2010-07-03

Android vs iPhone 4 signal strength display (FWIW)

In the light of the iPhone 4 Grip of Death fiasco, AnandTech reverse-engineered the signal-strength-to-bars mapping for the iPhone 4.  Tim Bray tweeted "Interested in signal-bar calculations? Android is open source, check updateSignalStrength() in http://is.gd/dd2Kh".  I used this source, combined with the 3GPP spec referenced in the Android source (thanks to @tweakt for the link) to produce the following graph comparing signal strength indicators on Android and the iPhone 4.

I want to stress that since the number of bars displayed for a given signal strength is just a subjective way of presenting signal strength info to the user, so this graph is only presented for what it's worth -- don't read too much into it or get hung up on the details :-)  (As pointed out in the comment below, "for all the millions of dollars in lost productivity spent discussing 'bars'...")


Observations:
  • The iPhone 4 consistently displays a greater percentage signal strength than Android (as defined by the fraction of bars lit).  However the signal-strength-to-bars mapping is not regulated or defined anywhere, other than the fact that Apple said in their open letter that AT&T recently came up with their own recommendation for this mapping on their own network.  Nothing necessarily says Android is more "right" than Apple.
  • Both Android and the iPhone 4 display the maximum number of bars (5/5 on iPhone, 4/4 on Android) for over half the usable signal strength range (as measured on the dBm scale).  The implication of this is probably that it's common industry practice to show full bars whenever the signal is strong enough that there are no real or noticeable connection problems.  So Apple may be inflating their signal strength status slightly for weaker signals in order to make it look like the iPhone 4 has excellent reception, but at least the practice of reporting full bars at -90dBm or greater appears to be the norm (based on these two data points), even though there's still a lot of signal strength headroom above that level.
  • Assuming AnandTech's measurements are accurate, it's possible to come to the conclusion that the Apple signal strength numbers appear manually fudged, accidentally or otherwise: note the short dBm range for 3 bars and the extra-long dBm range for 4 bars.  In other words the iPhone reports 4 bars at a much lower signal strength than it should relative to the other thresholds, the chosen thresholds don't follow a smooth curve.
  • The iPhone 4 is generally overreporting the number of bars relative to Android for lower signal strengths (under -101dBm or so), but is about in line with Android for the highest signal strengths (over -97dBm).  Assuming that a good set of thresholds were chosen in the Android source, and assuming that the radio in an average Android device and the radio in the iPhone 4 have similar characteristics, this supports Apple's point in their letter that weak signals were previously given too many bars.  (Note however that the 3GPP standard only reports signal strength at 2dBm intervals, so selecting thresholds is not exactly a precise science, and the Android numbers are probably chosen somewhat arbitrarily too...)
  • It's hard to say what mapping Apple will have to use to make it look like the Grip of Death isn't an issue on the iPhone, but based on AnandTech's testing the attenuation is high enough that they won't be able to hide it entirely.
  • Note that the Nexus One suffers from a similar problem to the iPhone 4, you can easily lose 3G reception if you grip the phone along the metal strip at the back.
Update: The comparison is made murkier by the fact that AT&T uses WCDMA for their 3G data connection and T-Mobile and other GSM carriers use standard GSM 3G standards.  The characteristics of different data transmission standards may be dissimilar as signal strength varies and across changes in environmental characteristics such as noise levels and terrain.

Follow me on Twitter <

--

Updates:

(1) an insightful comment by xtal on Slashdot:

For all of the millions of dollars being lost on productivity aimlessly discussing 'bars'..


Can someone please dissect the antenna and then connect it to a calibrated spectrum analyser? This is so mindbogglingly trivial to do it is beginning to hurt my soul. I do similar exercises at work with new, untested antenna designs. I am sure I am not the only one.


For comparison, do the same to other phones and publish actual measurements of received signal drops and the effect from the disturbance caused from closing your hand around the antenna. This is similar to how touching an old rabbit-ears style antenna effects the picture on a analog TV broadcast, if the effect is as I suspect.

Voila! An actual, meaningful assessment of what the phone bars mean in real numbers from a calibrated instrument.

An uncalibrated receiver, such as the iphone, is not a proper tool to do this.

14 comments:

  1. What is up with the Incredible? I have -92db right now and it is showing 1 bar. It should be at least 4. No?

    ReplyDelete
  2. I think this is GSM. There are separate sections in the Android code for GSM vs CDMA, so maybe it calculates differently for CDMA. Or maybe not.

    ReplyDelete
  3. John: you are correct. This is for GSM only. CDMA (the Incredible etc. -- any VZW or Sprint phone) is a totally different story.

    ReplyDelete
  4. The Nexus One doesn't have a metal strip along the back. Unless you mean the silver-looking plastic.

    I measured the dB drop a few different ways on an N1. It was necessary to grip the phone all the way around the bottom of the phone, merely holding it there and pressing to your head didn't give the full drop. I measure the drop as 10dBm.

    ReplyDelete
  5. Interesting graph. I took a look at my BlackBerry (Bold 9000 on Rogers) and it seems to have a fairer report. It doesn't report full strength until -70 dB. -89 dB gives me 3 of 5 bars. As you go further out it becomes 2 bars at -96 dB and 1 bar at -110 dB.

    As a result my phone does of course place phone calls just fine on 2 or 3 bars. However I do notice data speed differences with the signal strengths, so I think the wider steps are warranted.

    ReplyDelete
  6. Stranger: uh, yes the Nexus One does have a metal strip on the back, it's part of the aluminum case. The 3G antenna is in the smaller plastic cover below the strip.

    ReplyDelete
  7. Re: "... if the horizontal axis is distorted to correct for inverse square dropoff ..."
    dBm is a log scale. This is the appropriate way to graph power law data.

    ReplyDelete
  8. I do remember the time when Siemens phones used to have a hidden diagnostic menu. My old M35, from my memory - full strength - 5 bars at -60, 1 bar at -90, at -106 it will hop to another cell if one is in range and has better signal. At -118 the cell will not be attempted to be contacted at all.

    ReplyDelete
  9. You don't explain what happens at zero bars, where is the point that the phones stop working?

    ReplyDelete
  10. jamiekitson: I don't know at what point the phones stop working, but you can clearly see on the graph where the signal drops to zero bars.

    ReplyDelete
  11. ATT 3G is WCDMA -- Wideband CDMA-- not GSM. you're statement is incorrect. It is more the same as Verizon than GSM. Unless you turned off 3G in iPhone settings---then you get GSM.

    ReplyDelete
  12. Thanks for the clarification, Robert -- I'll update the post.

    ReplyDelete
  13. Android AIN'T a phone.
    What does this test mean?
    Android can be used on a great variety of phones.
    The correct title should be iP4 vs Nexus One, at best.

    ReplyDelete
  14. flapane: no kidding, Android isn't a phone. But this is the mapping encoded in the Android OS itself for all GSM phones, and all GSM radios have to return accurate dBm information. So this applies to all Android GSM phones, within the limits of variation in the peformance of individual radio firmware etc.

    ReplyDelete