Millivolts to Microvolts

Reading a Millivolt Span in Microvolts per ADC Count

A signal stops being a millivolt problem the moment you compare it with one ADC code. Divide a reference by 2ⁿ and the answer lands in microvolts, so that is the unit your sensor span has to be in before you can say how many counts it will move.

Conversion factor: 1 mV = 1,000 µV — multiply millivolts by 1,000. Worked example: a 4.7 mV bridge output is 4,700 µV. On a 12-bit converter with a 3.3 V reference one LSB is 3.3 V ÷ 4,096 = 805.7 µV, so that whole span moves the code by about 5.8 counts.

What the Two Units Are Doing Here

Spans arrive in millivolts

A 2 mV/V load cell on 5 V excitation gives 10 mV full scale, a current shunt 75 mV, a type-K thermocouple about 41 µV per °C.

One code lands in microvolts

LSB = V_ref ÷ 2ⁿ: at 3.3 V that is 805.7 µV on a 12-bit part and 50.4 µV on a 16-bit one — too small for millivolts to stay convenient.

Gain bridges the two

Gain is a ratio of spans: input range ÷ sensor span. Putting both sides in one unit first is what keeps a factor-of-1,000 slip out of the amplifier design.

From Datasheet Span to Microvolts in Four Steps

Type the span off the datasheet

Put the sensor's full-scale output in the millivolt field — 10, 4.7, 0,082. Comma or dot both work as the decimal mark and spaces are ignored, so a pasted table value converts as typed.

Divide the microvolt figure by your LSB

The µV field updates as you type. Divide it by the LSB in the table below for the count span: 10 mV = 10,000 µV is 12.4 counts at 12-bit/3.3 V, but 198.6 counts at 16-bit.

Work back from one code

Both fields are editable, so you can type an LSB such as 50.4 into the µV side and read the millivolt equivalent instead. The swap button (↔) flips the labelled direction.

Copy the bare number into your gain math

Each field has its own copy button and it copies digits only — no unit, no thousands spaces — so the value drops cleanly into a spreadsheet. Ctrl + C in a field does the same.

Going below a microvolt: the searchable unit menu on either side covers all 12 voltage units, so you can drop to nV or pV for a noise figure. Output carries 8 decimals and switches to scientific notation below 1e-6.

LSB Size in Microvolts for Common ADC Resolutions

Every figure below is V_ref ÷ 2ⁿ in microvolts, for the three references you meet most on a mixed-signal board.

Resolution	Levels (2ⁿ)	1 LSB @ 2.5 V	1 LSB @ 3.3 V	1 LSB @ 5.0 V
10-bit	1,024	2,441.4 µV	3,222.7 µV	4,882.8 µV
12-bit	4,096	610.4 µV	805.7 µV	1,220.7 µV
14-bit	16,384	152.6 µV	201.4 µV	305.2 µV
16-bit	65,536	38.1 µV	50.4 µV	76.3 µV
18-bit	262,144	9.54 µV	12.59 µV	19.07 µV
20-bit	1,048,576	2.384 µV	3.147 µV	4.768 µV
24-bit	16,777,216	0.149 µV	0.197 µV	0.298 µV

Two definitions in circulation: some datasheets quote LSB as FSR ÷ (2ⁿ − 1) — 805.9 µV instead of 805.7 µV at 12-bit/3.3 V, a 0.02% difference. What does matter: a bipolar or differential input spanning ±V_ref doubles the full-scale range, so one code is worth twice the microvolts shown.

What This Converter Does for That Workflow

Span in, microvolts out as you type

The microvolt figure follows keystroke by keystroke, so you can sweep candidate spans against one LSB without re-running anything.

Type an LSB, get the span

Both fields accept input and the swap button reverses the direction — what you want when the code size is known and the millivolt equivalent isn't.

Numbers clean enough to paste

The per-field copy button hands over digits alone, without unit or display grouping, so a gain spreadsheet gets a value it can compute with.

Readable down to the noise floor

All 12 voltage units on both sides, 8-decimal output and automatic scientific notation keep nanovolt-scale figures legible. Everything runs in your browser.

Resolution and Gain Questions

How many microvolts is one LSB on a 12-bit vs a 16-bit ADC?

On a 3.3 V reference, one code is 805.7 µV at 12-bit (3.3 ÷ 4,096) and 50.4 µV at 16-bit (3.3 ÷ 65,536) — a factor of 16 for four extra bits. On 5 V the pair becomes 1,220.7 µV and 76.3 µV, so the reference matters as much as the bit count.

What gain do I need to make a 10 mV sensor span fill a 3.3 V ADC input?

Divide the spans in one unit: 3.3 V is 3,300 mV, so 3,300 ÷ 10 = 330 V/V for an exact fit. Leave headroom for amplifier offset, reference tolerance and sensor over-range and a standard 200 or 250 is safer. At 330 all 4,096 counts are used and one count is worth 10 mV ÷ 4,096 = 2.44 µV at the input; at 200 the signal covers 2.0 V, about 2,482 counts, or 4.03 µV per count.

Why is my effective resolution (ENOB) worse than the bit count on the datasheet?

The bit count describes the code map; ENOB describes what survives noise and distortion, and follows from the measured signal-to-noise-and-distortion ratio: ENOB = (SINAD − 1.76) ÷ 6.02. A 16-bit converter specified at 78 dB SINAD gives about 12.7 effective bits, so its trustworthy step at 3.3 V is roughly 510 µV, not the ideal 50.4 µV.

Do more ADC bits help if my noise floor is already bigger than one LSB?

Only until the LSB drops under the noise. With 30 µV rms of input-referred noise, peak-to-peak is roughly 6.6 × rms, about 198 µV — a noise-free resolution of 3.3 V ÷ 198 µV ≈ 16,700 steps, near 14 bits. Going 12-bit → 16-bit still helps there; 16-bit → 24-bit does not, since the extra codes only dither the noise. After that the wins come from quieter gain, tighter bandwidth and a better reference.

How much does oversampling and averaging add to my resolution?

Averaging N uncorrelated samples cuts rms noise by √N — one extra bit per 4× oversampling. A 12-bit channel at 805.7 µV averaged 16× behaves like 14 bits (about 201 µV); 256× reaches 16-bit behaviour near 50.4 µV at one 256th of the throughput. It needs roughly 1 LSB peak-to-peak of noise or dither to work on, and never removes gain error, offset or non-linearity.