Hi All!

For the Analog devices OP amps the IMD is measured by ITU-R method, according MT-053 tutorial (http://www.analog.com/static/imported-files/tutorials/MT-053.pdf), but often in the other OP amps IMD is measured according SMPTE method, when use 60Hz and 7kHz signals combined in a 12 dB ratio. How can I recalculate IMD SMPTE to IMD ITU-R?

And for the AD8007/8008 according the datasheet IMD is 77dBc

( fC = 19.5 MHz to 20.5 MHz, RL = 1 kΩ, VO = 2 V p-p

Third-Order Intercept is 42.5),

but if I calculate in compliance with the MT-053 tutorial:

Why is the difference so big?

Thank You.

Hello serhiy,

I didn't write the AD8007/8 data sheet, but I believe I can shed some light on what's going on here. Problems are often encountered when applying RF specifications to op-amps and differential amps since RF systems typically operate in fixed-impedance environments (typically 50 and 75 ohms) with specifications that are based on power, while op-amps and differential amps operate in variable-impedance environments with specifications that are mostly based on voltage. A frequently encountered example of this occurs when an op-amp gain is calculated in "dB" as 20*Log(Vo/Vi) without regard to input and output impedances. This expression is not correct in the strict sense (hence the quotes) when the power gain of the amplifier is not equal to [(Vo)^2/RL]/[(Vi)^2/Ri], but it is used nonetheless since it's generally understood that only voltages are involved. The same type of approach can be used with IMD and IP3, which is useful when the op-amp is driving a high resistance load. If, for instance, the op-amp were at the end of a RF signal chain, driving an ADC, the voltage into the ADC is what's important, not the actual power. From this vantage point, it is useful to define IP3 "as if the amplifier were driving 50 ohms."

Looking at the +/-5V section of the AD8007/8 data sheet Specification Table we see that for for 19.5 MHz and 20.5 MHz input tones, IMD3 = -77 dBc for Vo = 2 Vp-p and RL = 1K. This is actual data measured on a spectrum analyzer with a 1K load. The problem occurs when we try to calculate IP3, since IP3 is based on power. (IP3 can also be measured directly by plotting the fundamental output power and third-order output power lines and extrapolating to the intercept point, but it is calculated from IMD3 for op-amps). To do this, we insert a "fictitious" 50 ohm load on the output and calculate the power delivered to this load. From the numbers in the AD8007 data sheet, it appears that a source resistance of 50 ohms was also included on the output as well, cutting the voltage into the load by a factor of two. In this case, the voltage out of the AD8007 was 2 Vp-p and the voltage into the 50 ohm load was 1 Vp-p. A 1 Vp-p sinewave into 50 ohms delivers +4 dBm (Po) into the load. We can now calculate IP3 from the following expression:

dBc = 2*( Po - IP3)

0.5*dBc = Po - IP3

IP3 = Po - 0.5*dBc

Substituting numbers:

IP3 = +4 dBm - 0.5*(-77 dBc) = +42.5 dBm

This agrees with the data sheet figure, and the method described here is most likely what was used in the calculation. I use a similar approach with the differential ADC drivers except that I do not include the source resistance. (The method is indicated in the data sheets.) If you're interested in more detail, I cover this and many other topics in the following ADC-Driving webinars:

http://www.analog.com/en/content/WC_DRIVING_ADCS1/webcast.html

http://www.analog.com/en/content/WC_DRIVEADCS_2/webcast.html

Please contact me if you have further questions.

Best regards

--Jonathan Pearson