Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Jack,
1) I was just using the same tolerance that Arlan Anderson used in the power number that he presented.
The 8% variation is at least a ballpark number and a starting point, but +/- 1% is probably way too tight for this application.
There is obviously some tolerance to which power supplies can be designed and manufactured.
Tight tolerances call for tight regulation, which is something that costs more.
Also, for the case of the negative resistance load, which we certainly will have, tight load regulation at the PSE is not good.
Our power experts should give us feedback on these numbers.
2) I was suggesting that having a current and a power limit, in the PD, set at the long cable lengths
can also take care of current imbalance at the short cable lengths.
For long cables: the 350 ma current dominates, but current per pin is balanced
For short cables: the max power dominates, and the total load current is reduced due to more voltage at the load.
Otherwise, some one could design a PD that always grabs 350 ma, then at short cable lengths, we go over the current
per pin limit of 175 ma due to line resistance mismatch.
thanks,
- Rick
-----Original Message-----
From: Jack Andresen [SMTP:jandresen@xxxxxxxxxx]
Sent: Tuesday, July 18, 2000 5:50 PM
To: Brooks, Rick [SC5:321:EXCH]
Cc: stds-802-3-pwrviamdi@xxxxxxxx
Subject: Re: power delivery question from Liaison report
Rick,
1) What is magic about 8% (or +-4%)?
2) Connector unbalance is more constrained than you suggest. While Cat3
is spec'd at .4 dB insertion loss that is not a resistive loss. Even at
the extremes of +/-5% wire resistance, the variation is less than hall
your supposition.
3) You are right, we may need to lower the available power is we look at
the worst case of everything.
Jack Andresen
*****************************
Rick Brooks wrote:
>
>
> I was reading in the Draft Liaison report from ISO/IEC JTC 1/SC 25/WG
> 3 to IEEE802.3 on power feeding that was
> handed out at the July Plenary.
>
> IEEE802.3af had question 4: Info on parameter limits (voltage,
> current, power, source impedance, ...) for world wide standards.
>
> i.e. what are the restriction beyond SELV.
>
> The response back was 48 VDC max, 175 ma max per pin.
>
> My question is:
> Is the 48VDC output from the port really 48VDC max as the response to
> the question indicates?
>
>
> If so, my thoughts are the following:
>
> We would have to spec our power output at the PSE as 48 VDC + 0%, -
> 8%, or something like that,
> so that it never exceeds 48 VDC continuously.
>
> This will further limit the available load power;
> it would be less than the load power that was discussed at the last
> meeting namely 14.6 watts.
>
> So, in that case the PD must be designed to draw at most 350 ma, as we
> discussed.
> And the power delivered at 100 meter cable would then be:
> Pwr = [44.2 - (12.5 x 0.35)] 0.35 = 13.9 Watts. (where 44.2 VDC is
> the lowest output voltage to still be in spec)
> For long cable lengths, the current per pin will be balanced, and we
> don't exceed the 175 ma per pin.
>
> For short cable lengths, we probably need an additional power spec, so
> that neither RJ-45 pin exceeds 175 ma.
> Say that due to connector imbalance, one pin is 175 ma, and the other
> is 20% below that, or 140 ma, which is a total of 315 ma.
>
> Then the power for a short cable would be (at least) 13.9 watts (44.2
> * 0.315).
>
> This would say that the PD device should be designed not to draw more
> than 350 ma,
> and at the same time not to draw more than 13.9 watts.
> That way we never exceed 48 VDC nor 175 ma per pin on a continuous
> basis.
>
> This puts the burden on the PD end to meet these current and power
> requirements.
> The PSE end would have a max voltage of 48 VDC, but it's current limit
> would be set slightly higher than 350 ma
> by some appropriate margin.
>
> If, on the other hand, we put the burden at the PSE end, then the
> available power goes down even more, but that may be OK also.
>
> comments?