Anyone tried new HG612 firmware?

As title, there is a new firmware out with a newer driver version for the BC chipset.

It's meant to perform a bit better.

http://huaweihg612hacking.wordpress.com/

I have tried it and got a lower sync

79999 to 79987 and no change in attainable rate.

the graph looks good for people on longer D-sides.

Edited by lockyatlrg (Thu 11-Apr-13 12:59:03)

Hmm, all a bit scary and I don't really want to be one of the first to try it! The difficulty is knowing what the modem would have resync-ed at had I just switched it off and on again - it my case it never connects at the same speed anyway.

I tried it for a few days.

My sync speed went from 26003 kbps / 5844 kbps to 26712 kbps / 5792 kbps.

Reverting to the original unlocked firmware as a test put sync speed back down to 26007 kbps / 5701 kbps.

I'll test it again in a few days to see if DS sync speed increases again.

It seems to move DS bitloading out of the slightly weaker 'Shared/Other' tones into specific DS tones.

Maybe it's of more use for users on longer connections like mine (around 1000m).

I tried it on a 2B and 3B modem (both same anyway)

It improved my upload about .5mbit (attainable rate)

People on lines syncing at 79999/20000 I wouldn't bother This will make no odds to you.

People on longer line, could be interesting.

Damn! Look at the HG622s he had. All sold .

I can't see any mention of it but I would use it if the attenuation bug is also fixed.

Yeah I would have had one of those.

In reply to a post by RobertoS:

Damn! Look at the HG622s he had. All sold .

Well they were all sold over 2 months ago so you didn't only just miss out!

In reply to a post by lockyatlrg:

I tried it on a 2B and 3B modem (both same anyway)

It improved my upload about .5mbit (attainable rate)

People on lines syncing at 79999/20000 I wouldn't bother This will make no odds to you.

People on longer line, could be interesting.

I'm on maximum upload speed anyway but download is below max, been synced at 68014 for the last month. I might give it a try next week.

In reply to a post by kasg:

In reply to a post by RobertoS:

Damn! Look at the HG622s he had. All sold .

Well they were all sold over 2 months ago so you didn't only just miss out!

I hope someone reports back on how they perform! (Preferably in a dedicated thread).

or just dodgy D sides.

your attenuation is higher then mine but have a 20mbit higher attainable. Fun and games on the BT lottery

I might try it out I guess see what happens.

On my line the HG syncs higher than everything else I have put on it, but the fritz (which matches dslam chipset) has about 30-40 errors a day vs about 1000-1200 a day on the HG. (on fast path).

Post deleted by Finguz

asbokid posted on another forum thread this morning that hopefully he should be getting more

I have now tried the new firmware and here are the results:

Before: 68014/20000, attainable 63196, SNR 4.3dB
(this had been stable for about 6 weeks)

After: 71606/20000, attainable 72736, SNR 6.3dB

This is good, but fairly inconclusive as there is no way of knowing what the modem would have synced at if I had just powered it off and on. I'll see how it goes.

TBB speed test
speedtest.net speed test

IP profile should be 69.3 Mbps, but the BT speed tester wouldn't tell me due to a "problem". My Plusnet profile is 72.4 (lucky) so will not be restricting speed.

Edited by kasg (Tue 16-Apr-13 10:11:20)

Now I'm back off holiday I'll flash my spare modem and give it a try sometime soon.

In reply to a post by R0NSKI:

Now I'm back off holiday I'll flash my spare modem and give it a try sometime soon.

Good idea, ISTR your speeds are not as good as they perhaps should be for your line length. I'm about 450m from the cabinet so am doing pretty well (estimate is 58.1).

You remember correctly, I'm also 450 meters from my cabinet, but only get a sync of around 40 to 45 with an attainable of about 52 IIRC.

So it will be interesting to see if it makes any difference.

Out of interest, what's the line like between you and the cabinet? Mine is entirely on old telegraph poles (apart from the nearest one to my house, which is fairly new and serves only my property; it was erected just to ensure sufficient clearance over the electricity cables from the pole on the other side of the road).

Edited by kasg (Tue 16-Apr-13 10:29:19)

In reply to a post by kasg:

I have now tried the new firmware and here are the results:

Before: 68014/20000, attainable 63196, SNR 4.3dB
(this had been stable for about 6 weeks)

After: 71606/20000, attainable 72736, SNR 6.3dB

This is good, but fairly inconclusive as there is no way of knowing what the modem would have synced at if I had just powered it off and on. I'll see how it goes.

That's a fair improvement!

My line (400 metres) is rather like yours - the SNR has gradually dropped below 6 over the last few months, so I'd probably get a lower sync if I rebooted. It is sync'ed at 80, with an attainable of 77ish.

It'll be another few weeks before I can re-flash one of my modems, but I'll try to remember to reboot to get new stats before I make the change.

whilst your sync may change just from a power cycle, its not normal for the attainable to jump around, so your result is certianly interesting.

I just flashed it.

Attainable reported slightly higher but actual sync is about 400kbit down.

Edited by Chrysalis (Tue 16-Apr-13 13:02:10)

In reply to a post by kasg:

Out of interest, what's the line like between you and the cabinet? Mine is entirely on old telegraph poles (apart from the nearest one to my house, which is fairly new and serves only my property; it was erected just to ensure sufficient clearance over the electricity cables from the pole on the other side of the road).

Mines all underground, and it may not even be in ducting, the estate was built in the mid seventies and some bright spark thought to save some money around that time they'd stop using ducting! Talk about short sighted!

and not forgetting that in the 1970's BT was in public ownership and was called the GPO.

So yet another example of how being in public ownership does not mean that the job was done to top quality standards...it was still a case of minimise the expense even back then.

In reply to a post by kasg:

This is good, but fairly inconclusive as there is no way of knowing what the modem would have synced at if I had just powered it off and on. I'll see how it goes.

Have you noticed any differences in your bitloading patterns?

I have tried the new BLOB twice now & each time there was a DS sync speed increase of just over 1Mbps on my low speed connection & DS bitloading seems to have moved from the 'Shared/Other' tones to slightly higher frequency tones:-


BLOB Comparison Graphs

Edited by deleted (Tue 16-Apr-13 18:47:42)

Well I do live in a new build with nice quality copper.

I move in 2 weeks to a bigger house on the same estate which just so happens to be 10 meters from the cab

In reply to a post by Bald_Eagle1:

Have you noticed any differences in your bitloading patterns?

TBH I haven't looked at them for a long time, once a connection is stable you kind of lose the incentive! I'm still collecting stats every hour or so using the original scripts so I might check it out.

In reply to a post by Chrysalis:

whilst your sync may change just from a power cycle, its not normal for the attainable to jump around, so your result is certianly interesting.

My attainable and SNR do tend to drift down over a period of time, but so far it's held up.

In reply to a post by Chrysalis:

I just flashed it.

Attainable reported slightly higher but actual sync is about 400kbit down.

Did you reboot immediately before flashing as well, i.e. would it have gone down (possibly by more) anyway? I didn't as I didn't want to upset things.

Edited by kasg (Tue 16-Apr-13 19:04:45)

yeah and it shows the huge variances.

I got 9.4 attenuation (similar to yours). With current sync of 65.9 on fast path, when I was on interleaved last I had a sync similiar to robertos and his line is at least 200m longer than mine.

I hope that cabinet near your new house is yours, I have a cabinet across the road from me, it isnt mine and initially I didnt care as I started off with great line stats, obviously I do care now tho.

On this new firmware whilst the sync speed boost didnt happen for me, I can report the error rate on the downstream has dropped significantly since I changed the firmware. For the first time on the HG I have less downstream errors than upstream on fast path. Still nowhere near as low as the fritz 3370 but is defenitly lower errors than the other hg firmware.

I didnt reboot no because of DLM.

I dont know what the disconnect trigger points are, and whilst I am ok with doing a single disconnection, I dont feel easy doing two in one day and especially close together.

The modem flashed and rebooted itself very fast I thought, downtime was barely 20 seconds. Painless process and didnt lose my ip either.

a 400kbit shift in speed is basically well within the realm of normal variance so yes its effectively no change. I have noticed now my line is always at a snrm above 6.5 (even right after a sync event) as if my target snrm has been raised slightly.

Edited by Chrysalis (Wed 17-Apr-13 04:49:28)

In reply to a post by Chrysalis:

The modem flashed and rebooted itself very fast I thought, downtime was barely 20 seconds. Painless process and didnt lose my ip either.

In that case, I suspect the firmware hasn't been fully updated.
A long reset whilst powering up the modem is required for it to take full effect & the modem should also reboot itself at the end of the upgrade process.

Altogether it should take around 2 minutes, being very unlikely for the IP address to remain unchanged (unless you are using a static IP address).

Do you see this in the device page?:-
Firmware verision A2pv6C035m.d22g

a 400kbit shift in speed is basically well within the realm of normal variance so yes its effectively no change. I have noticed now my line is always at a snrm above 6.5 (even right after a sync event) as if my target snrm has been raised slightly.

If the upgrade hasn't taken full effect then yes, you may well be seeing 'normal' variance here.

you suspect wrong, it was fully updated, just the hg reboots very fast.

It did reboot, it doesnt go offline during the flashing, so downtime is just for the reboot which is fast.

I did the emergency reset procedure but still same, sync speed within normal variance (this one did lose my ip as too long), In both cases the firmware version is new in the status page.

Edited by Chrysalis (Wed 17-Apr-13 09:07:08)

Just updated my spare modem, and my speed seems to have dropped by about 500kbps

12:00 midday today on original unlocked firmware

Max: Upstream rate = 10592 Kbps, Downstream rate = 51100 Kbps
Path: 0, Upstream rate = 10461 Kbps, Downstream rate = 42588 Kbps

19:30 with updated firmware

Max: Upstream rate = 10314 Kbps, Downstream rate = 51116 Kbps
Path: 0, Upstream rate = 10085 Kbps, Downstream rate = 42075 Kbps

And this is a line that's only ~450m

Seems my interleaving has shot up to 2713? I did have a resync this morning that went from depth of 709 to 1599, now it's 2713, never seen it that high.

All stats can be viewed here

its best to reboot the modem about dinner time, seem to get best sync then

I realise that but I'm at work during the day, and I doubt it'll make much difference on my line.

In reply to a post by bobbyboyuk:

its best to reboot the modem about dinner time, seem to get best sync then

Is that northern dinner time or southern dinner time?

That is rather a high interleaving depth... I'm about 1000m from cab and only have a 362 depth (old firmware). I also have 150m worth of ali on my line as well... I will update mine at some point tonight probably and post the results. Though I've disconnected my line a lot recently so it probably won't be accurate anyway.

It's best to restart the modem when the SNR is higher than 6db, this is typically around mid day though in rare cases you can get lower during the day instead of night but is very rare.

Edited by deleted (Thu 18-Apr-13 15:39:49)

That reply was obviously intended for R0NSKI. I don't have any interleaving at the moment on a 450m line, but I have had it up over 1000 in the past.

Just put the new firmware in...

Sync lost 400Kbps (up and down ), attainable reates up to 30Mbps from 26 - 28Mbps.

My interleaving depth has doubled to 734, since install it has sat in the 300 - 400's.

Also is the Attenuation still broken in this Firmware version?

It seems unlikely that, in the time taken to upgrade the firmware, DLM has decided on a new requirement for interleaving (ie the INP and max-delay values). At least unlikely for it to happen to multiple people.

More likely is that the INP and delay values have stayed the same, but that the new Blob chooses a different way in which to achieve interleaving and the associated FEC.

So, rather than just concentrate on the interleaving depth, it is probably better to see the full set of parameters:
- INP
- delay
- R
- D
- I
- N

The figures chosen by DLM are INP and delay. The modem takes these figures, and chooses the best way to achieve this... then sets R, D, I and N.

Interleaving works by filling blocks of size (D * I), so it is possible that I (capital i) has also changed.

FEC works by putting R bytes of parity overhead in blocks of size N.

It'd be interesting to see the interplay between all these values.

B: 34 161
M: 1 1
T: 64 5
R: 8 16
S: 0.0445 0.9965
L: 7732 1429
D: 734 1
I: 43 178
N: 43 178

I know the delay has always been 8, not taken note of I, N, or R in the past so maybe I will flash the firmware back in a couple of days. The INP is at the same as wel 3.25, care to explain what a few are for me as well please?

It's worth noting I can see 100, 000 FEC a min for hours on end though this normally ranges from 10k - 30k but in rare cases goes up to 100k. When I restart the connection it will go from 0 - 3k then after days just go up and stay there continuously without bursting.

Edited by deleted (Thu 18-Apr-13 19:22:16)

In reply to a post by ryant704:

I know the delay has always been 8, not taken note of I, N, or R in the past so maybe I will flash the firmware back in a couple of days. The INP is at the same as wel 3.25, care to explain what a few are for me as well please?

"delay" is easiest: Whatever interleaving is to be added, it is allowed to add only 8 milliseconds to the latency. You'll see ping times increase by this.

"INP" stands for impulse noise protection, and the value tells the modem how many "symbols" must be protected. This means that FEC must be able to self-correct a burst of noise that was long enough to corrupt 3 symbols, but doesn't need to be able to fix a bigger burst. (A symbol is a way in which the modem encodes digital data into an analog form, and usually encodes a few bits at a time.)

Generally, FEC can self-correct only a small amount of data, and a corruption of 3 symbols will be more data than can be corrected if it were all sent (and corrupted) in one consecutive batch. Interleaving is used to spread the bits around, so that when a corrupted section of interleaved data is (later) de-interleaved, the "single large corruption" is spread into "lots of small corruptions" in many different FEC blocks... each of which is small enough for the separate FEC processes to deal with.

So your modem has chosen the FEC blocks to be size N (43 bytes), of which R (8 bytes) are used as the overhead to help self-correct corruptions. This means 19% of your bandwidth is being used as overhead in case of noise.

If there is a fault in one of those 43 bytes, the modem can work out what the right value is.

The interleaving blocks are of size (I *D) or 734 * 43 bytes, or 31,562 bytes.

The modem can choose whether to offset speed against latency. Generally, the modem can attain the same level of protection by both increasing the FEC overhead size (so decreasing speed) and decreasing the interleaving size (and decreasing delay), or vice-versa.

It's worth noting I can see 100, 000 FEC a min for hours on end though this normally ranges from 10k - 30k but in rare cases goes up to 100k. When I restart the connection it will go from 0 - 3k then after days just go up and stay there continuously without bursting.

I don't follow this part. Are you saying that the number of FEC increases at a pretty steady rate, or that it doesn't change after the initial burst?

yeah it's on the same cab.

I don't really get much errors.

0 on upload and a couple of HEC on download, about it.

Send you a PM with a bunch of images.

From 12 midday before changing firmware

INP: 6.00 0.00
delay: 15.00 0.00
R: 14 16
D: 1599 1
I: 62 255
N: 62 255

From 12 midday after upgrading firmware

INP: 6.00 0.00
delay: 15.00 0.00
R: 8 16
D: 2713 1
I: 37 255
N: 37 255

In reply to a post by R0NSKI:

From 12 midday before changing firmware

INP:		6.00	0.00
delay:	15.00	0.00
R:		14		16
D:		1599	1
I:		62		255
N:		62		255

From 12 midday after upgrading firmware

INP:		6.00	0.00
delay:	15.00	0.00
R:		8		16
D:		2713	1
I:		37		255
N:		37		255

That's it precisely. Same input parameters, but a different way of achieving the result has been chosen by the blob.

The old interleaving is using a data matrix of 1599 width * 62 height, or 99 kbytes.

The new interleaving is using a matrix of 2713 width * 37 height, or 100 kbytes.

The FEC processing has then altered, so it works on blocks of size 37 instead of 62. The parity bytes added as overhead have changed from (14 out of 62) to (8 out of 37), improving the overhead slightly from 23% to 22%.

The result is that you have a 1% more interleaving (and probably 1% higher latency), but a 1% less FEC overhead (so probably 1% higher speed). A very different result from when you were looking at just the depth value.

However, the overall result you saw was a slight drop in speeds - but that is probably explained by different noise levels and bitloading at sync time, masking the impact of the FEC/interleaving changes. Perhaps you should try a resync at a decent time of day at the weekend...

BTW - those are hefty input requirements. 6 symbols & 15ms are both rather high!

In reply to a post by lockyatlrg:

yeah it's on the same cab.

I don't really get much errors.

0 on upload and a couple of HEC on download, about it.

yeah clearly the D side in your area is better built, whether its better twisting, lower density pair bundle, better joints, thicker cable or even your area having very low takeup its clearly better.

I wonder if BT can detect these changed BLOB versions on their side.

Latency went up quite a bit though, graph for the 17 April.

The one at 6am was the DLM triggered resync, not sure what happened at 2am though, my stats don't show anything unusual.

I'll try a resync at the weekend and we'll see what happens.

Edited by R0NSKI (Fri 19-Apr-13 06:32:28)

In reply to a post by R0NSKI:

The one at 6am was the DLM triggered resync, not sure what happened at 2am though, my stats don't show anything unusual.

The 6am jump in latency would look to be under 10ms, so would be consistent with DLM changing your "delay" value from a "more normal" 8ms to 15ms (presumably the "INP" value changed at the same time), and your interleaving depth changing as a result.

Did your stats show that sync was maintained normally at 2am?

I had an hour-long outage at 2am (just after 2am to just after 3am) on the 17th too - sync stayed, but PPP dropped, then failed to re-establish. That was on my ADSL line down south, rather than my FTTC line. When I rang Plusnet (Yes, I was using it at the time) they mentioned that there was some planned outages from BT, but that it seemed small enough that they hadn't mentioned it in their service status.

When I searched elsewhere, it seemed to be planned engineering work, with a BRAS upgrade having been planned in Guildford. I imagine that could have affected quite a widespread area - Guildford is one of the core 21CN locations for the South East, isn't it?

Latency went up quite a bit though, graph for the 17 April.

Around 2ms, I guess? A little more than I'd have expected from just the interleaving change alone.

The attenuation 'bug' is still in this firmware.

In reply to a post by WWWombat:

Did your stats show that sync was maintained normally at 2am?

Yes my stats showed everything was normal at 2am, so I guess it was related to the outages you mention. No idea if Guildford is a core 21CN location or not.

In reply to a post by WWWombat:

Perhaps you should try a resync at a decent time of day at the weekend...

Just turned modem off for about 15 minutes, these are the values now.

INP: 6.00 0.00
delay: 14.00 0.00
R: 8 16
D: 2713 1
I: 37 255
N: 37 255

And a very slight increase in speeds

Max: Upstream rate = 10269 Kbps, Downstream rate = 50979 Kbps
Path: 0, Upstream rate = 10212 Kbps, Downstream rate = 42340 Kbps

I don't know, I know it's 0.5 copper and the estate was built in 2005 and the cab PCB and wiring shortly after.

I know the cab population is 58%, BT ELC told me when I had a fault (on BT's core network)

There is 3 different routes cables take here so the bundles are not all in the same duct.

Not sure how old my part of BTs network is, alot of the properties round here were built in the 60s and 70s but not all were. It seems logical BTs network around here would have existed before then.

My cab is 3km from the exchange and my d-side is a 1.5KM.

Having done several tests (I have two hg612s) the newer firmware reliably reduces the downstream attainable rate by approaching 2Mbs and the actual rate by about 500Kbps.

It does add around 50Kbps to the upstream. My upstream is very slow, badly affected by d-side PBO.

I have reverted to the previous version.

Note I had to used the "long reset" menu to load the firmware, it wouldn't apply when loaded from the regular one.

when you compared the attainable was the SNRM also lower? if not it means its simply estimating lower. My view is the hg612 over estimates attainable anyway.

Whats PBO?

Under same conditions for me I got

Old Firmware
25Mbps DS
5.3Mbps US
Attainable: 26/27Mbps DS 5.3Mbps US

New Firmware
25.8Mbps DS
5.1Mbps US
Attainable: 29/30Mbps DS 5.1Mbps US

Normally synced around 5.5/5.6 US so lost a little upload

Thought I'd check my stats, had a resync this morning, and to my surprise I seem to be on fast path now

Max: Upstream rate = 10177 Kbps, Downstream rate = 47197 Kbps
Path: 0, Upstream rate = 10014 Kbps, Downstream rate = 47197 Kbps

INP: 0.00 0.00
Delay: 0.00 0.00

R: 0 16
D: 1 1
I: 240 255
N: 240 255

I changed the firmware on the 17 April.

Edit: Only problem is my Plus Net line profile hasn't caught up - I'll give it a few days to see if it changes.

Edited by R0NSKI (Mon 06-May-13 14:00:14)

Ronski, is your US speed normal at about 10Mbps?

I've been on the latest firmware for 4 weeks and have just had a re-sync which took my US from 16Mbps to 11Mbps . Not the end of the world but I have just gone back to the old firmware to see if it recovers in the next few days.

My DS is around 52Mbps which changed only slightly with this firmware.

Strange how my US has suffered but of course may be nothing to do with this firmware!

Yes my upstream speed is usually around 10Mbps. I don't think it is anything to do with the firmware, my upstream halved back in January, and at the same time my download increased, but eventually things went back to how they were.

I've not been on fast path (no interleaving) since last September.

You can view all my graphs and stats in Dropbox.

I'll generate some graphs for the entire 291 days I've had my FTTC connection in a moment, and then upload them.

you check stats right before flashing and after?

need to know more than the sync to have an idea whats going on, snrm, interleaving depth of US, and ideally before and after bitloading graphs.

the huge change in sync speed indicates line conditions changed which would make me guess your US had a low snrm prior to you changing the firmware.

I have posted images below of my line history that I have. Hope some of this help explains my drop in US - if I've managed to capture the relevant info

Stats Prior to Firmware upgrade

Stats just after firmware upgrade

Current stats - returned to original firmware post US drop

All Graphs from April 2013

All Graphs from 6th May 2013

your snr margins were low for both up and down prior to the first flash.

3.4 down 4.1 up.

this suggests you had a loss of dsl signal strength since the sync was originally established and with a target of 6db means even if you just did a modem reboot without touching the firmware you would lose sync speed.

The modem at that time you did the snapshot still estimated an attainable upstream rate of 16mbit vs the 18.6mbit sync but attainables are only estimated. Your downstream was also estimated to lose sync speed of attainable 50mbit vs the sync of 56mbit.

your stats after the upgrade show the 16mbit upload sync matching the previous attainable, and a 54.8 sync on the downstream.

So to me it looks like nothing changed on the upstream from the firmware (was due to line conditions), and you got lucky that the downstream gained 4mbit or so else you was looking at a 50mbit sync on the downstream.

However you then flashed the firmware back and lost a chunk of upstream sync speed. I dont see immediatly why thats happened. there is a chance power cutback has been applied at the dslam. Its not due to FEC or banding.

Edited by Chrysalis (Tue 07-May-13 04:05:32)

Thanks Chrysalis for your comprehensive reply. Will have to see what happens in due course now, does seem odd though that DLM has stepped in when the stats were looking ok. Oh well!

Ive just flashed mine,
Given me 3mbits extra down and 1 up..
I have quite a long line, and i had a lot of errors before, none yet!

In reply to a post by R0NSKI:

Thought I'd check my stats, had a resync this morning, and to my surprise I seem to be on fast path now

Even more surprising is that happened to me a few days ago (original BLOB) - also with Plusnet, but on a much longer/poorer connection.

I'm only at around 25Mb sync speed & have only ever previously seen Interleaving switched off when DLM has been reset (35Mb sync speed) & that only lasted for a few days.