like you, air cond, air flow, ducting, static air, velocity, volume etc, in the old days before flexi duct it was a science, now days they just wack a monster on the slab and away it goes. LOL

cheers.

I put the entire grill that was on the back onto the front.  I took the fan and plate from the front and put it on the back.
[/quote]

Ok, a couple of things to think about.

The original fan setup pics show that the fans are stepping in speed, this is a good thing, as the chip temp changes the fan adjusts. is the suction fan stepping, if its not and is constant thats not good.

The fan speed changes with pcb sensor, not completly sure if the temp chip sensor engages fan speed, could be secondary limit. Either way the drop in fan speed between the suction fan setup is due to lower pcb sensor temp, not necessarily from a better setup.

Have a look at the heatsink setup, the top of the unit has little restriction, and air can flow easily through this area, also the left side is board no 3 (bottom sensor) it has always the hottest readings, but it should be the coolest you would think, why because it flows more air near it, notice the gap between the board and the side case, a good 15mm, now have a look at the other side, board no 1 top sensor its always the second hottest and you would think it would be the hottest, it has less space than the rest, its jammed against the side wall.
This is a good example in the original setup of static pressure forcing air into the restricted board no 1, where as board no 3 the static pressure on the fins are lost and escapes around the 15mm gap in the side wall, more air passes but its hotter.

Im being objective here only to theroise what is happening with suction setup.

Nowing the air will suck easly through that big 15mm side gap and through the single fins at the top of the unit and through the gaps between the boards and swirl the air causing a pcb and chip temp drop. As i have suggested in the first post with the sensors being at the rear, The reason i suggest this is i have been able to reduce the pcb temps by redirecting the air and forcing it to flow closer to the sensors.

The suction setup could be good, i'm just not confident that its better due to all the factors in play,  some ways you could improve the unit without changing the fan setup, firstly remove the rear grill and put it aside. second, remove the fan grill, i made a new grill like a mouse cage and fitted it over the fan approx 50mm deep, also raise the unit off the table so the air can go through the cage bottom into the fan. third make a thin metal plate, best if i send a pic its a bit complicated, the plate improves pcb temps and brings down board no 3 now board no 1 is the hottest, its working well for me hope it does the same for you.

https://imgur.com/suA5T9q
https://imgur.com/EOK7lxd

I just took the Z9minis and put them into the garage with the fan on the back end.  The first two are of them running with the fan on the front pushing the air (original).  The second two are them running with the fan on the back pulling the air through.


---FAN ON THE FRONT---

https://i.imgur.com/8LTEUYf.jpg
https://i.imgur.com/3PrT67Z.jpg


---FAN ON THE BACK---

https://i.imgur.com/3j2azZU.jpg

https://i.imgur.com/6lnaiIB.jpg

[/quote]
What did you do with the rear grill and fan grill?.

What do you want to achieve, cooler chips, less noise or both, then put one of these on the fan end.

Look at this on eBay
https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com.au%2Fulk%2Fitm%2F192612612089

Furthermore if you look at the fan speed, chip and pcb temps they don't add up, the fan is running way to high for the  temps, I would suggest that the fan is not able to pick up the air, being blocked by the heat sink fins, this would create a vacume and in turn free reving with minimal load, or i am missing something else but i dont think so.

[/quote]
A/C units pull air because the coil is very close and larger than the fan and an even air flow through the coil is desired.

Also be carefull when pulling air over uneven coil as in the Z9 mini, the air will flow more with the least resistance areas, if you push air and the coil/ fins are close like the Z9, the static pressured air will not have time to find another course through the fins. In other words its better to push in on the Z9.

Sorry to burst your bubble, but you are creating false readings and you are putting stress on the unit.

 There are 4 chips on each board and  3 x boards, total of 12 chips, the sensor is on the last chip to the rear bottom of each board, the other 3 chips on the board are at the front near the fan, if you reverse the fan the sensored chip is getting cold air first and the other 3 chips on the board are getting hotter air, also the senor pcb is getting fresh air first, not good scenario for the 3 unsensord chips which is now hotter than the sensored chip and the fan is not activating to their real temps.

Normally they are always cooler than the sensored chip and have no need to be monitored.

Sorry to burst your bubble, but you are creating false readings and you are putting stress on the unit.

 There are 4 chips on each board and  3 x boards, total of 12 chips, the sensor is on the last chip to the rear bottom of each board, the other 3 chips on the board are at the front near the fan, if you reverse the fan the sensored chip is getting cold air first and the other 3 chips on the board are getting hotter air, also the senor pcb is getting fresh air first, not good scenario for the 3 unsensord chips which is now hotter than the sensored chip and the fan is not activating to their real temps.

Normally they are always cooler than the sensored chip and have no need to be monitored.

Cheers.

Bitmain site are now selling 50 units per order.