openbts/GPRS/notes.txt

7-13, uplink.persist=3000, 1 channel: speed 11.7 latency 1.8 
	uplink.persist=3000 3-down/2-up: speed 46.2 latency 1.3
	uplink.persist=3000 4-down/1-up: speed 41.5 latency 1.4
	again uplink.persist=3000 4-down/1-up: speed 27.8.5 latency 1.1
	again: uplink.persist=3000 4-down/1-up: speed 38.1 latency 1.3

Fri Jul 13 04:52:16 PDT 2012
7-11: Multitech modem 4-up/1-down.  This is after fixing the downlink USF bug.
	These are with T3192Code=2 (1500ms)
	downlink speed from mobilespeedtest.com: 22Kbps
	In these there is downlink traffic on the same shared PACCH:
		uplink ftp: total 1023652 bytes in 324s = 25Kbps
		Raw (from looking at a single tbf, 3062B in 0.9s) 27Kbps
		U412 96.5-93 4624b = 11Kbps.  Yuck.
		uplink ftp 50025 in 15s = 26.7Kbps.
	for 3-up/2-down: uplink ftp 50025 in 30s = 13Kbps.
	for 2-up/2-down: uplink ftp 50025 in 18s. second test: 18s = 22.2Kbps
	I checked the log, and we are definitely getting blocks on all 4 channels.
	For the 3/2 test that was slow, it looks like the phone is having to RACH in to
	start nearly every uplink.  For 4/1 the downlinks are so slow that they are still
	running when the uplink ends.
	These are with T3192Code=0 (500ms)


7-5: 4 channel multislot, T3192Code=0, got 32.8 Kbps.
7-5: 4 channel multislot, T3192Code=3, got 25, and cause=105 errors (TBF killed by RACH).
	I suspect setting T3192 too high can force the MS to RACH for an uplink TBF.
	

7-4-2012:
	Tried ftp.
	download 6:40
	upload: 20 seconds "to calculate the time"
	then sent a 0 byte file.  Gotta love that.
6-27-2012:
	With new channel allocator that guarantees 3-down/2-up assignments:
	mobilespeedtest.com: 25Kbps, latency 2.2
	Had to set GPRS.ChannelCodingControl.RSSI to -50 to get it though
	Setting GPRS.Codecs.Uplink,Downlink to 4 got 25.6Kbps
	Setting T3192Code to 3 gives 29.4Kbps
6-21-2012:
	iphone 001690000000002
	iphone speedtest.com 20.6KBps, latency 2.6s 100K in 39s.
	The phone sent XID type 5 (set U and UI frame size) to 1520
	and XID type 11 (L3 params): LLC:LLC XID xidtype=11 xidlen=3 value=256;  LLC:LLC Sending XID command:03FB1605F02F000100
	After power cycling the bts,
	this phone does not accept the DetachRequest or PdpContextReject ImplicitlyDetached thing.
	if you manually request the operator again with the phone, it sends a RoutingAreaRequest with a new TLLI
	and after receiving the RoutingAreaUpdateReject does an immediate new attach.
	--
	This phone lost connection temporarily in log: iphone/1 around 07:54:46.
	An uplink TBF#127 was (correctly) cancelled, but I subsequently saw many 'received uplink data block after expiration'
	How is that possible?
	--
	After power cycling the bts, the carrier selection reported ATT and T-MOBILE only; tried 3 times.
	It did not do anything to gprs, so I think this is a normal registration failure.
	Then when power cycled, it did an immediate gprs-attach.
	When I did the "Carrier Selection" again, it showed only 1001, not ATT and T-MOBILE.

	Samsung GT-I5500 Baseband version: I5500NEJP1

Put gprs tests on the wiki.

6-18-2012: Davids galaxy error log:
	res 303714: TBF#170 uplinkacknack - received
	res 303715: TBF#171 downlinkassignment - unanswered
	res 303735: TBF#171 immediate downlinkassignment - received
	res 305577 TBF#177 sendass PACCH - unans (at 583)
	res 305585 TBF#177 sendass PACCH - unans (at 591)
	res 305593 TBF#177 sendass PACCH - unans (at 599)
	at 5598 res 305601 TBF#177 sendass PACCH - unans (at 607)
	at 5606 res 305618 TBF#177 sendass imm on TN1. 617 PDCH#51:1 has no usf.
	at 5623 res 305636 TBF#177 packetuplinkacknack

6-13-2012:
	After RLC modification to send only negatively acknowledged blocks,
	now at 22.2Kbps, latency=2.0, 100K in 36.1s.

6-11-2012:
	After fixing nstuck bug, setting nstuck to 1000, fixing the negatively acknowledged resend bugs:
	15.6Kbps, 19.8Kbps, 19.3Kbps

With 2-down/2-up I typically get 19Kbps.
Tried setting Poll1 to 30, got 11-13Kbps on two different tests
with 8 total cause=3 errors.
Restarted with Poll1 at 10 (default) got 19Kbps with one cause=3.
then 10Kbps with one cause=3.

Restarted, removing the the needy USF fix entirely,
setting Multislot.Max.Downlink/Uplink =1/1 in sql and turning on GPRS.WATCH.
	18.5Kbps
Turned off GPRS.WATCH 14.7Kbps with 2x3105 retries
Turned on GPRS.WATCH 19.7Kbps, did not see any retries.
Turned off GPRS.WATCH 17.5, no retries, then 16.7 with 1x3105
Turned on GPRS.WATCH 19.3Kbps, 16.3Kbps
Turned off GPRS.WATCH 12.6Kpbs, 17.6Kbps (no retries), 17.2Kbps.
I think it may depend on how long you wait between re-running the test,
ie, maybe there are still old tcp packets coming that bash the current test.

6-8-2012 GPRS-Beta-4 3-down/2-up with needy USF turned off:
	Your speed: 18.4
	Your latency: 1.8s
	Transferred 100KB in 43.5
Changing GPRS.USFMode (needy USF) didnt change a thing.
with 2-down/2-up: 13.9Kbps, 2.2s, 100KB in 57.3s, and it had alot of retries.
again: 13.5Kbps, 2.3s, 100KB in 60s

6-7-2012 with 3-down/2-up working multislot:
	Your speed: 19.8Kbps
	Your latency: 1.8s
	Transferred 100KB in 40.3

6-3-2012 with 2-down/2-up working multislot:
	Your speed: 20.78Kbps
	Your latency: 1.8s
	Transferred 100KB in 38.5s

mobilespeedtest.com:
6-3-2012 with 2-down/1-up buggy multislot:
	Your speed: 18.29Kbps
	Your latency: 2.86
	Transferred 100KB in 43.74s
	

5-22-2012, with continuous tbf, but the speed test stops for maybe 5 seconds right in the middle -why?

	Your speed: 7.08Kbps
	Your latency: 2.702
	Transferred 100KB in 113s


	google.com -  8 secs
	pats home page (www.menu1.org/pat) - 15 secs
	cnn.com - 35-40 secs  (note that this page is not invariant.)
	wikipedia dinosaur  1:52 

After bug fix, but it still hung half way through, but the hang looked correct (saw fbi) so whats up?
Your speed: 8.08Kbps
Your latency: 2.6
Transferred 100KB in 99s

After setting RLCMsgPacketDownlinkAssignment->mControlAck
Your speed: 11.17Kbps
Your latency: 2.8
Transferred 100KB in 71
wikipedia dinosaur 1:40

After fixing the mSP bug:
	wikipedia dinosaur 1:35

============