|
The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> RF Design >> Method/algorithm to design an RF power amplifier https://designers-guide.org/forum/YaBB.pl?num=1385547136 Message started by baab on Nov 27th, 2013, 2:12am |
|
Title: Method/algorithm to design an RF power amplifier Post by baab on Nov 27th, 2013, 2:12am Hi, I would like to get advice about the method to design RF power amplifier. I was supposed to design a power amplifier with some specifications as follows: power gain: 50dB Pout: -100dBm I have started for a while and got nowhere. Now I need your help about the approach or method/alogrithm to design. Here is what I have in mind: Step 1: Using load-pull analysis to find Rload at which the output power sastifies the required one. Step 2: From the Rload determining the output matching components that are necessary to achieve the maximum desired power. And what should I do next? I am thinking about input matching. |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Nov 27th, 2013, 5:22am I think I've told you before, your spec of -100 dBm output power is very unlikely. Its way too low. For a first pass approximation, you can calculate the load based on the output power and the supply voltage. i.e. P = VDD2/(2Rload). I'm assuming single ended and your pk-pk output is 2VDD. Then you need to size your device to get the required gain based on the load. After you have your first pass approximation, you can start looking at load-pull, and all of your parasitics, and optimizing for efficiency, linearity etc. regards, Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Nov 27th, 2013, 11:55pm Thanks. Yes, you have told that before. I am confused. The power amplifier is used for GNSS receiver. Input power of the receiver is -130dBm. After that is LNA, I am assuming that its gain is 10dB. Next block is power amplifier. Its input power will be -120dBm. Now with the gain of 50dB => Power amplifier's output power is -70dB. Do you think it is too low now? What is the typical output power? I read somewhere that it is about 20 to 30 dBm. That is really high compared to my specification above. Another question: Pout = Vdd^2/2Rload. My specifications: Vdd = 1V. Pout = -70dBm. Then I get the load resistance wanted. And get that resistance, I need to use impedance transformation network. Could you tell me why it is so bad for such a low output power? |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Nov 28th, 2013, 12:37am Hi baab, you are designing a receiver, not a transmitter. Traditionally, a power amplifier is found in the transmitter chain. What you need to design is simply an RF amplifier, not a power amplifier (PA). You are right, a PA might output 10 dBm (10 mW) or more. Anything less than that and it can hardly be called a PA (maybe in some cases). For PA design, efficiency is very important, and it requires you to "design the load". For instance, if your efficiency is 20 %, and your PA is the most power hungry circuit, then you need to burn at least 50 mW from your DC supply. Based on your battery, you can estimate how long your circuit needs between charges (for example, a 1.5 V battery with 1 Ah rating will last 30 hrs ~ 1 day). There are of course a lot of other considerations...I'm just trying to give you an idea. Your output power is -70 dBm (that's 100 pW). That means it would take 100 million times longer to output the same amount of energy as the 10 mW PA. It absolutely doesn't make any sense at all to talk about efficiency in this case, as even the least power hungry block in your circuit would consume orders or magnitude more power than your amplifier's output power. For a VDD of 1 V, that means your load is 10 GOhm. This is not even close to being close to being practical. For your design, you have other considerations. As you only have 10 dB gain in front, you will still need to concern yourself with noise. Apart from that, stability will be very important, as a 50 dB gain RF amp is not trivial. In fact, I wonder why you put so much gain at RF? You may still need to have an output matching network, depending on the interface, but your design considerations will be totally different from a PA. I suggest you try and read up more about receiver/transmitter system design to give yourself a better idea of what you're designing. regards, Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Nov 28th, 2013, 12:54am Well, you are right. I read a paper in which it is RF amplifier. I was confused it with RF power amplifier. Could you tell me the difference between them, RF amplifier and RF power amplifier? I asked this but didn't get it because I couldn't express it well. I need that high gain because of the input power for mixer. |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Nov 28th, 2013, 1:17am Hi, an RF amplifier is just an amplifier that works at RF. A power amplifier is just an amplifier that amplifies power, and it is usually designed to output a certain maximum power which is quite large. Why do you need a certain minimum input power for your mixer? Isn't it linear? regards, Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Nov 28th, 2013, 1:31am Hi, I am still confused. Quote:
Then, RF power amplifier is also an amplifier. But what does RF amplifier amplify, voltage or power? I can't figure out clearly the difference. The only thing that I guess is that RF power amplifier is a subset of RF amplifier. And if I is said to design an RF amplifier, what should I need to amplify, voltage or power? Quote:
I don't know much about mixer. I just know its function in general. Did you mean that mixer is linear and then we don't care much about its input power and we can make its gain as much as needed? |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Nov 28th, 2013, 4:16am Hi, The RF Power Amplifier must amplify power. An RF Amplifier could amplify either power or voltage depending on the output interface. Usually in an SoC, it amplifies some of both. For a discrete design with 50-ohm input and output, it of course only amplifies power. When you get some experience designing it, you will understand it. One key difference is that as I mentioned before, a PA is usually required to output a lot of power efficiently. For a generic RF Amp, efficiency may not be important. For the mixer, as far as I know, there isn't usually a minimum input power (except for noise limitations). The thing is, you haven't revealed anything about what technology you're using, or what is your operating frequency, so I don't know if you're working under limitations... Quote:
Nope. When you said your mixer requires a certain input power, I assumed you meant it wouldn't work below that power. That means either its NF is really bad, or its non-linear. Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Nov 28th, 2013, 5:17am Thanks for help. I need to read your posts several times to thoroughly get it. In the mean time, I'd like to be help with another question. I didn't know that you need more information. Here is what I need to design: - RF power amplifier - Pout = -70dBm (as you said it is too low and I will have to reconsider it again) - Power gain = 50dB - Zin, Zout = 50 ohms - Operating frequency: 1.7GHz - Technology: CMOS TSMC 130nm This is my quesion: As you said in the previous post, with Pout = -70dBm and VDD = 1V then Rload = 10Gohms. And you said that is not practical. The output of the power amplifier is hooked up to mixer that has 50ohms input resistance. The 50ohms will be transformed to 10G ohms through impedance transformation network. Is the resistance 10G ohms impractical even the fact that it is the value that we get after impedance transformation? |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Nov 28th, 2013, 7:08am Hi baab, the purpose of the 10 Gohm impedance is to get good efficiency in your PA as it will allow you to maximize your output swing for the given output power. If you don't need good efficiency, then you don't need 10 Gohm. In your case, you don't need good efficiency...Besides, if your load was really 10 Gohm, the gain of your amplifier would be tremendous. Practically speaking you will probably end up transforming your 50-ohm load to between 100 ohm and 1 kohm (just a very rough figure). It is not possible to design an on-chip matching network to transform 50ohm to 10Gohm. You can try and work out the component values and the component Q you would need for an L-match. You would need a component Q on the order of 150,000 (practical Q might be less than 10), and an inductor size on the order of tens of microhenries (practical L might be less than 10 nH). And anyway such an inductor would never be able to work up to 1.7 GHz. I suggest you read up on L-matching, and if you can see what are the practical inductance sizes and Q you can get. regards, Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Nov 29th, 2013, 6:45am Thanks. I read about L and Pi matching network some time ago. As you said, it is impractical to design with that big resistance. In RF receiver, the input and output powers are very small (as in my calculation output power is about -70dBm). I am wonder what type of RF power amplifer is used for that case and the method to design such a circuit. Could you share your opinion? I got stuck as you said in my previous post that my approach is impractical. |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Nov 30th, 2013, 5:16am Quote:
Hi, I want to ask for help again. Please confirm if my understanding is correct. + First, I need to calculate load resistance based on the formula: P = VDD2/(2Rload) For example, with my specifications VDD = 1V and Pout = -70dBm => Rload = 10G ohms. After got the load resistance, put it into the circuit and do the next step. + With the load calculated above, put it into the circuit and sizing the component to get the output power. And what is the best way to get sizes of components? I know that we need to base on formula but what is needed beyond that? |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by tm123 on Dec 1st, 2013, 7:10am Hello baab, I have been following this thread and may see where you are getting confused. You said you are designing this amplifier for a GNSS receiver and this amplifier will be after the LNA which you assume has 10dB gain, but I think we need more information to help you. Are you designing this GNSS receiver on an IC? In other words, is your circuit on the same IC as the LNA and mixer? Do you have a block diagram of the GNSS receiver that you are willing to share with us? I think I understand your confusion with the term 'power amplifier' because your specifications are in terms of input power and output power. However as Aaron pointed out, the term 'power amplifier' generally refers to the last RF amplifier in a transmitter and has output power of something like 10dBm or higher. Is it true that you only have a specification for power gain and input/output impedance? I would expect that you also have specifications for noise figure and linearity (IP3, IP2). Please let us know what your noise figure and linearity requirements are for this design. Tim |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Dec 1st, 2013, 8:20am Thank you, tm123. Quote:
Yes, but I only has to design RF power amplifier. Quote:
Yes, of course. I attached it below. Quote:
Here is my specification: + output power: -70dBm + power gain: 50dB + input/output impedance: 50 ohms. + noise figure: not so important but there is no specific number. + linearity (IP3, IP2): I didn't get that. I was supposed to determine that myself. PS. The image attached is in the paper called "Reconfigurable and Simultaneous Dual Band Galileo/GPS Front-end Receiver in 0.13µm RFCMOS ". I wanted to attach it but it seems impossible here. |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Dec 1st, 2013, 8:20am Here is the paper: |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Dec 1st, 2013, 5:47pm Hi baab, Quote:
nope you are not correct. This formula applies if you are designing a power amplifier. From your system diagram, you are clearly not designing a power amplifier. You are simply designing an amplifier that works at RF. In fact from the paper you attached, you are really designing an LNA. You should concentrate your reading on LNA design. Quote:
don't use the load calculated from above. 1. For a 50 dB gain amplifier your will need at least 2 but more likely 3 stages. 2. You probably have a power consumption specification. This will roughly determine how much gm each stage can have. From there, you can calculate the load impedance you need at each stage to get 50 dB gain. 3. As it is a multi-stage design, you will need to optimize the gain distribution to get the best tradeoff between noise and linearity. You also need to pay careful attention to stability. However, in the paper you sent, the RF amplifier has a gain of only 13 dB. This is a much more reasonable number than the 50 dB gain you seem to be designing. Why do you want such high gain? Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Dec 1st, 2013, 11:40pm Well, thanks. You said that from power consumption we can infer the value of gm. And from gm, we need to calculate load impedance of each stage to get 50dB gain. I got stuck at this. The product of un*cox ( electron mobility*oxide capacitance) about 500uA/V^2 for my NMOS. I tried everything such as increase VGS, W/L in order to get a large value of gm but it is still very small, about 3 S. And therefore, voltage gain is also very small. What is the typical value for gm to get a gain about 20dB (not to mention about power disspation here)? And following what you said, if gm = 3 S then is it possible to find a load impedance so that its gain is 20dB? Assuming that gain = gm*Rload => Rload is very large and impractical! |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by aaron_do on Dec 2nd, 2013, 1:54am Hi baab, have you tried putting in the numbers? 3 S is very large. If Rload = 50 ohm, you have a gain of 150 (20*log(150) = 43.5 dB. You can also use your input transformation network to get more gain. I think you really need to do a lot more reading of the basics. Sounds like you might want to start off with Razavi's Analog book. regards, Aaron |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Dec 2nd, 2013, 2:18am Oh, sorry, I was mistaken I meant gm = 0.003 S. I put Rload = 1k and get gain G = gm*Rload = 3. That is far too low. Quote:
Did you mean "output transformation network"? Quote:
I know that, haha. I have read Razavi's book for half of a year and I am reading RF power amplifier by Marian K. Kazimierczuk. You know I don't want to proceed like that but there are no other choice. I have to read up and learn while doing. Regarding to gm = 0.003S, is it too low? What is the typical value you use to get a gain about 20dB? I know we can calculate it based on Rload but I don't know how small Rload (after impedance transformation network) can be. |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by tm123 on Dec 2nd, 2013, 7:58am baab, I have read the paper you posted. Note that in the paper they specify that the RF amplifier has the performance in Table 1 on Page 3. Based on this information it is not clear to me why you are trying to design an amplifier with 50dB of 'power gain'. At the risk of making things more complicated, I would argue that you should be designing for a certain amount of 'voltage gain' not power gain. The reason for this is that the mixers you are driving are on chip, therefore the mixer input impedance is arbitrary and there is no such thing as RF amplifier output power, only output voltage. Otherwise you will have to spend time and effort to impedance match the RF amplifier to the mixer which I would only recommend in specific cases, and this is not one of them. Also note that there is a noise figure spec of 3.8dB which is not difficult but you need to pay attention to that. All this being said, you are designing an LNA not a PA. My advice for you would be to take a step back and derive the gain, noise figure, linearity and input/output impedance for a simple NMOS differential pair with degeneration (resistive or inductive, your choice). I think you need to gain some insight into the basic circuit operation before you try to design anything more complicated. You may also find that a simple differential pair with inductive degeneration will do the job for you. It would be impractical for someone to explain this in a forum post, so I would recommend using references such as the text by Gray/Meyer to help you. As Aaron has said, trying to achieve 50dB of gain (voltage or power) in a single amplifier stage is unrealistic. Based on the paper you posted, I am not sure that is what you really need. Hope this helps. Tim |
|
Title: Re: Method/algorithm to design an RF power amplifier Post by baab on Dec 4th, 2013, 7:24am Thanks a lot for guiding me! I will switch to LNA, will come back later. |
|
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |