(Or how I stopped worrying and learned to love crappy power supplies)

Having seen a lot of circuits on the internet, I have discovered that a majority of them were designed to be as inexpensive as possible. There's nothing wrong with that! Having shelled out hundreds of dollars for parts myself, if I can save a buck or two, why not??

But there comes a point when you've put everything together and, lets face it, its quirky. It worked fine when you had it on the test bench, but when you went and plugged it into the rest of the circuitry, it doesn't seem to act the same! What happened?? No, it didn't have to do with the parts you used, or even the circuit itself. It's operating just the way it was designed. But perhaps it is lacking something…

Probably it is lacking a stable voltage reference.

Now most of us should know what a voltage regulator is, since most power supplies have at least one if not a whole chain of them (in the case of some vintage synths). A voltage reference is similar, in the sense that it is designed to keep a set voltage, but is not designed to handle the large amounts of current that can be drawn through a regulator! In fact, while a LM323 may be able to push 3A at 5V, an LM336-5.0V can only push up to 10mA!

So why would we really need voltage references, as opposed to regulators? Without the ability to source or sink a large amount of current, they can't be used to power circuitry. An obvious use is for A/D or D/A converters which need a stable voltage from which to scale its signals. But what about analog circuitry? What benefit can we gain by using them?

The first thing we should look at is the power supply. Now a power supply is typically very stable. Usually capable of 1% regulation over its load capacity and temperature range. However, there are some circuits which contain high precision circuitry, and obviously: the better the components; the better the operating environment; the better the circuit will work!

Obviously we can buy the best parts and end up with a better circuit. But I doubt that you will find me buying those Holco 0.1% 25PPM/C 1/4W resistors at roughly a buck a piece! It's not worth being that retentive. Besides, even with precision resistors won't solve the problems that a voltage reference will!

So what am I referring to? Well, mainly it is supply line noise and transient rejection. I don't care if you have caps on every integrated circuit on your board, the power supply is the worst place to find a stable voltage. This is because all of the circuits that the power supply is feeding are generating line/load transients.

What this means, is that if you have a VCO which is oscillating at 1kHz, if you take a look at the supply voltage under an oscilloscope, you will see a 1kHz signal on the supply line. Because the VCO's load characteristics are changing constantly, so does the power supply have to try to keep up. This problem is then compounded when you feed a lot of circuitry off of one power supply.

What this leaves us with, then, is a stable voltage with transients and/or noise that your circuitry may really not want to deal with.

Ever have two VCOs that seem to sync together when you tune them in close intervals? They shouldn't!

So how will using a voltage reference help? Well the only thing a reference can do is keep a very accurate voltage level. It wasn't designed to power circuitry. That presents only a minor problem. After all, we know that Op-Amps reject power supply transients pretty well and most can source enough current to run a small amount of circuitry. Take a look at the circuit below:

Some of the biggest problems that one can face is when there are more than one circuit being powered by a power supply. If all our circuits were doing nothing more than outputting DC signals (no audio or even sub-audio!) we'd be in good shape. But as circuits oscillate, their load characteristics change. Unfortunately, the power supply has to try to keep up with this (remember, it is regulating). And unfortunately, we tend to see signal bleedthrough onto the supply line.

No, it doesn't matter how much you filter and filter, its an unavoidable situation! However, there is some good news. First, Op-Amps (and most linear ICs) tend to have very good power supply rejection! This is great for those with noisy power supplies, but unfortunately, the bad news is that for most discrete circuitry there is little or no supply rejection built in!

The solution is to use a voltage reference (a good one, mind you!) to isolate sensitive circuitry from the garbage that sits on the supply lines. There are a number of good voltage references, my favorite being the LM4040 series from National. They have different grades available in a large variety of set voltages.

For the ultimate in stability, use an LM4040AIZ (A grade, TO-92 case) for 0.1% accuracy at 2.5, 4.096, 5.0, 6.0, 8.192, or 10 Volts. At worst, use a D grade (LM4040DIZ) for 1.0% accuracy, which will at least remove line transients from your sensitive circuitry.

Email the author: Tony Clark