# Max value of n-bits ADC

I'm still not sure about info I read on the internet. It said that to get the analog value also known as output bits of ADC use this formula:

Do you think the analog reference is the same as the voltage reference?

My real question is, do you think "Max Value of ADC bits" is same as 2^b where b is the number of bits?

I'm sure its maximum value of ADC bit must be (2^b)-1

I cropped that picture from here, that article assumes a 10-bit ADC so it uses a maximum value of 1024, while I think it must be 1023. I think what they mean with 1024 is the possibilites not maximum value of an n-bit ADC.

• $\left(2^{b}\right)\text{-1}$ – rdtsc May 14 at 11:57
• If it's 1023 at the denominator, you expect 1023 to be generated at EXACTLY Vref at input. But it is not the case in ADCs. – Mitu Raj May 14 at 12:34

There is a simple relationship between the voltages and digital values involved in a ADC: $$\frac{V_{IN}}{V_{REF}} = \frac{DigitalOutput}{2^N}$$ where $$\N\$$ is the number of bits.

If $$\V_{IN} = 0\$$ you would certainly expect the digital output to be integer 0.

The step size (voltage equivalent of a change in LSB) is $$V_{LSB} = \frac{1}{2^N} \times V_{REF}$$

So the voltage that will produce the largest digital output value is $$V_{MAX} = \frac{2^N-1}{2^N} \times V_{REF}$$

So, for a 10-bit ADC with a reference voltage of $$\3.3\,\text{V}\$$, the integer output value of 1023 (0x3FF) corresponds to a voltage of approximately $$\3.297\,\text{V}\$$. Any voltage above this value will also give a digital output of 1023.

There is one additional wrinkle. Some ADCs will offset the switching thresholds by $$\\frac{1}{2}V_{LSB}\$$ to reduce the quantization error. These ADCs have slightly different behavior; check the datasheet for details.

• As an extension to that wrinkle, some ADCs are designed so that if they repeatedly read a signal that would yield a value of e.g. 123.4, they'll consistently read 123 (which would be off by 0.4 LSB), while others are designed so that they'll read 123 about 60% of the time and 124 about 40% of the time. Taking ten readings and averaging them together might yield a value less than 1233 or greater than 1235, but it would usually yield a value that was closer to the ideal one than an individual reading would be. – supercat May 15 at 3:59

10 bit can represent the numbers from 0 to 1023 OR from -512 to +511. That really depends whether the signal you are sampling is bi-polar (like audio) or just positive (like video).

The quantization step is the same in both cases, it's $$\V_{ref}/2^b\$$ where $$\V_{ref}\$$ is the reference voltage of the ADC and $$\b\$$ the number of bits.

To make things easy, let's assume we have 3 bit ADC and a reference Voltage of 8V. Then the output will increment in 1V steps representing either 0V to +7V OR -4V to +3V.

• ee.se uses \\$. I vaguely remember it's because of the dollar sign, – a concerned citizen May 14 at 15:02
• Thanks. It seems to be different for ever stack exchange site which can be confusing – Hilmar May 14 at 22:58

10 bits leads to $$2^{10}$$ values, i.e. 1024. But one of them has to be zero. Thus the maximum value equating to 3.30 V must be 1023. If you exclude zero, you'd need a 10.0014 bit ADC.