I'd add an extra output capacitor for less noise and better operation without much cost. Parallel capacitors make for lower ESR also.

Nice design. You become an KiCad expert!

Ioannis

[Demon]

EDIT:

At first I was tempted to make 2 prototypes; 2 layers and 4 layers; just to see if the extra cost is worth it. But the more I think about it, the more I'm leaning for the 4-layer design.

Two of the main objectives with this circuit is to reduce heat and stabilize the voltage (ripples). If the pros design their Eval board to that size with 4 layers, I can't really see why I should consider cutting corners.

But boy am I tempted. :D

[Demon]

...Have you checked how much extra cost is added by JLCPCB? ...

Not yet, I wanted to get as close to a "finalized design" before using JLCPCB's quote generator.



... I'd add an extra output capacitor for less noise and better operation without much cost. Parallel capacitors make for lower ESR also. ...

I didn't know that about the lower ESR.

EDIT the CAPS section

I just noticed the datasheet has a range of output capacitance at 5V from 25 to 100uF (table 4). So I just added 10uF and 1uF ceramics alongside the 47uF.

These guys have a nice bla-bla about "why there can be too much output capacitance on SMPS circuits":
https://www.cui.com/blog/understanding-power-supply-output-capacitance-limits



... Nice design. You become an KiCad expert!

Ioannis

Thanks, but I can't really take credit for the design. I feel like Mr. Bean copying the test answers from the guy next to him. I really did put a lot of effort to do as close to what TI recommends.

As for "Kicad expert", that remains to be seen. :D

[Demon]

Well, if I'm going to have all that unused real-estate at the top, might as well do something with it:

- DC jack,
- power switch,
- multiple output headers,
- amp and volt test points.

It's all parts I have on hand, so it's not really costing me more. I don't see how it would disturb the cooling, since TI placed all the ventilation holes on the bottom section. It anything, it's extra cooling fins. :D

9848


Now I just have to remember to slap a jumper on the amp test point for the circuit to work. :)

[Demon]

... Have you checked how much extra cost is added by JLCPCB? ...

Ioannis

4 layers with 2-1-1-2 ounces with assembly, the area of my circuit jumps from 63mm x 50mm up to 73mm x 70mm.

2oz $32.20
1oz $16.07

Another issue, I can't use Economy with 0201 SMD, i have to use Standard. I had to swap 0201 resistor and capacitor for 0402.


Final bill for 5 PCBs:
- PCB $55.45
- Assembly $83.89
- Total $165.21 (with UPS shipping), $33.04 each PCB.


Plan B:

1oz all layers

PCB $23.15
PCBA $67.92
Ship $ 25.87
Total $116.94, $23.39 each PCB.


Tempted to go with plan B. It's not like I'm going to push 5 amps, most likely 2.5A at best.

[Ioannis]

Ten dollars is quite enough. I'd select that 1oz too.

Maybe and more caps too.

Ioannis

[Demon]

I ordered 2 models:

- 4 layers, all 1 oz, 5 for $147 including shipping and duties, $29.40 each
- 2 layers, all 1 oz, 5 for $108 including shipping and duties, $21.60 each

I was able to stay on Economic price by avoiding 2oz layers, and remained at recommended area size by TI.

Output caps, 47uF (recommended), 10uF and 1uF.

I was tempted to add 0.1uF and 0.001uF but I thought I might be exaggerating. I can still add them in the final version if you guys think I should; I have room and it's not like they cost anything.

At least I'll have something to see how it ripples using PWM from transistors and mosfets.

I can also push it for heat test by PWMing a whole roll of LED strip.

[Demon]

I can get the price down a bit more by using another inductor:

Current: XAL1010-332MED by Coilcraft, $4.4191 x 5 = $22.10

Alternative: SRP1265A-3R3M by Bourns, $1.5387 x 5 = $7.70

That's a difference of $14.40 ($2.88/unit)


I didn't use the Bourns cause I'd have to order a minimum of 10 from Mouser, and that takes 8-15 lead days, and I wanted my PCB for testing yesterday.

There was 73 inductors in stock from Coilcraft, but that's already down to 38 (I only used 10 in my 2 orders).


Ioannis, have you had any experience ordering from Element or Onlinecomponents?

[Demon]

Once I get my order and done some testing, I'll be in a better position to know what configuration of layers I'm going to need.

I'm always getting emails from PCB assembly suppliers from alibaba, so I'll to make a RFQ for that "winning" configuration.

Just to see if I can get anything better than JLCPCB.

That'll also give me a chance to find a good competitor for my PCBs in the future. But I gotta admit, JLCPCB's online system is pretty slick once you get used to it.

[Ioannis]

Ioannis, have you had any experience ordering from Element or Onlinecomponents?

No. My country adds a lot of import taxes so I select suppliers within European Union or Mouser that ships tax free.

May main suppliers are Mouser, TME, SOSelectronics and local resellers for larger quantities.

Regarding the inductors, I think either Coilcraft or Burns (or other reputable brands) will make no difference, as long as you select one that has the specs.

At testing time you will see if adding more caps on output of the power supply makes any significant difference. Can't tell right now.

For mass production of assembled pcb JLCPCB may be more expensive. I usually do this for no more than 20-30 pcs, depending on the component count too. Greater production is cheaper for me if I do it locally and just order only the pcb's from JLCPCB.

Ioannis

[Demon]

Assembled, checked each header and basic test on v1; 2 layer, 1-1oz.

9864


Now I need to finish my load tester using this puppy.

9865


The plan is to PWM the power resistor while monitoring current, and using E02 thermal camera to watch heat spots on PCB.

Still undecided if using N or P channel logic-level MOSFET. I have yet to find a good tutorial to determine the base resistor.

[Demon]

Tested on prototype board; 430mA.

9866


Top side, TPS56637 step-down IC is the hottest at 25C, followed by inductor and switch bracket. I just touched power jack at far left, so that's not actual temp.

9867


Bottom side, center is at 23C, the max spots are my finger (can see through top left mounting hole too) and LED strip in background.

9868


That's insane compared to 7805 or LM1084. Humungous difference in efficiency and heat (never expected it to be so much difference).

I can touch this with my fingers.

[Demon]

The LM1084 was maxxing out near 124C, and that was only at 300mA with heatsinks on front and back.

9869


I'm both flabber and gasted. :D

[Ioannis]

This is not a heatsink. It is a sample of heatsink!

Ioannis

[Demon]

Tested the TPS56637 against my prototyping board. It's operating beyond expectations for heat. The only thing I could improve is the ripples in the lines. I think max peak-to-peak was 22mV during preliminary tests, and the TI's Webench power designer tool said 11.73mV.

It took a lot of trial and error playing with capacitors,, but I've found a happy medium that runs under 11mV except for the LED strip.

https://youtu.be/pJla-D89PyQ?si=tGKfbWM9Xjsw8NYJ


As I say in the video, I'm going to reorganize the board so the TPS56637 is in the middle, and all breadboard rails get individual VDD/VSS lines. the only drawback is that I'm going to lose that centralized row of input capacitors, but I've already got an idea to solve this as I'm typing.

And I'm not sure what I'm going to do with that bank of capacitors on the breadboard. There's about 18 of each (I just emptied my box compartment). I did notice that it costs some mV to feed all those caps.

I'd really like to keep a MOSFET for the LED strip cause right now I'm limiting how much current it consumes (I think it's 180mA). But if the TPS can supply so much more efficiently, I'm thinking of letting the PWM rate go up a bit higher. It'll all depend on how the TPS behaves with 3 LED strips connected.

[Ioannis]

I am glad you got the results you hoped for.

With all those ground loops and the probe cable also, I'd expect much more noise of that!

Usually to measure noise on a regulators rail, you would ditch the cable with the crocodile clip and use a coil spring to earth the probe as close as possible near the tip.

Adding so much capacitance, from a certain point on, is not worth it as much as have good grounds.

Congrats,
Ioannis