Transformer:Matching for Sigle-End to Diff Signals
A 1 GHz clock coming from a signal generater is splited by a transformer into differential clock signals, which come to a chip's differential clock input. These clock input pins should be connected to nmos gates (I guess).
Here is my understanding:
1 GHz has a wave length of 30cm, which is much longer than the trace from the transformer output to the chip's clock input. On the other hand, the chip's clock input sense voltage instead of power, so it is not necessary to do matching in this part (from the transformer output to the chip's clock input).
On the other side, the clock signal comes a long way from the signal generator to the transformer input, thus matching is needed.
My question is: is my understanding correct? how to do the matching?
Assume that the transformer turn ratio is 1:n for both positive and negative outputs.
Thanks.
yolande_yj
dont't think to power match as related to subsections stand alone.
If you consider the balun lossless, your generator sees the input impedance of the circuit (divided by transformer impedance ratio).
1GHz signal usually it's better to match on a board, you can control better what happens in circuit subsections.
In your case, I suggest to control the differential input impedance of your circuit (if intrinsic cap is negligible, just solder a shunt resistor on it of a desired value, for example 100 Ohm).
So, if you choose a proper transormer (1:2 impedance ratio), you'll be matched up to your IC.
Use transmission line of 50 ohm characteristic impedance in between of the generator and the balun, and between balun and IC.
Roughly it should work.
I hope it can help.
Mazz
Actually, the wavelength depends on the medium. If you are using a
microstrip, stripline, CPS, CPW,etc. then the wavelength is NOT 30 cm.
Roughly it is about 50 to 60% of the wavelenght on the air. So, I would
say the wavelength is about 18cm. A 1/10 of that would be 1.8 cm.
Now we can say that anything greater than 1.8 cm should be consider
a transmission line and needs matching.
For other side, for a 1 GHz clock the cycle is 1 ns. Assuming that 5% is
for rise time or 50 ps, the bandwidth to consider would be about 7GHz
(BW= 0.35 /(Tr)). In others words, the wavelength to consider would be
18cm/7= 2.57 cm and 1/10 of it would be 2.57 mm. So, again anything
longer than 2.57 mm is a transmission line.
Other consideration, for BALUN (or transformer) made inductively the
bandwidth is roughly 750 MHz at the best. So, I doubt that the main problem
you will face is the matching but the transformer itself.
Hi Mazz, thanks. Do you mean connecting like this?
Added after 3 minutes:
[quote="jallem"]Thanks jallem. If transformer can not be used, how people split the signal? Any suggestion?
You should be careful with the transformer. The ration 1:2 is
for the whole thing. That means the 100 ohm is for "differential"
output not single ended. The way the picture shows and
assuming the transformer is 1:2 then is incorrect.
There are many ways.
1)BALUNS
Among these there are:
Transmission lines transformer
Lumped elements
Ceramic Baluns
2) Hybrids
3) Linear Phase Filters
Some of them are limited in bandwidth. For example,
TxL Transformer is very good but it does have limitation in
low frequency.
Take a look at this link:
http://www.arrl.org/qex/2005/qx9roos.pdf
What is the criteria of chosing the transformer turn ratio?
For my case, if i still want to try on the transformer since I already have a pcb with the transformer foodprint, how to verify that the differential signal are properly generated?
I would say the only criteria is the output impedance. In other words,
if you have 50 ohm at the input and 50 ohm at any of the output, then
the ratio is 1:2 and so on.
I think the way to verify is to check the phase and insertion loss of the two
path of signals. As long the difference in phase is not more than 5 degrees
and the difference in insertion loss is not more than 0.5 dB then you the
transformer is OK. Of course one more check is the bandwidth of the
transformer itself. The net effect of the transformer on the pulse of the clock
is going to be to smooth the edges of it, decreasing the rise time, lowering
the eye pattern and probable decrease the probability of detection. Detection
is a term used in communications but applies here as well. The limited
bandwidth of the transformer has the same effect as channel in communications.
In other words, the pulse wil be deformed and will there willbe not guarantee
that the pulse have enough amplitude or same phase than the original.