Slow Speeds And Strange Hsdpa (light-blue/dark Blue Led) Switching

[techmind]

Hi,

During the past week (since around 19th Sept) my downlink data throughput for large files in the evening has been often less than the usual 1.4Mbps, often down to a few hundred kbps. Is this due to excess load on my base station (CB22 5BP) due to the students returning to university, or is there a network issue?

Another odd thing is that in the past 3-4 days, my dongle (E220) has sometimes been remaining in dark-blue 3G rather than jumping light-blue 3G+ (HSDPA) when I'm using the connection. Sometimes it'll be in light-blue, then switch to dark blue when I start a download or watching a streaming video (which then, unsurprisingly, breaks up and becomes stuttery). I'd never seen that before this week. (This has been happening in the evenings, and now at 11:50 on Saturday morning.)

I'm still in the same place, a hundred yards or so from the base station, with 3-4bars of signal.
Any ideas?

Thanks,

Andrew

[Retired-Helen]

Hi techmind

It's a shame that you've been experiencing problems with your modem connection.

I've had a look at the cell site nearest to your postcode and it seems to be taking traffic for both calls and data as usual, there are no reported issues with the site.

We have had issues with 3G in London in the last few days, but I wouldn't have thought that would impact you in your location.

It could be due to increased traffic on the site due to the start of term, but again the current stats for the site in your area show no problems or dropped connections.

It might be worth taking the SIM from your modem and putting it into a phone and doing a manual roam on the handset to make sure that the SIM is all up to date with the network.

Let me know how you get on,


Helen

eForum Team

[heady]

Hi,

During the past week (since around 19th Sept) my downlink data throughput for large files in the evening has been often less than the usual 1.4Mbps, often down to a few hundred kbps. Is this due to excess load on my base station (CB22 5BP) due to the students returning to university, or is there a network issue?

Another odd thing is that in the past 3-4 days, my dongle (E220) has sometimes been remaining in dark-blue 3G rather than jumping light-blue 3G+ (HSDPA) when I'm using the connection. Sometimes it'll be in light-blue, then switch to dark blue when I start a download or watching a streaming video (which then, unsurprisingly, breaks up and becomes stuttery). I'd never seen that before this week. (This has been happening in the evenings, and now at 11:50 on Saturday morning.)

I'm still in the same place, a hundred yards or so from the base station, with 3-4bars of signal.
Any ideas?

Thanks,

Andrew

The jump from 3G to and from HSDPA is controlled by the network. The modem must get the OK from the network before it can jump to HSDPA; normal behavior is to camp on 3G.

The OK from the network is based on a couple of things:

• The RF load on the cell. For 3G and therefore, HSDPA - a cell cannot go above about approx. 75% of max theoretical load (signal to noise margin). More than 75% puts the cell in unstable run-away - so this needs to be finely controlled. Even if subscribers are throttled.
  
• Available codes. As 3G & HSDPA use codes for subscriber seperation - codes must be available. If not then either a code swap or tree re-pairing (defragment & garbage collection) is required. If this cannot be done then your left were you were or moved. The subscribers place on the code tree relates to max throughput for that subscriber.
  
• Signal quality. The modem is contineously monitoring the BER for the current cell(s) as well as adjacent cells and giving status reports to the network. The network uses these reports to try to keep the BER target. If an adjacent cell has better values - then the network can initiate a handover.
  

All of these are defined by network parameters and can be "optimized" as per the operators wishes. This means that only the RNP/RNO staff at Vodafone would be able to say for certain whats being triggered.

From your comments above - I'd have an educated guess:

• The load on the cell has been rising in your area over the last many months and is now getting busy. (any Vodafone tech should be able to confirm)
  
• That the signal quality (not signal strength) has also degraded within the cell. (Only the RNP/RNO guys will have an idea)
  

As high traffic rates make the BER worse and increase the cell load. The network will try to:

• Move subscribers around adjacent cells to distribute the load. If the cells are quite busy then this gets quite hard. (dropped connections)
  
• To keep the BER target values a "downgrade" back to 3G might be considered by the network as required.
  

The problem with the Huawei E220 is that it does not report the current BER value so it is hard to diagnose from the modem side. However, you should be able to tell if the modem is doing handovers.

So to help answer you initial question above - you'll need to answer the following:

• Is the modem doing handovers?
  
• See if you can try another modem (from a friend) that reports BER values and see if they spike before the downgrade to 3G.
  

[techmind]

The jump from 3G to and from HSDPA is controlled by the network. The modem must get the OK from the network before it can jump to HSDPA; normal behavior is to camp on 3G.

...
From your comments above - I'd have an educated guess:

• The load on the cell has been rising in your area over the last many months and is now getting busy. (any Vodafone tech should be able to confirm)
  
• That the signal quality (not signal strength) has also degraded within the cell. (Only the RNP/RNO guys will have an idea)
  

I see what you're saying. It had occurred to me that if the cell is becoming heavily loaded (as I had suspected) it wouldn't make much sense to allocate one user the extra bandwidth of HSDPA.

I live in a semi-rural area, the nearest 3G base station (100yds away) is an 11metre high "lamp post" (and I'm probably within the nearest 10 houses), then the next nearest is a bigger installation on a proper tower about a mile away across open fields plus a few trees. Given the ratio of the distances my intuition is that it'd be unlikely to be doing much with the further base-station - but I'm not familiar with the real deep down details of 3G. (I have read significant chuncks of the ETSI 2G GSM spec!).

It would be fascinating and revealing to see the total site loading in real-time 🙂 but without being a Vodafone employee that's not going to happen!

[techmind]

The jump from 3G to and from HSDPA is controlled by the network. The modem must get the OK from the network before it can jump to HSDPA; normal behavior is to camp on 3G.

...
From your comments above - I'd have an educated guess:

• The load on the cell has been rising in your area over the last many months and is now getting busy. (any Vodafone tech should be able to confirm)
  
• That the signal quality (not signal strength) has also degraded within the cell. (Only the RNP/RNO guys will have an idea)
  

Thanks Heady!


I see what you're saying. It had occurred to me that if the cell is becoming heavily loaded (as I had suspected) it wouldn't make much sense to allocate one user the extra bandwidth of HSDPA.

I live in a semi-rural area, the nearest 3G base station (100yds away) is an 11metre high "lamp post" (and I'm probably within the nearest 10 houses), then the next nearest is a bigger installation on a proper tower about a mile away across open fields plus a few trees. Given the ratio of the distances my intuition is that it'd be unlikely to be doing much with the further base-station - but I'm not familiar with the real deep down details of 3G. (I have read significant chuncks of the ETSI 2G GSM spec!).

It would be fascinating and revealing to see the total site loading in real-time 🙂 but without being a Vodafone employee that's not going to happen!

[techmind]

(double-posted due to timelags - please ignore)

[heady]

...
I see what you're saying. It had occurred to me that if the cell is becoming heavily loaded (as I had suspected) it wouldn't make much sense to allocate one user the extra bandwidth of HSDPA.

The network will still allocate the user HSDPA resources - but only when available and only when "needed".

My modem for example - camps on 3G. If I send pings - I can actually see that the first many pings are sent while the modem is still on 3G. (By both the light and times). After a bit of time the modem jumps to HSDPA and the light and times change. The time duration observed between first packet and jumping to HSDPA could be anywhere from parts of a second to many seconds. (how busy the cell is)

This jumping back and forwards from 3G to HSDPA kills interactivity; and as a result if I require interactivity I leave ping running in the background. When I'm using the VPN the keep-alives do the same job - keeps the HSDPA link active.

I live in a semi-rural area, the nearest 3G base station (100yds away) is an 11metre high "lamp post" (and I'm probably within the nearest 10 houses), then the next nearest is a bigger installation on a proper tower about a mile away across open fields plus a few trees. Given the ratio of the distances my intuition is that it'd be unlikely to be doing much with the further base-station - but I'm not familiar with the real deep down details of 3G. (I have read significant chuncks of the ETSI 2G GSM spec!).

What I'm about to try to explain depends on how Vodafone structures it's macro (coverage) and micro (capacity) radio networks; how the NodeB is physically configured (DSP resources & number of antennas) and how the RNC is configured and optimised - so it may well be different. Its also been a couple of years since I've been deep in the access side of wireless.

The uplink signal in your situation would be received by both towers. So 3G allows the codes used to separate subscribers and assigned by the network to your modem to also be assigned to more than one cell (actually DSP resource). Multiple cells per tower; one or more antennas per cell; providing an uplink system gain. The network (RNC) can then decide (in the simplest case) to use the best "quality" (error free) data block received from any of the cells. This is typically used to increase the distance the modem (Ue) can still successfully communicate for a given transmit power level. (Very useful to conserve phone/laptop batteries)

The downlink signal "could be" split between multiple codes - which can then be distributed to one or more cells. Therefore, increasing the total power available to you within the air interface - more power allows less protection bits and therefore, faster overall bit sending. (Well kinda)
The same can happen for the uplink signal - using multiple codes but you only have one low power transmit source being the modem.

The record-breaking "lab" speeds you hear about for 3G/HSDPA download speeds are from using all the DSP resources for one subscriber; all the possible air interface resource (transmit power) with the cell load at a whisker below run-away; with a perfect direct line-of-sight path and non-moving modem (Ue); delivering a very constant packet stream; from multiple cells.

So its quite possible that you are sending and receiving from both towers.

However, depending on how high the antennas are and how they have been installed/set being too close can be just as bad as being too far away as you could be in a shadow.

RF is also a black art - it is also totally feasible that the exact place the modem is currently sitting is suffering from a super-position null for the signal from the closest cell - meaning that the signal strength could be stronger for the far cell.

So position can matter - and signal strength is not the same as signal "quality". What the modems/phones use to provide the depiction of signal strength can also differ.

The Huawei E220 follows the 3GPP spec and using specific ATT commands will give you a signal receive level which can be converted to dBm. I know the E220 doesn't provide BER values - but I have not looked to see if it provides noise margin values.

It would be fascinating and revealing to see the total site loading in real-time 🙂 but without being a Vodafone employee that's not going to happen!

I've only ever seen animated models. Cell loading data and operator optimisations are many times closely guarded secrets. But yes it would be fascinating to see in real-time; especially network wide.

If you had a modem that provided signal levels; current power transmit levels; BER values; noise margin values; you could deduce how busy your surroundings actually are. That would be quite accurate.

However, if you ever get your hands on a mobile with test software or network monitoring software - the phones keep track of how busy the surrounding cells are internally and on at least the Nokia phones with test software will list all reachable surrounding cells and the current measured noise values. The higher the noise typically can be translated to busier the cell.

[SurfSilver]

I also have noted the reported behaviour in recent time (light/dark blue alternation, download speeds below the range I'm accustomed to seeing, plus the occasional disconnect) - but I can't put a date on it with any confidence and it's only now that Techmind and Heady have given so copiously of their wisdom that I'll be watching out for it.

Home postcode is LE4 8LJ, I quite frequently see download speeds in excess of 2Mbps - sometimes sustained over several minutes, with transients right up to 2.7 or 2.8 Mbps - and the basestation ID I'm normally registered on is ID 50349 (I think). Speeds (and usage) are monitored by Hootech Net Meter. Can't do too much investigation right now though as I'm currently out in the boondocks of LE67 in a GPRS-only coverage area (oh well, green makes a change from blue, whether light or dark).

The dongle is an E172. Don't suppose there are too many BER or other diagnostics I can do with that?

Anyway will be back on home territory tomorrow evening onwards and will report back if I see any further oddities.

Thanks again to those two wise men for taking the time and trouble to write this up.

[techmind]

The response to AT+CSQ is
+CSQ: 17,99

and I gather signal level in dBm = (rssi*2)-113
= -79dBm

If I string the modem up a bit higher (I'm in a ground floor flat so somewhat limited) I can get 19,99, i.e. -75dBm.


I should have a laptop to play with later in the week, in which case I can go into the garden to where I can see the top of the nearest mast (it's through one building from where I am now) and see how strong a signal I can get from there. I could also write my own little program to monitor the status chatter from the E220 rather than having to watch a crude terminal display 🙂

Maybe for debugging purposes I should take up VF's offer of a newer modem. Doubt they'll promise any particular model though, and I don't want to commit to another six months and get one which isn't an improvement on what I've got!

[techmind]

If I clip the modem to just the right spot on the curtains I can get an RSSI of 22, i.e. -69dBm. (or even 25, i.e. -63dBm)

Heady, should -79 to -69dBm be "plenty" for good high speed data? VMC shows 3-4 bars at comparable strength.

And would that be consistent with a signal from a microcell 100 yards away (but perhaps sheilded by a rather large house)?
And if the modem is potentially receiving from more than one base station, which figure is reported? (the strongest, I guess?)


BTW Heady, you didn't used to work at a company with postcode xxx5HA at some point did you???