Log in

View Full Version : More Vodafone HSDPA testing



Ben
7th September 2006, 08:43 PM
Well, much to my surprise, I'm currently in a Vodafone 3G area that has HSDPA enabled. It's nice to see they've expanded beyond the central London area.

It's good news, very good news. I'd already told you that HSDPA looked promising, but the fact that I'm currently using my Vodafone 3G Broadband card over a WiFi connection should really indicate to you the quality of service. As you're about to see, the speeds I'm getting are better than the ADSL speeds most people in the UK are signed up to:

Downstream 1,430.0 Kbps ( = 1.4 Mbps )
Upstream 353.6 Kbps ( = 0.3 Mbps )

I am seeing some handovers between HSDPA and standard 3G, but the connection still seems to be stable. I only have a couple of bars signal, so this performance is very promising indeed. If I had a laptop with integrated HSDPA, where the antenna were built into the screen, I'd expect the performance to be even better.

Latency pinging www.google.co.uk is about 100ms. This is quite a lot worse than ADSL, but it's far far better than anything I've experienced over cellular networks to date. It's in the area where games become playable and telnet and remote desktop become usable.

Ben
7th September 2006, 11:01 PM
HSDPA mode is indicated on my card by the blue and green LED's forming a sky-blue colour. Very cool. Anyway, it is becoming very clear that HSDPA disactivates after a few minutes of being idle and then kicks in again when I start to use the connection. Very curious! I wonder why on earth this happens?! It doesn't affect me in the slightest, but I have to wonder - is there some sort of overhead to having a HSDPA user active on the network?

bsrjl1
8th September 2006, 09:16 AM
A cell can only handle 16 HSDPA users & there's no point your card wasting power listening to the HS-DSCH & regularly uploading measurements when you're idle.

So after a couple of timers expire you'll drop down to a Cell-PCH state (if the Ericsson RAN supports it), where the card just listens out every so often for any data, and as soon as you start doing anything more than keep-alives or pings, you'll jump back to a DCH then quickly back to the HS-DSCH.

Ben
8th September 2006, 12:24 PM
Interestingly even a ping is enough to get it back on HSDPA. As it times out very quickly, about 5-10 seconds, and as there's a slight delay of about half a second while it hooks back up, I've taken to leaving a ping running.

A cell can only support 16 HSDPA users? That's outrageous!

bsrjl1
8th September 2006, 01:38 PM
They must have their threshold set very low then, or you're using very large pings? ;-)

How often will 16 HSDPA users turn up in the same place & all be on the same cell?! Sounds reasonable to me, considering a cell can normall only support 3*384kbps connections.

Hands0n
8th September 2006, 06:58 PM
Funny you should ask that - my local 50-member HUG (HSDPA Users Group) often converge on a local mast now that the darker evenings are closing in :p

Ben
8th September 2006, 07:37 PM
Lmao! ¬_¬

I'm just thinking to the future - if HSDPA were to become the connection method of choice for laptops then they'll need to be able to support more than 16 in a cell. I guess it's not as unreasonable as I first thought though, especially if a cell can only support three 384kbps connections without HSDPA!

Hands0n
9th September 2006, 08:00 AM
IF the Mobile Ops get that yellow streak off their backs and sort out the Data Tariffs I reckon that [within a couple of years] there won't be a mobile computing device that does not have HSDPA/HSUPA integral. I observe how [firstly] Bluetooth and now WiFi are becoming almost ubiquitous in such devices, and increasingly so as we move forward.

Maybe I'm not fussy :rolleyes: but I'm reasonably impressed with receiving bog-standard 3G's 300Kbps+ to a mobile device. But maybe that is because I recall only too well GSM's 2.4Kbps when mobile data first became a [hideously expensive] reality. At that time we had established a small number of mobile users with tn3270 access to an IBM Mainframe through some quite convoluted mobile dial-up means (the MF boys wouldn't hear of rigging the Big Blue box out with an IP stack and an Ethernet connection).