mixaloybas
Junior Member level 1
Hello from a new member. At first i would like to send my congratulations about the forum! (Please forgive me about my english).
I am trying to make a tranceiver design based on Chipcon CC1020 at 868Mhz ISM band. Due to many reasons I must not follow the reference design, at least the pure rf part. That means that I will almost copy the layout which deals with the low freq signals and power delivering, but I will have to design my own matching networks for RX and TX.
In vesrion 1.5 datasheet , Chipcon included info about lna and pa matching. There is a smith chart which shows the lna impedance for a freq sweep from about 300MHz to 1000MHz, and another one that shows the optimal PA LOAD impedance. However the quality of the print is bad, and only on 868MHz the load impedance is given in text form (and it is 27.997 + j 65.430).
Guessing that optimal means most possible power delivered to the load, I suppose that I will have to design a matching network so that the amplifier sees the 50ohms load(the antenna) as 27.997 + j 65.430 at 868MHz. However the impedance of the amplifier might not be 27.997 - j 65.430, is it right?. (I suppose that chipcon found the value 27.997 + j 65.430 using load pull analisys, right?)
So I designed a matching network (using RFdude initially, it gives insight i think) that uses a shunt inductor (neccessary for the biasing also) and a series capacitor, and I took account for the transmition lines present. I verified the design with rfsim99 and later with eagleware Genesys (I also placed TEEs and VIAS in genesys).
However this network is highpass by nature, so I needed to add a simple lowpass filter (shunt cap, series ind, shunt cap), to attenuate possible harmonics of the Power amp (chipcon recommends using such a filter). The filter was designed to have 50ohms terminations, so I just cascaded it to my matching network and t worked OK. I did some optimisations using Genesys and it really became a nice design, that means I watched the s11 and s21 graphs versus freq (10MHz to 3000MHz) and i had the desired attenuation at the stopband, and the desired matching at 868MHz.
But then I realized that all the frequency points (except 868MHz) possibly have got no meaning and may be totally wrong, because neither the power amp impedance is known at these frequencies (and I suppose that genesys exrapolating uses the value 27.997 - j 65.430 for all the frequencies - which is definitelly not the case), nor the antenna impedance will remain 50ohms even a few MHz beyond 868MHz.
So is finally all the analysis I have done any usefull?? What the graphs show me about matching, is it at least correct at 868MHz?? And about the filtering characteristics, do the graphs show me a rough estimation of the filter shape or could the whole picture be inverted and for example on the third harmonic 2604MHz the whole stucture is delivering energy to the load whith minimum attenuation (which would be very bad..)...
Is there also any case of intstability arrising due to the freq dependent terminations??
EDIT
I made some experiments: Firstly the input port had only one frequency point (CC1020= @868 27.997 - j 65.430) and the output too (Antenna= @868 50 j0). I used a feature called design centering that genesys has, and then running a monte carlo analysis (with realistic cmponent tolerances) I had a perfect yield (the target was s11<-20 arround 868, and s21>-0.5 arround 868, and s21<-20 beyond 1300MHz).
After that I changed the input port to have more frequency points (I used the datasheet's smith chart to estimate the impedance at other frequencies). I did the same for the output port (the antenna) using imaginary values eg
# MHZ Z RI R 1
750 35 -60
800 40 -30
868 50 0
900 60 40
950 65 100
Now the graphs of s11 and s21 were changed a little, but the return loss at 868 was exactly the same. It seems obvious, right? (The filter shape was changed - some new zeros for example, but it remained very good, even a little more sharp!)
I ran again a monte corlo analysis and the return loss at 868 MHz was all the times <-20 , yield 100% again. The other frequencies changed a lot I would say, but the tendancy was the same.
Is this some sort of guarranty that whatever is the impedance out of 868 (both at cc1020's power amp and antenna) the matching will be ok at 868?
It is a very long post. I am really sorry (but Icouldnt stop writing). Would it be better to split it ot smaller ones?
Thank you in advance!
I am trying to make a tranceiver design based on Chipcon CC1020 at 868Mhz ISM band. Due to many reasons I must not follow the reference design, at least the pure rf part. That means that I will almost copy the layout which deals with the low freq signals and power delivering, but I will have to design my own matching networks for RX and TX.
In vesrion 1.5 datasheet , Chipcon included info about lna and pa matching. There is a smith chart which shows the lna impedance for a freq sweep from about 300MHz to 1000MHz, and another one that shows the optimal PA LOAD impedance. However the quality of the print is bad, and only on 868MHz the load impedance is given in text form (and it is 27.997 + j 65.430).
Guessing that optimal means most possible power delivered to the load, I suppose that I will have to design a matching network so that the amplifier sees the 50ohms load(the antenna) as 27.997 + j 65.430 at 868MHz. However the impedance of the amplifier might not be 27.997 - j 65.430, is it right?. (I suppose that chipcon found the value 27.997 + j 65.430 using load pull analisys, right?)
So I designed a matching network (using RFdude initially, it gives insight i think) that uses a shunt inductor (neccessary for the biasing also) and a series capacitor, and I took account for the transmition lines present. I verified the design with rfsim99 and later with eagleware Genesys (I also placed TEEs and VIAS in genesys).
However this network is highpass by nature, so I needed to add a simple lowpass filter (shunt cap, series ind, shunt cap), to attenuate possible harmonics of the Power amp (chipcon recommends using such a filter). The filter was designed to have 50ohms terminations, so I just cascaded it to my matching network and t worked OK. I did some optimisations using Genesys and it really became a nice design, that means I watched the s11 and s21 graphs versus freq (10MHz to 3000MHz) and i had the desired attenuation at the stopband, and the desired matching at 868MHz.
But then I realized that all the frequency points (except 868MHz) possibly have got no meaning and may be totally wrong, because neither the power amp impedance is known at these frequencies (and I suppose that genesys exrapolating uses the value 27.997 - j 65.430 for all the frequencies - which is definitelly not the case), nor the antenna impedance will remain 50ohms even a few MHz beyond 868MHz.
So is finally all the analysis I have done any usefull?? What the graphs show me about matching, is it at least correct at 868MHz?? And about the filtering characteristics, do the graphs show me a rough estimation of the filter shape or could the whole picture be inverted and for example on the third harmonic 2604MHz the whole stucture is delivering energy to the load whith minimum attenuation (which would be very bad..)...
Is there also any case of intstability arrising due to the freq dependent terminations??
EDIT
I made some experiments: Firstly the input port had only one frequency point (CC1020= @868 27.997 - j 65.430) and the output too (Antenna= @868 50 j0). I used a feature called design centering that genesys has, and then running a monte carlo analysis (with realistic cmponent tolerances) I had a perfect yield (the target was s11<-20 arround 868, and s21>-0.5 arround 868, and s21<-20 beyond 1300MHz).
After that I changed the input port to have more frequency points (I used the datasheet's smith chart to estimate the impedance at other frequencies). I did the same for the output port (the antenna) using imaginary values eg
# MHZ Z RI R 1
750 35 -60
800 40 -30
868 50 0
900 60 40
950 65 100
Now the graphs of s11 and s21 were changed a little, but the return loss at 868 was exactly the same. It seems obvious, right? (The filter shape was changed - some new zeros for example, but it remained very good, even a little more sharp!)
I ran again a monte corlo analysis and the return loss at 868 MHz was all the times <-20 , yield 100% again. The other frequencies changed a lot I would say, but the tendancy was the same.
Is this some sort of guarranty that whatever is the impedance out of 868 (both at cc1020's power amp and antenna) the matching will be ok at 868?
It is a very long post. I am really sorry (but Icouldnt stop writing). Would it be better to split it ot smaller ones?
Thank you in advance!