L
Admin
Two-tone third-order intercept is one of the most quoted amplifier and mixer numbers and one of the easiest to measure wrongly. The usual mistake is to measure the intermodulation of the test setup instead of the device.
1. The basic setup
Two signal generators, a power combiner, the DUT and a spectrum analyser. Tones at f1 and f2 with spacing small compared with the DUT bandwidth (for example 1 MHz). The third-order products appear at 2f1 - f2 and 2f2 - f1. OIP3 = Pout(tone) + delta/2, where delta is the difference in dB between the tone and the IM3 product at the output.
2. Keep the sources from mixing each other
Each generator sees the other through the combiner and generates its own IM products in its output stage.
- Use a combiner with good isolation (a hybrid or Wilkinson with 20 dB or more), and add isolators or attenuators (6 to 10 dB) on each generator output.
- Running each generator at higher level into a pad is better than running them hot into the combiner.
3. Keep the analyser out of compression
The analyser mixer has its own TOI. If the IM3 you see moves with the input attenuator setting, it is the analyser.
- Increase the input attenuation by 10 dB. If the IM3 products drop by 10 dB along with the tones, they come from the DUT. If they drop by about 30 dB (three times the attenuation change), the analyser was generating them.
- Use the analyser's IP3 or mixer-level optimisation if it has one, and keep the mixer level well below its TOI.
4. Verify the 3:1 slope
Sweep tone power in 1 dB steps. Genuine IM3 from a weakly non-linear device rises 3 dB per 1 dB of input. If the slope is not near 3:1, you are near compression, at the noise floor or measuring something else, and the intercept you compute is meaningless.
5. Other traps
- Unequal tones: the calculation assumes equal tone power at the DUT output; level them there, not at the generators.
- Memory effects in PAs make IM3 depend on tone spacing; sweep the spacing if the device will see wideband signals.
- Noise floor: narrow the RBW, but check the generator phase noise is not filling the IM3 bins.
How do you document IP3 in your test reports: the tone spacing and power, the slope check, or just the number?
1. The basic setup
Two signal generators, a power combiner, the DUT and a spectrum analyser. Tones at f1 and f2 with spacing small compared with the DUT bandwidth (for example 1 MHz). The third-order products appear at 2f1 - f2 and 2f2 - f1. OIP3 = Pout(tone) + delta/2, where delta is the difference in dB between the tone and the IM3 product at the output.
2. Keep the sources from mixing each other
Each generator sees the other through the combiner and generates its own IM products in its output stage.
- Use a combiner with good isolation (a hybrid or Wilkinson with 20 dB or more), and add isolators or attenuators (6 to 10 dB) on each generator output.
- Running each generator at higher level into a pad is better than running them hot into the combiner.
3. Keep the analyser out of compression
The analyser mixer has its own TOI. If the IM3 you see moves with the input attenuator setting, it is the analyser.
- Increase the input attenuation by 10 dB. If the IM3 products drop by 10 dB along with the tones, they come from the DUT. If they drop by about 30 dB (three times the attenuation change), the analyser was generating them.
- Use the analyser's IP3 or mixer-level optimisation if it has one, and keep the mixer level well below its TOI.
4. Verify the 3:1 slope
Sweep tone power in 1 dB steps. Genuine IM3 from a weakly non-linear device rises 3 dB per 1 dB of input. If the slope is not near 3:1, you are near compression, at the noise floor or measuring something else, and the intercept you compute is meaningless.
5. Other traps
- Unequal tones: the calculation assumes equal tone power at the DUT output; level them there, not at the generators.
- Memory effects in PAs make IM3 depend on tone spacing; sweep the spacing if the device will see wideband signals.
- Noise floor: narrow the RBW, but check the generator phase noise is not filling the IM3 bins.
How do you document IP3 in your test reports: the tone spacing and power, the slope check, or just the number?