External antenna does not improve but degrades speed

I just took a LPDA antenna to a friend's place who has been struggling with poor speeds from his local tower. Speeds rarely exceed a few hundred kbit/sec from this tower. We wanted to see if he could get higher speed by using an external antenna to get better signal from any of his local or distant towers.

We put his Huawei E367 with a Cell C sim (modem bought at 8ta shop) on the UMTS2100 band only in order to deselect his local tower (the tower nearest to his place only has UMTS900). The modem then selected a more distant tower on the UMTS2100 band.

Standing outside with the laptop and modem and without the external antenna plugged in MDMA reported -87 dBm. Plugging in the external antenna the modem reported -79 to -81 dBm. (I suppose the improvement would have been better without the 7 meter cable and 30cm pigtail which is more lossy at 2100 MHz than at 900 MHz. On UMTS 900 the improvement on the local tower was 12 dB)

Now with the external antenna plugged in we got 2 mbit download and 0.3 mbit upload. We were however gobsmacked when we unplugged the antenna and got 5.5 mbit download and over 1 mbit on upload. We thought it was a fluke and retested with the same results, confirming the tower ID and speeds.

I theorised then that somehow the smallish difference in the signal strengths caused the modem to "become confused" which receiver to use - the internal or external. So we retested by first unplugging the external and holding my hand over the modem first to degrade the internal antenna's signal to around -97 dBm. Then, while holding my hand over the modem, we plugged the external antenna in and confirmed the -81 dBm signal with the external antenna. Bob's your uncle we got very much the same high speed we got when the antenna was disconnected!

So what on earth is going on here? It would seem that one should enclose the modem with foil to get better speed from the external antenna

The questions I now have is what is the "receiver confusion threshold" between the signal strengths of the internal and external antenna, if there is such a thing?
Does this happen on all networks and with other modems? Would be great to get some results from others, especially on Cell C with an 8ta E367 with/without an external antenna on UMTS2100 only.

Are you trying to get Cell C on 2100MHz?

Cell C operate UMTS on 900MHz and GPRS/EDGE on 1800MHz.

Forgetting that for a moment, the antenna "socket" on most modems is not provided primarily for external antenna use but for factory alignment and testing. Some of these sockets develop faults in that they do not switch out the internal antenna when the plug is inserted and thus leaves the modem with double termination. This in turn effects the impedance and thus the VSWR at the socket.

A crude but effective work around is a passive coupling to the antenna...

I found better results just extending the USB cable and putting the MODEM higher and away from the PC

JimM said:

Are you trying to get Cell C on 2100MHz?
Cell C operate UMTS on 900MHz and GPRS/EDGE on 1800MHz.

Click to expand...

In areas with many subscribers they add UMTS2100 towers, like Stellenbosch with all the students for example.

JimM said:

Forgetting that for a moment, the antenna "socket" on most modems is not provided primarily for external antenna use but for factory alignment and testing. Some of these sockets develop faults in that they do not switch out the internal antenna when the plug is inserted and thus leaves the modem with double termination. This in turn effects the impedance and thus the VSWR at the socket.
A crude but effective work around is a passive coupling to the antenna...

Click to expand...

Evidently the internal antenna is not getting switched out on the E367. The inductive coupler is an option but it introduces more loss.

I'm guessing the E367 has two internal antennae like the E1820, although unlike the E1820 its external connector switches out the primary antenna.
I wonder if what you are experiencing is caused by the external antenna and the secondary antenna seeing different towers.
Maybe both antennae seeing the same tower is better than the external antenna seeing a different tower.

ajax said:

Evidently the internal antenna is not getting switched out on the E367.

Click to expand...

No. Nature of such switch is that if faulty it would give poor contact or no contact at all to the internal antenna when unplugged. When plugged, disconnecting internal antenna works always flawless.
I suspect something different. Is it poynting antenna or some different? There are antennas on the market designed originally for 900-1800MHz, now sold as 2100MHz. I saw characteristics of one of such antennas, it shows sharp drop around 2000-2100MHz. Working on such slope would affect seriously decoding to that level that modem would be unable to negotiate higher order modulation. It could be explanation for this behaviour.
Did you observe SNR in MDMA? Signal strenght can be still high (say 3dBm lower than expected), but it doesn't say anything how useful it is. We don't know much about SNR, but higher value is better.
On the side note, your hand can act as an antenna too, tin foil must be wired to the ground to be effective.

@sajunky, thanks. I was waiting for your input.

The antenna is the one Uniterm Direct sells, LPDA11A on their site. I might just add that I borrowed this antenna from a friend and since he got this antenna his connection has been solid and with much better speeds, on both bands, and that with a E367 on Cell C. Nevertheless, I will see if I can get a VSWR or S11 graph from Uniterm.

I only took a screengrab of MDMA during one of the tests with the antenna, not without. With the antenna plugged in MDMA reported a SNR of -83,-2 and the signal strength was -83 dBm (tower 44004, UMTS2100). I am not sure how to interpret the SNR reading.

ajax said:

I only took a screengrab of MDMA during one of the tests with the antenna, not without. With the antenna plugged in MDMA reported a SNR of -83,-2 and the signal strength was -83 dBm (tower 44004, UMTS2100). I am not sure how to interpret the SNR reading.

Click to expand...

It doesn't show any abnormality. 2nd figure -2 looks nice as well, suspected some sort of noise ground level. In my location at the same signal level and internal antenna it never goes above -6.

Edit: You say "receiver confusion threshold", funny , it could be attempts switching to higher order modulation. If you compare results using different E367, you could eliminate a small possibility that modem is faulty. Different modem type, not 21mbps but 7.2Mbps would tell us whether higher order modulation has something to do with. We know for sure that 7.2Mbps modem is unable to negotiate QAM-64. This simple check would tell us more.

Some time ago ginggs sent me some data of SNR_A, SNR_B and RSSI that he had collected with his modem. I plotted it (as shown here http://i.imgur.com/3LyTb.png) and what I think it is, is the following.


'SNR_A': The instantaneous RSSI.
'RSSI': The averaged/smoothed RSSI (much more presentable to the user, note the delays in change when 'SNR_A' changes).
'SNR_B': The actual SNR but scaled so that it's way above what 'SNR_A' or 'RSSI' could ever be. Note the high correlation between it and 'SNR_A' (the slow sags and humps of multipath). The noise floor will be approximately flat (AWGN) with minor variations in time. I don't know what the offset would be but I think adding anywhere between 20 or 30 would make it more-or-less accurate. Note that by SNR I mean baseband (which is what SNR means anyways) so this is after adaptive coding and modulation (ACM).

That's my theory anyways.

Hi UnUnOctium, nice plotting. How did you obtain it? If you were observing ^RSSI indication, you should be aware that it report signal changes greater more than 5dBm (this is from Huawei documentation). And it is what I actually see from the graph. If you query RSSI using +CSQ command you should receive more accurate results (still delayed thought).

I like this plotting but was not convinced to the conclusion that RSSI is an averaged SNR_A. I delayed this posting to the moment I could confirm my own observation regarding SNR.

Here it is. Computer is located in factory building covered by steel corrugated roof, about 10 meters from the window facing tower 2100MHz. With a different modem (before I could see SNR), I found that signal strenght is the best when modem is positioned to receive reflection from the roof. In this position I could see more CellIDs than in other positions. It looks like steel roof works as a large reflector, but there are problems: even I can establish connection, there are many lost packets making WEB browsing difficult. When I position modem to face directly the tower, signal stranght is lower by 3dBm, but connection is stable.

Now I can meassure SNR in this location with E367. When modem is positioned to receive reflections from the roof, SNR_A is lower by 1.5 to 5dBm than RSSI. When modem is facing directly the tower, SNR_A is following very closely RSSI which is around 91dBm. SNR_B is better too.

Conclusion:
1. RSSI is not just an averaged SNR_A.
2. SNR A/B carry additional information how useful signal is.

sajunky said:

It looks like steel roof works as a large reflector, but there are problems: even I can establish connection, there are many lost packets making WEB browsing difficult. When I position modem to face directly the tower, signal stranght is lower by 3dBm, but connection is stable.

Click to expand...

Would it be correct to say that the effect you're seeing is similar to having an old bunny ears aerial on top of a tv receiving an analog transmission - multiple reflections or multipath effects causes ghosting in the picture. This ghosting in the digital domain could make some packets drop as some packets arrive slightly later in time than the "correct" packet therefore "confusing" the receiver?

sajunky said:

Hi UnUnOctium, nice plotting. How did you obtain it? If you were observing ^RSSI indication, you should be aware that it report signal changes greater more than 5dBm (this is from Huawei documentation). And it is what I actually see from the graph. If you query RSSI using +CSQ command you should receive more accurate results (still delayed thought).

I like this plotting but was not convinced to the conclusion that RSSI is an averaged SNR_A. I delayed this posting to the moment I could confirm my own observation regarding SNR.

Here it is. Computer is located in factory building covered by steel corrugated roof, about 10 meters from the window facing tower 2100MHz. With a different modem (before I could see SNR), I found that signal strenght is the best when modem is positioned to receive reflection from the roof. In this position I could see more CellIDs than in other positions. It looks like steel roof works as a large reflector, but there are problems: even I can establish connection, there are many lost packets making WEB browsing difficult. When I position modem to face directly the tower, signal stranght is lower by 3dBm, but connection is stable.

Now I can meassure SNR in this location with E367. When modem is positioned to receive reflections from the roof, SNR_A is lower by 1.5 to 5dBm than RSSI. When modem is facing directly the tower, SNR_A is following very closely RSSI which is around 91dBm. SNR_B is better too.

Conclusion:
1. RSSI is not just an averaged SNR_A.
2. SNR A/B carry additional information how useful signal is.

Click to expand...

Ginggs just provided me with the data. I've only had that one data set. From what you described I think it could be that SNR_A is the SNR after equalisation and RSSI is the CNR (pass-band SNR) before equalisation. Of course, the only way to know for sure is if Huawei actually told us.

ajax said:

Would it be correct to say that the effect you're seeing is similar to having an old bunny ears aerial on top of a tv receiving an analog transmission - multiple reflections or multipath effects causes ghosting in the picture. This ghosting in the digital domain could make some packets drop as some packets arrive slightly later in time than the "correct" packet therefore "confusing" the receiver?

Click to expand...

Correct in a way, multipath manifests as non-uniform channel gains in the frequency domain. That's why the transmissions have pilots which can be used to interpolate the approximate channel frequency response. Not sure of their implementation (and haven't had time to read up that much on HSPA releases) but it's possible that the multipath gets too bad at some point and too many Hybrid ARQs occur (if they use a relatively new release) or even normal ARQs (which correct some errors instead of asking for a retransmission for all errors).

ajax said:

Would it be correct to say that the effect you're seeing is similar to having an old bunny ears aerial on top of a tv receiving an analog transmission - multiple reflections or multipath effects causes ghosting in the picture. This ghosting in the digital domain could make some packets drop as some packets arrive slightly later in time than the "correct" packet therefore "confusing" the receiver?

Click to expand...

Yes, ghosting is a major obstacle in digital communication confusing receivers. UnUnOctium brings some more technical explanation. You are correct, the same data stream arriving shifted in time, it is what makes a big mess.

UnUnOctium said:

From what you described I think it could be that SNR_A is the SNR after equalisation and RSSI is the CNR (pass-band SNR) before equalisation. Of course, the only way to know for sure is if Huawei actually told us.

Click to expand...

We know more about RSSI. At least this value is passed back to the tower before deciding on the type of modulation (so it cannot be taken after equaliser). Besides, it works both in 2G and 3G mode.
I tend to agree that SNR_A is taken after equalisation, but then it should be 3G specific. My observation is that it works also in 2G, with some irregularities, so not sure.

sajunky said:

We know more about RSSI. At least this value is passed back to the tower before deciding on the type of modulation (so it cannot be taken after equaliser). Besides, it works both in 2G and 3G mode.
I tend to agree that SNR_A is taken after equalisation, but then it should be 3G specific. My observation is that it works also in 2G, with some irregularities, so not sure.

Click to expand...

Hmm, I think SNR_A would work for both 2G and 3G given that they both support adaptive coding and modulation. I think this should actually get passed as feedback to the tower since the modulation is decided by the SNR and not the RSSI. Granted you can infer the SNR from the RSSI but why go through the extra effort when it's already there.

I think that if anyone could do a test to check that RSSI does not change when modulation scheme changes would confirm for us whether it's the CNR or not. I don't have a modem though so unfortunately I can't help.

Question is, what would SNR_B be for then if SNR_A is the SNR (unless it's maybe SNR_A = before demodulation and SNR_B = after demodulation).

UnUnOctium said:

I think that if anyone could do a test to check that RSSI does not change when modulation scheme changes would confirm for us whether it's the CNR or not. I don't have a modem though so unfortunately I can't help.

Click to expand...

From the brief observation I can say that SNR changes (increases) in the moment of disconnecting. RSSI stays. One thing can give a clue: "CSNR" is the name of the command which bring these values. Sounds similar to CNR.

UnUnOctium said:

Question is, what would SNR_B be for then if SNR_A is the SNR (unless it's maybe SNR_A = before demodulation and SNR_B = after demodulation).

Click to expand...

Going to far with speculation. Rather interpretation from this site:
http://knol.google.com/k/modem-based-3g-signal-analysis#