In my other thread I was struggling with the MPS Delta 40. I now have it working as a level sensor, but I can't use it for it's intended purpose (to trigger alarms and warnings) since it is spiking to 0%.

I wonder if anyone can tell me if there is some way I can cause an alarm to only activate after a condition exists for a period of time. I don't believe there is, but maybe I've missed it.

The problem is that when my pump PWMs up from low to high speed, the pressure drop that it causes in the res is making the MPS Delta overshoot, and spike right down to 0%, so if I set up a fill level alarm it will trigger every time the pump increases speed. Once flow stabilises at the new higher rate, the fill-level reading recovers to a correct value and remains stable.

I have already set the time based sensitivity of the MPS delta 40 to the lowest possible value to try to smooth the output, but that seems to smooth over maybe 2 seconds, but the spike to 0% I'm seeing lasts more like 10 seconds.

Sure, I could run the pump at a constant speed, but that's not what I brought an aquaero for!!!

I need to either:

1. Further smooth the output of the MPS delta 40, over a longer time period

or

2. Trigger an alarm only if the fill level reading goes low and stays low for a period of time.



I know you can set periods of time that fan speed readings can be absent/0 for, before an alarm or warning is triggered, but there doesn't seem to be a way to do that with a fill level alarm.

It would also be great to be able to do exactly the same thing (have a setting for the number of seconds before the alarm triggers) with all the other alarms, to eliminate spike problems generally.

Just to add to this, coincidentally, right as I posted this, my aquacomputer high flow meter has started spiking to 0 and triggering alarms as well. I've had it up and running fully for about a week, and it has never done this before, reading reliably right down to 0.2gpm. If I put my ear up against it, it doesn't sound like it can be an impeller stall as I can still hear it turning during the alarm.

I know it's a beta, but this is a major usability issue - the alarms are one of the main features, and in my install right now they are basically broken.

Devs, please can we have an adjustable time delay before alarms trigger, so spiking can be managed.

Thanks

Zitat von »LarryWill729«

Boris, a quick question, is your MPS 40 level sensor connected to its own, private hose to the Res?
.
If it is connected to ANY hose that has flow through it, the readings will spike due to water flow past the connector for the sensor.
.
The Delta 40 is intended to measure a difference in pressure, so for use as a level measurement, one side needs to be connected to the top of the Res in open air, and the otherside connected to the bottom of the Res. And neither high or low side can have any flow past the connector. So if you connected the wet side to a "T" connector at the inlet of your pump, it will do exactly what you describe.
.
For the high flow sensor spiking down to zero, I suspect a wire problem. Try changing the cable.

No, it's not connected with flow through it.

It's connected, with a single (small, 6mm/4mm) hose from a port on the bottom of the res, then at the top of the res there is a pressure equalisation membrane (http://shop.aquacomputer.de/product_info…roducts_id=2906)

The spiking only happens when the pump speeds up. It's not a problem with the readings - if I leave the pump on low (or high) flow, there is no spiking over hours of monitoring, but as soon as the pump speeds up there is an immediate spike to 0%. Within about 10 seconds, the level recovers somewhat, then after about 30 seconds it recovers fully, to the new lower level (correctly reading, as when the pump is at full speed the res level is about an inch lower than when it's at lowest speed).


EDIT: I just, as I made this post, thought what it might be and I've figured it out!!!

It occured to me that rather than being a drop in pressure at the high side of the sensor, it could be the air in the res getting pulled to a lower pressure and taking a while to equalise. Sure enough, I have removed the pressure equalisation membrane so that there is just an open hole in the res, and it immediately started working - when the pump PWM's up, the level reading drops straight to the (new, lower) correct value and stays there without any spiking or recovery time.

The pressure equalisation membrance clearly takes too long for my application.

I have some very free-breathing dust filter material here, and an unused G1/8 to G1/4 plug that I used with a tubemeter that went wrong, so I think I'm going to go and get hacking and see what I can lash together to make a fast-equalizing membrane of my own.

Note to the devs - although I have worked out what my problem was I do want to ask still;

- Please can we have the ability to "delay" alarms, so they trigger only if a condition exists for a period of time, rather than instantaneously, then we can manage spiking problems.

Ok,

I tried my G1/8 to G1/4 adapter with some of this: http://www.samokleykin.com/ filter material wadded up and shoved in it, and no spiking. But it was ugly as hell.

So, I took my aquacomputer pressure membrane ( which is a Gore PMF200128 ) and started hacking - the top part (the plastic part) comes off if you push a 6mm or so socket into the metal part to pinch the clips equally from all sides and then it comes apart, however I later realised that's completely unnecessary.

If you just pull at the top around the edge, it's just clipped on, and you can remove the top plastic cap.

I then cut the supplied membrane out with a stanley knife, and replaced it with some of the samokleykin dust filter material, and now no more spiking.

I notice at http://www.gore.com/en_xx/products/venti…-in_holder.html that the PMF200128 flows 450 litres per minute of air at a delta p of 70 millibars, but there is also a PMF200484 that flows 2500lph at the same pressure difference.

To retain most of the resistance to water ingress that the stock membrane provides, it might be worth trying one of them (they are rated for about half the water resistance). In my application, I'm not sure I'll bother - I can't forsee any situation in which the membrane should get wet, and it works for me now.

Perhaps the aquacomputer team could source and offer the PMF200484 as an option, since a quick google search reveals it to be made of pure unobtainium, so it's probably necessary to order through a trade account.

Zitat von »sebastian«

You can move the move the filter speed scroll bar to slower.

Hi Sebastian,

I'd already done that and set it to the slowest setting possible, but that seems to filter the readings down to about one every 2 seconds, but my spikes were lasting up to 30 seconds until the reading fully recovered and stabilised.

Perhaps using a PMF200484 membrane, instead of http://shop.aquacomputer.de/product_info…roducts_id=2906 which is a PMF200128 would be better, it has much lower resistance to air flow.

Thanks Shoggy,

I am controlling pump speed via PWM (MCP35x). I originally planned to do this because of pump noise, but once I received my pump, I realised it's very quiet even at full speed. I still like having it slow down when there's no system load though, since it should reduce wear and increase the life of the pump. I find that slowing down to 0.2gpm = 45lph still keeps everything cool at idle, and it even seems enough flow for 2 x 580GTX in SLI under load, the pump only really needs to speed up when the CPU is loaded.

I think a lot of people with loud pumps would want to slow them down when the system isn't under load.

I agree that the low pressure measurement the delta 40 was reporting is real - it was the pressure drop in the top of my res (in the air) pulling on the water and lowering the measured pressure.

If you did implement a delay for alarms, it wouldn't have to cause a long delay in the case of an alarm. The spikes I was seeing in my flow meter (which seem to have gone now) were only for maybe 5 seconds, so setting alarms to trigger after 10 seconds for example would have easily managed them, without adding a long delay.