Recommended single supply rail-to-rail jellybean op-amp?

[Tac Eht Xilef]

What's the go-to, cheap, jellybean, single-supply rail-to-rail output op-amp these days?

I've got a variable 1.5v-6.5v DC signal (VFO fine tune in an old receiver) that I need to convert to 0-5v for the internal ADC of an ATmega. No great difficulty; a simple differential amp referenced to 1.5v will do the trick. Both +11v & +5v supplies are available so I don't need something that can hit Vcc, but I do need something that can drive its output reasonably close to gnd (I'd prefer to avoid having to generate a -ve rail just for this).

For bonus points: Ideally, I'd like something that can be (ab)used elsewhere as a comparator to convert a < 5MHz 0.6v sine wave output to a +5v square wave

Suggestions?

[TVman]

Digikey?

http://www.digikey.com/

[Tac Eht Xilef]

Digikey?

http://www.digikey.com/

Thanks, but I already know where to buy stuff. I'm looking for recommendations for particular parts...

[TVman]

Oh, sorry I thought you wanted to know where to buy it.

[T3sl4co1l]

Not worth using an op-amp as comparator, certainly not at 5MHz.  To make a passable square wave, you'd need GBW >> 50MHz and slew rate >200 V/us.  Okay if you can find one that's uncompensated, but who even makes those anymore?

LM339 won't really do 5MHz very well, but LM319 will, as well as numerous other, much newer and better parts.  Or you can simply use a couple 2N3904s.

For generic sorts of purposes, I like TLV2372 and the like.  Good to 16V (so 12V single supply is practical, but more than that is pushing it), reasonable performance near the rails (recovery from saturation not too bad) and general purpose Vos / GBW (compare LM358, though without the RF rectification susceptibility of a bipolar amp).

As for your particular case, since you already have so much voltage swing, why not just divide it against 0V and live with the offset?  It's not like you're losing many bits -- or that you (evidently) care about many bits to begin with (the internal ADC is 10 bit with some 4 LSBs of INL/DNL error -- if you require absolute, uncalibrated accuracy, it's more like 8 bits anyway).

Tim

[poorchava]

LM358 and LM2904 swing to negative rail. Datasheet quotes 20mV max. They don't swing to positive rail though - about 2...3V below positive rail is maximum swing.

Microchip also has some nice opampa, rather widely available. They have very good rail-to-rail performance, but low power supply voltage range.

[luky315]

The TS912 is a good start if you don't need something special. Rail to Rail in and out, 2,7 to 16V and low input bias. GPB is "only" 800kHz, but this should be enough.
Never "abuse" an OpAmp as comparator. The schematic symbol is the same and maybe the input stage is similar, but the rest is not. Use a comparator to to a comparators job.

[mariush]

I wouldn't rely too much on that 5v as a reference for the Atmega.  Input voltages can change with the load, with how much the regulator heats (if it's linear).  I would either use the internal voltage reference (I think it's 2.56v, though I'm not sure how precise it is) or use an external voltage reference (like tl431 or better).

Or you could just divide the voltage by two (and you'll have 0.75-3.2v) and use a better adc, for example the 16bit MAX11205 is only 2 dollars : http://www.digikey.com/product-detail/en/MAX11205BEUB%2B/MAX11205BEUB%2B-ND/3479539  (you may have to add a 3.3-3.6v ldo as well if you go with this one)

 

[T3sl4co1l]

I would either use the internal voltage reference (I think it's 2.56v, though I'm not sure how precise it is)

Eww..

First rule about internal references is, we do not use them on ATMega.
Second rule about internal references is, we do not use them on ATMega.
 

Even under load variation, a 7805 is literally several times better spec.  Their analog bits are embarassing.

Tim

[Rory]

OP:

We are using LM6132 and LM6134 and have used LM6211 in previous projects.  I will leave it up to you to determine if they are suitable for you.
 

[Tac Eht Xilef]

Thanks everyone for the suggestions (& sorry I couldn't reply earlier - internet died yesterday arvo ). I'll check them all out to see what specs / packages / availability suit best. For some reason I forgot the LM358 went down that far - I should've checked the datasheet before posting...

As for your particular case, since you already have so much voltage swing, why not just divide it against 0V and live with the offset?  It's not like you're losing many bits -- or that you (evidently) care about many bits to begin with (the internal ADC is 10 bit with some 4 LSBs of INL/DNL error -- if you require absolute, uncalibrated accuracy, it's more like 8 bits anyway).

Or you could just divide the voltage by two (and you'll have 0.75-3.2v) and use a better adc ...

True, and I'm using a divider at the moment. In this case I'm less concerned with accuracy/repeatability (beyond knowing where the middle and ends of the pot rotation are) than resolution, and using a divider essentially throws away ~20% of what little useful resolution there is. Between that, dealing with the ADC quality, & including a small dead-band around the pot's centre rotation (to aid tuning), the resulting fine tuning of the digital VFO is just a little too "step-like" and digital.

I've tested it with a 0-5v divider pot and the extra resolution I gain from going all the way down to 0v is enough to make things feel suitably analogue. However, this is designed to hook up to existing expansion connectors in an existing receiver, and I'd rather not mod the receiver to change the fine tune signal from 1.5-6.5v to 0-5v. Hence the need for a suitable op-amp...

Not worth using an op-amp as comparator, certainly not at 5MHz ...  Or you can simply use a couple 2N3904s.

Yeah, you really don't want to see what I'm using in my breadboarded proof-of-concept. Hint: why use "a couple of 2N3904s" when a single BC548 works j-u-s-t well enough? 

Never "abuse" an OpAmp as comparator. The schematic symbol is the same and maybe the input stage is similar, but the rest is not. Use a comparator to to a comparators job.

Yup, well aware that it's not the done thing, and I even know why - but I'd hardly be the first to do it, would I?

Point taken though; I'll use a couple of suitable transistors. I need one anyway for a low-impedance driver on the VFO output.

Again, thanks all!

[fmaimon]

Eww..

First rule about internal references is, we do not use them on ATMega.
Second rule about internal references is, we do not use them on ATMega.
 

Even under load variation, a 7805 is literally several times better spec.  Their analog bits are embarassing.

Are you sure you are not talking about the ATXMega? I've never heard of anyone complaining of the mega series and I've never had any problems myself. The only "problem" is the accuracy of the voltage reference that needs to be calibrated but, once you know it's value, it's very stable.

[Tac Eht Xilef]

The only "problem" is the accuracy of the voltage reference that needs to be calibrated but, once you know it's value, it's very stable.

I think that was T3sl4co1l's point.

• The internal reference in an ATmega328 is IIRC spec'd as 1.01v +-0.1v i.e. ~10% without calibration, and from my (admittedly limited) experience there's a pretty random distribution in that range.
• The ancient LM7805 is spec'd at 3%~4% out of the box, and in my experience is typically ~1.5% or better for branded devices.
• An LM1117-x is spec'd at 1.5% and is typically <1%.

The internal bandgap reference should (and, AFAIK, does) have better stability than a regulator - but then again, so does a proper voltage reference. You can either calibrate to the internal reference on an part-by-part basis, or you can supply an ARef that is accurate and stable to start with .

[T3sl4co1l]

Xmega are a little better, as well as other companies that actually know analog (e.g. MSP430s), but the main point is, if you're actually needing an ADC and not just a gimmick, just put in a MCP3208 or whatever.  Way faster and 1-2 LSBs total error out of 12 bits.  Or even 16 bits if you don't mind lily-gilding.

Speaking of analog, I've heard their BOD/POR circuits suck too (bad stability / doesn't work on slow startup).  So throw a MCP120 in there too..

If you're so hard pressed on space or cost that you can't do this, then you'll have to contrive something with dithering (PWM channel DAC + LFSR?) and a substantial reduction in bandwidth for lots and lots of averaging.

As for refs, TL431B is better still (0.5% is it?), let alone any of the "precision" ones that extend all the way out to real bits of guaranteed untrimmed accuracy for your ADC.  Don't forget that you need precision amps down there too; the ADC might only be $2 but need a $4 buffer amp for your interface!

TL431s are excellent to have on hand anyway; don't think adjustable zener, an obvious lie; think op-amp with Vos = 2.50, +in = GND and open collector output.

Tim

[CyberWalker]

Hi to all!

 I have been a member of this forum for quite some time now but this is the first time that I reply, mostly because I am quite enthusiastic about this topic.

For lab projects, rapid prototyping, scientific research, proof of concepts, instrumentation, demos, etc. (jelly bean honors here) I would highly recommend the use of an MCP6022 Op Amp. Actually, the whole series, as it comes to various packages and formats.

- Single Supply
- Rail to Rail
- Bandwidth 10 MHz
- Supplied from 2.5V to 5.5V
- Quite inexpensive (less than two dollars per piece, and each piece includes two of them)

Regards,
Ioannis Mandourarakis

[SiliconWizard]

Not worth using an op-amp as comparator, certainly not at 5MHz.  To make a passable square wave, you'd need GBW >> 50MHz and slew rate >200 V/us.

Yeah. Signal fundamental frequency is one thing but doesn't mean much for signals other than pure sine waves. So for this particular application, I would rather use a fast comparator than a general-purpose opamp that would have to be real fast (and thus draw a lot more power while being more expensive). Many out there. The LM319, already listed, would fit the bill here and it's pretty cheap too. There's a lot others that are much faster and much more expensive, but that would be overkill here. I could also have listed a number of LT parts, but those are in general significantly more expensive too.

Lesson is, do not bother trying to always get one-size-fits-all parts, because that's not necessarily going to work.

[EPAIII]

This thread is from 2014. I suspect the OP has long since found a solution.

I do wonder if all that he needed to do was drop 1.5V, why a couple of series diodes would not have worked.