Help on AC and DC coupling at wide range of frequencies

[alimjoco]

I need to check AC PSRR of an ADC, one thing First is i need to ride an AC sinusoid to a DC level. Usually it can be done with blocking cap series on Signal gen and an inductor series on a DC supply meeting at a point on power input. However, frequency range is from 10Khz to 500Mhz, it may degrade AC magnitude in terms of Accuarcy specifically on high frequencies(im using fixed inducor-1mH and fix capacitor-1oouF)? Is there an effective way to couple AC and DC across wide band frequencies? Is it possible to use bias tee??

 

[jiripolivka]

Yes, a bias tee is the device you need to superimpose an AC signal on a DC line.
There are some important conditions to meet:
1. Bias tees are designed to cover certain frequency bands, and also for a specific line impedance, like 50 Ohms. Inductors and capacitors in such a bias tee must be carefully selected, mainly if your AC bandwidth extends from 10 kHz to 500 MHz. I guess no single inductor can meet it.
2. Your DC line also carrying the superimposed AC signal must be treated as a high-quality RF line along its length if you need it to carry DC as well as an undistorted AC signal.

 

[alimjoco]

Thanks!!I found a bias tee(picosecond pulse lab, 5547) extending from 5kHz to 15Ghz. I guess that will do the trick.

On no#2 condition, What if my DC line is just a simple wire coming from a DC power supply(signal and ground)? Will it effect or do i need also a 50 ohm coaxial cable perhaps coming from the DC source..But i guess my DC soure is high output impedance. My signal gen(AC) has an RF out of 50ohm so i think that will not be a problem...

 

[jiripolivka]

Congratulations! Picosecond makes very good bias tees!

To the #2 problem: now as you use your line to carry both DC and AC including RF up to 500 MHz, you must make sure all components can be transmitted undistorted.
Your "one bare wire" with a ground return is only good to carry DC. To achieve a good response from 10 kHz to 500 MHz, your line must be a perfect coaxial, without any inhomogeneities. If you need e.g. to make a tap, you must use a wideband power splitter which is matched to 50 Ohms again from 10 kHz to 500 MHz.
In addition to the coaxial line, you can use a symmetrical twin-line if you wish, but then you would need bias tees for both "wires". And again, your twin line must stay perfect from start to end. No close objects, for instance.

 

[alimjoco]

Thanks, your very good in RF. ANother question, is the bias tee really good to use on testing AC power supply rejection?The AC plus DC port of the bias tee will serve as my input to ADC. Thus the device input should see the exact DC level in order to power up. Since I'm also limited on accessory right now, i can just put the standard power supply on DC in port plus my signal generator(50ohm out) on RF in port, thus i will only used one bias tee. I'm hoping to buy one and its pretty expensive..just want to make sure that this will be a valid set-up..

Thanks!!!!

 

[jiripolivka]

From your writing I do not know what is the DC current you will send through its specified DC current or voltage.
Due to the very low frequency limit for AC, 10 kHz, one of the inductors in the bias tee must be of the order of > 100 mH, thus made of a thin wire. The DC current may be limited to < 0.1 A.
On the DC end of the bias tee, the AC component is suppressed by the inductors and capacitors in the tee. If you still feel the DC power supply may be affected by AC, you can simply add a capacitor across the DC port.
You have not mentioned the DC or AC power to be transmitted over your combined line. Again, some limitations may exist in the bias tee.
On the other end of your line, the load should be 50 Ohms to maintain the line matched for RF, but this load may consume the DC power, too. If you do not want the DC to be loaded with 50 Ohms, you can use another bias tee to separate them, or, if DC is not to be used on line end, you can use a simple DC block, or series capacitor, to stop the DC current. The capacitor should be, however, large enough to pass the 10 kHz to the load. Its reactance should be < 10 Ohms at 10 kHz, smaller than 50 Ohms of line impedance.

 

[alimjoco]

The DC supply input current of the device is 390mA with 1.8v. The bias tee i prefered has a 500mA DC max current. I guess this will be ok.

The load for my combined AC and DC will be my input of the device. Thus in order to power up the device, it dut should have the 1.8v which is injected on the DC input port of the bias tee. The AC injected in the input port of the bias tee will also ride on the combined output port. So it's like im simulating the noise on the input of the device with a use of AC signal while the device is powered(actually ideal application, there should be no AC riding). I want to simulate my device on how it reject to its input noise(AC signal) across wide range of frequencies. So i guess the bandwidth of the bias tee such as 5khz to 5GhZ is appropriate when I sweep frequency from 10Khz to 500Mhz while measuring it's rejection.

Will my plan be achieved with the use of bias tee? I guess i just need to have my input on the dut an AC riding to a DC. DC level will be my needed input power on the device. Thanks!!!!