Sunteți pe pagina 1din 16

Need for impedance matching in electronics - a general view

Olli Rajala <olli.s.rajala@tut.> 1st February 2009

CONTENTS

Contents
Preface 1 What is matching? 2 Why is matching important? 3 What about lower frequencies 4 What role does wavelength have in this equation? 5 Matching and noise performance 6 How about splitting a signal? 7 Eects of excessive loading 8 What about video? 9 Final conclusions References 4 5 6 7 8 9 11 12 13 13 16

Olli Rajala <olli.s.rajala@tut.>

LIST OF FIGURES

List of Figures
1 2 3 4 5 A really basic circuit . . . . . . . . . . . . . A basic circuit with a noise source . . . . . . A reworked basic circuit with a noise source Splitting audio signals . . . . . . . . . . . . An example of matching problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 10 10 12 14

Olli Rajala <olli.s.rajala@tut.>

LIST OF FIGURES

Preface
This report was made for the Active RF-circuits -course (ELE-6256) in Tampere University of Technology. The purpose of this text is to give some additional information about matching in electronics. Matching in RF circuits was discussed quite heavily during this and other RF courses, but there are some dierences between RF and baseband electronics when regarding matching. I hope that I can give some food for thoughts through this paper. Even though I try to keep this as generic as possible, I will mainly concentrate in audio systems, because I have quite much experience in dealing with big sound reinforcement systems and some experience with large scale video production systems. When you are designing and installing that kind of systems, you need to think really well all electrical things and how will they aect the particular system. You may have hundreds of devices located in various spaces and connected to each other. It is not uncommon to have hundreds of meters of cable between a source and a load, for example. I also will discuss somewhat about video systems, because even though they are not true RF systems, their bandwidth is still quite large when compared to audio systems. My examples are from sound and video systems, because I think that those examples will emphasize the purpose of this report quite well. I also want to thank at the beginning of this report Jim Brown from Audio Systems Group1 , Ray A. Rayburn from SoundFirst & K2Audio2 and Bill Whitlock from Jensen Transformers3 for their comments and writings regarding this topic.

http://www.audiosystemsgroup.com/, referenced 2008-11-11 http://www.soundfirst.com/, http://www.k2audio.com/, referenced 2008-11-17 3 http://www.jensentransformers.com/, referenced 2008-11-11
2

Olli Rajala <olli.s.rajala@tut.>

1 What is matching?

What is matching?

Even though matching and dierent matching techniques have been discussed in other seminar reports, I think it is useful to revisit the basics of that topic here. In Figure 1 we can see a very basic circuit. There is source Vin with some source impedance Zs , and then there is a load which has some load impedance ZL .

Figure 1: A really basic circuit

According to the power transfer theorem, the source and load impedances need to be conjugate matched if we want to transfer maximum power from the source to the load.[1] This is the situation commonly found in RF systems and RF engineers are taught to match everything very carefully. Lets check if this really is a useful theorem (Im using R instead of Z to keep things under control, but the basics are the same even with complex impedances) [2]: RL Rs + RL And the power delivered to the load: VL = Vs V2 PL = L = RL RL Vs Rs + RL
2 2 Vs2 RL 1 RL 2 = 2 = Vs RL RL (Rs + RL ) (Rs + RL )2

(1)

(2)

When we want to know the maximum power delivered to the load, we

Olli Rajala <olli.s.rajala@tut.>

2 Why is matching important?

use the following equation:


2 Pl 2 (Rs + RL ) 2RL (Rs + RL ) = Vs RL (Rs + RL )4 (Rs + RL ) 2RL Rs RL = Vs2 = =0 3 (Rs + RL ) (Rs + RL )3

(3)

As we can very easily see, (3) is true when RL = Rs . So, the power delivered to the load is maximized when the load impedance is matched to the source impedance. In addition, if we are dealing with complex quantities, source and load impedance must be conjugate matched, such that Zs = x + yi and ZL = x yi. As every RF engineer knows, impedance matching can be made with many dierent techiques. We can use lumped components, transmission lines and even transformers. One very good tool for designing a matching circuit is the Smith Chart4 [1][p.84-92]. It shows what is the input/output/whatever impedance of your network and you can calculate from that what is the required matching circuit. These days there are powerful computer programs for this, so you do not have to calculate anything by hand. You just place dierent components and can see what is the result from those. Of course you still need to know what you are doing to be able to use this tool as well as just is possible.

Why is matching important?

As we saw in the previous section, impedance matching is important if we want to transfer maximum energy from the source to the load. The law of the preserving energy states that energy can not be created or demolished, only transferred to some other form. So, if some energy will not ow to the load, it will reect back from the junction. This may or may not be a problem, it depends on the situation and of course, on the amount of reected energy. If the mismatch is great, much energy is reected back to the source, and it can even brake the source if the amount of the reected energy is great enough. The reected power will create a standing wave with input power and that is not usually a good thing. On the other hand, especially in the mobile or other battery operated low power systems, it is important to be able to transmit/receive as much energy
http://www.microwaves101.com/encyclopedia/smithchart.cfm#smith referenced 2008-11-12
4

Olli Rajala <olli.s.rajala@tut.>

3 What about lower frequencies

as just possible. This also means that we need to have good impedance matching, so we would not lose any power due to mismatch losses.

What about lower frequencies

So, now we have discussed what is impedance matching in general, and especially why it is important in the RF circuits. The reader may now think, what is the situation in the lower frequencies? Do we still need to impedance match everything very carefully? In RF, we are trying to transfer power and in that case we need to match everything. On the other hand, in lower frequencies, we are transfering voltage, and impedance matching in the RF sense is not that important anymore. We need to take source and load impedances in the account also in lower frequencies, but the situation is not as strict as in RF systems, so I do not consider it matching in the same way. Actually we are doing kind of voltage matching, not impedance matching as it is the case with RF systems. More info will follow. Probably the most familiar example of voltage matching is the power distribution system we have. When it was invented, many people doubted that it would never work, because in their thoughts you were wasting half of the available power at the load. Fortunately this is not the case, because our electrical system is kind of voltage matched. There is a constant voltage source and the loads are placed parallel up to the point when the maximum allowed current is owing. At this point a fuse will blow and protect the system. As we have seen in the basic electronic courses, an ideal voltage source is such that the source impedance Zs = 0 and load impedance ZL = . This can be seen from the following paragraph. If we again use the circuit in Figure 1, we can calculate the output voltage with (4), which actually is the basic voltage division equation: RL Vin (4) Rs + RL As we can very easily see, the bigger the load impedance is compared to the source impedance, the greater the voltage across it will be. It is quite easy to see that the ideal situation really is like we just read. On the other hand, if the load would be matched to the source, we would have only half of the available voltage delivered to the load, as can be easily seen from the previous equation. That is probably not something we want to have if we are trying to maximize the amount of delivered voltage. So, a big load impedance with a small source impedance is the key for success here. Vo = Olli Rajala <olli.s.rajala@tut.>

4 What role does wavelength have in this equation?

When we are using this kind of circuits, we can use term voltage matching. It means that we have a voltage source and then some loads connected to it. These days solid state (and even vacuum tube) low frequency ampliers want to see a load that is much greater than their output impedance. They are meant for transferring voltage, not power. If we try to reduce the load impedance so that it would match the source impedance, more current will be sourced from the amplier and because it can not source as much current as is needed, the output signal will be distorted. In addition, when we try to source more current than is available, the output signal will get distorted so our amplier becomes non-linear. The end result is that the model in Figure 1 is no longer valid, because the model assumes linear condition. That also means, that the maximum power transfer theorem no long applies in this case.[2] Modern ampliers want to see a load that is 20 to 100 times greater than their source impedance. When that is the case, RL is said to be bridging the line because it is more than 10 times higher than the source impedance.[2] As we just saw, this is the situation we want to have when we are transfering voltage.

What role does wavelength have in this equation?

One other thing we need to also consider is wavelength and the transmission line theory that depends on the wavelength. According to Brown[3], a rule of thumb is that if we have a system with a cable with more length than /10, transmission line analysis needs to be applied. On the other hand, if the length of the cable is less than /20, we do not apply transmission line analysis. It is usually practical to take /10 as our limit, and do calculations based on that. So, if we are dealing with a system that has broadband signals, it is crucial to do some calculations to nd out the wavelength. Depending on that we need to use a proper analysis method. The equation (5) tells how the wavelength in the vacuum can be calculated. The wavelength in a cable is somewhat smaller, but the ratio is similar if our cable has the same propagation velocity for dierent frequencies. = c0 f (5)

So, to have some examples from the real life, I thought to use following frequencies: f1 = 1.5 GHz and f2 = 20 kHz, where f2 is considered to be the Olli Rajala <olli.s.rajala@tut.>

5 Matching and noise performance

upper limit of the audio frequencies. 1 20 cm 2 15 km (6) (7)

So, as we can see from (6) and (7) there is quite a big dierence between these wavelengths. As Brown said, we can now calculate how long cables can be without requiring transmission line analysis. A cable for 1.5 GHz should be smaller than 2 cm but a cable for audio frequencies must be under 1.5 km. It is quite easy to see why we do not have to use transmission line analysis with audio cabling. At least in most systems. Analog phone lines were, and are if they still use them somewhere, dierent animals, but they also require some very interesting equalization techniques to get at least somewhat understandable signal through. In those the matching impedance is 600 , and an interested reader can found more information from [3]. One of the byproducts of the transmission line eect is that we need to start to consider matching much more, because we can not transfer voltage anymore. But when we are transferring voltage, this matching is not that important as we just saw.

Matching and noise performance

In RF systems it is sometimes necessary to have non-optimum matching for better noise performance. It means that even though we have a 50 system, we may need to match the input of our amplier for some other value, like for example to 45 j 12 . The exact source impedance value for the lowest noise gure can be found from the manufacturers data sheet. It is totally dependant of the used transistor. In audio systems this it not so big thing. As long as the load is bridging the source, things are well. Although, we want to have as low source impedance as just possible, as can be seen from the following example that was constructed by Brown[2]. Lets consider a circuit from Figure 2 where Vs = 1 V, Rs = 20 , RN = 1 M, VN = 100 mV and RL = 10 k. Our noise is coming from magnetic or capacitive coupling, and the impedance can be that high. Of course it is not that simple in the real life, but I just try to keep things under control here. We can rearrange the circuit so that it looks the same as in Figure 3. Now we can see that the noise is attenuated by the voltage divider formed Olli Rajala <olli.s.rajala@tut.>

5 Matching and noise performance

10

Figure 2: A basic circuit with a noise source

by RL and RN . So, in this case 20 Rs VN 100 mV = 2 V (8) RN + Rs 1 M where VN,load tells us how much the induced noise voltage is when it reaches the load. Rs is so small when compared to RN that it can be neglected. VN,load =

Figure 3: A reworked basic circuit with a noise source

On the other hand, it can be said that the low source impedance has attenuated the induced noise voltage by Rs = 94 dB RN We can also calculate the signal to noise ratio by 20log10 SN R = 20log10 Vs VN,load = 20log10 1V = 114 dB. 2V (9)

(10)

Olli Rajala <olli.s.rajala@tut.>

6 How about splitting a signal?

11

So, our SNR was quite good, but what if we would change our Rs = 1000 ? Lets do the calculations again and see what will happen. Rs 1000 VN 100 mV = 100 V RN + Rs 1 M Now the SNR is only VN,load = SN R = 20log10 Vs VN,load = 20log10 1V = 80 dB 100 V (11)

(12)

SNR just got 34 dB poorer just by having a somewhat bigger source impedance, so it is now quite easy to understand why we want as low source impedance as is practically possible. So, it is again that kind of a situation, where we do not have strict limits for matching as would be the case with RF systems, but we need to do some thinking about the topic. In addition, resistance and capacitance in the system behave as a RC low pass lter, so with high output resistance the bandwidth of the system is reduced.

How about splitting a signal?

Pozar [1][p. 384] tells us that you can not have a three-port network which is lossless, reciprocal and matched at all ports. In audio systems these requirements are much more relaxed. In general we can say that as long as the total load impedance is big enough, there is no additional problems with splitting signals. Remember, we are trying to transfer voltage, and voltage is the same over every branch of a parallel circuit. Of course current is divided between the loads but we are not interested in it at the moment. An example would probably be a good idea. In Figure 4 we have two loads, ZL1 = ZL2 = 10 k, and our source impedance is again Zs = 20 . Now we have ZL1 ZL2 ZL1 ZL1 Z2 ZL1 = = L1 = = 5 k (13) ZL1 + ZL2 ZL1 + ZL1 2ZL1 2 This is still quite much more than 10Zs , so this splitting will not cause us any problems. If we connect too many loads, or loads with lower impedances, it is possible to get us into troubles. So, it is again necessary to have some idea about this thing, but again it is not that big a deal than with RF systems. Even though hard wire splitting is possible, in well engineered sound reinforcement systems distribution ampliers, often some kind of active splitters ZL,tot = Olli Rajala <olli.s.rajala@tut.>

7 Eects of excessive loading

12

Figure 4: Splitting audio signals

or transformer splitters, are used. The biggest reason for this is that we are trying to avoid the biggest problems in plain hard wire splitting. One of the main problems is ground loops. When equipments are located in dierent places, there are big dierences between the ground potentials of the said equipment. If we use hard wire splitting, ground currents may ow through the signal ground wires and cause this 50 Hz (or 60 Hz, depending on the local electrical system) hum that is so common. With active splitting it is also possible to totally isolate outputs. If there is a short in some output, other outputs have no idea of that. Even though this topic is very important and interesting, it is beyond the scope of this report. If the reader wants to know more, Brown and Whitlock have published an AES paper that concentrates on microphone signal splitting[4].

Eects of excessive loading

It was already discussed, and it is also commonly known, that if we try to source too much current or voltage from an transistor, the output will be distorted because the transistor can not supply necessary amount of voltage or current. In the worst case the transistor can even break down, but that is not very commmon. In sound reinforcement systems the most common splitting occurs from microphones to mixing consoles. In reality it may be possible to split signal to three dierent mixing consoles and there can be even more than 100 m of cable between the stage and the main sound mixing console. On the other hand, if we have one or more consoles with too small input impedances and a mic with high output impedance, troubles may occur. There are microphone+mixing console combinations which allow only one split and even that can be troublesome. Olli Rajala <olli.s.rajala@tut.>

8 What about video?

13

A microphone is a device that converts acoustical energy to electrical energy. Every microphone has some maximum sound pressure level it can handle without distortion. One of the symptoms of excessive loading is that this maximum SPL gets lower. It can also alter the frequency response of the used microphone. Symptoms depend on the basic working method of the used microphone, but as was the case in the previous section, more info can be found in[4]. To put it shortly, signal gets distorted in one way or another. It really depends on the used systems what kind and how much distortion will occur.

What about video?

This is not a proper place to discuss the deeper properties of a video signal, we can just take granted the following. If the reader wants to know more about video systems, Hodges has written a very good book for beginners[5]. In analog video systems the required bandwidth is quite much larger than in audio systems. It is usually said that audio requires 20 20 kHz and an old fashioned broadcast level video 0 7 MHz, although these days we have high denition video with much bigger resolution, and it can require frequencies up to 100 MHz. Although, that DC requirement is not very strict, video has a DC component but receivers can generate the required DC component from the rest of the signal if the DC component is lost. The upper limit of the video bandwidth appears to be so high, that RF type of matching starts to be important, again. Video systems are 75 systems and depending on the used equipment one needs to take care of terminating the line or not. Some equipment take care of it and some not. So, the designer must again know his/her stu. Although in antenna cabling it is quite the same whether you use 50 or 75 cables and connectors[6], in video systems it is crucial to use the right ones. Reections will be seen and they do not look good. In Figure 5 we can see the eects of improper matching[5]. The upper picture has been matched properly, and the lower picture is not. In this case it means that the line is not terminated to the used characteristic impedance, so the level of the signal is too big.

Final conclusions

I have tried to show that even though you need to think about matching in low frequencies, it is not as important as with RF systems. We are transfering voltage, not power, in these systems and the wavelength is so huge that there

Olli Rajala <olli.s.rajala@tut.>

9 Final conclusions

14

Figure 5: An example of matching problems.

Olli Rajala <olli.s.rajala@tut.>

9 Final conclusions

15

is no transmission line eect happening. As was discussed, a designer needs to check the used wavelengths and how long cables there are in the system and do further analysis based on those results. I also showed some problems that will occur due to improper matching in low frequencies. As Mr. Rayburn once said during one email conversation I had with him: Everything can be done poorly. That is so true in this case. We need to think matching but it is quite hard to make it wrong. On the other hand, in these systems there are many other properties that will cause head aches for designers. Those properties are out of the scope of this report, though, but one can nd good material from Browns website[7] and from the website of Jensen Transformers[8], for example.

Olli Rajala <olli.s.rajala@tut.>

REFERENCES

16

References
[1] D. Pozar, Microwave Engineering. Addison Wesley, rst ed., 1991. [2] J. Brown, Termination, impedance matching and the maximum power transfer theorem in audio systems, Syn-Aud-Con Newsletter, vol. 10, no. 1, 1982. (http://www.audiosystemsgroup.com/ SAC-MaxPwrTransfer.pdf, referenced 2008-11-04). [3] J. Brown, Transmission lines at audio frequencies, and a bit of history, (http://audiosystemsgroup.com/TransLines-LowFreq.pdf, referenced 2008-11-17). [4] J. Brown and B. Whitlock, A better approach to passive microphone splitting, AES Convention 118, no. 6338, 2005. (http: //www.audiosystemsgroup.com/AESPaperSplittersASGWeb.pdf, referenced 2008-11-12). [5] P. Hodges, And Introduction to Video and Audio Measurement. Focal Press, third ed., 2004. [6] J. Brown, Which coax for wireless mics, Syn-Aud-Con Newsletter, vol. 30, no. 1, 2003. (http://www.audiosystemsgroup.com/Which_ Coax_for_Wireless_Mics.pdf, referenced 2008-11-12). [7] http://www.audiosystemsgroup.com/publish.htm, referenced 200811-12. [8] http://www.jensentransformers.com/apps_wp.html, 2008-11-12. referenced

Olli Rajala <olli.s.rajala@tut.>

S-ar putea să vă placă și