What's your opamp dual supply from single supply jellybean solution?

[dusan]

What's your opamp dual supply from single supply jellybean solution? If you need dual supply (either postive to negative, or positive to positive and negative) and maybe some extra regulation afterwards if you need it. My current consumption is few mA.

[rvalente]

For my simple op-amp circuits: virtual ground (kinda inefficient, works ok for audio stuff) or voltage inverter with a TC7660 (noisy).
Liked the topic, good stuff will come up from the seniors

[langwadt]

this is clever, https://www.edn.com/efficient-voltage-doubler-is-made-from-generic-cmos-inverters/

[pqass]

Three possibilities, see here.

[reboots]

I have heard that the TI "rail splitter" ICs (e.g. TLE2426) are considered too noisy for some applications, and that there are discrete active virtual ground designs which perform better. But that's all I've heard about that.

[ArdWar]

TLE2426

I'm not exactly sure why would anyone still use that today over the much more ubiquitous opamp and 2 resistors solution. It might have their merit back in the day (low area/part count, guaranteed stability?), but I don't think it offer much of an advantage today over BOM consolidation etc.

[PCB.Wiz]

What's your opamp dual supply from single supply jellybean solution? If you need dual supply (either postive to negative, or positive to positive and negative) and maybe some extra regulation afterwards if you need it. My current consumption is few mA.

You have not given starting and end voltages.

Already mention is the 7660, that is the classic 8-pin simple voltage inverter pump, made by many vendors.

Less conventional is to use MAX202 / variants as a charge pump, that starts from 5V or 3v3 and gives RS232 supplies circa +/- 9V. These series parts are widely available and cheap.

There are also LCD driver pumps similar, that take 3-5V and give +/- 5.75V

TI have SOT23 parts that take 3-5V in and give negative mirror voltage.

[PCB.Wiz]

TLE2426

I'm not exactly sure why would anyone still use that today over the much more ubiquitous opamp and 2 resistors solution. It might have their merit back in the day (low area/part count, guaranteed stability?), but I don't think it offer much of an advantage today over BOM consolidation etc.

Yes, thirty+ years ago, getting sub 1% typical and 25ppm in a single TO92 package had some appeal.
These days, 1% or better, and 25ppm resistors are mainstream, and the TLE2426 now has a maintenance price.

[reboots]

TLE2426

I'm not exactly sure why would anyone still use that today over the much more ubiquitous opamp and 2 resistors solution. It might have their merit back in the day (low area/part count, guaranteed stability?), but I don't think it offer much of an advantage today over BOM consolidation etc.

The "rail splitter" parts were widely promoted and used at one time, and might still seem like an obvious choice to a novice searching the Web for virtual ground solutions. Just a note for any such person who might be reading this thread.

[voltsandjolts]

XL6012 example:

https://www.eevblog.com/forum/testgear/juntek-dpa-2698-10mhz-signal-amplifier-digilent-analog-discovery/msg3649978/#msg3649978

[rvalente]

Inverting regulator, like 34063 https://www.onsemi.com/pdf/datasheet/mc34063a-d.pdf, there are better choices out there, as the 34063 is also noisy.

Inverting charge pump, similar to TC7660 or MAX232 as suggested, but driven from a oscillator (often a uC pin) or simple 555
https://blog.mbedded.ninja/electronics/components/power-regulators/charge-pumps/

Possibilities are endless, i guess

[dusan]

Inverting charge pump, similar to TC7660 or MAX232 as suggested, but driven from a oscillator (often a uC pin) or simple 555
https://blog.mbedded.ninja/electronics/components/power-regulators/charge-pumps/

Why driven from oscillator?

[rvalente]

Lets say you have a simple board with a microcontroller and a simple LCD, that needs -3V for improved contrast (like a few HD44780 clones), you could use an output from a uC clocking at A few KHz and use a charge pump to creat a very cheap negative power supply with only a few caps and a fast diode.

https://product.torexsemi.com/en/technical-support/techinfo/doc_2095

[exmadscientist]

Three possibilities, see here.

Just please don't make the idiot mistake of calling the output of a rail splitter "ground" like this article shows. It's not ground, unless it was floating and you bonded it to ground somehow. This needs to be called "common" or something equivalent. This can become critical to safety in certain circumstances, and is all too often critical to function and/or EMC.

I am willing to overlook loose terminology when there's only one reference voltage in a circuit. But once there's more than one (and a rail splitter always creates another reference voltage, so there definitely is now), you need to get this right.

Anyway, for heavier lifting, a sync buck controller rigged as an inverting buck-boost does excellent work for negative rail generation. But probably overkill for a few milliamps!

[electron_plumber]

Yet another split rail solution, when isolation is needed, is the push/pull topology. Very simple to implement.

[exe]

I have heard that the TI "rail splitter" ICs (e.g. TLE2426) are considered too noisy for some applications

I don't think this is true. Switching regulators are noisy, but this one is linear.

Yes, thirty+ years ago, getting sub 1% typical and 25ppm in a single TO92 package had some appeal.
These days, 1% or better, and 25ppm resistors are mainstream, and the TLE2426 now has a maintenance price.

I don't think precision is needed for a rail splitter that supplies opamps. Opamps don't care.

I'm not exactly sure why would anyone still use that today over the much more ubiquitous opamp and 2 resistors solution.

Most opamps cannot drive capacitive loads.

I used TL431 as a rail splitter (beware of capacitive load and stability regions, different TL) for "virtual ground", but I didn't need symmetrical split. I just needed to "lift" ground for 2V or so.

For my next project I'd just use two positive LDOs: first to stabilize total output voltage  (say, to 10V), and then the second one to split it (say, -5V from V+). Just beware most LDOs cannot sink current.

[Phil1977]

If you allow using a boost converter you can do two setups:

a) boosting e.g. 3.7V supply voltage to 15V. You can add a charge pump to the boost converter with two diodes and two capacitors that "mirrors" the output voltage to a negative -15V. It´s not very stable, but if the load on +15V usually is higher than on -15V it works.

b) boosting e.g. 5V to 12V. Define the 5V potential as the virtual ground with relatively -5V and +7V. This setup delivers quite stable voltages if the boost converter is strong enough.

[rvalente]

Yet another split rail solution, when isolation is needed, is the push/pull topology. Very simple to implement.

I believe you could use an ethernet transformer for that, easy to source

[HalFoster]

Three possibilities, see here.

Just please don't make the idiot mistake of calling the output of a rail splitter "ground" like this article shows. It's not ground, unless it was floating and you bonded it to ground somehow. This needs to be called "common" or something equivalent. This can become critical to safety in certain circumstances, and is all too often critical to function and/or EMC.

I am willing to overlook loose terminology when there's only one reference voltage in a circuit. But once there's more than one (and a rail splitter always creates another reference voltage, so there definitely is now), you need to get this right.

Anyway, for heavier lifting, a sync buck controller rigged as an inverting buck-boost does excellent work for negative rail generation. But probably overkill for a few milliamps!

It's not a mistake - "Ground" is simply a reference to a common potential.  For bonding for safety reasons this is typically earth ground, power line or chassis ground.  I agree that better naming is important for understanding the purpose of the reference potential (digital ground, analog ground or some other descriptive term) but they all only refer to some specific voltage level.  There are distinct schematic symbols for things such as the earth, power line, chassis (defined by specification or at least convention) and other types of ground that indicate their typical use.

[tooki]

Three possibilities, see here.

Just please don't make the idiot mistake of calling the output of a rail splitter "ground" like this article shows. It's not ground, unless it was floating and you bonded it to ground somehow. This needs to be called "common" or something equivalent. This can become critical to safety in certain circumstances, and is all too often critical to function and/or EMC.

I am willing to overlook loose terminology when there's only one reference voltage in a circuit. But once there's more than one (and a rail splitter always creates another reference voltage, so there definitely is now), you need to get this right.

Trying to get people to not say “ground” when it’s just the 0V reference in a circuit is… an uphill battle. That usage is thoroughly and deeply entrenched in electronics, and you know this. So I concur with HalFoster that this is not a mistake at all, never mind not an “idiot mistake”.

What is absolutely true is that one needs to be more specific when it matters, as HalFoster also said. In the article, the author refers to the ground as an “artificial earth” and explains some limitations. If anything, what the article should probably say more explicitly is that the DC supply itself needs to be floating. Remember that that website’s focus is on audio circuits, and in the context of audio gear, it’s more likely than not that the grounds of different devices get connected, so the virtual ground created here will likely connect to a true ground somewhere, and thus the PSU must not already ground one of the outputs already.

[HalFoster]

One of the most confusing issues, at least for me, is remembering that all of those different "grounds" are not necessarily at the same potential and can have significantly different impedances and other properties.  For example, a single supply op-amp circuit that uses a artificial ground:  I just sketched out a circuit a few days ago for a RF amplifier and caught myself about to bypass the power terminals of the op-amps to the generated ground. Oops (I was using another HF op-amp as a buffer for the V+/2).   Anyway, there are some very good application notes floating around about (what is the best name? Artificial? Generated? Virtual to me is reserved for inverting op-amp circuits) the artificial grounds from TI, AD and others - there are a lot of subtleties that come into play depending on the application.

On a completely unrelated note: ever notice that the slew rate you need is always about 10-20% more than the slew rate of the op-amp that you have? Sigh.

Hal

[jwet]

Just a note- several mentioned the 7660.  This is a great problem solver for making a low current negative rail but has been improved mightily over the years.  Unfortunately, these old devices ran at very low frequencies as low as 5Khz for some and required large caps, originally 10 uF.    Newer pin compatible parts from all the standard sources run at much higher frequencies, use small caps and are available in SOT's, etc.  A nice example is the ADI MAX1852/3 series, they run at 50/200 Khz, make 30 mA and use .68 uF caps and come in a SC-70.  Go on DigiKey and do a search for Charge Pumps.

[Buriedcode]

Jellybean?

Charge pumps, usually good for <15mA, couple of schottky diodes and caps driven from a TTL oscillator (74xx14) or 555 timer.  That either gives positive, or negative or both.  And if you don't connect the doubler to VCC you get fairly symetrical outputs.  If you can accept lower currents, you can add two stages on the negative side for Vout = -(2*Vin - 2*DVf).   I have a strip of 100 dual schottky diodes in SOT23 specifically for this - so 3 components (dual diode + 2 caps) gives me a negative rail good enough for a few opamp buffers.

It's quick and simple, not great efficiency, but not bad EMI, and the output ripple is often fine for opamp circuits given their PSRR.

There are however "better" solutions.  The next step up is a boost converter with a charge pump for generating negative output, which is much more convenient than a negative buck-boost with a positive charge pump.

[beenai2018]

The attached  idea pops up a few times in the EDN archives from Linear Technology .I have not tested it personally ..

[beenai2018]

Here is one more from the High-Priest . I imagine he  said " Hold my beer"..

Single inductor 5V to +- 15V.


Full app note 
https://www.analog.com/en/resources/app-notes/an-29f.html