I was actually working on this 2 days ago, trying to work out why my S7 wasn't mining. Turns out my PoE had died but in the hour or so of debugging I had the "hot exhaust" issue. Exhaust temps were up a lot compared to stable mining but heatsinks weren't up anything significant.
So when you say the heatsinks weren't up - do you mean what is reported through the web interface? I just wonder how that would be possible, since hotter air pretty much == hotter temps... I almost think it stops reading the sensors and/or controlling fan. Either way, it was 100% using full power when it had no ethernet connection going, so something is up with it.
With my temp probe gun, was unable to connect to S7 as my network was fubar'ed. The full equation for forced convection heat transfer is as follows:
q = hc A (Ts - Ta)
where
q = heat transferred per unit time (W)
A = heat transfer area of the surface (m2)
hc = convective heat transfer coefficient of the process (W/(m2 K) or W/(m2 ° C))
Ts = Temperature of exhaust air
Ta = Temperature of intake air
In this scenario we can take a steady state snapshot at the end of the board closest to the exhaust. Comparing normal operation to weirditsstillminingbutitsreallynot operation:
q = constant
A = constant
Ts = up
Ta = up
hc = pretend its constant for now
Ie q1 = q2, A1 = A2, hc1 = hc2
In the above scenario:
q1 = hc1 A1 (Ts1 - Ta1)
q2 = hc2 A2 (Ts2 - Ta2)
So Ts1 - Ta1 = Ts2 - Ta2
or Ts2 - Ts1 = Ta2 - Ta1. Ie if the difference between exhaust and intake temperature remains constant.
Earlier I said set hc as a constant - its not. It varies on lots of properties but we can simplify it to hc = 10.45 - v + 10 v1/2 for air, where v = the relative speed of the object through the air (m/s). When you graph this its a nice easy
curve.
If the velocity halves then you drop from ~32 to 28 (hc1, hc2) for an example, quite a small change [hc2 / hc1 = 28/32 = 87.5%]. Adding this back in above we can get:
hc1 (Ts1 - Ta1) = hc2 (Ts2 - Ta2),
hc1/hc2 (Ts1 - Ta1) = (Ts2 - Ta2),
0.875 (Ts1 - Ta1) = (Ts2 - Ta2).
Even though we halved our effective air velocity, we only have to increase our difference between exhaust and intake air temps by 14.2% (1/0.875). Putting some example numbers in:
If exhaust temps were 50C and exhaust temps 30C usually (20C diff)...
We can maintain the same heat removal with half the air speed, with the intake remaining at 30C and the exhaust.... but with an exhaust temp of 30/0.75 = 57C (27C diff)
If you want to look at the relationship between heatsink (chip) temperatures, its varies based on the average air temp. So if we raise our average air temp by 3.5C ([27 - 20] /2), we also raise our heatsink temp by 3.5C.
tldr: Decreased air velocity = smaller increased increased exhaust air temperatures = even smaller chip temperature increase.
