Click here to go to our main page on phase shifters
Click here to go to ourt page on basic network theory
Remember when we told you that passive reciprocal two-port networks will have the same transmission loss in both directions? It turns out there is a major caveat to that statement. In many networks, it does not hold for large signal operation. Sure, a waveguide bend will have the same loss in both directions at one watt or 1 kW, but what about a digital phase shifter or attenuator? For the record, we are talking about networks that are considered reciprocal, in other words, no ferrites allowed!
Lets look at example, without the benefit of any real data. Starting to sound like politics? Not really, we are going to apply some thought here, let's call this what it is: a gedanken.
The TGP1439 is a five bit 18-20 GHz (K-band) five bit phase shifter. If you hunt around you can find a data sheet on the web. It was a TriQuint (now Qorvo) part and was obsoleted in 2014. You can still find some stock at distributors if you are interested, try RFMW. Qorvo did not pay for this page so we will not praise its performance here, but as an Englishman might say, "it's not bad".
TGP1439 phase shifter
You can learn a lot more about the design if you read the following paper:
Charles F. Campbell and Steven A. Brown, "A Compact 5-Bit Phase-Shifter MMIC for K-Band Satellite Communication Systems", IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, Vol. 48, No. 12, Dec. 2000.
It is one of the classic papers on phase shifters, if you want to be a phase shifter designer you should study it closely. Also, study the mug shot of one of the authors below, he is the Symposium Chair for CSICS for 2015! Phase shifter designers are often shifty characters. Perhaps they have played at Microwaves far too long without a helmet...
If you look at the RF ports on the left and right, you will see the left side is labeled "RF IN" and the right side is labeled "RF OUT". It's settled then, you always want to put the power into the left side... or why would they label the ports like that?
Before we get to far, someone is going to yell at us that a digital phase shifter is not a "passive" network. For all intents and purposes it is, and it is treated as such. With FETs or PIN diodes, there is some energy added to the system, but not in the sense that the part is "active". The devices are merely used as switches, not as gain elements.
The reason they label the ports IN and OUT is that this eliminates a lot of phone calls from customers. It also eliminated having to put any thought into the differences between driving the two ports which would require taking some actual large signal data on a small-signal part, ore disclosing more information about the design such as the actual FET peripheries and schematic.
Often when a phase shifter is used in a TR module, and RF passes in one direction during transmit, and the other in receive. It does not matter which direction the RF goes in receive, the loss (and the same noise figure). But in the transmit chain, it acts like a limiter: after you exceed a certain RF power level, the output saturates. This is because of two effects on the switch FETs: they current limit when on, and they voltage limit when off. Of course, you can put a ton pf gain after the phase shifter in the transmit chain, and not spend too much time thinking about its nonlinear behavior. But why not configure it for maximum cap[ability (and least stress?)
Voltage nonlinearity
his circuit was processed on a 0.25 um GaAs pHEMT process, so you can imagine that the breakdown voltage is around -10 volts and the threshold voltage is around -1 volts. If you drive the devices at -5 volts, you can expect the voltage nonlinearity contribution to be small, and the devices that are off will handle almost 1 watt. But wait, teh data sheet shows that you can operate the device from -2.5 volts to - 5 volts. You might want to use -2.5 volts if you used CMOS driver circuits. At -2.5V, the threshold is but 1.5 volts away, and the off-state devices will go nonlinear at 11 mW. Yikes! But no matter what voltage you drive it at, all five bits have devices in the off state that are in the direct path of the RF, and all will behave equally, so there is no advantage of driving either end.
Current nonlinearity
What about the FETs that are turned on and are in the direct path of teh RF signal? Each bit has series FET that are turned on in the refernce state, the 180 bit (a switched HP/LP network) has two in each state. In this case, the size of the FET matters (although we really hate that size meme unless it is applied to male SUV owners...) The smallest bit (11) on the right has the largest series FET, followed by the 22 bit, followed by the 45... it is obvious that the phase shifter will pass more power from the right side to the left side, given that the voltage nonlinearity is overcome by driving it with -5V.
Conclusion:
- Passive networks can be nonlinear
- In the case of MMIC phase shifters (and their cousins, digital attenuators), the P1dB or Psat can be different depending on which side you drive
- You need to consider all sources on nonlinearities, i.e. voltage and current in switch FETs
- You need to read beyond the data sheet and give some thought to how parts are used if you want to maximize your chances of success!