[kriskalish]

I'm trying to decide how thick to make my traces in a power supply I'm building using Radioshack's two sided PCB.  The thing is, it doesn't say on the packaging what the fill weight is.  Here's the product I have:
 
http://www.radioshack.com/product/index.jsp?productId=2102495
 
 
 
Does anyone know from past experience of have a good estimate?
 
 
Thanks,
 
Kris

 

[jcx]

2 oz/ft^2 used to be standard, 1 oz often used today for finer resolution smt parts, can measure with a decent micrometer

 

[wakibaki]

This only really becomes an issue where currents are high.
 
Google 'pcb trace width calculator', enter the anticipated current and permissible temperature rise and see what you get for 1 oz (35 microns). Retailers are unlikely to supply anything less than 1oz. (0.5 oz is as low as it goes AFAIK)
 
You can double the recommended trace width to be sure. It's unlikely that there will be insufficient room to accommodate oversize traces in a PSU, and you can always make room if forewarned.
 
There are other considerations such as frequency and skin depth, ambient and available cooling, but check first whether your application is demanding.
 
w

 

[kriskalish]

Ah, thanks for the insight guys.  It actually looks like my initial guess of 65mil traces wasn't far off.  With 1oz pour, and permitting only a 10 degree temperature rise necessiates traces of 55mil. 
 
Is there really any reason I shouldn't just make traces as thick as I possibly can while maintaining decent clearance (such as the clearance recommended by these calculators)?  I know making traces thicker makes them more capacitive, but that doesn't seem like a problem in a variable power supply as it would be in an audio signal path.
 
-Kris

 

[wakibaki]

Quote:

Is there really any reason I shouldn't just make traces as thick as I possibly can while maintaining decent clearance (such as the clearance recommended by these calculators)?

No. Often in professional designs you will see irregular areas joining pins as opposed to straight chamfered tracks. These contribute to cooling and reducing temperature gradients across boards and are employed where there is no contra-indication, such as stray capacitance.
 
w

 

[kriskalish]

As promised, here is the initial design.
 
 
Schematic

---

It's basically the exact same thing that can be found on page 7 of the LT1033 datasheet, where it is described as a "dual tracking 3A supply."
 

 
Components

---

C1-C4 are 2200uF electrolytics.
R1,R2 are 1.3ohms rated for 5W
C5,C6 are typically 10uF tantalums, but I have 100uF low ESR electrolytics on hand
R3, R4 are 12k ohms
R5, R6 are 120 ohms
R7 is a 10-turn 10k ohm pot.
C7, C8 are 220uf 100mOhm ESR electolytics
 
 
Justification

---

The choices for C5,C6,C7, and C8 are based on the following research:
http://www.tnt-audio.com/clinica/regulators2_impedance1_e.html
 
 
C1-C4, R1 and R2 form a CRC filter with a corner frequency around 60Hz to filter the power before it even hits the regulators. This is somewhat overkill and requires large resistors, but I had the space.
 
I don't bypass the LM317 or LM337 with a small ceramic capacitor as the indication for those is being more than 3 inches from the input filter capacitors. 
 
The other resistor values are simply necessary to have a board which accepts a 10k pot (a common value) and uses its entire range to vary from 1.25v to 37V. I should make a better writeup of selecting these values as it is somewhat tricky and results in a slightly non-linear voltage curve as you turn the pot.
 
 
 
Board Layout

---

I kind of just trusted gEDA PCB when I imported my schematic to get a set of ratlines. I can't help but feel like like either D5 or D8 is oriented in the wrong direction. Some thought needs to go here. I also haven't really taken advantage of my ability to make really thick traces for the power lines.  They are currently 85mil and the signal lines are something like 40mil.
 

 
 
One thing that's difficulty to understand from this (since it will be self-etched) is the grounding scheme.  The board has copper on both sides, and what we're looking at is the bottom layer. I intend to use the top as a ground plane, and the 5 vias you on the midline of the board are places where the ground plane attaches to the bottom layer. The other two vias at the extrema of the midline are actually just holes for terminal blocks.  The ground plane will be tied to the polygon fill and the chassis in one spot where I will put a standoff.  I haven't decided the best place for this.
 
I'm really looking for some advice about my grounding scheme. I've read loads about the subject here and on diyaudio but I just see a number of conflicting ideas.
 
 
Thanks,
-Kris
 
 

 

[Avro_Arrow]

I've built that design before using LM317/337 and it didn't track very well.
At high voltages it wasn't too bad but when you got down below +- 10
volts it started to get ugly. Down around 5 volts it could be off by a couple
of volts.
 
Maybe your results will be better than mine...
I just use dual independent regulators...
 
You can use polygons for your power traces as well as your ground plane.

 

[kriskalish]

 
Quote:

...it didn't track very well. At high voltages it wasn't too bad but when you got down below +- 10 volts it started to get ugly. Down around 5 volts it could be off by a couple of volts.

 
Damn, this is kind of sad to hear.  I might just build it anyway since I still have enough board left to etch a second design. If it ends up being terrible like yours I can come up with something different. Thanks for the heads up though; now it will be a lot less infuriating if it doesn't work as expected.
 
 
Quote:

I just use dual independent regulators...

 
Do you just control them with separate knobs or do you have an opamp match them?
 
-Kris

 

[Avro_Arrow]

Here is an example:
 
Top Layer...Ground plane
 

 
Bottom Layer...Power
 

 

 

[wakibaki]

Tracking regulator idea...
 

 
I can't claim credit for the basic idea, I was talking about tracking regulators with a friend, he sent me a schematic he found for a fixed voltage one with a 7805. I modified it to give a variable voltage. This is good for ~1.5A depending on regulator type, and input voltages up to +/-18V, you can get more current with pass transistors and more voltage with a different opamp, say OPA454. See LM317 datasheet for RV1, R3 value calculations
 
w
 
p.s. I haven't tried it, but it looks OK.

 

[kriskalish]

I tried etching the board I posted, thinking I would build the power supply and test how well it tracks, then redesign later with big polygon traces everywhere. However, long story short, I didn't reserve enough space for my heatsinks, so I couldn't assemble the board. It was also the first board I've ever etchedso the results left something to be desired.
 
 
This ended up being a nontrivial redesign as I had to squeeze some extra space out of the design, make a real footprint for the heatsink, and figure out how to attach lines to polygons in gEDA PCB.
 
Here's the resulting layout:

 
I ended up getting rid of the ground fill because I simply couldn't figure out how to get gEDA pcb to put clearance around the other polygons. Does anyone think this is a real problem?
 
-Kris
 
 

 

[Avro_Arrow]

I would say that the ground pour is more important the the power planes.
R5/R6 are usually tantalum caps...you have footprints for electrolytics there.