Please critique this layout?

[ampdoctor]

I'd just like to post this layout of a fairly simplistic power supply to get a few extra eyeballs on it and find out if I've done anything uniquely stupid as well as getting any comments or suggestions for improvements. The layout is still in need of some touch-ups and the silkscreen needs a few adjustments as well.

A few minor points of interest that I should mention because I know I'm going to get called out on them;

1. The low voltage power supply is fairly under designed, however it's only being used to drive a few relays and switching transistors.  The load won't exceed 600mA so I don't see the need for anything much more complex than what I've got.

2. the alt-ht connections were added due to the occasional need to use this supply with a transformer designed for use with a full wave bridge rectifier and a few minor modifications.

[krivx]

The multiple RC filters to make lower voltage and use of "HT" suggest this is a high-voltage design. Maybe check if the distances between your traces meet minimum creepage distances?

[ampdoctor]

What voltage will the 7805 see at it's input when you're drawing your 600mA?

600mA is going to warm it up nicely at pretty much any voltage input above the minimum it needs.

The regulator is typically going to see about 8.9vdc so it'll have to dissipate just under 3 watts. While it'll get warm it shouldn't pose any reliability issues with even modest heatsinking.  I'm pretty much stuck using an off the shelf power transformer which limits me to a 6.3vac winding. I considered using a simple unregulated supply here as well but with mains voltages that can vary wildly, performance is going to get really sketchy!

I would also always use 100n's on both input and output of a voltage regulator, and at least a 10uF on the output too. A lone electrolytic will provide no real HF decoupling for the poor little regulator.

I've bodged this circuit into a number of pre-existing designs and never had any in service issues, so should it become a problem on the prototype I'll add it but I don't forsee any issues...famous last words!

The multiple RC filters to make lower voltage and use of "HT" suggest this is a high-voltage design. Maybe check if the distances between your traces meet minimum creepage distances?

Yup, it's a supply for a valve audio amplifier, so I'm looking at 460v on HT1 and HT2, and a little under 400v on HT3 and HT4. One of the problems with many commercial boards used in 'pro audio' instrument amplifiers is that they're often running right at the margin so I was quite mindful of that issue and kept my distances substantially greater than 200 thou.  The only place that might be a bit iffy is where I've necked down the trace on HT4 where I've got 75 thou of creepage. With a solder mask I'm cutting it close but don't see any other solutions without going double sided.

My biggest concern is whether or not I'm going to pick up any appreciable noise on the audio ground by tying into the digital ground.  Part of me wants to just run with a double sided board and keep the grounds fully separate up to the chassis ground point. However, I'd like to keep the board single sided if possible.

[c4757p]

The only place that might be a bit iffy is where I've necked down the trace on HT4 where I've got 75 thou of creepage. With a solder mask I'm cutting it close but don't see any other solutions without going double sided.

Make it "double sided" with a jumper!

[Bertho]

The regulator is typically going to see about 8.9vdc so it'll have to dissipate just under 3 watts. While it'll get warm it shouldn't pose any reliability issues with even modest heatsinking.

That is still a problem. A TO220 version has a Rjc of ~5 deg/W. A "modest" heatsink is 18..20 deg/W.

Dissipating 3W will see a temperature rise of 70..75 degrees C. That means a junction at room temperature (25) of about 100 degrees C. With some margins and often poor convection from the heatsink you'll end up at a junction temperature of 120..130 degrees at room temperature. That is awfully close to the maximum and will wear down the component pretty fast. It will also degrade neighbouring components at such high temperatures. You will already reach the maximum limit if the internal case temperature rises by 20..30 degrees. Not an unseen thing, especially when you have tubes in the neighbourhood.

[ampdoctor]

That is still a problem. A TO220 version has a Rjc of ~5 deg/W. A "modest" heatsink is 18..20 deg/W.

Dissipating 3W will see a temperature rise of 70..75 degrees C. That means a junction at room temperature (25) of about 100 degrees C. With some margins and often poor convection from the heatsink you'll end up at a junction temperature of 120..130 degrees at room temperature. That is awfully close to the maximum and will wear down the component pretty fast. It will also degrade neighbouring components at such high temperatures. You will already reach the maximum limit if the internal case temperature rises by 20..30 degrees. Not an unseen thing, especially when you have tubes in the neighbourhood.

Nice catch! To be candid, I never really bothered to calculate the junction temps. Having worked on these things for close to 20 years I don't often see regulators with heatsinks and most have effectively the same circuit. They just...work.  I think I may have to really wring out all the numbers to try to figure out how the major mfg's get away with it and maybe even measure some ambient case temps of a few different amplifiers to see exactly how warm they do get on the inside. My gut tells me that it wont be much more than around 40C, admittedly I might be in for a surprise.

If it starts to look really ugly I might wind up mounting the regulator vertically with a large sink, bolt the little bastard right to the chassis and forgo mounting it on the board all together, or as a worst case scenario rework the regulator design and incorporate a pass transistor.

[kripton2035]

never use a 7805 (i.e. linear regulator) if you draw more than 100mA with more than 3V of voltage drop
or you will heat the room ...
switching regulators are easy to use nowadays.

[DutchGert]

never use a 7805 (i.e. linear regulator) if you draw more than 100mA with more than 3V of voltage drop
or you will heat the room ...
switching regulators are easy to use nowadays.

Or a bit more 'modern' LDO instead of a rusty 7805 (so u get better Tj->Tcase performance.

But still, (Vin-Vout) * Iout is allwaysdisipated inside the device.

Buy a drop-in switching regulator and use that i would suggest  (http://nl.farnell.com/xp-power/tr05s05/dc-dc-converter-0-5a-5v-sip/dp/2319830)

[ampdoctor]

I ran some back of the envelope calculations and assuming an ambient of 40C and a Pd of 2.4 watts I'll need a heatsink with a thermal resistance less than 16.3C/W.  Even if I were to introduce a fudge factor and reduce that by 1/3, that's not what I'd call a massive heatsink by any means. May have to just build it up and do some empirical testing.

As I said earlier, I'm pretty much stuck with using a 6.3vac transformer winding so there's really no practical means of reducing the input voltage to improve the Pd or implement a LDO regulator, and changing the regulator to something like a LM317 won't really provide any real advantages over the 7805.  Even though it's a moldy oldie, It's still a workhorse part that I don't foresee dissapearing any time in the near future

That drop in converter looks interesting and I may give that a shot just out of curiosity.  Having never used them, I'm going to assume that with those I'll need to address high frequency switching noise on the output and possibly a few other quirks in one way or another.