Transmission line terminating at high impedances--suggestions? 1

[tlmassey]

Hi,

I'm working on an RF (~500 MHz) impedance transformation circuit. On the chip side of the circuit, I have a 3cm microstrip transmission line connecting the 50k ohm-impedance chip to the 50k ohm-impedance side of the transformation network. The characteristic of impedance of the line is (within a factor of two or so) bound to 150 ohms.

My question is this:
Are there any tricks that I can use to make the transmission line look less like a transmission line or otherwise improve the signal coupling to the chip?

My limitations:
1) The chip is provided by another group, and cannot be changed. I'm stuck with the high impedance and small package.
2) I can't shorten the transmission line by moving the components closer, because the chip is too small to fit eight inductors and other components.

Thanks in advance for any help that you can provide!

-TMassey

 

[Comcokid]

If you are dealing with a chip that is capable of inputs at 500 MHz, I don't think it's input impedance at 500 MHz is 50,000 Ohms. Maybe 50 K is its DC resistance or impedance measured at some low (<100kHz) frequency.

Can you provide more data? Do you have any info on its complex (real/imaginary) or Smith chart input data?

 

[tlmassey]

Thanks for the prompt reply, Comcokid! The parameters were all given to me by others, but I'll see if I can get more accurate values at RF. 50k was actually an estimate on my part; all I heard from the physicists designing the chip is that the impedance is "very large." Anyway, I'll try and get back to you soon with better information.

-TMassey

 

[Comcokid]

Good. On a transmission line with a standing wave, or a resonant circuit you could talk about an instantaneous point in a waveform or along a line where measuring the voltage and current the impedance calculated to 50K, but as a practical impedance to deal with at 500 MHz it doesn't make much sense.

As you probably found, it's hard to create a transmission line with an impedance above about 180 Ohms. Impedance of free space is about 377 Ohms. You've made me wonder when dealing with RF, just what is the maximum impedance you can ever expect.

 

[tlmassey]

The DC impedance was originally quoted to me as 10G ohms, with 10 ohms LC DC impedance. I probably won't be able to verify the 50k until Monday, but it seems a bit more reasonable in light of the 10G DC.

(Yes, I know these values sound outrageous; they are.)

 

[logbook]

The reason why 50,000 ohms impedance at 500MHz is unrealistic is because 1pF at 500MHz is 318 ohms. If the chip is unpackaged and flip chipped the lowest package capacitance would have to be of the order of 0.1pF (3180 ohms) without worry about the actual circuitry involved.

A quarter wave transmission line transformer will transform low impedance to high impedance as required, but you have to use the exact length to suit the frequency of operation.

 

[tlmassey]

Logbook, I minimized the package capacitance by removing the copper pour below the chip and grounding the package. The reactive impedance of the on-chip circuit is expected to be about ten ohms; it's the real impedance that I'm worried about.

Just taking the area of the silicon from a copper pour (had it not been removed), you would have been about right on the capacitance. 3.5pF through 62 mil FR4 (er=4.35), neglecting fringing (a bad approximation at this thickness, but oh well).

I'm not trying to say that I think 50k is reasonable, but I haven't been convinced yet why it can't be on the order of 50k. Until then I have to assume the worst case.

 

[logbook]

If you are talking about a shunt reactance of 10 ohms and a shunt resistance of 50,000 ohms that is a maximum Q (when considered as a component) of 5000.

A parasitic capacitance will be a lot more lossy than that.

 

[tlmassey]

The RF impedance IS 50k, but the conventional model for transmission line termination breaks down at these values, so it turns out that it will work well enough anyway. There may be a fair amount of loss due to the reflections, but the instruments will still pick up what does come through.

Thank you both for your suggestions, Comcokid and logbook, and we can now consider this discussion to be closed.

 

[Higgler]

If you do end up with a low impedance to high impedance mismatch. Try a transformer from MiniCircuits. Cheap, maybe ask for samples to start. 4:1 impedance ratio's are most common, maybe use two of them to get to 16:1.

https://www.minicircuits.com/products/transformers_sm_a.html#config_a

kch

 

[tlmassey]

Thanks for the tip, but I'm aiming for the highest possible transformation ratio (several hundred, at least--limited by component Q), so I've designed my own circuit. I'm having the board manufactured now, so I'll just have to see when it comes in and I populate it how the circuit behaves.

 

[biff44]

The impedance of the transmission line is not important at all in this case. This is becuase the transmission line is so short. A wavelength at 500 MHz, depending on what your board dielectric constant is, will be somewhere around 1/3 meter. So a couple of cm of line hase very little phase length, and as such will not transofrm your source impedance much.

It sounds like you have the standard problem of using a fixed impedance source, like a 50 ohm output amplifier, into a high impedance load. In that case, you do not have to do anything to match the input to the source. The source, with a high impedance terminating load, will act like a voltage source and give you the best output voltage swing at your load.

 

[nbucska]

Lambda=60 cm. in air, for an epsylon of 4 it is 30 cm.
-- but how can we discuss impedance transformation
without knowing the bandwitdh ?

How much loss can you afford? If not critical, you can
just terminate the line with Zo and neglect the parallel
50k (?) .


Plesae read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>

 

[tlmassey]

T-line is 1/10 lambda; right. According to the rule of thumb that I learned, that's about borderline for what can be considered short, and what can't.

The bandwidth is unknown, but I do know at the moment that the devices I'm working with resonate at 400 to 500 MHz, and later devices may resonate at higher or lower frequencies. I'm designing primarily for these frequencies but trying not to completely ignore the possibility of others, as well.

How much loss can I afford? No one knows. If I knew, I would tell you. I am confident, however, that the instruments will pick up whatever comes through.

Quite frankly, I find it rude that you're telling me to read the "how to ask a meaningful question" thread. I have read it. I gave you all of the information available, and added new information as it was aquired.

Furthermore, the content of your post indicates that you did not read the previous posts. If you had, you would have seen that the length/lambda was already discussed, as well as the issue of loss and the fact that I've requested that this thread be closed because the board's off being fabricatd.

 

[nbucska]

Sorry, no insult was intended -- the notice is automatically appended for the ones who need it
-- and most of then need it...

You haven't gave us all the info you have -- what
is the signal, what does the device do, how is
the information -- if any -- encoded, S/N ratio,
required error rate -- if applies etc. etc..

and I could list pages of guestions which may
be applicable but we don't know enough do tell
if it is.

E.g. if the bandwith is narrow you can make an
impedance xformer with concentrated LC elements
or capacitivelly loaded shotrened xmission line.

Finally, speaking about rudeness, the number of people
who try to help you should be allowed to expect
all of the relevant information without a tedious
20 question game.








Plesae read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>

 

[MacGyverS2000]

nb,


I've been busted by your auto-added thread a few times when I didn't pay attention to who was replying. Maybe adding in a couple of blank lines and a line of "-----------------" directly above the link will help people separate the signature from the actual message.


Dan - Owner

http://www.Hi-TecDesigns.com

 

[nbucska]

Ok, Mac.

Plesae read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>

 

[nbucska]

is this better?

----------------------------

Plesae read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>

 

[MacGyverS2000]

Much... though I would remove the blank line between the dashes and the FAQ, make it even more evident (and my analness would require me to fix the mispelled word "please" ;-) )


Dan - Owner

http://www.Hi-TecDesigns.com

 

[tlmassey]

Nbucska, I apologize for reacting so strongly earlier to the link. I couldn't tell that it was a signature before, but the dashed line separates it well from the body of your message. Thanks for noticing the problem and making the suggestion, Macgyvers.

Anyway, you're right; I didn't tell you what the device is. It's a new type of resonator being researched. I could have mentioned this, but it's not important; it might as well be a black box (in fact, it is being presented as a black box to me). The question was specifically in reference to the transmission line. I've found that extraneous information tends to send people on tangents, so I prefer to omit information that is not applicable to the problem at hand.

The applied signal will be a sinusoid swept through a wide band of frequencies in order to characterize the frequency response of the device. Of course, I'm only interested in how it performs at its resonant frequency--approximately 400-500 MHz. There is no encoding of information, and no required SNR as long as we can clearly make out the signal.

The transformation network is basically a tapped capacitor resonator with high-Q components.

Hopefully this answered some of the questions that were bothering you. Unfortunately, the problem wasn't as well defined as it would have been if this was a development task. This is why I tried to phrase my original question as a general electromagnetic theory question as much as possible.