hEX Lite: PoE input massively lower efficiency than ac adapter?

Hardware
1

I measured power consumption of my hEX Lite including ac adapter in 3 different ways and get consistent 1.75W across all measurements (idle). Very impressive.

Assuming a typical efficiency of 80%-90% (I'm using 86% for round numbers) for the ac adapter the the hEX itself is just 1.5W !!

I was pretty excited to replace the ac adapter and use the PoE input. Shockingly I found 2.6W used now! Nearly 1W more.

1.5/2.6=57.7% only and this doesn't even include the AC-DC losses for the on the other end.

Is there a reason powering via PoE input can be so much worse than the ac adapter?

2

I have no idea what the wart voltage is on this device, but if it is lower than the POE voltage, there will be a difference in losses across the voltage regulator.

3

This is almost certainly measurement error.

The dc-dc efficiency does change with voltage, but not as dramatically as these results would imply.

4

Which power adapter are you using?
The default 24V 0.38A one coming with the device?
What happens using the same as PoE power supply with an injector?
How (exactly) are you measuring the power consumption?

The declared one is:
Max power consumption 2 W

So 1.5W idle sounds reasonable.

If you calculate as (2.6-1.5)/1.5=73.3% the increase is so large that it sounds impossible, as lurker888 said different voltages and the different path the power takes (from jack or from ethernet) may well change something but not so much.

5

Yes, the default 24V 0.38A.

I measured power in 3 different ways:

  1. Kill-A-Watt instantaneous: jumping between 1.7W and 1.8W
  2. Leaving Kill-A-Watt running for ten days and dividing the resulting 430kWh by 240h ... which results in 1.79W
  3. Using Eaton EMAT08-10 ePDU and averaging the monitored power waveform over a day. To avoid inaccuracies at lower loads, I added a constant baseload (6.53W) and subtracted the load with hEX. Result: 1.74W

For PoE, I am using TP-Link TL-SG1016PE. The values reported by the switch are generally accurate for other devices. In table below, the hEX is Port 8, attached via this converter. That converter changes 48V (or 52.5V per switch) to the 24V which can also result in losses. But still, 1.5/2.6=57.69% efficiency for this doesn't sound realistic (even a simple dissipative load/LDO would come close to this)

image
image1572×526 63.6 KB

6

Seems like a better test would be to use the same 24v wart into a passive POE injector, and remove all the variables that you have introduced. Not sure what else you own, but the MikroTik injectors that come with some AP's work perfectly for this . . . purely passive, so no loss.

Then you will know if it's the off brand/offshore converter, POE natively, planetary alignment or possibly a burnt cheese pizza . . . :slight_smile:

7

Hmm, yeah I don't own any of them sadly.

I'll check if I find any of those in my network for a test.

Assuming I can get one of them and I can confirm that it's ~1.7W too, this means the issue wouldn't be hEX.

Any suggestions how to determine if the issue is the 48V-24V PoE converter or the TL-SG1016PE?

8

That converter ouputs max 24V 1A , I.e It Is a 24 W device.
If It Is very inefficient when giving out less than 1/10 of its target Power I would not be much surprised.