R
Member
Dielectric resonator antennas (DRAs) keep showing up in my mmWave work because they avoid conductor loss at frequencies where a patch starts to struggle. But tuning one on the bench is less forgiving than a patch — the resonant frequency depends on the dielectric constant and the geometry together, and there is no copper to trim.
Here is what I actually check, in order.
1. Start from the dielectric constant, not the datasheet headline. The vendor quotes epsilon_r at 1 MHz or at a single spot frequency. At 28 GHz the effective value can be several percent off, and a 3% error in epsilon_r moves the resonance by roughly 1.5%. I measure a sample puck with a split-post resonator before committing to a geometry. It costs an afternoon and saves a board spin.
2. Aspect ratio sets the mode and the bandwidth. For a cylindrical DRA, height-to-radius ratio decides which mode dominates and how wide the resonance is. A taller puck (higher aspect ratio) gives narrower bandwidth but a cleaner radiation pattern; a squat puck broadens the match but excites higher-order modes sooner. In my last design I settled on h/a around 0.8 after sweeping — the return loss dip was 300 MHz wide at 24 GHz, which matched simulation within 5%.
3. The ground plane is part of the antenna. DRAs radiate partly through the ground plane edges. A ground plane smaller than about 1.5 wavelengths across distorts the pattern and pulls the resonance down. I learned this the hard way when a prototype on a small coupon measured 400 MHz low and the "fix" was a bigger ground, not a smaller puck.
4. Feeding: aperture coupling beats a probe for repeatability. A probe feed is easy to prototype but its inductance varies with every solder joint. Aperture coupling through a slot in the ground plane is more repeatable across builds, at the cost of a slightly narrower match. For anything going to production, I take the slot.
5. Measure the pattern, not just S11. A good S11 dip says nothing about efficiency if the energy is going into a surface wave. I do a quick Wheeler-cap style check: compare the input impedance with the antenna in free space versus inside a small metal cavity. If the radiation resistance barely changes, most of the power is not radiating.
What is your go-to DRA feeding approach for mmWave prototypes — probe, slot, or microstrip coupling? And have you found a reliable way to source pucks with tight epsilon_r tolerance?
Here is what I actually check, in order.
1. Start from the dielectric constant, not the datasheet headline. The vendor quotes epsilon_r at 1 MHz or at a single spot frequency. At 28 GHz the effective value can be several percent off, and a 3% error in epsilon_r moves the resonance by roughly 1.5%. I measure a sample puck with a split-post resonator before committing to a geometry. It costs an afternoon and saves a board spin.
2. Aspect ratio sets the mode and the bandwidth. For a cylindrical DRA, height-to-radius ratio decides which mode dominates and how wide the resonance is. A taller puck (higher aspect ratio) gives narrower bandwidth but a cleaner radiation pattern; a squat puck broadens the match but excites higher-order modes sooner. In my last design I settled on h/a around 0.8 after sweeping — the return loss dip was 300 MHz wide at 24 GHz, which matched simulation within 5%.
3. The ground plane is part of the antenna. DRAs radiate partly through the ground plane edges. A ground plane smaller than about 1.5 wavelengths across distorts the pattern and pulls the resonance down. I learned this the hard way when a prototype on a small coupon measured 400 MHz low and the "fix" was a bigger ground, not a smaller puck.
4. Feeding: aperture coupling beats a probe for repeatability. A probe feed is easy to prototype but its inductance varies with every solder joint. Aperture coupling through a slot in the ground plane is more repeatable across builds, at the cost of a slightly narrower match. For anything going to production, I take the slot.
5. Measure the pattern, not just S11. A good S11 dip says nothing about efficiency if the energy is going into a surface wave. I do a quick Wheeler-cap style check: compare the input impedance with the antenna in free space versus inside a small metal cavity. If the radiation resistance barely changes, most of the power is not radiating.
What is your go-to DRA feeding approach for mmWave prototypes — probe, slot, or microstrip coupling? And have you found a reliable way to source pucks with tight epsilon_r tolerance?
