Use a hybrid 90 degree to feed the patch probes. A square "thick" patch is the way to go and it'll have enough bandwidth (typically 20%) to cover both frequencies you need.  

Antenna will be about 6.5" square and spaced off the groundplane by about 2.5" and the location of the two probes will be closer to the outer edge of the patch, with coax. protruding thru the ground plane by around 2" to reduce center conductor inductive losses.

If this is the article you found to put four slots in the patch; http://www.eurojournals.com/ejsr_32_2_06.pdf, I'm certain you don't need any slots for your patch antenna. Those are only to extend the bandwidth much further than you need and are an extra complication in the manufacture.  

Thank you Higgler for your help, I will try it during this week! I was following the attached pdf, because I need good S11 values for this two frequencies and to keep circular polarization.  

• http://www.google.es/url?sa=t&rct=j&q=a%20circularly%20polarized%20

That article, if made with one wideband patch would require 30% bandwidth for the two frequencies 2.2 and 3.0 ghz. That's more of a challenge with good circular polarization over a wide angle in a simple printed patch antenna. Your case is much easier with only 5% bandwidth requirement to cover both frequencies.

If you have a hint on what gain or pattern coverage area you need, that'd help the solution.

One other option is a notch antenna with some simple wings to achieve CP over your 5% bandwidth and a gain value between 0 dBi up to +8 dBi could be chosen simply by changing the size of the antenna (little brother vs. big brother). It's something I came up with years but I tried to make it ultra wide band. The notch would protrude outward though, i.e. not be a flat circuit and be about 5"x5"x0.01" size (thickness 0.01 inches not too critical, even one inch thick would work) for zero dBic gain and about one foot square x very thin for +8 dBic gain. That antenna would have 2:1 frequency bandwidth if you need to go from 0.8-2 ghz.

 

Today I have been testing with your suggestion, but I didn't obtain too much good results. I am sure that I am doing something wrong. In spite of the fact of I obtained bad results, I could notice that you were right respect to bandwith. I have upload a photo of the antenna that I am using. I am feeding the probes with 90º phase to obtain CP (that simulates the hybrid).
Thank you in advance again!

I don't see the upload?

Sorry for the mistake! Here is the image :)
http://files.engineering.com/getfile.aspx?folder=ef6062fc-6262-4fc4-ad98-d1372c1fc5e9&file=antenna.jpeg

If you analyze the probes with the center conductors extending the entire thickness of the path (i.e. from bottom plate all the way to the top plate), the inductance of the center conductor is really high and your VSWR is horrible. Hence you need to have coaxial cable protruding thru your bottom plate, roughtly 80% of the way to the top plate, then your center conductors extend the final 20% distance to connect to the top plate. Then your VSWR will be good.

I think that's the problem.  

There are broadband circular polarized antennas, but they are not patch types.

A rectangular patch will never do the bandwidth, but will do the circular polarization.

An E-shaped patch, like this:

http://www.google.com/#sclient=psy-ab&hl=en&source=hp&q=e+shaped+patch+antenna&pbx=1&oq=e+shaped+patch+antenna&aq=f&aqi=g1g-v1&aql=&gs_sm=e&gs_upl=641l11267l0l11517l36l18l3l9l9l2l984l5781l0.2.11.3.0.1.1l26l0&bav=on.2,or.r_gc.r_pw.,cf.osb&fp=e9af6eeddbe4bad8&biw=1136&bih=687

might just do the bandwidth, but it will only do one polarization.

You should probably look at some sort of spiral shaped antenna .

www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting

Ooops, sorry wrong link.  E shaped patch:

http://www.google.com/url?sa=t&rct=j&q=e%20shaped%20patch%20antenna&source=web&cd=1&ved=0CEcQFjAA&url=http%3A%2F%2Fwww.jpier.org%2FPIER%2Fpier75%2F24.07061909.Chung.A.pdf&ei=-QfYTreFG-rp0QGTh-HpDQ&usg=AFQjCNE6Lf64DTo6HYqDLFRa5i-bilqi4g

www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting

Thank you for your answers. During this time I was testing with the size that higgler proposed me. The size of the patch it´s ok, but the ground distance is too big, so I am using different distances, about 10mm from the substrate.

Bandwidth is proportional to height. Hopefully it works, but if your bandwidth is too small, just increase the height. Optimum bandwidth is at about 6x the height you've chosen. See uploaded file for thick patch bandwidth vs. thickness from page 77 of Pozar's book. It's a great book, I think our company has 6 copies of it. I suggest purchasing it.  

• http://files.engineering.com/getfile.aspx?folder=e0d1682f-5cbf-49d0-864b-f7

Hi again, thank you for your help first of all. I was making some tests during this months, but the results were not as better as I expected. So, I would like to make my question again, but with another words.
What would you recomend me to do, to achive a Circular Polarized patch antenna, between 870 mhz and 920mhz (It is no necessary to have a bandwith of 50 MHz, only if that works in both frequencies is enough).
My last design is as I mentioned before,a square patch antenna with two probes 90°.
Thank you in advance!

I am following at the moment the attached paper, but now I have another problem. With the two probes with 90º phase, I achieve perfect circular polarization and enough gain, but the Reflection Coeficient is not enough (less than -7db). With the new design, gain and RC are very good, but I can not achieve a CP range bigger than 20 MHZ.
*Higger, I can send you some graphs of my results to explain me better.
Thank you to all!
 

• http://www.google.es/url?sa=t&rct=j&q=sucw%20single%20broadbadn&

With two probes at 90 degree phase it's impossible to not get excellent return loss.
You have a problem with your 90 degree hybrid most likely. I hope you put a 50 ohm load on the unused fourth port of that 90 degree hybrid, if not your return loss will be bad.

A good 90 degree hybrid, with nothing connected to the ouput, and a good load on the fourth, unused port will get -20 dB rtn loss, so your -7 dB is very odd/something done wrong.