I use an IQ-0307LXP mixer to create up-converted microwave pulses with single side band(SSB). I apply a few GHz CW(continuous wave) microwave to LO port and 10-150MHz CW or pulse to I and Q port. I and Q has the phase shift of 90 degree. I normally calibrate the IQ mixer in such a way to get minimum LO feedthrough and minimum suppressed sideband. To do so, I sweep DC voltage to I and Q port alternatively to find the DC offsets that minimize LO feedthrough power. After that, with the optimal DC offsets applied and fixed, I sweep the amplitude imbalance(ratio of I to Q amplitude) of I and Q and also the phase skew(deviation from 90 phase shift) of I and Q to find the optimal values that minimize the suppressed sideband power.

I found that the higher the IF frequency, the larger the phase skew. However, the phase skew goes beyond the typical value specified in the datesheet (~3 degree), even if IF is within the datasheet value, DC-500MHz. For example, the following shows the parameters used and results of my mixer optimization.

LO : 5.1 GHz, 10-13 dBm

IF: 10 MHz, Vpp=0.5V –> I offset=-0.008 V, Q offset =-0.023 V,  phase skew=3.9 degree
IF: 100 MHz, Vpp=0.5V –> I offset=-0.017 V, Q offset =-0.023 V, phase skew=-17.9 degree
IF: 141 MHz, Vpp=0.5V –> I offset=0.004 V , Q offset =-0.012 V, phase skew=-44.1 degree

It seems that the the phase skews go out of specification when IF is 100 or 141 MHz. So I wonder why they deviate so much and how the phase skew(or I/Q quadrature phase deviation) in the datasheet is measured at Marki. Perhaps, it’s normal due to the way I do mixer calibration?


  • You must to post comments


This is a very good question. You have a sophisticated setup for calibrating out the LO feedthrough and sideband suppression.

First the phase deviation with IF frequency. The phase will always vary more with higher IFs than with lower IFs, simply because the same time delay becomes a greater phase delay at high frequencies. For example, a 5 degree phase shift at 100 MHz becomes a 50 degree phase shift at 1000 MHz since the period is 10 times shorter. We always test with a hybrid around 60 MHz.

Second, the unit variation. The IQ-0307LXP, as with any part number that begins with ‘IQ’ (as opposed to ‘MLIQ’ or ‘MMIQ’), is a legacy, hybrid, handbuilt mixer. As such it has inherent unit to unit and lot to lot variation. The performance can and will change depending on the circuit lot, diode lot, assembler building the part, and whether that assembler is having a great day or only an okay day. These are low volume products as well, so it is difficult to specify them precisely. Additionally phase balance is a difficult thing measure, so we typically do not measure it in production. This is why the only specs that we guarantee are uncompensated image rejection and LO-RF isolation. Since the image rejection is a combination of phase and amplitude balance, it is possible that the phase could be significantly different from the typical phase, and we might not notice in production. This is the second time recently someone has mentioned the phase balance of the legacy IQ mixers, so we plan to investigate it more carefully. However, if we cannot quickly find a root cause for the problem then we will only be able to change the typical number on the datasheet.

The long term solution for this problem is to use a much more repeatable MMIQ mixer. For your frequency plan I recommend the MMIQ-0520LS.


  • You must to post comments
Showing 1 result
Your Answer

Please first to submit.