Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Rectifier - RFID Tags - Design Issues

Status
Not open for further replies.

shaikss

Full Member level 4
Joined
Jun 18, 2007
Messages
229
Helped
1
Reputation
2
Reaction score
1
Trophy points
1,298
Activity points
3,319
Hi,

I am designing a RFID rectifier. I am referring a IEEE paper for this purpose.
The paper has shown results for both 130nm and 350nm technologies. I have chosen 180nm technology.
So, I expected results in between the above technologies. To my surprise, the results are wierd - undesirable.

Let me explain you the stuff:

I have used the Widths as mentioned in the reference paper. The rectifier rectifies the input ac voltage from 350mV with an efficiency of 50%. So, I expected that I will be achieving atleast 20-30% of efficiency for 500mV to 700mV of input voltage. To my great surprise, I achieved efficiency less than 8%. I tried to use multi stage of rectifier; even then efficiency is less.

The current at the load is around 4uA. I varied the load from 10k to 100K. When load is 100k, current is 4uA. I varied the input voltage from 100mV to 1V (Peak values). The input current is increasing as the voltage is increasing. That is true. But I expected that the current at the input is same, when I go for 2-stage rectifier. I was shocked to see an increase in the ac current at the input end for 2-stage rectifier.

What should I do in order to increase the efficiency? DC load current is always around 4uA. I am able to see the voltage conversion efficiency. But I couldn't achieve power conversion efficiency. What should I do in order to improve PCE for the attached circuit?

Attached is the circuit of 2-stage rectifier.
 

Attachments

  • 2.jpg
    2.jpg
    79.3 KB · Views: 102

I think for a good efficiency you need to isolate not only the PMOS bulks, but also the NMOS bulks from each other, meaning you'd need an SOI process. You didn't mention this.
 

Take frequency response analysis and what is Q value of this circuit , what's fabrication factory you plan to connect (because many factor depend on Quality of Process)? :???:
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top