I have designed PCB Traces with a Tee connection. Port one input is for AC signal input (with DC block) and another is for DC signal input (with inductor). Both AC and DC signal combines and enters my DUT. Similar concept as in Bias tee. However, the frequency response is damping. I do not get smooth/flat response as shown in the attached file. I have simulated the circuit prior of designing and i do not find any damping problem.
May i know what is the possible reasons resonance occur? My assumption is the DC path will not bring effect to my AC signal. DC source is an AC ground (please correct me if wrong) and thus an inductor is placed at the DC source (Vcc).
This is very likely because your inductor has substantial shunt capacitance and creates a resonant circuit above which the capacitor is essentially connectd to the DC source. THe DC source probably has a lot of variation in its impedance as you go up in frequency. I designed the bias tees in the HP 8753 and I can tell you that it is very complicated to make an indcutor that is broadband. In that case, there were a total of about 9 different inductors used to create a single effective inductor to cover a range of about 100 kHz to 6 GHz.
Thank you for your reply. I have tried to put this circuit in s parameter and DC simulation and it turn out to be fine. Everything seems good till the product is out. Now, i am clueless on what approaches i should use to solve this problem. Do you have any literature for references?
FIrst I would disconnect everything and measure the thru response with nothing connected to be sure the analyzer is OK. If this is mostly flat and 0 dB, I would connect the system and use the VNA internal bias tee (instead of the oone you ahve) to send voltage to the DUT.
See what the response is then. Post back with results and we can take it from there.
You say the s parameter simulation looks fine, but what are you using for the L and C in the simulation? As Dr. Joel mentions, an actual capacitor or inductor are far from ideal, and you need to account for their actual responce. You can try measuring the C and L and putting those S parameters in your simulation.
Normally you would not assume that the DC supply is a AC ground, and you would add capacitors to ground at the Vcc feed point to make it a good AC ground.
If cost is not a factor, then the easiest way I have found to make a good broadband bias tee is to use some exotic components. Piconics and Coilcraft make some very broad band conical inductors, and AVX makes an "ULTRA-BROADBAND" capacitor. I've used these in a test fixture for a broadband (50-3000 MHz) power amplifier where I did not want to have to go through the trouble Dr. Joel refers to in "making" a broadband inductor. I wasn't going past 3 GHz, but I had a good experience with these parts. The cost of them put them out of the question for high volume product production.
Is this a measurement of the bias tee ? Or did you calibrate after the bias tee and this is a measurement of the DUT? Is this a good measurement and all is well, or a bad measurement and all is terrible? Please re-read your post and think about someone who does not know what you are doing, and then post back if you have a question, but please be more clear about that you are doing or there is no way I can provide any help.
Sorry for the unclear information. That is the measurement using bias tee to my DUT. It is a good measurement. There is no oscillation. But when i use my built in T-Connection in PCB as oppose to the external bias tee, the frequency response becomes oscillating as shown in my first post. As mentioned earlier on, s parameter simulation does not see the problem. So i do not know how to solve the oscillation problem. Looking forward your advises.
your built in bias tee does not match the simulatution because parasitic effects in the elements are not accounted for. You will need to measure the impedance over a wide freqeuency up the maximum frequency of operation of your DUT for all the elements in the bias chain, and then use that information to create a more complex model of each element before you simulation matches the real world.