Proper op amp power supply?

[ftransform]

So I am wondering how are dual supply op amps powered in the industry? I know of a dual tracking linear regulator, the http://www.ti.com/product/LM325#

This is now obsolete. So I am left to wonder, what are the best solutions to power low noise precision dual supply op amps?
I'm having some trouble figuring this out.... I found a few hobbyist circuits that will do this but how is it done professionally? Dual supply op-amps seem so common and I can't seem to find a commonly used solution to this. Why did they stop producing that IC? IT seems like a treat...

Any other tips relating to op-amp power supply would be appreciated

[Rerouter]

in most of these cases, the supplies are a fixed voltage so it vastly simplifies the design, for precision work it is almost always linear, or linear follower, as for stability and precision its generally going mad on decoupling, and depending on the speeds, planning out your ground and power traces (even ground has resistance and thus a voltage drop)

if you look at most of the older test gear supplies, they where done with an op amp and pass element and slightly excessive bulk capacitance on the output, thus a stable low noise rail,
every op amp had seperate decoupling from rail to ground as close as possible, but no rails under the pins (crosstalk from rail into high impedences!)

if you do want to do tracking though, its still pretty much the domain of the lm317/338, or similar topologies, where they are a matched pair, using the same dual potentiometer for a reference,

[ftransform]

Sorry I forgot to clarify I did mean a fixed voltage, not a variable voltage supply.
Are there good +- 10v linear regulators?

Using two is not a big deal I suppose, I was just curious.

[SeanB]

Most opamps are run from 15V rails, commonly using a 7815/7915 regulator set and big input and output capacitors. I have seen rails of up to +-24V to get a 15V output from the opamps, using standard industrial opamps.

[Mechatrommer]

So I am wondering how are dual supply op amps powered in the industry? I know of a dual tracking linear regulator...
I found a few hobbyist circuits that will do this but how is it done professionally?

to name a few, i guess...

RC4194 http://www.datasheetcatalog.org/datasheets/120/354845_DS.pdf
http://www.ebay.com.my/itm/1PC-RC4195NB-Dual-tracking-Voltage-Regulator-XR4195CP-/221135419460?pt=UK_BOI_Electrical_Components_Supplies_ET&hash=item337cb2b844
ebay module http://www.ebay.com.my/itm/Mini-Power-Dual-Supply-Step-down-Voltage-Regulator-Output-1A-/221075949903?pt=LH_DefaultDomain_0&var=&hash=item7918663fe6
http://www.mornsun-power.com/UploadFiles/pdf/A_S-2W%20&%20B_S-2W_EN.pdf

[ftransform]

How could you guys describe the capacitors on the input and output, other then big and excessively big?

[Mechatrommer]

read the datasheet

[ftransform]

I figured that big and excessively big meant something above and beyond the datasheet.

[ftransform]

Serves me right for not using digikey...http://www.digikey.com/scripts/dksearch/dksus.dll

[Rerouter]

there are 1001 different sites sporting a ripple vs current calculator for capacitors, to that respect you work out what value you need to get the performance your chasing and go past the mark to as far as you want to,

e.g. for 10mA, a 1uF might suit most electronics perfectly, but for a very precise analog segment you might up it to 100uF, aka going overkill,

in addition to this and in your favour a little, there is a factor on your datasheets known as supply rail noise immunity, apart from when the supply dips too far for a given output, it may well be anywhere from 20 decibels for a really poor one to 120 decibel for a precision one, while this does vary with frequency it gives you a good idea of what its capable of, this however means that the rail variations should not effect the inputs or outputs

if you where doing something like a voltage reference with an op amp output, that noise immunity works in your favour, however if your using it with a pass element, it changes the whole ball game, as the op amp is then having to correct for every variation on its input, note if it was directly on its output a few uV of noise, if using a control element there is correction noise (when the op amp output tries to vary it) and the noise on the control element supply it wasnt fast enough to correct for, this could be say 1-10mV of noise, it is this case primarily where a really stable rail is required, or when you are working very close to the limitations of your op amps inputs and outputs,

[ftransform]

For reference of other people this parameter is known as PSRR and it is given in decibels as you have said.

http://en.wikipedia.org/wiki/Power_supply_rejection_ratio

[Mechatrommer]

I figured that big and excessively big meant something above and beyond the datasheet.

10,000 µF should be good enough.

[ftransform]

can an audio cap double as a heat sink for my linear regulator?

[Rerouter]

NEVER!, seriously never use a capacitor near a large heat source, best case it fails in a year, worst it explodes or leaks onto the regulator and causes some streif for whatever its powering,

[free_electron]

I figured that big and excessively big meant something above and beyond the datasheet.

10,000 µF should be good enough.

and it can have an esr and esl so large it actually does more harm then have any effect at all...

capacitors need to be selected and selected right. just slapping them on does more harm.

dual supply is not a problem. simply use paired regulators 7805 7905 etc lm317 lm337. they only thing you want to avoid is having intermediate nodes beeing pulled under the rails during powerup powerdown. so put a diode antiparalel across each regulator. that avoids nasty surprizes... and check your biasing circuitry for stored charge  that could do the same ... any refrence voltage generated that has a cap attached to it should have a bleed-off diode to discharge the cap at power down.

[Mechatrommer]

I figured that big and excessively big meant something above and beyond the datasheet.

10,000 µF should be good enough.

and it can have an esr and esl so large it actually does more harm then have any effect at all...
capacitors need to be selected and selected right. just slapping them on does more harm.

you know its a bluff right? but since the OP want something more than what the datasheet advices. and oh, if you concern about the esr, make it 10 x 1000µF paralled

[free_electron]

and too low an ESr can have other repercussions... any idea what the peak current is to charge a 10uF X7R 0805 at powerup ?

[Mechatrommer]

too large capacitance will pegged the power source  (i'm not talking mains supply, i talked about limited wattage psu)

[ftransform]

im sorry that capacitor thing was a complete troll. Btw I am interested in temp sensing not audio so I am not a audiofool.

[ftransform]

Can anyone explain to me how PSRR works for an op amp?

http://en.wikipedia.org/wiki/Power_supply_rejection_ratio

I'd like to be confirmed.
If I have a opamp with a gain of 2, which comes out to be -6 db and it has 120db PSRR.

Lets say my voltage ripple on the power supply is 1V.

So is the formula:
120 - (-6) = 126
1V * 10^(-126/20) = .5 microvolt ripple

And assuming the power supply ripple is 1.5mV peak to peak (spec for a linear power supply I am thinking of buying)
0.0015*10^(-126/20) = 0.75 nanovolt ripple

This seems adequate for testing my op-amps right? 0.75 nV ripple seems really small. Especially considering that the noise of my op-amps output can be as large as 100nV without causing a problem on my ADC's least significant bit..



Now something odd...
When I looked at a few linear regulators I found that they have a output ripple voltage of around 80uV, which is significantly better.
Correct me if I am wrong but these fixed linear regulators seem to offer better ripple output then high end Agilent supplies?? The new agilent supplies offer ~200uVrms noise and ~1mV p-p.
What am I missing? Are bench top supplies that much nosier then in circuit linear regulators?

Can anyone give an example of a type of sensor system that would be effected by 1nV ripple noise?

[Rerouter]

psrr simply means that internal to the chip the crosstalk between the supply and the output

every 20db is another power of 10, so 120-6, (larger gains suffer more crosstalk) comes close to 1/500,000th of the supply noise

so your 1V of noise would be 2uV of noise at the output, now i would like to remind you that when uV's matter to you you really need to rethink designs, as even a massive thick ground plane is going to shift by more than that over a balanced star ground any current flowing through any trace will need to be planned out and such, even leakages between traces begin to matter,

to take it a little bit furthur, for a normal linear supply your noise is generally around 10mV, so 20nV effective noise on the output, and just about everything else in the circuit would far outway this, even parasitic capacitance variences over temperature begin to matter at that level,

so for what type of sensor would be effected by it, it comes down to what your doing, i would imagine accurate voltage or current measurements, or voltage references and current references,

as for why the lab power supplies may well be near that 1mV of noise, using a pass element cripples that psrr, so short of using one heck of an op amp to supply 3-4A your op amp has to correct for every small fluctuation on the pass elements input, equally to get a fast response of the output voltage and the ability to current limit generally requires a fairly small output capacitance,

on top of all that pass elements have capacitances, turn on delays and turn off delays, which limits how fast the loop can respond to the noise,

[ftransform]

Now I feel much better about my old ass power supply! 

And that spec was without any amplifier gain. This stuff has so many damn factors it makes my head spin!

[Rerouter]

the other big tip, always have a minimum load on your supply rails, not for discharging while off but so the loop is constantly in control, should your control element become saturated or turned off, your out of control and funny things can happen with your circuit, equally it minimises the recovery time of overshoots,