Hi, Aaron.

Here are S-parameters for the circuit after adding output matching network:

S11 = -99.19 dB
S12 = -35.86 dB
S21 = 23 dB
S22 = -12.43dB

I think S12 is good here. There is only one problem is S22. It is not really good.

Hi baab,


so after improving S22 further, S11 degrades down to -2 dB? That sounds quite odd. Seems like I've missed something. I'll think about it, but let me know if you figure it out...

BTW, I assume you are checking the same frequency every time. Also, -12 dB isn't bad...


Aaron

aaron_do wrote on Mar 12^th, 2014, 2:46am:


Hi, Aaron.

I intended to mention it before but didn't.
I did input/output matching as follows.
1. Do output matching first.
After this I got really good S22. I don't remember now but it is very small.
2. Do the input matching by choosing the appropriate values of Ls and Lg.
After this I got a very good S11, it is about -99dB. However, S22 now degraded.

Now, input matching is really good while output one is not so good.
When I wrote that post, I used this circuit along with all values calculated above.
I remeasured S parameters like this:
Before adding output matching network:
I removed the output matching and measure S parameters.
After adding output matching network:
I recovered the output matching calculated previously and measured S parameters.
That explained why S11 is good after adding output network.
I think the replacement of TSMC components will cause many changes and I need to optimize again.
By the way, I just measure the power consumption. It is about 8mW. Do you think it is too large?
I am thinking about reducing it by lowering bias voltage Vgs.  I choosed Vgs = 0.6V.

Hi, Aaron.

I replaced ideal components by the real ones. With Ls and Lg TSMC components, there is no significant change in S11 and S22. However, I am struggling with the inductor in series with bias voltage Vgs. It is 1H. But as you said, it is impossible to do it in TSMC.
I tried to do replace it by an inductor in TSMC with the same inductance. The result is that S11 and S22 changed completely. It is almost unacceptable. S11 is about 0dB while S22 is -6dB.
I hope you could help me out.

I just replaced it by a resistor 10K and the S11 and S22 are good at the orignal. However, resistor 10K is also impossible.

Hi Aaron.

I am glad that we can use the 10K resistor.
However, I am having a problem with the inductor at the output (the one in parallel with 1K resistor). I tried to replace it by an inductor TSMC with the same inductance 1H. However, the results are terrible. NF increased significantly from 1dB to 15dB. S11, S12 degraded to about -2 or -3 dB.
I am wondering why 1H inductor TSMC 0.13um and 1H ideal inductor from AnalogLib can cause a great difference like that.

Hi, Aaron.

I think that will change almost completely my components values.

Now. I'd like to ask some questions before making any changes.

1. In the circuit posted above, there is a resistor 1K in parallel with an ideal inductor 1H at the drain of cascode transistor.
As you said before, the resistor is a parasitic component of the inductor not actually a separated resistor itself.
I am wondering why don't we put a high resistor at drain to get a very high voltage gain.
Av = gm* Rd => If we put a high Rd, for example, 10K, then the voltage gain or S21 will be very large.
As I know, if we use Rd from parasitic component, it won't be as large as the separated one.

2. Is 1F capacitance implementable in TSMC 0,13um?
If not, then I also need to replace all input/output coupling capacitors with the real ones.
And, for example, because 1F capacitor doesn't exist. I will have to use capacitors of much smaller capacitance.
Therefore, these coupling capacitors will affect significantly to the input/output matching networks. I will have to take them into account.
They are part of input/output matching networks.
3. In my case fc is about 1.5GHz and bandwidth = 20MHz. I am considering about bandpass matching network at the output.
I attached my schematic below. Could you confirm if this is right?
Thanks.

Hi, Aaron.
Now, I don't know where to start. The ideal case seems much simpler. I think first, we need to choose the width of transistors to get noise matching. However, with this practical case, the inductors and capacitors also include internal resistance and they affects significantly to optimum noise impedance. The problem is that I can't know in advance these internal resistance and can't simulate to get NFmin.