badoogie wrote
Thanks for the replies guys. I'm building these in separate enclosures, so each will need its own protection/filtering.
When choosing caps for the filtering is it best to breadboard and try a couple of values?
induction, would Vref be pin 3? What sort of value cap would be required?
First question:
Filter caps are used in a
low-pass filter arrangement. Look at R1 and C1 in Frank's PSU Filter 5. If you assume your hum is 60 Hz and higher, then you can choose C1 to make sure the corner frequency is below 60 Hz and you'll be fine. I don't know what layout you're using, or if there is an equivalent to R1 on your board, though. The wire itself will have some resistance, so if you don't have R1 in your circuit, it will still work, but I'd increase C1 to maybe 470u, or just try values on the breadboard, as you say. I find you can often get away with anything 10u or above. I like to use 47u because I have a ton of very small, 24V 47u electros.
Generally it's best to use a lower value on Vref than on V+. It often doesn't matter, but sometimes it does. When you pull the power, the bigger the cap, the slower it drains. If Vref drains slower than V+ you can sometimes have DC current running backward through the circuit, which some components don't like. Voltage regulators are particularly troublesome, unless you give them their own protection diodes.
Second question:
Yes, Vref in this case is found on pin 3 of the original OS schematic, but you can't just put the cap there without making some other changes, because it will bleed your signal to ground, treble-first. PSU Filter 5 is the right one for this job. (I wouldn't use Filter 6 for anything, by the way. That's just a bad design. C1 feeds them hum directly to Vref. It's a recipe for hum and oscillation.) Replace R2 and R3 in Filter 5 with 390k and 470k, respectively. These are the bias resistors connected to pin3. The trick, though, is to put a 220k between the Vref node in Filter 5 and pin 3 of the op-amp. The 47n cap should still go directly to pin 3, but it should be separated from Vref by the added 220k resistor. Why 220k? 390k and 470k in parallel give 213k. The closest standard resistor value to 213k is 220k. The 220k makes sure that your input signal sees the same resistance between itself and ground as it did in the stock OS. This whole arrangement is called 'low-noise biasing' (or sometimes 'noiseless biasing' - that name is maybe a bit optimistic, but believe me, it works really well). The good part is that once you learn this trick, you can apply it over and over. It's very useful. The bad part is that it's often hard to modify a layout to allow for it. Unless you modify the layout, it's often necessary to do some point-to-point work to separate the Vref node from the input signal with the 220k resistor.
In case this is confusing, here are some pictures that should clear it up. Compare C4, R4, and R8 in the
OS schematic (with the original biasing setup) to R1, R2, R3, C1, and C2 in this
buffer schematic (with low-noise biasing). The component values will depend on the circuit, but the low-noise biasing topology is fairly universal.
For what it's worth, I use some variation of Filter 5 on every circuit I make. I'll usually use an inline protection diode instead of a parallel one (D1), and sometimes R1 has to be adjusted downward or removed completely if it drops too much voltage. I use low-noise biasing almost exclusively, whether it's in the original circuit or not. (Some circuits like the COT-50 and the Fuzz Face don't allow for it, so it's either a battery or a regulated adapter for them. I use my own regulated power bricks for all of my pedals, but I still use low-noise biasing and decent V+ filtering no matter what, because I'm anal about these sorts of things.) Sometimes you don't need a Vref node. In that case I still include the V+ filtering and polarity protection from Filter 5, I just don't include the biasing network.
Locked
