What 9v regulators are out there? Looking for better performance.

[azrael]

The 7809 is something that's commonly used, but the LM317 has lower noise characteristics.

Is there anything better than that? I've tried searching, but most companies only have stuff up to 5v. Texas Instruments has a terrible search engine and their parametric search seems to be broken at the time of this post.

So, what regulator is best to power our pedals? Maybe even a discrete design?

[Lizard King]

Not to jack your thread, but I'm looking for - I'm not sure what you call them - low overhead regulators....one's that will pass 9v even with a 9v input.  I've heard of them but haven't been able to find any.

[Mark Hammer]

The 7809 is something that's commonly used, but the LM317 has lower noise characteristics.

Is there anything better than that? I've tried searching, but most companies only have stuff up to 5v. Texas Instruments has a terrible search engine and their parametric search seems to be broken at the time of this post.

So, what regulator is best to power our pedals? Maybe even a discrete design?

1) WHAT noise?  Do you mean ripple, or do you mean some other sort of noise?

2) I'm not sure what you would consider better performance to be.  In many instances, I suspect people are expecting the regulator to do something on its own that realistically requires a little bit of help from the rest of the circuit.  For instance, a wallwart that feeds a regulator 12v that meets the barebones minimum to quaify as "DC", is not going to bring out the best outcome unless that regulator has some assistance from other components.

[R.G.]

So, what regulator is best to power our pedals? Maybe even a discrete design?

Let's talk about that word "best" and the idea of diminishing returns.

First, have you personally ever heard the difference between a battery, a 7809, and an LM317? If so, what was the audible difference?

Second, "best" is a meaningless word unless you have some way to measure it. Does "best" mean lowest noise, smallest variation from the set DC level, smallest sag under load, fastest recovery from a load transient, or even some specific internal impedance? Or does it mean the sound coming from the pedal is somehow better, and if so, how?

Diminishing returns is an issue in almost all technical designs. The idea is simple - getting the majority of the "goodness" from some design may be easy, but each additional unit of goodness obtained by more extensive, detailed design and elaborate circuitry costs more and more to get each additional step of goodness. As an example, I once designed power supplies and power regulators for a living. We did a lot of discrete design work. I have personally designed all-discrete voltage regulators. I can tell you that it is not simple and easy to beat the performance of a 7809 or LM317 on many of the ways their performance can be measured, including most particularly stability under varying loads.

So - how do you measure "best", and how much are you willing to spend to move a little closer to that. Remember that one measure of "goodness" for DIYers is the ability to stuff in one part and have it work first time. 

Not to jack your thread, but I'm looking for - I'm not sure what you call them - low overhead regulators....one's that will pass 9v even with a 9v input.  I've heard of them but haven't been able to find any.

They're called "Low Drop Out" regulators, or LDOs. http://focus.ti.com/paramsearch/docs/parametricsearch.tsp?family=analog&familyId=400&uiTemplateId=NODE_STRY_PGE_T&DCMP=hpa_pmp_ldo&HQS=ldoparametric-bb

[azrael]

Well, to clarify, I have a transformer with several secondaries that I'd like to make into a power supply.

As for noise, no, I haven't measured the difference between a battery and different regulators.

I'm just looking to design a nice regulated power supply. So when I so say best, I mean fairly fast transient response, and low output noise. I was set on the 317, but I just wanted to know if there's something better out there - my goal when building stuff if to do something different or better than what's on the market. Otherwise, why not buy it? So I'd like to outperform or match my Voodoo Labs Pedal Power II.

[Chris Brown]

Slightly off topic... It would definitely be cool to run a pedal board on 3 rechargeable 18650 li-ion batteries with an appropriate 9v regulator (perhaps the 3A LM350)... has anyone tried this?

[R.G.]

... when I so say best, I mean fairly fast transient response, and low output noise. I was set on the 317, but I just wanted to know if there's something better out there

Let me preface this with the fact that I have not done a wide survey of the available chips recently, and semiconductors keep getting better all the time. But the last time I did look at the market, I decided to use the LM317 family. They are very hard to beat in terms of the most performance for the money and time spent designing.

my goal when building stuff if to do something different or better than what's on the market. Otherwise, why not buy it? So I'd like to outperform or match my Voodoo Labs Pedal Power II.

I think the pedal power uses the LM317L inside. Could be mistaken, but I think that's what's there.

[jh9067]

I'm attempting to modify one of Jack's (www.muzique.com/) designs for a similar purpose - maybe he can clarify.
When you buy one of his multi-purpose boards, among the schematics he sends is a discrete 5v regulator.  Since it has 9v input, it seems hard to get 9v output, but my simulations seem to get a regulated 7.5v output, maybe 8v. (I'm still a very novice designer, thus the maybe) If you were willing to use 12v in, it seems like no problem to get regulated 9v out from a small discrete circuit.  Hoping to build it this week and find out!

I'm certain Jack could illuminate further!

I agree that "best" is a very loose term at best, but one of the reasons I'd like to put regulators in my designs is to avoid wild fluctuations in volume and sound quality when different or wacky power is used.  I found that the Behringer Super Fuzz (not to be confused with the real thing!) develops a strange-but-cool gated-fuzz quality when 12v is applied.  While this is a happy accident and gives the Behringer some usefulness (where it otherwise might end up in a junk pile), I'd like to avoid such accidents in my designs!

[R.G.]

I'm attempting to modify one of Jack's (www.muzique.com/) designs for a similar purpose - maybe he can clarify.
When you buy one of his multi-purpose boards, among the schematics he sends is a discrete 5v regulator.  Since it has 9v input, it seems hard to get 9v output, but my simulations seem to get a regulated 7.5v output, maybe 8v. (I'm still a very novice designer, thus the maybe) If you were willing to use 12v in, it seems like no problem to get regulated 9v out from a small discrete circuit.  Hoping to build it this week and find out!

I'm not familiar with that particular design, but I'm fairly familiar with linear regulators in general.

A big limitation to all regulators, and one thing that determines the drop-out voltage in general, is the smallest voltage the output device can make happen from input to output. Let's talk as though all regulators were positive output, and regulate on the positive side of the circuit, which is most of the standard type.

The standard version of a regulator uses an NPN as an output device, with its collector connected to the input voltage and its emitter connected to the output voltage; that is, it's set up as an emitter follower. So the smallest drop out voltage that the regulator can possibly do is determined by how hard it can turn on that NPN. Since the input voltage on the collector is the highest possible voltage in the circuit, there has to be an active device of some kind determining how much current flows from the input voltage (which equals the collector voltage) to the base of that NPN.

NPN power devices can often saturate their collector-emitter voltage to a small fraction of a volt at modest currents, but the base has to be raised higher than the collector to do this, as the base-emitter voltage never gets below the Vbe "on" voltage of about 0.7V at saturation. So with the base of the NPN shorted to the collector, the input to output voltage would not be less than about 0.7V. It gets worse because there has to be another device controlling the current into the base (otherwise, it's not regulating!) so the "on" voltage of that device is added to the minimum base-emitter voltage for the output NPN. By the time you get through getting enough current gain (the output is usually a darlington to get more current gain) and enough base drive/control, you get to the 2.0V minimum that is what the standard three terminal regulator gives. The limitations of using all NPNs on the same wafer with gain limits involved means that you might do better with discrete devices, but you're still going to be over about 1.5V best case.

The fundamental limitation is you don't have enough base-drive voltage available to saturate the output device hard enough.

The situation changes if you can make good PNPs or good P-channel MOSFETs on the same die as the NPNs used to make the rest of the regulator circuit. With a PNP, you tie the emitter to the input voltage, the collector to the output, and pull down on the base to turn it on. Now you have the whole input voltage to work with and you can saturate that PNP output pass device HARD and get its voltage down to a small fraction of a volt. This is what LDOs do - they use PNP or P-channel MOSFETs so they can drive the input-output differential down to 0.1 to 0.2V at reasonable currents.

Why aren't they all like this? It seems ideal.

It almost is. The problem that has historically kept this from happening is the poor gain and poor frequency response of PNPs that can be diffused on to the the same wafer as good NPNs.  Poor gain means you need a lot of silicon area to get current flow, and waste a lot of base-drive power. But the thing that really killed this approach was the poor frequency response. A PNP device in this configuration has voltage gain from base to collector, which is now the output. The NPN style had an emitter follower output where the gain was unity, and the frequency response was the maximum possible from the device. A poor-response PNP means that keeping the amplifier/feedback portion of the regulator stable is ... um, well,  a challenge. 

Better silicon has given us LDOs with P-type output stages. But even these have critical regions lovingly noted in the datasheets where if you load the output with too much or too little capacitance and certain output currents, it will start oscillating.  You get low input-output differential, but now it's no longer simple, and may start screaming at you.

So you may be able to get a regulator with an emitter-follower output down to about 1.5V or so from input to output with discrete devices. But you're not going to get to 0.2V, which LDOs really can do. You can build your own LDOs, but having done this a few times, you're going to learn more about stabilizing feedback amplifiers than you may want to; worse, getting it stable means giving up the feedback that makes it quiet and accurate. May as well just use more BFCs on a standard regulator to make it quiet.

Literally every possible permutation of how to hook up NPNs, PNPs, FETs, etc. has been tried for making linear regulators work well. Regulators are so important to the industry that lots of Smart Guys have been paid to look for this for decades.

I agree that "best" is a very loose term at best, but one of the reasons I'd like to put regulators in my designs is to avoid wild fluctuations in volume and sound quality when different or wacky power is used. 

Kewl. To you, one of the things in the list for "best" is that it doesn't change how it operates when odd input voltages are applied. Good for you! Good objective, and one of the reasons voltage regulators were invented.

[PRR]

> fairly fast transient response, and low output noise

I'm sorry we are bursting so many bubbles, but....

Many pedals don't put transients on the power rails.

Many regulators have goofy transient response unless there's a good cap on the output.

With cap, most regulators show good transient response for load, but do _not_ handle kick-back from the load. (It's a one-way valve.) Kickback is rare, but a few circuits will.

In audio, IMHO, the best supply is some impedance (typically resistance) then a BIG capacitor.

I've run quite dirty DC directly to audio path via 4,700uFd 100r 4,700uFd. Dead clean, hum buzz and hiss. (Clean not just to ear: I've run those recordings through audio analysis enuff times to know if there was any electric residual above room acoustic ambient.)

BTW: no regulator. Many audio circuits have a wide range between blow-up and don't-work. As long as voltage variation is _slow_ (BIG caps), it does not get to the output.

[duck_arse]

azreal, have a look at the micrel MIC5209 family of regs. they have a drop out V  of 100mV at 50mA. I use the fixed 5V sot223 cause I found some cheap, but they have adjustables as well.

[Jazznoise]

Outside of the scope of standard FX pedals, but would anyone here ever parallel regulators for better load handling? I'm always fearful of using large caps because of the low impedance being presented to the regulator upon startup. Or is that why you'd use a load resistor in the first place, PRR?

[R.G.]

Paralleling regulators can be done, but is tricky, and hence a bad idea for neophytes.

Regulators do their very electronic best to force the output voltage to be dead, rock stable. Because of the variation that's always there in electronic parts, each regulator may be some fraction of a volt different than the others. The highest voltage regulator, even by millivolts, winds up supplying nearly 100% of the load current until it runs into current limiting. Well, if it has current limiting. If it doesn't, it often dies.

When the highest voltage regulator can no longer supply the current, it sags and the next regulator starts supplying current. This relieves the load on the first one a bit, and it comes running back in, then goes into current limit again, then sags, then...

The issue is current sharing of voltage sources. If you have N parallel regulators, you want each regulator to supply 1/N of the total current all the time to spread out the thermal loading and not have funny changeover points in the loading.  Current sharing can be done with independent regulators, but it complicates things. It's easier if you can segment the load down so that each regulator supplies an independent load. This is why 3-terminals have become so popular.

If you just can't change the load to be many small ones, then it's better to make one big regulator with many parallel output devices inside it, all run by one feedback amplifier setting the voltage. There are still current sharing issues with this, but it's much easier to handle, typically a ballast resistor per output device.

[PRR]

> ever parallel regulators for better load handling?

No.

There are better ways.

FWIW, I am advocating NO regulators in many many cases. But this does assume you know the circuit being powered, well enough to know how it will work.

[jh9067]

Paul -

You always seem to have great answers, so I'm inclined to trust your judgement.  However if you have a sec to expand on why you don't like regulators, I'd love to hear it.  It just seems like a simple, low parts count regulator could go a long way in helping a circuit perform at it's best since you could tweak all the values for that specific voltage.  I've measured many popular 9v DC supplies at up to 12v, which makes certain pedals "better" or "louder," but makes building something very specific difficult.  Your thoughts?

Thanks,
JH