Loading problem in the Tx chain

[wccheng]

Dear all,

I am design a up-mixer with variable gain in the Tx path. In the design, I use passive mixer architecture. It is quadrature input and differential output. Afterwards, the up-converted signal will enter the buffer with transformer loading and give out single ended output. The single ended output will connect with driving amplifier input. The problem is appeared. In the preliminary design, I just using 3pF as output loading in the design. After my colleague finished his DA design, I connect my upmixer and do the simulation. It is found that his loading is 5pF // 50ohms. It makes my upmixer gain drops so much even I have tuned back to the same frequency range. How could I fix the problem in order to get back the gain?

Thanks,

wccheng

 

[chuckey]

If you add a bit of series inductance in the connection with the same impedance as 2pF at your working frequency, you will have taken this 2pF out of your colleagues input circuit.
Frank

 

[wccheng]

I have tried to add a series inductor in between. However, it could not work. Moreover, it influences the other side of transformer. How could I solve it?

 

[chuckey]

I am afraid you have made a mistake some where. Inductors act like negative capacitors ( +J v -J). I suggested a series inductor, how about if you put a parrallel inductor across his input, at exact resonance, you will have made his 5pF, look like infinity.
Frank

 

[wccheng]

Hi chuckey,

I have tried to add a parrallel inductor in the input of the driving amplifer. It could not work.

Do you think the input loading of DA is too large and makes it hard to obtain a gain in a high frequency?

Thanks,

wccheng

 

[chuckey]

Ah!, I've re-read your first posting, if you make the load for your mixer 3pF, then you are trying to make some resonant circuit, because the next stage must consume some power so it must have a resistive component. I think you should have designed it for 50 ohms, with a small adjustable capacitor so you can tune out any extra cable capacitance. This is an age old problem with interconnected RF modules that perform to spec on the bench with a certain cable into a certain meter or generator, then perform differently when interconnected with different cables in the actual rack.
Frank

 

[wccheng]

Hi chuckey,

Thanks for your reply. Actually, my upmixer output connects the driving amplifier input. So, I could not put 50 ohms as my upmixer output. I have tried to simulated the DA input and find the equivalent input circuit. It is 4pF // 60-ohm. I have tried to connect this equivalent circuit one by one and see the effect. As I put 4pF, the gain will shift down. After I added 60-ohm in parallel with 4pF, the gain drops 8dB. I find that if the resistor value is increasing, the gain is increasing too. However, the DA is designed by my colleague. I could not find any method to solve it. Do you have any suggestion?

Actually, the transformer is quite hard to use it. As I tune the 2nd coil, it will affect the 1st coil side. Why do I say that? As I make the lost of the transformer is ~ -5dB, the buffer gain will drop extremely (just could get 3dB only). However, as the buffer gain could get 17dB, the transformer will lost -17dB. It is hard to tune it. Does it any technique to tune it? The circuit is same as found in ieee paper. The passive mixer connects with a DA. But for my case, I make DA as buffer and the transformer output connect to my colleague DA input. Why do I need to do that way? It is because my mixer requires to give out different gain stage.

Thanks,

wccheng

 

[chuckey]

Without a circuit, and any clue to the frequencies involved it is difficult to proceed. You do not even say how your transformers are coupled to each other inductive or capacitive? I think in general the output impedance of passive mixers is low - 20 -200 ohms. So what you need to do is to put a bit of selectivity between your mixer and the buffer/DA stage, or it will get overloaded by unwanted mixer components. So you have low Z O/P (from mixer) feeding a low Z (input) buffer. So you want a couple of high Q tuned circuits, loosely coupled. So you mixer output must be tapped down across the first circuit, either by an inductive tap, or a capacitive tap. This tuned circuit is then coupled by a small capacitor to the second one, where again its output is tapped down to feed the DA. If you are finding that tuning your transformers interact, put a resistor across the one you are not tuning, then swop it over and repeat, try 10K for starters. You might still have to repeat the process a couple of times. Years ago I actually did this on a top of the range communication receiver. I put 1,000 pf straight across a 50 line, with a 100 pf going to a inductance which had an earth on its other end. It worked beautifully. The original design had an inductive tap across this IF output from a mixer, feeding the tuned circuit. But at very high frequencies, its leakage inductance rose so the effect of a resonant IF filter was lost. The capacitor did not show this effect of course.
Frank