Setting a high SNR with the Billion 7800

I've been using a Billion 7800 in a G.dmt setup for some six months now - and generally speaking I've been quite pleased with the 7800's performance. It's one of the few around where you can set up the line SNR from the user interface. That's essential for me, as I'm on a long line and need to have the downstream SNR running at between 12dB and 15dB in order to keep the line maintained steady.

Setting the SNR is done via an extension to the router's address in the Web address field, which then takes you to a sidelined part of the 7800 user interface where all you need do is enter a decimal value that corresponds to a required SNR value, eg. entering 50 is reckoned to get you a target line SNR of 3dB, ie. 3dB below BT's default SNR. Various websites, and including within one or two forums on this one, have published a table of these values for entering.

However, none of the tables indicate the values you need when you want to go positive in SNR. Instead, the tables are designed primarily for those users wanting to increase their downstream sync by lowering their SNR. Despite this, some months ago I was passed some values by one or two helpful souls for increasing the SNR. These didn't seem to work properly but after many attempts I managed to obtain the sort of SNR I needed pretty much by trial and error.

Since then my line's SNR has dropped a small amount and occasionally now I'm losing sync, particularly in streaming applications. So I've tried bumping up the SNR by another 3dB (to around 15dB total). The point is that the enterable values for increasing the SNR that I was given don't actually fit with the results. I've just made around a dozen attempts to home in on 15dB but none of the values I tried - which included 150, 200, 250 - would seem to get near 15dB. I got 18dB. I got even 21dB. And at the other end of the scale I got 6dB and 3dB, all with sensible corresponding sync speeds. But I just couldn't go from 12dB to 15dB. I'd have thought that entering the value 150 would have done the trick but it certainly didn't.

For the positive SNR values above 6dB there seemed to be no consistency. For example, I'd have thought that entering 100 would have produced a 0dB change, 150 a 3dB positive change, 200 a 6dB positive change, and 250 a 9dB positive change, but this was not what I found. Instead, the results swung from one extreme to the other, and seemed almost arbitrary. One would think that the target SNR values here would be relative to 6dB but since that's not how it worked out, I thought that perhaps instead the target value would be with respect to the last SNR value. But no, that didn't fit either.

So, has anyone managed to fathom a set of values that actually work for going positive in SNR, rather than negative? And if so, are the values always with respect to 6dB or instead with respect to the last SNR line value achieved?

As matters stand at present I'm running at an SNR of 15.7dB, giving me around 4.6M bps sync, but I only got that after a dozen attempts with all sorts of values, as the recommended ones just didn't seem to work properly. I'll now leave it at that, but if I ever need to tweak the SNR again it'd be nice to know which values will actually render the required SNRs.

Edited by meditator (Sun 23-Dec-12 12:13:28)

Decibels are measured using a logarithmic scale. Which is why you are getting nowhere.

It is sheer coincidence that the "50" setting, which is in fact 50%, that you cite gives (about) 3dB result, as it would intuitively give if the scale were linear. 3dB happens to be the figure that actually double or halves the effect. I've not explained that very well, but a couple of examples should do it.

Starting again at the 6dB, a setting of 200% will give about 9dB result - that being double 6dB. Conversely if your line has been set to 9dB by BT's DLM then the 50% setting would reduce it to 6dB, not 4.5dB. If the DLM has set it to 12dB, then 50% will reduce it to 9dB. 25% would reduce it to 6dB from 12dB.

300% from 6dB will give 12dB.

Have a browse around a couple of these google results.

In reply to a post by meditator:

So, has anyone managed to fathom a set of values that actually work for going positive in SNR, rather than negative? And if so, are the values always with respect to 6dB or instead with respect to the last SNR line value achieved?

Rather than edit my previous post, as this is a separate issue, in the case of using BT Wholesale circuits the percentage is always in relation to whatever the DLM has set for your line. The DLM tries to impose that, and your percentage setting then overrides it.

I'm no longer using a 7800N (using a Draytek 120/Cisco EA4500 combo instead) but if you've got the time & patience read this.

Except it doesn't say the values are based on the DLM/DSLAM setting.

FWIW 4.6meg sync with a 15db TSNRM doesn't sound like a "long line" to me!

Those numbers you enter the interface are percentages but the SNRs/NMs are measured in deciBels on a logarithmic (base 10) scale. E.g. +3 dB = +0.3 Bels = log (2) Bels = log (200%) Bels. Is that clear?

So enter the percentage as 10^((desired NM -6)/10) *100.

To have a 12 dB NM when the DLM has set a Target NM of 6 dB enter:
10^0.6 * 100 = 400%

To have a 15 dB NM when the DLM has set a Target NM of 6 dB enter:
10^0.9 * 100 = 795%

Roberto, XRay, et al,

Yes, as a retired professional engineer I was well aware that the values you enter into the configuration box in the 7800's interface are decimal values representing percentages but that the SNR is measured in dBs, which is a logarithmic scale.

Roberto, I seem to recall that +6dB is a doubling of LEVEL but +3dB is a doubling of POWER. So, the values you need to enter might depend on whether the router works with voltage levels or instead with power levels. If those quoted figures on that australian forum are to be believed - and I reckon that those are the same figures that I was given originally - then we are talking about levels here, and so an entered decimal value of 200 (percent), let's say, should have increased the SNR by 6dB from its default value. As far as I'm aware, the default value on BT lines is usually 6dB. Similarly, 100 would be 0dB change, and 250 (two-and-a-half times) would produce an increase of 9dB from the default SNR. Be it actually right or wrong, that was my understanding of the mechanism (taken from others), when the other day I tried to bump up my SNR by 3dB.

My starting point was in fact just over 12dB. But I didn't get 15dB when I entered the value 150. Instead, I got over 18dB (which just doesn't seem to make any sense at all). At that stage I thought that maybe you have to always return the situation to the line having an SNR of 6dB and perhaps then and only then put in the requisite value. But doing that didn't work out either. In the end, I just kept using values in the range 50 to 200 until I finally achieved 15dB or thereabouts. It took me about a dozen frustrating attempts before I happened to land on 15dB.

If the decimal figures represent 'percentage changes of the dB values', then the figures I used should have been correct. But it's obvious that they're weren't. Those australian figures surely can't be correct? Or at the very least they must be couched in a misleading way. There seems to be some confusion over 'percentage of what?' (if you see what I mean).

Incidentally, the 7800 always has a default value of its own sitting in the SNR configuration box in its Web interface, that being 50 (even after using the facility).

Assuming that this mechanism in the 7800 does work properly when aiming in the positive SNR direction, what also needs clarifying is whether the figures used are always with respect to 6dB SNR or instead can work with any existing SNR. For example, in XRay's sample numbers it's been assumed that you subtract 6dB in the 'raised-to-the-power'. But what if your starting point is 12dB SNR? Using XRay's formula, if I'm starting from 12dB and I'm wanting 15dB, the decimal value you seem to need to enter is 200, ie. 10 to the power of 0.3. Is this correct, XRay?

In reply to a post by meditator:

My starting point was in fact just over 12dB. But I didn't get 15dB when I entered the value 150. Instead, I got over 18dB (which just doesn't seem to make any sense at all).

The point is whether the 12dB was the noise margin at that point in time, or the noise margin being set by the DLM at sync time. The DLM sync time value is what the percentage is applied to, not the running value.

Furthermore, given a lot of bit-swapping and possible frequency "write-offs" by the modem, results can change with a modem reboot, due to those being brought back into play.

At that stage I thought that maybe you have to always return the situation to the line having an SNR of 6dB and perhaps then and only then put in the requisite value. But doing that didn't work out either. In the end, I just kept using values in the range 50 to 200 until I finally achieved 15dB or thereabouts. It took me about a dozen frustrating attempts before I happened to land on 15dB.

There's a high chance that in that process you caused the DLM to think the line was unstable, and bump its sync time value up by 3 or 6dB anyway. That could explain your getting inconsistent results.

Before you endeavour to change the SNR, how can you determine what the DLM's SNR value is?

From what you say, it would seem impossible for the user to set the line's SNR.

Back in the days when I used the graphical utility DTM, I never had this kind of problem, so my view is that what the DLM does has a minimal effect.