Being negative can still be positive

Other Parts Discussed in Post: LMH6554, LMR70503

The first operational amplifiers (op amps) used what is commonly called split power supplies, meaning the power for the amplifier was symmetrical around ground with both positive and negative polarities.  Since most power supplies used transformers to convert 120 V mains electricity, a simple center tapped secondary winding gave easy access to a negative power supply. 

Today, many devices now run on battery power, and even mains powered equipment is trending towards having only positive power supplies. But, I don’t recommend ruling out having a negative power supply. 

Many signals are still ground referenced and carried on coaxial cables.  High performance components, like analog to digital converters, may run on 1.8 V power supplies and require an input signal centered at 0.9 V.

Even the lowest threshold MOS technology cannot generate signals that are truly rail-to-rail and the base-to-emitter voltage (Vbe) on bipolar technologies has not changed significantly from the 0.6 V commonly associated with silicon PN junctions.  This means that if an amplifier is going to recreate a signal that swings close to ground, or even more difficult, is ground referenced, a negative power supply is necessary.

Fortunately, generating a negative power supply is very inexpensive, especially if the current requirements are low.  A device like the LMR70503 SIMPLE SWITCHER can generate a negative power supply with only 7 external passive components. 

The LMH6554 is a differential amplifier designed for both single supply and split supply configurations.   Using the LMR70503, an existing  3.3V supply, can be converted to a +3.3 V and   -1.7V split supply.   With this supply configuration, the LMH6554 output voltage can now swing from  -0.6 V to +2.2 V.  This range is adequate for driving most analog-to-digital converters and many ground referenced signals such as composite video.  Other combinations of supply voltages, like +4 V and -1 V would be ideal for an ADC such as the ADC081500 shown below. 

With the advent of switching power supplies there is no reason to limit yourself to only positive power supplies, even when designing battery powered equipment.  Don’t be afraid to specify a negative power supply in your next high system design. 

  • I had in mind applications where high frequency linearity is a key specification.  The LMH6554 has an OIP3 over 40dBm at 250MHz, but to get the rated linearity it is necessary to set the output common mode very close to mid supply.  

    Sometimes, an inexpensive switching regulator can give that extra flexibility that makes the design come together.  

  • Bonnie, I agree that the input rail to rail is pretty well covered in single supply op amps. Good point on the output but your analysis is quite generous with regards to the dropout, depending on what kind of load we're talking about (and class) of course and assumes a single stage, now imagine taking that 5mV into a second stage to boost up the gain (isn't this a common practice in medical insrumentation). It's no longer 4 codes you're losing is it? The good news, as Loren says is that there's a solution, a monolothic one too called the LM7705 which prevents the loss of the codes. On the other hand, the counter argument would be "well charge pumps are noisy",  sure they can be but how noisy is noisy (for the overall system) and if you're doing a single supply design, it's probably to reduce power no? Well doesn't power reduction mean higher tolerance for noise?

  • You have written about a topic that has been on my mind for years. This is an extremely good point and well written. I am kind of tired of catering to single supply crowd for almost everything however, the single-supply chip designers have done a superior job in trying to tackle the obvious problems with this technology. These guys have effectively got the input rail-to-rail issues under control. The low power change pump on the input stage tackles this issue. The output is still a challenge, but the rail-to-rail output behavior is under control by keeping the swing in the millivolt region as opposed to the 10s or 100s of millivolts. With that said it is important to come to terms with the concept of how important is this rail-to-rail behavior. As you travel to the output rail of amplifiers, the limitation is towards the rail, not across the entire output range. A loss of 5 mV at the rail with a following 12 bit converter is the loss of 4 codes (assumes 5V supplies with 5V {or 2.5V} reference to ADC). So instead of having 4096 usable bits with the converter you only have 4092 usable bits. The impact is 0.00977% on your dynamic range. I guess that is a livable error.