Broadband Related > FTTC and FTTP Issues
Upstream SNRM/bitloading weirdness
GaryW:
Hi all,
I'm seeing a weird inconsistency between the reported upstream SNRM and what I see when I look at SNR/tone and bits/tone. I'm using a Billion 8900AX-2400 and I'm on an ECI DSLAM.
My modem is only using US0 (tones 6-31) - probably due to the long line. The SNRM in the basic stats is around 6db, but when I look at the SNR/tone and bits/tone data it shows around 10-12db margin for each and every tone... Overall I'm only getting a little over 500 kbps upstream...but if the "missing" 2 bits were added to each tone I'd be getting somewhere over 600kbps.
What's even weirder, is that when I used to use a (Lantiq-based) BT HH5a I did get over 600kbps, but with every Broadcom based modem I've used (BT Smart Hub, TP-Link VR2600, and now the Billion) I only ever get a little over 500kbps!
Has anybody seen similar behaviour? Would switching to a Lantiq-based modem restore the upstream performance? (but not the HH5a as it's lacking in other ways and downstream performance was much worse than the broadcom modems)
You can see my stats on MWDS, username GaryW.
Thanks,
Gary
roseway:
Welcome to the Kitz forum.
The 'inconsistency' is actually quite consistent. The SNRM is the SNR margin but the SNR per tone graph shows the actual SNR for each tone. SNRM is the difference between the actual SNR and a base value which is defined in terms of the error rate expected at that level of SNR.
GaryW:
--- Quote from: roseway on December 13, 2017, 10:43:05 PM ---Welcome to the Kitz forum.
The 'inconsistency' is actually quite consistent. The SNRM is the SNR margin but the SNR per tone graph shows the actual SNR for each tone. SNRM is the difference between the actual SNR and a base value which is defined in terms of the error rate expected at that level of SNR.
--- End quote ---
Hi,
It does indeed show the actual SNR, and if you subtract the 3db per bit you should be left with the SNRM for that tone...which is where my figure of 10-12db SNRM for each tone comes from. In contrast, for downstream the SNRM per tone is fairly consistent with the overall stated SNRM. Trust me, if you look at the graphs there's a huge gap above the bitloading for each tone on the upstream...whereas for downstream there's a 6ish db gap above most tones. (It's hard to see directly on MWDS as you can't overlay the graphs, but on dslstats it leaps out of the page).
roseway:
--- Quote ---if you subtract the 3db per bit you should be left with the SNRM for that tone
--- End quote ---
I don't know where that 3 dB figure came from, but it's a false assumption. In fact you can't sensibly talk about SNRM on a tone by tone basis - as I said in my previous message, the SNRM base level is defined in terms of the error rate for the connection as a whole. The SNRM base level differs between different modems and different connections. Just for comparison you can see my bitloading graph with SNR per tone superimposed. The downstream gap is about 2 dB, and the upstream gap is about 5 dB.
GaryW:
Hi,
Well, that 3dB figure came for a number of places, including this web-site! It's also confirmed by the chart that you included in your post: compare the left-hand scale (bits) and the right-hand scale (dB) and you'll see that each bit corresponds to 3dB.... For the SNRM for each tone you've actually quoted the number of bits (left-hand scale) rather than the dB (right-hand scale). Your graph shows a typical SNRM for downstream tones of 4-6dB (and according to MWDS your downstream SNRM is about 4dB; it also shows that the typical SNRM for upstream tones is around 12dB (and according to MWDS your upstream SNRM is about 10dB).
In other words, your upstream and downstream make sense...whereas my upstream doesn't (I've got a similar upstream SNRM per tone to yours on my graphs, but my modem is saying my overall SNRM is about 6dB rather than the 10dB that yours is saying).
I notice that you're on a Huawei cabinet with a Broadcom modem, whereas I'm on an ECI cabinet with a Broadcom modem. Hence my original question - will I get some of the "missing" bits back with a Lantiq chipset?
Navigation
[0] Message Index
[#] Next page
Go to full version