Published Nov 14, 2026 · Last updated Nov 14, 2026 · 4 min · IndieRF

Op Amp Level Shifter Design: 0.6 V/V, −3 V, Four Resistors

Two points give the line, RF/RG sets the offset and a divider sets the slope. Worked for a DAC into a GaN gate bias, with five neighbouring specs on E24 and E96.

A 0 to 2.5 V DAC into a −3.0 to −1.5 V GaN gate bias is the line Vout = 0.6·Vin − 3 V: RF = 12 k over RG = 10 k on a 2.5 V reference makes the −3 V, and a 7.5 k over 20 k divider makes the 0.6, all on E24. With 1% resistors and a 0.1% reference the gate moves ±63 mV at code 0 and ±68.5 mV at full scale.

m = 0.6 V/V, b = −3 V. Non-inverting with offset (SLOA097, m > 0, b < 0).
  1. RG 10k, R1 20k, R2 7.5k, RF 12kBest E24
    0 mV at 0 V, 0 mV at 2.5 V
  2. RG 10k, R1 22k, R2 8.2k, RF 12kOption 2
    0 mV at 0 V, −6.6 mV at 2.5 V
  3. RG 10k, R1 20k, R2 8.2k, RF 12kOption 3
    0 mV at 0 V, +99.3 mV at 2.5 V
Error is the output minus the target with an ideal op-amp, ±15 V rails, RG fixed at 10 kΩ.Open in Op-Amp Lab

Type 3.3 into Vin, full scale for a 3.3 V DAC: the best E24 set becomes 27 k, 6.8 k and 12 k, and it lands 39.4 mV low at full scale.

Write the line, then let its signs pick the circuit

Two points give the slope and the offset. From (0 V, −3.0 V) and (2.5 V, −1.5 V), m = 1.5 V / 2.5 V = 0.6 and b = −3 V. Every level shifter is this straight line, and the signs of m and b decide which one-op-amp circuit builds it: TI's SLOA097 gives one circuit for each of the four sign cases.

A positive slope with a negative offset is the circuit here. The DAC goes through R1 into the + input with R2 to ground, and the reference goes through RG into the − input with RF back from the output:

Vout=Vin R2R1+R2(1+RFRG)−Vref RFRGV_\mathrm{out} = V_\mathrm{in}\,\frac{R_2}{R_1 + R_2}\left(1 + \frac{R_F}{R_G}\right) - V_\mathrm{ref}\,\frac{R_F}{R_G}

The reference arrives inverted, so a positive 2.5 V reference makes the negative offset and the board needs no negative reference. The op-amp does need a negative rail: its output sits at −3 V. On +5 V and −2.5 V rails the wizard refuses this spec and asks for 0.5 V more on the negative rail.

Fix RG, solve RF, then the divider

At Vin = 0 the divider term drops out, so the offset belongs to RF/RG alone: RF/RG = |b| / Vref = 3 / 2.5 = 1.2. With RG fixed at 10 k, RF is 12 k.

The slope then sets the divider. The stage's noise gain is 1 + RF/RG = 2.2, so the divider has to be 0.6 / 2.2 = 3/11. Its scale is free, so spend it on bias current: make the resistance at the + input equal the resistance at the − input, so the op-amp's two bias currents drop the same voltage and cancel. RF ∥ RG is 12 k ∥ 10 k = 5.45 k, and a 3/11 divider with 5.45 k of source resistance is R1 = 20 k over R2 = 7.5 k. That's four E24 values with no rounding, which is luck, and the table below shows what happens when the luck runs out.

The Op-Amp Lab's Rescaler wizard does those steps from the two points, then tries each resistor's two nearest series values and ranks every combination by its worse endpoint error.

The Rescaler wizard on 0 to 2.5 V in, −3.0 to −1.5 V out: three E24 sets and their error, then the best one placed and solved. Open this spec in the Op-Amp Lab

Code 0 belongs to RF/RG

Change the DAC or the window and the four values stop landing on the series. Each row below is the wizard's best set, and its error budget is the circuit solved in the Op-Amp Lab.

Scroll sideways for more columns

The same circuit on other DAC ranges and gate windows: 2.5 V reference, RG 10 kΩ, ideal op-amp. Worst case is the linear sum at code 0 and full scale with 1% resistors and a 0.1% reference.
DAC → gateSeriesR1, R2, RFOff at code 0, full scaleWorst case at 1%
0 to 2.5 V → −3 to −1.5 VE2420k, 7.5k, 12k0 mV, 0 mV±63.0, ±68.5 mV
0 to 3.3 V → −3 to −1.5 VE2427k, 6.8k, 12k0 mV, −39.4 mV±63.0, ±70.4 mV
0 to 3.3 V → −3 to −1.5 VE9626.1k, 6.98k, 12.1k−25.0 mV, +13.8 mV±63.5, ±71.0 mV
0 to 2.5 V → −5 to −2 VE2418k, 12k, 20k0 mV, 0 mV±105.0, ±101.0 mV
0 to 2.5 V → −2.8 to −2.2 VE2447k, 5.6k, 11k+50.0 mV, +8.9 mV±57.8, ±61.9 mV
0 to 2.5 V → −2.8 to −2.2 VE9646.4k, 6.04k, 11.3k−25.0 mV, −11.7 mV±59.3, ±63.7 mV

Read code 0 first. It moves only when RF moves, because at Vin = 0 the output is −Vref·RF/RG whatever the divider does. The 3.3 V DAC keeps the −3 V offset, so RF stays 12 k and code 0 stays exact; the divider takes the rounding, 39.4 mV at full scale. E96 trades that the other way. It has no 12 k (12.1 k is its nearest), so code 0 moves 25 mV and full scale comes in to 13.8 mV.

The narrow −2.8 to −2.2 V window is where this bites. RF/RG = 1.12 wants 11.2 k, which neither series has: 11 k puts code 0 50 mV high, 8% of a 600 mV window, and E96's 11.3 k puts it 25 mV low. When code 0 matters most, shop for RF: 10 k plus 1.2 k in series is 11.2 k exactly.

What the 1% spread costs at the gate

With 1% resistors and a 0.1% reference the worst corner at code 0 is −3.064 to −2.938 V, a 126 mV spread, and at full scale it is ±68.5 mV, 4.6% of the 1.5 V window each way. Tighten to 0.1% resistors and the code 0 spread drops to ±9.0 mV.

These numbers assume an ideal op-amp. Its offset voltage and bias currents add on top: set Vos and Ib on U1 in the Op-Amp Lab and the error budget prints their share next to the resistors'.

Next step

The solved case, with its parts table, worst corners and the 0.1% run, is the op amp level shifter example page. To change one number on the real circuit, open the gate-bias spec in the Op-Amp Lab, press Place, and the sheet solves with its DC sweep, Bode plot and error budget. The same method with a positive offset, ±10 V into a 0 to 3.3 V ADC, is the non-inverting amplifier with offset page.

References

  • R. Mancini, "Designing Gain and Offset in Thirty Seconds," Texas Instruments SLOA097, 2002. The four sign cases and their circuits. The table's numbers are the Op-Amp Lab's, checked against exact-fraction hand calculations of the same circuits.

Questions? Contact

Discussion

No comments yet. Start the thread with a measurement, a correction, or a worked example.