I'm building a small audio amp (a current controlled varient of the szekeres design) and it apparently needs a very clean, regulated power source. Due to minimum parts order requirements, i'm going to end up with, amongst other things, a spare lm317 and others. As i understand, both the standard transformer-> wheatstone bridge bridge rectifier and switching power supplies are relatively noisy, so since i have parts spare, i'm wondering if following the reference design to build a voltage regulator would make a difference, or if i should just use a switching power supply to power it - i should be able to find a 15-20 v one which i'm planning on using anyway, and not using the voltage regulator at all for power supply

For that matter does using a voltage regulator reduce rippling and noise, or is it just a complicated voltage divider?

  • 2
    \$\begingroup\$ standard transformer-> bridge rectifier is actually very quiet, for the most part. As you load it, you will get some 120 hz noize, but there is little to no high-frequency noise. However, the output voltage will vary significantly as the load changes. \$\endgroup\$ – Connor Wolf Nov 8 '10 at 5:27
  • 2
    \$\begingroup\$ Do you mean "bridge rectifier"? \$\endgroup\$ – endolith Nov 9 '10 at 15:22
  • 1
    \$\begingroup\$ @Fake: Hmm? I was talking to the questioner. Said "Wheatstone bridge" (used to measure resistances), when I think they mean "bridge rectifier". \$\endgroup\$ – endolith Nov 10 '10 at 16:49
  • 2
    \$\begingroup\$ the layout of the grounds in the circuit will probably cause more noise issues than a switching power supply. Just use a high frequency (~400khz->2mhz) buck-boost switcher and put a decent LC filter on the output. If you find you need it "ultra clean" use a higher order LC filter on the output. \$\endgroup\$ – Mark Nov 10 '10 at 17:45
  • \$\begingroup\$ All the answers are equally good, but well, its impolite if i don't select a best one. Thanks all - i'll be keeping this in mind when i build my amp ^^ \$\endgroup\$ – Journeyman Geek Nov 12 '10 at 5:45

Switching power is noisy, there's no doubt about it - typically anywhere from kilohertz up to the megahertz range, both CM and DM.

I also think you're getting your terms mixed up. I presume you meant to say transformer and bridge rectifier, not Wheatstone bridge.

You can use a switcher to feed a 317 regulator and get the benefit of cleaner output and less efficiency loss than a fully linear solution (mains frequency transformer et al)

A 317 is active and will reject ripple as a function of the voltage feedback network controlling the series-pass transistor in the device.


The voltage regulator isn't a glorified voltage divider - it has internal references and feedback circuitry that allow it to actively maintain the power level. It's not perfect (so you still get ripple) but it is much better than a voltage divider.

Power supplies are generally rated by two criteria: line regulation and load regulation. Line regulation is the ability of the supply to compensate for changes in the input voltage. This helps to reduce ripple - if you're using a transformer and wheatstone bridge you'll still have some ripple after low-pass filtering. Good line regulation (low delta Vout / delta Vin) means that the output voltage resists these ripples and is cleaner. Load regulation deals with how well the supply can deliver current while maintaining voltage. If you were to draw no current from the supply, chances are you'd get the correct output voltage. But as you draw more and more current, most supplies lose voltage. A good load regulation characteristic (low delta Vout / delta I load) means your power supply maintains the correct voltage despite changes in the load.

Voltage dividers have terrible line or load regulation - they're just feed forward devices that don't compensate their outputs at all. Any regulated supply will be better than that. Look at the regulators data sheets and see what numbers they provide for line and load regulation. Also important will be bandwidth. The regulator has a little control loop inside that can only account for ripple that it sees. Any ripple that falls outside of its bandwidth is invisible and it can't fix. If you see a particular problem frequency you may need to switch regulators. There are also some application notes from some manufacturers that contain information on how to extend the bandwidth of regulators sometimes.

Good luck.


Regulated power (voltage) supply operation:

To start from the big, more simplistic perspective, the job of the voltage regulators, both switching and linear is to act as ideal voltage source. That is to provide constant voltage even in case of varying load and/or its own supply.

Usually it is being achieved by using feedback loop. In such setting output voltage is being sensed and in case it drops below set value, something is done to provide more current to output which should result in output voltage going back to set value (and reverse if voltage is above set value). This "something" in linear regulators is to make pass element *1) to conduct more current from input to output by adjusting base current or gate voltage. In switching regulator usually "something" is to adjust frequency and/ or duty cycle to achieve the same goal. So in sum both linear and switching regulators main job is to reduce output voltage variations.

Now, nothing in life is perfect, and both realizations of the same goal have (severe) limitations. There is a lot of factors to to take into account (line, load regulation, regulation speed, stability, output noise, operating input/output voltage/current range and many many more) but for sake of (over)simplification linear regulators are better at providing ripple-less output then switched at the cost of the efficiency (this is beacouse switched regulator introduce its own ripple, but in turn are more efficient and can do things that linear regulators cannot do - such as stepping up voltage).

For the case from question:

A) In this application one really need good, regulated power source, as 50Hz/60Hz (100Hz/120Hz) ripple from mains are audible (so called power line hum). Also beacouse linked amplifiers trade supply variation immunity for simplicity.

B) LM317, from its DS, has 80dB *2) typical ripple rejection at 120Hz (power line x2). That is if you have 1V pk-pk ripple at the input your output ripple should be 0.1mV (10k times attenuation). Do not quote me on that (as there are a lot of factors to take care of) but it look like this should be more then enough for this application.

C) The switching regulator/power supply may be good enough providing that it does reject 100Hz/120Hz very well (80dB as in case of LM317 would be nice). Even as it produce more ripple (finding one with less then 5mV output self-ripple is difficult), if those are above 20kHz (and for most of the switching supply it is the case) you do not have to worry about it as those ripple would be outside frequency range that human ear can hear.

BTW you can look at linear regulators as "complicated voltage dividers"), which in fact is quite good analogy (as the pass element could be treated as "adjustable" resistance). Note, however, this "complication" gives you 80dB of ripple rejection :)

*1) pass element - usually it is transistor, BJT or MOSFET, connected between regulator input and output. The feedback loop will adjust it towards more "open" or "close" state such this element will pass more or less current in order to maintain output voltage.

*2) you need to design it in correctly, that is provide sufficient decoupling caps, make sure that it will work with appropriate drop to maintain regulation etc. Documentation is your friend.

  • \$\begingroup\$ I find when I write something this long I have to use headings to get people to read it. The headings allow people to see organization and it looks like a shorter answer while still getting the message across. \$\endgroup\$ – Kortuk Nov 8 '10 at 15:31

A linear regulator is a voltage regulator based on an active device such as a bipolar junction transistor (BJT) or field effect transistor (FET) operating in its linear region. It is very inefficient compared to a switched-mode power supply, since the difference between the input and output voltage is dissipated as heat.

The LM317 has three terminals: input, output, and adjust. The regulator develops a nominal 1.25V reference voltage between the output and adjust terminals. This constant voltage is applied across a resistor, causing a constant current to flow. This constant current flows through a second resistor tied to ground. By varying the value of the second resistor, the voltage across it will vary and hence the output voltage can be set.

Although the LM317 can be used without capacitors, adding a 1 uF capacitor on both the input and output will give a cleaner output.

This page is useful for calculating the values of the resistors. Here is another one.


The voltage regulator compares the output voltage against a reference voltage (often built into the regulator itself), so if there were no any imperfections, the output voltage would be totally independent from the input voltage, output current, temperature etc.

A switching regulator may not necessarily be a bad idea, if you can make sure the switching noise is always out of frequency band of your signal of interest (here 20Hz~20kHz) - then it is just as good as a linear voltage regulator. In practice that might not be so easy to verify (the noise is modulated by the loop response etc.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.