Choice of Stable Voltage Reference Circuit IF YOU THERMALLY POTTED IT

[minime72706]

Hey guys!

I just bought some thermal epoxy for some testing of Cree LED light bulbs as the dimmers in my house were making them vibrate a lot and I'm also mulling around the idea of building a power supply for myself as well and it occurred to me... I could probably make a pretty nice voltage reference by potting a circuit in this epoxy. One example of a type of circuit I've used in the past involved a LM334 constant-current source IC and a silicon diode for cancelling the temperature coefficient. It didn't work well if you blew on it or something because the two devices were not in thermal contact, but the idea was sound. I figured if I potted it, it would work much much better. Any other suggestions on simple circuits that make good references, especially if thermal contact between components is guaranteed? I kind of want to go somewhat overboard, but also old-school, with my design.

My current BK Precision triple-output supply is nice, but it has some short-comings: I wish it wasn't so heavy. I wish the fan wasn't on 24/7 at full blast. I wish there was one more digit on the current and voltage meters (I'll use 12-bit ADC/DACs). The list goes on.

If you're curious about the light bulbs: The circuit design is legit but it sits in the base almost entirely loose. There's quite a bodge in the design where a fairly large inductor is sitting on the board at an angle because a surface mount diode is mounted below it ... I kind of think that's the main source of noise; however, there are a total of 3 inductors. I drilled a hole in the side and pumped in potting compound and then stuck it in the oven. This didn't work so great as the potting compound seeped out and the film on the glass dome was damaged, but the proof of concept was successful in that the bulb was quieter. Next time I'll probably drill a hole in the glass (carefully!) and just wait for it to cure naturally (... I think it takes like 5 days).

[Jon86]

Depends how much accuracy you want, I guess. How far above ambient do you want to go? Personally I'd just go with an off-the-shelf shunt regulator with some tiny tempco. Makes it a lot more reliable than building your own and having a million and one things that can go wrong...

Also, potting? Potting compound? or are you referring to the thermal epoxy you've bought?

[minime72706]

It smells like epoxy but it's not marketed as such. It's the black stuff you'll find things like electronic ballasts and other power electronics caked in. I bought it primarily for the LED bulb experiment that I mentioned.

I will probably just use the TL431; I had no idea that it's tempco was as good as it is. It seems to beat the LM334 and the LM385, in the very least. I also have a couple of 4.096V references from Linear that I intended to use for a "Dave Supply" which I never even began working on. I kind of want to keep this old-school, so the TL431 is the likely choice. Going to keep SMT parts out of this if I can. The 4.096V reference is good quality but I really just got it because it pairs well with 12-bit ADCs (in the sense that it merely satisfies my mild OCD). For fun I might buffer the reference with a chopper-stabilized operational amplifier.

Right now I'm just throwing together a power supply ... for my power supply. (INCEPTION) I plan to have a separate low-current transformer for the control electronics and two nice torroidal transformers for the dual supplies. The nice thing I realized about using separate transformers is that I can put them in series, parallel or "center-tapped" series while sticking with "positive" regulator topologies. Transformers are wonderful ... sometimes.