The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> on chip LC filter
https://designers-guide.org/forum/YaBB.pl?num=1297719173

Message started by kelly on Feb 14^th, 2011, 1:32pm

Title: on chip LC filter
Post by kelly on Feb 14^th, 2011, 1:32pm

Hi All,

Most of the LC filter calculator needs the termination resistances at input and output. Do you need that if you were to design the filter on chip? Or how do you design an on chip LC low pass if the output of the filter goes into a high impddance node?

Thanks.
Kelly

Title: Re: on chip LC filter
Post by wave on Feb 14^th, 2011, 5:38pm

For a Hi-Z output driving another Hi-Z input, you typically want to buffer it.
ie, buffer Amp, or maybe a simple Source Follower, depending on your situation.

Title: Re: on chip LC filter
Post by vp1953 on Feb 15^th, 2011, 9:13am

Hi Kelly,

If the output of the filter is connnected to a high impedance node and this node is onchip without the signal travelling a significant distance, then you can just connect it directly assuming of course that this high impedance node does not load your filter.

Title: Re: on chip LC filter
Post by kelly on Feb 15^th, 2011, 11:19am

Hi VP1953.

But how do you decide the LC values if the filter output is connected to a high impedance? Say I calculate the LC values for RS=50 and RL=100. When I use that set of values and desconnect the RL (which is the same as connecting the filter output to a high impedance node), the filter peaks by around 10dB. I assume it's because the LC values are not chosen for RL=infinite. But if you look at the equations, the L will go infinite for RL=infinite. That's where I am stucked.

Thanks.

Title: Re: on chip LC filter
Post by vp1953 on Feb 15^th, 2011, 2:27pm

Hi Kelly,

Why not set RL to whatever value that gives the desired filter response and then connect the high impedance node to RL (which hopefully will not impact the filter) - there is no need to completely remove RL?

Title: Re: on chip LC filter
Post by kelly on Feb 15^th, 2011, 5:32pm

Hi VP1953,

It's basically noise. There is a noise requirement for the stage (it's a VGA) after the filter. The VGA input is biased to a voltage through pull up resistors, and they are in the Kohms range for minimum noise. So as far as the filter is concern, the Rl is pretty high. Not to mention the corner of the LC filter needs to be around 500MHz, so the inductor is huge. Since I am not too familiar with the LC filter, I am not sure whether you can build the LC for a infinite RL condition.

I guess you are saying the LC filter needs to be designed with a specific RL then? If that's the case, I can look at what's the maximum R VGA can use (trade that off with the noise) and design the filter based on that R then.

Thanks.

Title: Re: on chip LC filter
Post by RFICDUDE on Feb 15^th, 2011, 7:53pm

Well, the first question is why do you want to perform LC filtering on chip?

Inductors are big and somewhat lossy for frequencies below 3GHz. A single LC resonant tank is easy enough to deal with, but designing coupled LC tanks is not common for integrated RF designs. The losses are significant to the point that it is not worthwhile to consider higher order on chip LC filters.

Title: Re: on chip LC filter
Post by kelly on Feb 16^th, 2011, 10:41pm

To filter out te LO+RF from the mixer before it goes to the on chip VGA. Originally, the mixer out just goes straight to VGA (high input impedance), now if I need a LC low pass filter in between, it kind of force a specific termination on both ends. Doesn't that also introduces voltage loss too? Since the the input of VGA is no longer high impedance , but say 100 or 200 ohms (to keep the L reasonable). Is there a better way to filter out the LO+RF?

Title: Re: on chip LC filter
Post by rfidea on Feb 16^th, 2011, 10:56pm

Hi Kelly!

What is the problem with designing a single terminated LC filter with high-Z load? I got this from FilterFree from Nuhertz.

As you say, the inductor value will be huge. Of course depending on the source resistor, but anyway high. Can you use a single pole RC filter instead?

Title: Re: on chip LC filter
Post by kelly on Feb 16^th, 2011, 11:54pm

Hi rfidea,

Nothing that's what I need! I just didn't know how to do it. All the papers I saw are using remination on both ends.

Can you please send me the link or papers on this, thanks.

Title: Re: on chip LC filter
Post by rfidea on Feb 17^th, 2011, 10:33am

Sure, its a free version of a filter program from Nuhertz. The Filter Free program is free.

http://www.filter-solutions.com/

Title: Re: on chip LC filter
Post by kelly on Feb 17^th, 2011, 5:58pm

Hi RFidea,

Thanks. I simulated the one you posted, the output looks really peaky (>5dB). Maybe I am not looking at this right?
Thanks.

Title: Re: on chip LC filter
Post by kelly on Feb 17^th, 2011, 6:18pm

sorry my bad, it looks fine now. Thanks.

By the way, the just an RC doesn't give me enough rejection at 1G.

Title: Re: on chip LC filter
Post by RFICDUDE on Feb 17^th, 2011, 6:22pm

Your original post was regarding onchip LC.
If the inductor is on chip then the Q will be very low at 500MHz (2 or less).

The series resistance will be (assuming a Q=2 at 600MHz) about 65 ohms and the self resonant frequency will likely be around 1GHz.

So under these conditions is the filter still useful?

You could consider an offchip inductor where the Q will be much better, but then you have to offchip and back onchip.

Title: Re: on chip LC filter
Post by kelly on Feb 19^th, 2011, 4:52pm

Hi RFICDUDE,

yeah, you are right about the practicality of the actual implementation.

I was hoping the inductor need not be that big. It seems like if I use the single terminated case that RFidea has, with a smaller RS, say 25ohms, the inductor can be 4 times smaller. I actually need the corner to be around 400M, so that'll incease the inductance too.

I am still waiting for the filter program place to send me the trial version. Has not heard from them..........

Thanks.

Title: Re: on chip LC filter
Post by rfidea on Feb 20^th, 2011, 12:35am

Hi, you can do this dimensioning in the free version, FilterFree, of NuHertz software. You do not need to use the trial version of the purchase version.

http://www.nuhertz.com/register/reg_free.html

I agree with rficdude, it can be hard to get a good filter with a low-Q integrated inductor. Usually the knee at the filter edge will not be as sharp as the ideal filter response, but you maybe can survive anyway.

Title: Re: on chip LC filter
Post by kelly on Feb 20^th, 2011, 3:01pm

Hi RFidea,

This where I went, it asks you to fill out information in order to send you password and license key. Now they say they are evaluating my request. Did I do something wrong?

Thanks.

Title: Re: on chip LC filter
Post by rfidea on Feb 20^th, 2011, 11:31pm

I just pressed the link above and downloaded the new version 13.1.0 of FilterFree without any waiting for keys and passwords. You are probably on the wrong webpage trying to download the trial version of the purchase version of the software

Title: Re: on chip LC filter
Post by kelly on Feb 21^st, 2011, 12:32am

got it, thanks! I was at the filter solution instead of filter free :)

Title: Re: on chip LC filter
Post by rfcooltools.com on Feb 25^th, 2011, 5:57pm

Hi Kelly,

Is it possible that at 500MHz you may be better off with a RC filter or to figure out a way to use the VGA as part of the feedback loop for an active RC filter. The reason I ask is that as stated by RFICDUDE the inductor will not be economical and you may want to look for alternatives. I am not sure what your area or current budget is or even the process you are using, but if you are using an expensive less than 90nM process then do something else because the inductor will be expensive. Also I f you want a free program to play around or even simulate a filter go to my site at http://rfcooltools.com and select the filter design button or you could try a free product AADE filter design, or another free product Elsie which can generate a prototype that you desire. but remember the Q will be small around 2 to 4 at 500MHz and this will be your biggest issue in implementation.

http://rfcooltools.com