Hi,
After read some text, and watch videos about voltage level shifter, and many of these text or video recommend to avoid voltage divider for high frequency, I decide to simulate some circuits to check that happens.
We need some information before start design a level shifter, as operation frequency, input capacitance, input resistance, input inductance, and output impedance, normally very low. And others.
I not found RPi electrical characteristics.
Critical information is input capacitance, similar devices have near 5pF per pin. I used three times this values, or 15pF.
Input resistance is very high, 500kOhms or more. And low inductance. Both not take on account.
As simulation software I used NI Multisim 11.0 (windows XP).
I tested two version of voltage divider, one at 10mA and other for 1mA.
http://i.imgur.com/L302HGH.jpg
C1 and C2 is a input capacitance.
R1 and R2 is 10mA divider, and R3, R4 is 1mA divider.
Signal is 10MHz.
10mA is usable on most applications.
1mA, can be not if need to measure time on high or low state.
Same circuit at 100MHz
http://i.imgur.com/2RidBR0.png
1mA is useless.
10mA can be used on some situations.
Inserting a compensation capacitor parallel to R3 or R1.
http://i.imgur.com/3dvCuCq.jpg
Now, 1mA version is usable at 10MHz, better that 10mA version.
1mA version at 100MHz, without compensation and with.
http://i.imgur.com/MTn3ktL.png
Compensated signal is very nice.
Test Square wave signal have a 1ns of rise/fall time.
In my opinion, voltage divider can used for frequencies up to 20MHz without compensation capacitor.
Most uses is above 1MHz (i2C, SPI, 1W, etc), that voltage divider works very well.
Also, active shifter can be improved if take in consideration input capacitance, and others characteristics of input and output pins of devices.
I not simulated parasitic inductance of resistor and others parts of circuit.