Increasing Transmit Power

This does seem to be an area which somehow causes people to break out in a sweat. I actually get quite angry about this, so please excuse me if I occasionally ridicule some of the forum posters - but it seems very difficult to otherwise get people's attention. There is just so much utter nonsense published on the web by people who have no idea whatsoever what they are talking about. Running "higher power" CAN and DOES improve your performance and depends on known laws, and not the mumbo-jumbo or magic apparently entertained by many forum posters.

Now - first of all, let's get real about this "high power" crap. We are talking about a ridiculously smalll amount of power used in Wifi routers - I would normally even hesitate to refer to it as a transmitter - it is simply PUNY. The very term "HIGH POWER" to me conjurers up images of at least a 1,000W transmitter, cooled by high speed centrifugal fans. That is 8,000 times more power than our little 30 dollar router, OK? Even a mobile phone runs 20 times the power!

So - we are not talking here about running "High Power". We are talking about an increase from a puny amount of power to a *slightly less* puny amount of power.

In this article I am not telling anyone to just go and use "higher" power, that would be against sound engineering principles and create a lot of problems in an already congested band. Higher power goes further and therefore potentially causes more interference to other people using the same channel. And not to mention in your country, it may be illegal. I'm just trying to dispel some myths which seem to have taken root over the years, which are just bloody ridiculous to a communications / RF engineer.

So I will try to explain here why so many people are talking out of their backsides.

IMPROVING THE ANTENNA

Of course, improving the antenna is one way to get better performance. But you must first understand how antennas can "improve" performance. Many people insist that using a higher gain/more directional antenna will always be better, as it improves data flow in both directions. But there are many occasions when using a directional aerial will not improve the signal, such as in a multi-storey building. Antennas cannot magically provide power from nowhere! Energy from other directions is concentrated in one particular direction to provide gain. It is taken AWAY from those other directions in order to concentrate it in others.

For example, a typical add-on router antenna would be a vertically stacked collinear array of coaxial dipoles, with a relatively small gain over the standard antenna. This kind of antenna will concentrate the signal into a very flattened omnidirectional donut, reducing the signal severely both above and below the router. This may be fine for a single floor, but it will actually make things much worse for everyone on floors above and below the router. This is one scenario where increasing transmit power or even better, adding additional AP's, is a better approach.

And of course, people do seem to forget that the signal back from a client to the router can ALSO be improved by simply using a USB wireless adapter with a higher power output, and usually an external antenna connector with a 3 or 4dB gain antenna. Several of these have appeared on the market recently which provide about 200mW. Thus, the improvement can easily be made reciprocal. But be careful not to break the appropriate laws in your own country where applicable.

There is another myth that has taken hold, due to people with no knowledge trying to explain something they don't understand themselves. This one really makes me see red. And that is, that increasing power above the tomato defaults adds NOISE. This is also nonsense. The supposed "noise" can actually be generated when the power amplifier in the transmitter becomes overloaded and nonlinear in operation. This may or may not happen when the level exceeds the router's defaults, depending on who set those defaults in the first place and many other issues such as what transmit method is in use. In general, the defaults have been set very low to satisfy local regulatory requirements, and the power level can be increased considerably without the transmitter being driven into nonlinearity and producing interference or noise. If performance drops when power is increased, generally there is another, simpler, explanation.

Why an increase in power can and does improve your download speed even if the client's signal back to the router is not simultaneously improved.

LET US FIRST CONSIDER A CASE WHERE THERE IS A FIXED SPEED IN BOTH DIRECTIONS

There is a direct relationship between the numbers of dropped packets versus packet size. Small packets (acks) from the client are far less likely to get dropped because of poor signal or interference. The larger packets used by the router to send incoming data streams to the client, are much more subject to packet loss, and thus, we have an unequal situation. We often break down large packets into smaller ones to increase the link reliability and reassemble them at the other end. For anyone who has worked on long distance links over marginal paths, this is common practice.

We can reduce these packet losses by improving the signal to noise ratio at the receiver. Increasing the router's transmit power does just that. It can - and does - increase the speed and reliability of downloads to your computer IF your signal was marginal.

The replies from the client remain unaffected of course - but as stated, they are mostly small packets and subject to much less loss. Of course, if the client cannot communicate properly with the router after it has associated, then this is no longer true, but for most marginal signals a very significant benefit will often be seen. The fact that the client associated in the first place means that it can talk to the router.

BUT ROUTERS ACTUALLY NEGOTIATE SPEEDS !!

Actually, however, the router does not have a fixed speed in both directions. It negotiates a link speed with the local computer in both directions - if a weak signal is available, a lower link speed will be set up. A strong signal will support higher speeds. The transmission method and even a change from G to B speeds will be done, whatever is necessary to set up a reliable link at the new speed.

Let us consider a PC to Router link which is fairly poor, and the link has settled at negotiated speeds of 1 Mbps UP (to the router) and 1.5Mbps DOWN (from the router). The 1Mbps uplink speed is the lowest available - it is a B speed. Notice anything? Most ADSL uplinks are 1Mbps - so actually, this isn't really a big problem anyway! As the signal gets even weaker, packets will begin to be dropped and the speed might fall to say 100Kbps … but so what? It is still adequate for most browsing purposes.

Now let us increase our router's transmit power from 42 to 150mW, about 6dB.

Our new negotiated speeds for upload remains the same at 1Mbps, it may even be as low as our hypothetical 100Kbps, but the download improves to 14 Mbps !. So our downloads have increased by 10 and our uploads (mostly small ACKS) have remained exactly the same as they were before. The difference seen by the user, who is mostly concerned with his download speeds, is HUGE.

The fact that we have not improved our upload performance is quite irrelevant - because we are only sending a small amount of data in that direction. We don't NEED to improve it.

This big improvement is most easily seen by using a firmware which has simultaneous display of UP/DOWN wireless connection speeds to the clients. Once you have seen this, you will realise that most of the posts on this subject maintaining that increasing transmit power is pointless, are complete rubbish. If this fact isn't obvious, go back to the beginning and read it again.

To facilitate this I have added connection rate display to Victek's RAF 8515.2 and you will find a link below to download it. Teddy Bear's USBmod / USB-RT mode was the source.

So, to recap - increasing transmit power on the router often makes a tremendous improvement on a marginal link, because:

[A] it is the DOWNLOADS from the router that are the largest size packets, and the ones most likely to suffer losses.
[B] Whereas the smaller ACKS from the PC are much less likely to be lost, and
[C] the downlink speed FROM the router will also be negotiated at a higher rate then the uplink.

And if the client really can't communicate well with the router, you can always use a higher power USB adapter, as mentioned above. The choice is entirely up to you! There are many such adapters which run 200mW which makes the link broadly reciprocal.

NB - This isn't something that is up for discussion or argument, as it is very basic communication theory. For anyone who doesn't like it, then perhaps moving to a different universe where different laws of physics apply might be a good idea. Now, I am not going to apologize here for upsetting anyone. I am an RF and communications engineer and I do know what I am talking about. The forums are full of complete nonsense and it is high time someone put it right by explaining why many of the commonly held beliefs are utter garbage. To those people who argue in the face of all the facts, I would ask you to either go to University and learn the theory and practice by yourself, then go work in industry for at least 20 years, or perhaps believe someone who knows a tad more than you do about the subject, OK?

I agree totally that there is no justification for increasing power IF it is not necessary. However, for many people with insufficient signal at the receiver, it can mean the difference between a relatively poor, lowspeed service, and ultimate happiness. If you live in a multi-floor building made from steel reinforced concrete, for example, you will find it very useful indeed. For most people it isn't at all necessary, and may even result in a decrease in your download speed because of overload of one or both receivers (you may be too close to the router). More to follow on this in another post.

I hope this helps some people, because trying to make a technical subject understandable by others who don't work in that field, is not easy. Check elsewhere for confirmation by all means, but be very careful before accepting anything you read on the forums as gospel. Most posts are by laymen who are not experts in the field. Search through communications handbooks, research papers, specifications, papers, from sources such as IEEE, Cisco or the other manufacturers, professional bodies, universities and research establishments - these are the best sources of information. In particular, avoid magazine articles from professional journalists … remember, if you need an operation to remove an appendix you would not ask the local baker, would you?

Next, I will dispel some more myths by posting some details of the WRT54GL's transmitter, and it's power levels and performance.

Is the WRT54GL (and the ASUS RT-N16) really capable of more output ?

The ASUS RT-N16 uses the same power amplifier chip mentioned in this article and is therefore subject to the same arguments. See this link:

http://www.linksysinfo.org/forums/sh...64&postcount=3

Firmware

The original Linksys firmware is said to have had the transmit power level fixed at 28mW. Since many people wanted to increase the power for various reasons, probably to annoy the doomsday brigade, most third party firmware allows an increase in the level. The aim of most mods is to allow the use of full power on all channels.

Tomato firmware, like most others, has a power setting selection up to 251. This is usually taken to mean the transmit power output in milliwatts, but I believe this should sensibly be viewed not as the actual transmit power output level, but only a relative number. The actual power output will often be considerably less than this figure, and is dependent on the wireless driver version in use. What that level actually is, isn’t easy to determine without proper equipment. So far, I haven't found any article in which it has been tested by an accurate method and a result posted which can be taken as authoritative.

Different versions of the wireless driver result in different output levels, as the country selection and some other parameters can prevent operation at the full power and set internal maximum defaults. In the original Linksys/Tomato driver, an override parameter existed which would allow full power operation regardless of this, but this override no longer functions in the later "ND" drivers. No source code is available for the drivers, so we cannot investigate further.

Hardware

This router has a Broadcom BCM2050 wireless chipset driving an external SiGe 2528L power amplifier.

http://www.sige.com/uploads/briefs/D...ep-10-2008.pdf

The BCM2050's transmitter can provide +5dBm (3.16 mW) output. To increase this power, an external SiGe 2528L three-stage amplifier chip is used. The chip has internal temperature compensation and can also withstand a high level of antenna mismatch. In common with most modern devices, it is quite rugged. I could not find any figure for heat dissipation, but bearing in mind that the chip has a relatively low duty cycle, it does not get very hot even at the "full" setting of 251. The data sheet does not suggest that a heat sink is necessary. Nor would there be any practical way to attach one due to the chip size and packaging. The tried and tested "finger" method results in no burns, it is possible to leave one's finger on the chip without any pain. [Here I refer readers to consider the many recent laptops in which the graphics chips get so hot that the solder is melted and the chips drop off].

According to the data sheets, and in accordance with normal practice, the PA is switched off when not in use by a pulsing an “enable” input. Therefore it can produce no thermal noise when not transmitting. This is so that noise generated in these three stages does not increase the noise floor at the input to the router’s own receiver.

Many people who insist that the router will "only produce more noise" if the power is turned up refer to the noise measurement on the router itself. This is not an indicator, because the noise measured by the router obviously cannot come from the router's transmitter - because both the transmitter and it's PA are switched off when receiving. Your transmitter cannot cause an increase in noise figure on the router's own receiver.

If your router’s noise reading changes, it must therefore be due to something else.

The noise floor of a channel is the amount of noise that the receiver sees on that channel when a client is not transmitting. This noise can come from of many different sources. Noise can of course be created by our transmitter, but what you should check for would be an increase in noise at other nearby receivers on the same or adjacent channels.

The Crunch …

The 2528L has a maximum power output, according to the data sheet, of 24dBm/251mW (mode b) and 21dBm/125mW (mode g).

So let us quickly dispel the myth that the router will fail if used at a setting of 251 - The data sheet shows clearly that it is DESIGNED to run at a maximum power level of 251 mW, and that, obviously, is why that figure was originally patched into the firmware. It will NOT get unduly hot, nor will your router melt or explode if you increase the power, as maintained by some forum posters.

Tests

So let's see what happens in practice when the transmit power is varied, using Victek's mod 1.25 version 8515ND (which uses the 4.158.4.0 driver) and the country set to Singapore.

I used the normal antennas on the WRT54GL, and a D-Link DWA110 USB adapter at the client. I selected channel 6, with only one other access point noted at a level of -88 dbm +/- 3dB approx. To avoid any misleading results caused by receiver overload, I used a line of sight path to the client at a distance of 20m. Since the intention of this test was only to verify operation at levels up to 251, I did not try a greater range and a low signal level.

I decided not to try to measure download speed because this depends on so many other factors, but instead, the error rate (packet loss) of the download. Before beginning the tests, with the client switched off, I recalibrated the firmware's noise floor level, which on this channel was -95 dbm. Signal levels reported by the DWA110 were found to closely follow the readings from another WRT54GL, also calibrated, so these measurements are something everyone can relate to. Bear in mind they are only approximate figures calculated by the firmware from information supplied by the wireless hardware, and that every wireless card is different.

I used mode (b) to begin, at a power setting of 10mW. The client associated at 11Mbps at -66dbm and a signal to noise ratio of 29dB. I started download of a large file and noted the level of errors at different settings. I checked settings of 10, 20, 42, 84, 150, and 251. There was no significant difference in the mean error rate at any setting. Noise floor stayed constant at -95dbm, the signal to noise ratio improved steadily until it reached 42dB at a setting of 251 and -53dbm. During these tests, I kept two WRT54GL’s on channels 1 and 11 running 2m from the router, which showed no change in their respective noise floors. I also measured the noise floor on Channel 6 using another router in the same room - there was no increase in noise level. The link connection speed remained at 11Mbps during the tests.

Mode (g) connection at 20m was initiated at 54Mbps at a power level of 10mW, -70dBm, s/n ratio 25dB - quite a good signal. Using the same steps in power, there was little difference in the error rate at any speed, reaching -58dbm, s/n ratio 37dB. It did in fact improve slightly with settings over 150. The link connection speed remained at 54Mbps during the tests.

PLEASE NOTE - Placing the client next to the router, at a distance of under 1m resulted in a high number of errors. Watching the error rate while moving the client machine showed the error rate increasing as it approached closer than about 2.5-3m to the router. This indicates receiver overload, and is a probable cause of most of the posts in the forums where people have found increased power to result in lower throughput.

*

It is interesting to note that the reported signal strength on mode (b) was about 4dB lower than in (b) mode. So the router does seem to be adhering to the datasheet which indicates that maximum power should be 125mW in mode (g).

WRT54GL - Test Results

Here are some recent measurements on three different setups, tested in Mode G at 54Mbps all under the same conditions. Measurements by a TP-Link TL-WN422G USB adapter using InSSIDer. Please note that I make measurements only at levels under -60dB or so, where the accuracy is not compromised by receiver overload or the nonlinear RSSI firmware.

1) DD-WRT v24sp1 with ND driver 4.150.10.5
2) Tomato 1.25 with standard driver 4.130.19.0
3) Tomato 1.25 RAF with ND driver 4.158.4.0

RT-N16 - Test Results

You can see from this that the RT-N16 is pretty similar to the WRT with ND driver. Though the actual level settings are quite different. A setting of 60 gave the maximum signal strength. I established that the B, G, and N plots with 20 and 40MHz bandwidths followed almost exactly the same curves. Statistics showed no undue packet loss and speed was no different between 17 and 60 when we did a quick file download from my webserver. Signal strength was the same on any channel 1-11

Conclusion

These quick tests confirmed that there isn’t usually any problem in running a WRT54GL at any level you like. You are very unlikely to have any problems. I’ve always kept mine at or under the 150 setting, purely to be on the safe side, and I have never had any problems with overheating, increases in noise floor, or appearances of demons during the night. It is quite normal in parts of Europe for Tomato users to use 250mW! If you believe you have a situation where increasing transmit power might help, perhaps with the addition of a better wireless adapter at the client end, don't be put off by the merchants of doom. I have some clients who live in "blindspots" (rooms behind the lift shafts, for example) and they benefit immensely from the increased power levels on the router. This has encouraged some of them to try using something like the TP-Link 200mW USB adapter with external 4dBi antenna, which has improved things even more. Long - distance links using the same model router at both ends on higher-gain antennas also benefit from both routers being set to use a higher power - this would make the change reciprocal of course.

http://www.tp-link.com/products/prod...el=TL%2DWN422G

Your mileage may vary, but if it does, keep an open mind, don't panic, and look carefully for the reason. There is always a reason. You may not see or understand that reason - that is the purpose of schools and universities, research laboratories, etc - nobody expects you to be an expert. But just don't accept the word of a forum poster who happens to be an airline pilot or a building worker - because it isn't his field and what you're getting is only his uninformed opinion.

Please note that being closer than maybe 2-3m to the router may result in receiver overload and increase the error rate. Many of the people on the forum claiming better results by lowering the transmit power, have subsequently mailed me to say that it was because their client was too close to the router, and it all worked fine 30m away!

Additional teaser

When setting up a bunch of AP's for a new project, I always test them in my apartment for a week before installation. It so happens that I have just bought the hardware for a new installation. So, just to see what will happen, I put 28 new AP's on the same channel all running at a setting of 251. The noise floor understandably suffered a little, -91 instead of a more normal -94/96. However, I am still able to associate with the AP's in this residential block - even the weaker ones - on the same channel - and send this post. QED !

EDIT - December 2009

It is now almost two years since I installed over 200 routers here in different blocks, all running 150mW and overclocked to 250MHz. Not a single failure. There are now over 350, I don't have an exact count. My only router failures so far have been 2 bricks due to my own experiments. Mea culpa! Both recovered with JTAG.

WiFi Performance

Here are the best throughput rates normally achievable with typical Wifi equipment using different connection speeds and modulation type. The conclusion to be drawn from this chart, is that it pays to set your router to (g) only if you need fast WLAN speed!

[IMG]http://img27.imageshack.us/img27/8256/throughputfigures.jpg[/IMG]

The next table shows the probable best range and signal strength requirement for unobstructed line-of-sight path, that may be expected from a Wifi installation (source - Intel). The figures are a bit on the optimistic side but are the ones most often quoted whenever Wifi is discussed. You can see that mode (b) is capable of much greater range at lower speeds, the range at 1Mbps is quite remarkable. A 14-20dB stronger signal is required for fullspeed operation at 54Mbps (g) mode. There is another table commonly circulated on the internet which is even more optimistic, showing a threshold of -94 dBm at 1Mbps and -71 at 54Mbps, but I don't believe those figures can be achieved in practice.

I have marked the 80211(b) speeds in red.

Next is a chart showing typical packet loss at various signal-to-noise ratios, on a 2Mbps “b” link. This mode was used to best illustrate loss, because “b” mode uses no forward error correction and hence packet loss is not masked by the firmware. You can see that a level of approx. 15dB SNR was necessary to initiate a connection, this corresponds pretty well with the last table assuming a low noise floor. Error rate dropped rapidly with increasing strength until at around 20dB SNR the error rate was almost zero. Some random scatter data above the baseline was actually due to co-channel interference.

This scatter chart shows the signal strengths of a large selection of typical clients over time, versus distance from the router. From this you can see the range will extend significantly if you can increase gain by, say, 10dB.

And this one shows the difference in PER (error rate) with small packet size of 10 bytes against that of 1000 bytes. This uses 8011b at 1Mbps, and is of particular interested for those attempting very long distance links. Down in the weak-signal zone between 3 and 8dB SNR, you can clearly see the error rate for small packets is much lower than that of the larger packets. This phenomenon is what allows the use of asymmetric power levels, the signal from the client to the router (consisting mostly of smaller packets and ACKS) is less likely to suffer from dropped packets as the main data stream from the router TO the client.

Setup

When setting up a wifi installation, you should aim for a signal-to-noise ratio of >20dB. This will allow operation at the full 54Mbps connection speed with some margin to spare. A further safety margin above this is desirable to allow for laptops being moved while in use, etc. and 25db SNR is my own personal "minimum".

Unless you positively have to support some older (b) clients, turn off support for (b) mode.

If you are trying to set up a large site with few AP's, or operate over long distances with marginal signals, setting (b) only may give you better throughput. Consider also enabling CTS protection mode.

NEGOTIATED SPEEDS

Actually, however, the router does not have a fixed speed in both directions. It negotiates a link speed with the local computer in both directions - if a weak signal is available, a lower link speed will be set up. A strong signal will support higher speeds. The transmission method and even a change from G to B speeds will be done, whatever is necessary to set up a reliable link at the new speed.

Let us consider a PC to Router link which is fairly poor, and the link has settled at negotiated speeds of 1 Mbps UP (to the router) and 1.5Mbps DOWN (from the router). The 1Mbps uplink speed is the lowest available - it is a B speed. Notice anything? Most ADSL uplinks are 1Mbps - so actually, this isn't really a big problem anyway! As the signal gets even weaker, packets will begin to be dropped and the speed might fall to say 100Kbps … but so what? It is still adequate for most browsing purposes.

Now let us increase our transmit power from 42 to 150mW.

Our new negotiated speeds for upload remains the same at 1Mbps, it may even be as low as our hypothetical 100Kbps, but the download improves to 14 Mbps DOWN. So our downloads have increased by 10 and our Uploads (mostly small ACKS) have remained exactly the same. The difference seen by the user, who is mostly concerned with his download speeds, is HUGE.

The fact that we have not improved our upload performance is quite irrelevant - because we are only sending a small amount of data in that direction. We don't NEED to improve it.

This big improvement is most easily seen by using a firmware which has simultaneous display of UP/DOWN wireless connection speeds to the clients. Once you have seen this, you will realise that most of the posts on this subject maintaining that increasing transmit power is pointless, are complete rubbish.