Another bipolar power supply for low power opamp.

Hi.

Working on yet another PIC development board (this time for analogue) I've ran into the age old problem of getting a full range out of an opamp. I have a 10-bit DAC on board, MCP4812, and altohugh its output is buffered, I would like to add a voltage follower on the output, so I can draw just a little bit more current from it before it's error becomes too much.

For ease of use, an LM358 springs to mind. My board will be powered from 3.3/5V (selectable by user) so, in order to increase the rail output, I'm happy to use a charge pump voltage doubler, to provide the opamp with ~6/10V. But, given the resolution of the DAC, at the lower end I will never get true 0V out, even the LM358 bottoms out at 40mV... which is a few LSB's! The top end sohuldn't be too much of a problem, as the DAC has an onboard 4.096/2.048V reference, which means that is the DACs full scale output. I could of course regulate the doubled voltage to 5V for the DAC, so it can output 4.096V when running at 3.3V. Or just switch the voltage down to 2.048V.

So, a negative supply is required, albeit a very small one ~ -1V.

As you can see, its starting to all add up. Normally I would just use the +V and -V outputs from a MAX232, but I do not have one of these in my design. Not only is its quiescent current quite high for a 'low power' system, but also, many only run at 5v, or 3.3V, not both. I'm trying to avoid having two seperate voltage converters, one for double Vcc, and one for -Vcc. Only solution I have so far is the MAX680, but that runs at 8kHz (bit too low, bit noisey eh?) and its current output is dreadful. Another option is a dual charge pump with a 555 timer. However, the NE555 doesn't run <4V, and its CMOS equivilent has a very limited current output (crap for charge pumps).

So, any idea's for simplicity? I realise rail-to-rail opamps are available, but these are never true rail to rail. Also a half decent current output is needed, we'll say 10mA+ for the positive supply, so I can draw *some* current from the opamp buffer, without it sagging.

Once again, I'm over complicating things, since the problems with the negative rail only occur for <100mV for the LM358 (with a current load). But at 10-bit, with a Vref of 2.048, thats 50 LSB's out.

Blueteeth

hi Bt,
Happy 2011

Look at the MCP6002 from Farnell, only about 50p and its rail2rail.

https://uk.farnell.com/microchip/mcp6002-e-p/op-amp-dual-1mhz-dip8-6002/dp/1332117

EDIT:
Why would you want to draw 10mA from DAC.??

You might have a look at my PIC tutorials, where I had to overcome the same problem.

Even rail-to-rail opamps don't go 100% to the rails, so it's easier to provide a negative rail and a precision reference voltage to avoid all problems.

Hey guys, thanks for the replies!

Eric: Thats actually a very nice 1MHz opamp! Within 20mV of GND for 5mA output. Given the price I've added some to my farnell basket for next order But, as I read over my question, as well as reading nigels post (and also refamiliarising myself with his 'analogue board on his tutorials) I guess for 'true' 0V output with accuracy for an ADC/DAC, a negative rail will be needed regardless of rail capabilities.

Nigel. Yep, as I said above, alas its true. A bipolar supply generally makes life a LOT easier for DC apps. Although for many apps a single supply is convenient and does the job, in my exerience, one must sacrifice certain parameters. As the ground reference is true ground, the power supply rejection means the power supply can be a little noisier.

Once again, I guess you've noticed I take things way too far, perfectionism is counter productive, and the '10mA draw' from the DAC was probably an extreme estimate - 5mA would be more than enough, which the DAC's themselves can cope with. However, I just realised that the opamps on my board used for ADC buffering would also benefit from a bipolar power supply, so even without a DAC buffer, looks like I'm going to need this anyway.

Given the stupidly tight specs, Vin: 3.3/5V, output: +6.0V min and -1V min @ 20mA max. Along with relatively high efficiency (for low power battery powered apps) its a bit of a bugger. But going through my bits box, I believe I've found a half decent solution, I've used them before for extra headroom for analogue circuits.

**broken link removed**

Using a couple of diodes, and a couple of caps, I can 'tap' the negative side of the pump cap (which is roughly a square wave, 0-VCC @ 11kHz) and make a crude charge pump for a negative voltage. Efficiency is still quite high, and it'll easily provide -(VCC-1) more than enough for my -1V target. The current output for the above chip is actually quite high, and should satisfy my requirements.

I know theres a sticky on dual opamp power supplies, thanks to Nigel, but I may post some schematics of 'cheap and simple' low power dual supplies for those who run things of batteries, or USB. Adding charge pumps is cheap and cheerful If anyone is interested, I'll compile a list. Since downloading LTspice, I've wasted far too much time playing about with it, and making cheapo little solutions for power supplies.

Sorry if I've used this thread to 'think outloud', your advice is always welcomed and re-assuring, I hope I can provide the same for others.

Blueteeth

Blueteeth said:

Eric: Thats actually a very nice 1MHz opamp! Within 20mV of GND for 5mA output. Given the price I've added some to my farnell basket for next order But, as I read over my question, as well as reading nigels post (and also refamiliarising myself with his 'analogue board on his tutorials) I guess for 'true' 0V output with accuracy for an ADC/DAC, a negative rail will be needed regardless of rail capabilities.

Click to expand...

True, the amps don't go completely to ground if they have to deliver current. But if the load you are driving is also connected to ground and doesn't draw any current at 0V then the op amp likely will go very near ground with a 0V input.

Another bit of the puzzle

Hi,

Well, I've sorted out a switchable bipolar opamp power supply using an LM2766. Using a single two way jumper/switch, one can either have an output of V- = GND, and V+ = VCC - 0.3... OR..... V- = -VCC(-1), V+ = 2*VCC(-0.1). As a surface mount 'module', its pretty handy to add one some smaller test boards - namely my three dev boards I'm designing. When in the first position, it consumes around 420uA (measured), so it just adds some versatility when one requires biploar voltage measurement.

I'm not trying to hijack my own thread, because this is, again, another analogue question, one that highlights my ignorance.

I have knocked up an 'analogue front end', similar to the ones used in Nigels PIC tutorials (not copying you Nigel!) to work on my experiment/dev board for PICs. Along with buffered inputs and available Voltage references, I added some resistors which can be switched to ground to provide selectable attenuation, and/or gain. Since each switch is single toggle, one can use DIP switches, which saves space, and makes for easy configuration. Four switches can provide 3 levels of attenuation, and 3 levels of gain. It's all very basic, but is there anything wrong with adding a resistor in the feedback path of a voltage follower? for when all switches are off? (basically a buffer) I realise this may add 'noise', but for 10-bit measurement, this shouldn't be much of an issue.

Criticisms welcome. Note: I haven't worked out resistor values for specific attenuation/gain, but I'm planning on basic gains of 1,2,2.5 (for a Vref of 2.5) and attentuation of 0.5 and 0.1. It all depends on the Vref used, to make ADC measurement calculations easier.