PCB Design Critique, Simple Regulated Linear Power Supply

Hello All,

Here is a PCB design I did with Express PCB, and the associated schematic I used to create it.  I am new at board design yet I have had some exposure to web-based design tips.  Will you offer some criticism?

For example, please tell me if I have errors in layout, trace locations, improper anything..

Thank you for your time and consideration,

Andrew 

Here is the schematic:

There are some errors, like the way the 10VCT xfmr is connected, it delivers 20Vac, so the unreg voltage is ca. 28Vdc. If you want to power LED's, you don't need a regulated supply at all. LED's want to see current, so they only need a series resistor connected to a reasonably filtered voltage.
Both +/- regulators should have a 22-100uF electrolytic at their outputs. The smaller caps are somewhat useless. 4700uF after the bridge has no justification. A rule of thumb is 1000uF for each Ampere of current.

Andrew

2 things to look at:

The 18v secondaries need to be  in series; OK, now I've put my glasses on, I see the 36v legend in RED

The protection diode on the -18v reg is the wrong way round

Peter

peterc said:

The 18v secondaries need to be  in series;

The protection diode on the -18v reg is the wrong way round

Click to expand...

Hey Peter, what do you mean regarding the 18v secondaries?  I'm using a dual complementary set up.  The transformer is 36V CT Series, 18V parallel. 

Regarding the diode, I believe it is correct for one reason. The potential on the cathode is approx -24, and anode approx -18, therefore the diode is under normal conditions being held in reverse bias, or not conducting. 

The potential on the cathode is approx -24 V, and anode approx -18 V, therefore the diode is under normal conditions being held in reverse bias, or not conducting.

Click to expand...

If the cathode is at lower potential (-24 V is less than -18 V, right? These are absolute units, not magnitudes!) than the anode this is a forward-, not reverse-biased condition.

Samuel

Regarding the diode, I believe it is correct for one reason.

Click to expand...

Your diode for the (-) leg is backwards. 

If you don't believe us, download the DATASHEET for a 337!

abbey road d enfer said:

There are some errors, like the way the 10VCT xfmr is connected, it delivers 20Vac, so the unreg voltage is ca. 28Vdc. If you want to power LED's, you don't need a regulated supply at all. LED's want to see current, so they only need a series resistor connected to a reasonably filtered voltage.
Both +/- regulators should have a 22-100uF electrolytic at their outputs. The smaller caps are somewhat useless. 4700uF after the bridge has no justification. A rule of thumb is 1000uF for each Ampere of current.

Click to expand...

You are correct that I could use an unregulated supply for the LEDs.  

The transformer is a 5v parallel 10v series secondary type.  I'm using the secondary as the 10v series type.  The approx rectified/filtered peak/dc is 12.8v.  I've got a lot of LEDs running at near 20mA per LED. The total current demand will be about 1A, with all LEDs on.  The LED voltage is low, because the design is using drivers that do not require series LED resistors to drop current/voltage.  

I used data sheets/app notes to set up the design, thats why you see the small value film and tantalums.  

Can you tell me why it is common to see larger value caps at the output of the regulators? The app notes I used didn't really explain the need for the small value caps around the regulators.  

Concerning the filter caps, there are trade-offs.  The higher value means lower ripple voltage, and higher inrush current possible among other things.   I used the Elliot Sound C = ( I L / ΔV) * k * 1,000 uF to figure the values.  However you must consider "k" when using this formula, and for the US, "k" =6. So, the actual requirement for 1A @ 1Vpp is 6000uf. This is true if you want 1 Vpp of ripple voltage. To drive the ripple lower, you can increase the cap value.

Samuel Groner said:

The potential on the cathode is approx -24 V, and anode approx -18 V, therefore the diode is under normal conditions being held in reverse bias, or not conducting.

Click to expand...

If the cathode is at lower potential (-24 V is less than -18 V, right? These are absolute units, not magnitudes!) than the anode this is a forward-, not reverse-biased condition.

Samuel

Click to expand...

You got me.  I am stupid. It is forward biased.

My apologies to all for my stubbornness.

Afarawayland said:

abbey road d enfer said:

There are some errors, like the way the 10VCT xfmr is connected, it delivers 20Vac, so the unreg voltage is ca. 28Vdc. If you want to power LED's, you don't need a regulated supply at all. LED's want to see current, so they only need a series resistor connected to a reasonably filtered voltage.
Both +/- regulators should have a 22-100uF electrolytic at their outputs. The smaller caps are somewhat useless. 4700uF after the bridge has no justification. A rule of thumb is 1000uF for each Ampere of current.

Click to expand...

You are correct that I could use an unregulated supply for the LEDs.  

The transformer is a 5v parallel 10v series secondary type.  I'm using the secondary as the 10v series type.  The approx rectified/filtered peak/dc is 12.8v.  I've got a lot of LEDs running at near 20mA per LED. The total current demand will be about 1A, with all LEDs on.  The LED voltage is low, because the design is using drivers that do not require series LED resistors to drop current/voltage.

Click to expand...

Then are you sure the smoothing cap is large enough? How much will you draw from the 12.6 voltage? And the LM317 will get hot, dissipating nearly 10W.

I used data sheets/app notes to set up the design, thats why you see the small value film and tantalums.  Can you tell me why it is common to see larger value caps at the output of the regulators? The app notes I used didn't really explain the need for the small value caps around the regulators.

Click to expand...

These caps are needed for stability. The 78/79 regulators are prone to oscillations. The output cap should be physically close to the regs, and if the PC tracks between the smoothing caps and the regs are too long, it is advised to put smaller caps closeto the regs.

Concerning the filter caps, there are trade-offs.  The higher value means lower ripple voltage, and higher inrush current possible among other things.   I used the Elliot Sound C = ( I L / ΔV) * k * 1,000 uF to figure the values.  However you must consider "k" when using this formula, and for the US, "k" =6. So, the actual requirement for 1A @ 1Vpp is 6000uf. This is true if you want 1 Vpp of ripple voltage. To drive the ripple lower, you can increase the cap value.

Click to expand...

The xfmr's description is misleading; it says parallel but it is connected in series. Is 130mA the current you intend to run it? In that case 1000uF is large enough.

If you are planning to pull an amp out of each of the regulators then you should have large heatsinks. I don't see much room for attachment of heatsinks, what did you have in mind? With about 10V across the 317 at 1A you will be dissipating about 10W. If you have 6V across the 78/79 regulators then they will each be dissipating about 6W at 1A load.

You are mounting transformers directly on the PCB, however don't appear to have any mechanical fixing. The PCB looks like a single-sided PCB, so you are relying on some solder joints to hold it in place. It's quite common for heavy objects such as power transformers to vibrate loose from the PCB over time, depending on how the board is mounted. Either add some mechanical support or use off-board transformers. You could get away with a single transformer in this design to save space/cost and not have to have mains voltages on the PCB.

The output side of the regulators should have a 220-470uF cap near it. A 1uF tantalum won't be much help. The 0.1uF ceramic caps should be very close to the regulator pins, even closer than what you have now (btw 0.1uF ceramic decoupling caps are smaller than what you have shown on the PCB layout).

If your +/-18V supplies are meant to be low noise, you may be better using the LM317/LM337 regulators instead.

abbey road d enfer said:

Then are you sure the smoothing cap is large enough? How much will you draw from the 12.6 voltage? And the LM317 will get hot, dissipating nearly 10W.

I used data sheets/app notes to set up the design, thats why you see the small value film and tantalums.  Can you tell me why it is common to see larger value caps at the output of the regulators? The app notes I used didn't really explain the need for the small value caps around the regulators.

Click to expand...

These caps are needed for stability. The 78/79 regulators are prone to oscillations. The output cap should be physically close to the regs, and if the PC tracks between the smoothing caps and the regs are too long, it is advised to put smaller caps closeto the regs.The xfmr's description is misleading; it says parallel but it is connected in series. Is 130mA the current you intend to run it? In that case 1000uF is large enough.

Click to expand...

The ripple on the 317 side should be under 1Vpp, I figured that using 1.8A.

I was thinking that the regulator itself would drop some voltage too, around 3 volts.  Still, yes, there is a lot of wasted heat from reg in to out on the 317.  I am planning to use a large heat sink on it.  I do not know how to correctly size the heat sink for this amount of heat, and am relying on a guess, using the biggest one I have. I tried to leave more space on the board for it around the 317.  It is a column type about 2.5" high, with triangular fan ends.  It may not fit now, but I can check/change when I print the design out.

Sorry for the schematic goof, the smaller transformer is 130mA 18 V parallel, and 36 volts 65mA series.  I will need to check my calculations for max current demand.  This particular side is powering ICs, about 7.  I went over-kill on the filter caps for both sections. 

Do you think my pcb traces from filter to reg, and so forth are short enough?

gswan said:

If you are planning to pull an amp out of each of the regulators then you should have large heatsinks. I don't see much room for attachment of heatsinks, what did you have in mind? With about 10V across the 317 at 1A you will be dissipating about 10W. If you have 6V across the 78/79 regulators then they will each be dissipating about 6W at 1A load.

You are mounting transformers directly on the PCB, however don't appear to have any mechanical fixing. The PCB looks like a single-sided PCB, so you are relying on some solder joints to hold it in place. It's quite common for heavy objects such as power transformers to vibrate loose from the PCB over time, depending on how the board is mounted. Either add some mechanical support or use off-board transformers. You could get away with a single transformer in this design to save space/cost and not have to have mains voltages on the PCB.

The output side of the regulators should have a 220-470uF cap near it. A 1uF tantalum won't be much help. The 0.1uF ceramic caps should be very close to the regulator pins, even closer than what you have now (btw 0.1uF ceramic decoupling caps are smaller than what you have shown on the PCB layout).

If your +/-18V supplies are meant to be low noise, you may be better using the LM317/LM337 regulators instead.

Click to expand...

The +/-18 side wont supply much more than 130mA at most. So, I imagine that they won't providing much more.  I may not need them but I will be attaching small heat sinks to the 78/79.

As far as the xfrmers, yes they are PC mount.  The board will be single side.  I plan to use some silicone adhesive to fasten them to the pcb before soldering. 

I've heard this mentioned before, "keeping the mains off the pcb".  Why does one not want mains voltage on a pcb?

Would the added caps on the output (220-470) be acting as more filtering?

Owels ps design for example has 1000uf filters and an additional 1000uf on the output.  I am curious as to what this does.

I am using films for the decoupling caps.  Is this bad?

Your PCB traces are short enough and wide enough, for sure. But I'm still concerned about the location of the LM317. Dissipating 10W, you will need a large heatsink, so I suggest you move the LM317 to the side, so you can fit a larger one. Check the specs for heatsinks, you need one with less than 5°/W. The smallest i could find is 50mm x 46 x 33 (and I think it's gonna run too hot for my taste)
http://fr.farnell.com/abl-heatsinks/205ab0500b/dissipateur-thermique/dp/177011

Afarawayland said:

The +/-18 side wont supply much more than 130mA at most. So, I imagine that they won't providing much more.  I may not need them but I will be attaching small heat sinks to the 78/79.

Click to expand...

Sounds fair, the regs will be dissipating about 1/2W each.

Afarawayland said:

As far as the xfrmers, yes they are PC mount.  The board will be single side.  I plan to use some silicone adhesive to fasten them to the pcb before soldering. 

Click to expand...

OK, this should help.

Afarawayland said:

I've heard this mentioned before, "keeping the mains off the pcb".  Why does one not want mains voltage on a pcb?

Click to expand...

Only for safety when you have it on the bench and testing. It's easier to wire a switch, fuse and socket to the transformer and have no chance of making contact with it. On a PCB there are bare areas of mains that pose a safety risk. It also does not give you the flexibility of re-locating the transformer or mains wiring if there's some stray coupling in your signal wiring.

Afarawayland said:

Would the added caps on the output (220-470) be acting as more filtering?

Owels ps design for example has 1000uf filters and an additional 1000uf on the output.  I am curious as to what this does.

Click to expand...

It dampens the output of the regulator so you don't get overshoot with step load changes. A 1000uF is probably a bit high, however provided it's a low ESR cap then it's probably OK.

Afarawayland said:

I am using films for the decoupling caps.  Is this bad?

Click to expand...

What kind of film? Ceramic decouplings caps right on the regulator leads are better they have lower inductance.

gswan said:

What kind of film? Ceramic decouplings caps right on the regulator leads are better they have lower inductance.

Click to expand...

AVX radial metallized polyester stacked film

I found a heat sink, 3.7˚C/W, 1.5"x1"x1.62"

http://www.alliedelec.com/Search/ProductDetail.aspx?SKU=6190109&MPN=529802B02500G

Definitely have to move the 317, but will it work?

Afarawayland said:

gswan said:

What kind of film? Ceramic decouplings caps right on the regulator leads are better they have lower inductance.

Click to expand...

AVX radial metallized polyester stacked film

Click to expand...

A multilayer ceramic capacitor is a better choice for this application.

alright, I'll order a change for that cap.

Thank you all for your input, and again, my apologies for my stubbornness.

Anything else?