Nutrient Pulse Reactor (DIY DyMiCo filter)

[saltyclaus]

Hi,

Allow me a brief introduction:
In 2013 we decided to get into the marine aquarium hobby. Due to renovations (sounds familiar dartier?) and professional changes we paused the implementation of the planned system just after the tanks were delivered. After reading through this thread, seeing the first test setups and others in the making, I decided it was time to (re)register with RC and chime in.

Afterall, the system we are creating was planned as a traditional Berlin with a twist: an emphasys on heavy feeding and additional nutrient export with an ATS, wet skimming water change and carbon dosing. During the long pause the dymico system came to my attention when the hobby beta was announced. I have been following it ever since with the idea to create a diy version some day as additional or even as replacement for export but also as an alternative to a proportional pricey CR.

My attempt at a dymico won't be around soon. I have to finally get the system up and get some fish and invertebrates in the house :lolspin: There is a lot of remaining work before that'll happen. Hence, the dymico will be integrated later in the running system, probably as soon as water changes can't keep up with Ca demand. In the meanwhile I'ld like to offer my insights in the dymico system so far as well as any translations if new info pops up (I'm Flemish = Dutch native speaker). Lastly, I am a .Net architect professionally (big word synonym for nerdy programmer) so could assist with the programming logic and flow. I envision a pseudo-program that is readable by many and can be translated in whatever needed language (python, Arduino C, C#, reef angel, ...).

As a first contribution I'ld like to share the following (which you all probably have seen but I haven't seen it in the thread). It's a post by Tim Wijgerde (phd involved with the research behind dymico) on a dutch forum in which an (english) excerpt from the facebook page was posted:

1. Seawater is actively drawn through the sand bed with a return pump. This means the system is not dependent on diffusion of nitrate and carbon into the sand bed. The amount of water filtered is controlled by the redox value measured inside the sand bed by an IKS Aquastar computer. The lower the redox value, the less oxygen is present in the sand bed, and the more water is filtered. The opposite is also true, where higher redox values trigger less pumping time. This results in a sinusoidal redox value, which fluctuates around an average value of about -200 mV. This value is ideal for denitrification.

2. The sand bed receives a carbon source every hour, which the bacteria use to break down nitrate. This also ensures that carbon does not have to diffuse through the sand bed, like in conventional denitrification systems. Again, this process is controlled by the computer. The lower the redox value, the less carbon is injected. Higer redox values trigger more carbon injection into the sand bed, to stimulate oxygen consumption by bacteria.

3. The sand bed also receives carbon dioxide (CO2) injection, based on the requirements of the user. A pH of 7 is sufficient in most cases to result in high calcium, magnesium and KH/alkalinity levels. DyMiCo works as a powerful calcium reactor, with a capacity of about 10% of the total system volume.

4. Finally, the carbon source and CO2 are actively and horizontally pumped through the sand bed, to ensure that the sand bed is evenly affected. This prevents black patches of sulphate reducing bacteria that produce toxic hydrogen sulfide gas (H2S), a common problem with conventional sand beds.

5. Because the entire aquarium water volume is filtered only once per 24 hours (on average), it removes less plankton and feeds from the water. This is why DyMiCo aquaria are richer in plankton (algae, bacteria and crustaceans) and filter feeders such as sponges, when compared to Berlin systems. The system also uses very little energy, and requires no water changes or Balling additives (calcium chloride, magnesium chloride or sodium bicarbonate).

Quoted from: http://www.zeewaterforum.info/forums/archive/index.php/t-147132-p-3.html

I believe most of the info is still accurate (it predates the release of the hobby systems) except for the average mV of -200. The manual V13 mentions an operation range between -250/+250mV with going into alarm at -300/+300mV so the average is probably closer to 0?

Another thing of note for me was that carbon is dosed once per hour. So, not the frequency changes, only the dose fluctuates depending on redox value. Together with the sinusoïdal pattern of the cycles this makes me think of the system as a perfect candidate for a machine learning algorithm. Considering the bio-load of a closed system is fairly constant, the swings in redox value combined with the nitrate knee moments allow for a self adjusting system: let the program alter the parameters dynamically to achieve the desired min/max redox and cycles/hour. That would introduce an evolutive system that adapts to any bio-load and filter capacity combination as long as the system can sustain the desired outcome.

Ok, guess I should stop the rambe here for a first post (too late right?). I look forward to seeing all your updates and will eagerly contribute where desired

 

[OzIA]

Salty...

Salty...

This showed up in my email but not on the thread. Not sure what happened here so I am copying for everyone to see:

Hi,

Allow me a brief introduction:
In 2013 we decided to get into the marine aquarium hobby. Due to renovations (sounds familiar dartier?) and professional changes we paused the implementation of the planned system just after the tanks were delivered. After reading through this thread, seeing the first test setups and others in the making, I decided it was time to (re)register with RC and chime in.

Afterall, the system we are creating was planned as a traditional Berlin with a twist: an emphasys on heavy feeding and additional nutrient export with an ATS, wet skimming water change and carbon dosing. During the long pause the dymico system came to my attention when the hobby beta was announced. I have been following it ever since with the idea to create a diy version some day as additional or even as replacement for export but also as an alternative to a proportional pricey CR.

My attempt at a dymico won't be around soon. I have to finally get the system up and get some fish and invertebrates in the house :lolspin: There is a lot of remaining work before that'll happen. Hence, the dymico will be integrated later in the running system, probably as soon as water changes can't keep up with Ca demand. In the meanwhile I'ld like to offer my insights in the dymico system so far as well as any translations if new info pops up (I'm Flemish = Dutch native speaker). Lastly, I am a .Net architect professionally (big word synonym for nerdy programmer) so could assist with the programming logic and flow. I envision a pseudo-program that is readable by many and can be translated in whatever needed language (python, Arduino C, C#, reef angel, ...).

As a first contribution I'ld like to share the following (which you all probably have seen but I haven't seen it in the thread). It's a post by Tim Wijgerde (phd involved with the research behind dymico) on a dutch forum in which an (english) excerpt from the facebook page was posted:


---Quote---
1. Seawater is actively drawn through the sand bed with a return pump. This means the system is not dependent on diffusion of nitrate and carbon into the sand bed. The amount of water filtered is controlled by the redox value measured inside the sand bed by an IKS Aquastar computer. The lower the redox value, the less oxygen is present in the sand bed, and the more water is filtered. The opposite is also true, where higher redox values trigger less pumping time. This results in a sinusoidal redox value, which fluctuates around an average value of about -200 mV. This value is ideal for denitrification.

2. The sand bed receives a carbon source every hour, which the bacteria use to break down nitrate. This also ensures that carbon does not have to diffuse through the sand bed, like in conventional denitrification systems. Again, this process is controlled by the computer. The lower the redox value, the less carbon is injected. Higer redox values trigger more carbon injection into the sand bed, to stimulate oxygen consumption by bacteria.

3. The sand bed also receives carbon dioxide (CO2) injection, based on the requirements of the user. A pH of 7 is sufficient in most cases to result in high calcium, magnesium and KH/alkalinity levels. DyMiCo works as a powerful calcium reactor, with a capacity of about 10% of the total system volume.

4. Finally, the carbon source and CO2 are actively and horizontally pumped through the sand bed, to ensure that the sand bed is evenly affected. This prevents black patches of sulphate reducing bacteria that produce toxic hydrogen sulfide gas (H2S), a common problem with conventional sand beds.

5. Because the entire aquarium water volume is filtered only once per 24 hours (on average), it removes less plankton and feeds from the water. This is why DyMiCo aquaria are richer in plankton (algae, bacteria and crustaceans) and filter feeders such as sponges, when compared to Berlin systems. The system also uses very little energy, and requires no water changes or Balling additives (calcium chloride, magnesium chloride or sodium bicarbonate).
---End Quote---
Quoted from: http://www.zeewaterforum.info/forums/archive/index.php/t-147132-p-3.html

I believe most of the info is still accurate (it predates the release of the hobby systems) except for the average mV of -200. The manual V13 mentions an operation range between -250/+250mV with going into alarm at -300/+300mV so the average is probably closer to 0?

Another thing of note for me was that carbon is dosed once per hour. So, not the frequency changes, only the dose fluctuates depending on redox value. Together with the sinusoïdal pattern of the cycles this makes me think of the system as a perfect candidate for a machine learning algorithm. Considering the bio-load of a closed system is fairly constant, the swings in redox value combined with the nitrate knee moments allow for a self adjusting system: let the program alter the parameters dynamically to achieve the desired min/max redox and cycles/hour. That would introduce an evolutive system that adapts to any bio-load and filter capacity combination as long as the system can sustain the desired outcome.

Ok, guess I should stop the rambe here for a first post (too late right?). I look forward to seeing all your updates and will eagerly contribute where desired

 

[OzIA]

Based on what i have seen so far, I would have to agree with Salty, regarding the frequency of the cycle. I have not been able to get it to do a complete cycle in less than an hour, regardless of how far the ORP value was ranging.

Prior to this post, I was triggering carbon and flush cycles, based on the ORP value. I ran an hourly cycle for about a day (low spot on the graph) and then restarted dosing carbon at a fixed amount, just after the filter was flushed. This time the ORP cycle built back up on it's own but the swings are less.

I am going to let this run for a while and see what the outcome is. Since I don't have the ability to build a complicated algorithm with he APEX, I may have to manually adjust volume of carbon and flush as I go.

Aaron

 

[dartier]

This showed up in my email but not on the thread. Not sure what happened here so I am copying for everyone to see:

Hi,

Allow me a brief introduction:
In 2013 we decided to get into the marine aquarium hobby. Due to renovations (sounds familiar dartier?) and professional changes we paused the implementation of the planned system just after the tanks were delivered. After reading through this thread, seeing the first test setups and others in the making, I decided it was time to (re)register with RC and chime in.

Afterall, the system we are creating was planned as a traditional Berlin with a twist: an emphasys on heavy feeding and additional nutrient export with an ATS, wet skimming water change and carbon dosing. During the long pause the dymico system came to my attention when the hobby beta was announced. I have been following it ever since with the idea to create a diy version some day as additional or even as replacement for export but also as an alternative to a proportional pricey CR.

My attempt at a dymico won't be around soon. I have to finally get the system up and get some fish and invertebrates in the house :lolspin: There is a lot of remaining work before that'll happen. Hence, the dymico will be integrated later in the running system, probably as soon as water changes can't keep up with Ca demand. In the meanwhile I'ld like to offer my insights in the dymico system so far as well as any translations if new info pops up (I'm Flemish = Dutch native speaker). Lastly, I am a .Net architect professionally (big word synonym for nerdy programmer) so could assist with the programming logic and flow. I envision a pseudo-program that is readable by many and can be translated in whatever needed language (python, Arduino C, C#, reef angel, ...).

As a first contribution I'ld like to share the following (which you all probably have seen but I haven't seen it in the thread). It's a post by Tim Wijgerde (phd involved with the research behind dymico) on a dutch forum in which an (english) excerpt from the facebook page was posted:


---Quote---
1. Seawater is actively drawn through the sand bed with a return pump. This means the system is not dependent on diffusion of nitrate and carbon into the sand bed. The amount of water filtered is controlled by the redox value measured inside the sand bed by an IKS Aquastar computer. The lower the redox value, the less oxygen is present in the sand bed, and the more water is filtered. The opposite is also true, where higher redox values trigger less pumping time. This results in a sinusoidal redox value, which fluctuates around an average value of about -200 mV. This value is ideal for denitrification.

2. The sand bed receives a carbon source every hour, which the bacteria use to break down nitrate. This also ensures that carbon does not have to diffuse through the sand bed, like in conventional denitrification systems. Again, this process is controlled by the computer. The lower the redox value, the less carbon is injected. Higer redox values trigger more carbon injection into the sand bed, to stimulate oxygen consumption by bacteria.

3. The sand bed also receives carbon dioxide (CO2) injection, based on the requirements of the user. A pH of 7 is sufficient in most cases to result in high calcium, magnesium and KH/alkalinity levels. DyMiCo works as a powerful calcium reactor, with a capacity of about 10% of the total system volume.

4. Finally, the carbon source and CO2 are actively and horizontally pumped through the sand bed, to ensure that the sand bed is evenly affected. This prevents black patches of sulphate reducing bacteria that produce toxic hydrogen sulfide gas (H2S), a common problem with conventional sand beds.

5. Because the entire aquarium water volume is filtered only once per 24 hours (on average), it removes less plankton and feeds from the water. This is why DyMiCo aquaria are richer in plankton (algae, bacteria and crustaceans) and filter feeders such as sponges, when compared to Berlin systems. The system also uses very little energy, and requires no water changes or Balling additives (calcium chloride, magnesium chloride or sodium bicarbonate).
---End Quote---
Quoted from: http://www.zeewaterforum.info/forums/archive/index.php/t-147132-p-3.html

I believe most of the info is still accurate (it predates the release of the hobby systems) except for the average mV of -200. The manual V13 mentions an operation range between -250/+250mV with going into alarm at -300/+300mV so the average is probably closer to 0?

Another thing of note for me was that carbon is dosed once per hour. So, not the frequency changes, only the dose fluctuates depending on redox value. Together with the sinusoïdal pattern of the cycles this makes me think of the system as a perfect candidate for a machine learning algorithm. Considering the bio-load of a closed system is fairly constant, the swings in redox value combined with the nitrate knee moments allow for a self adjusting system: let the program alter the parameters dynamically to achieve the desired min/max redox and cycles/hour. That would introduce an evolutive system that adapts to any bio-load and filter capacity combination as long as the system can sustain the desired outcome.

Ok, guess I should stop the rambe here for a first post (too late right?). I look forward to seeing all your updates and will eagerly contribute where desired

Welcome SaltyClaus. I noticed that you were the last poster but that your post seemed to have vanished. It may have been due to the inclusion of an external link in a post from a new registration. Probably a form of anti-spam protection.

I read quite a bit about DyMiCo on the Zeewater forum, but always through Google translate, so a native dutch speaker is most welcome.

I have largely the same conclusion as you have posted from Tim. I had read that passage before and it was quite insightful. I have read few other things since then that seem to contradict Tim a bit, but that is not to say that Tim's explanation is not the most complete of the operation pattern of the filter.

I was leaning towards a PID loop (with safe limits), but a machine learning algorithm would also be fun to try. I am also a programmer by trade. Hopefully between all of us we can code up a flexible algorithm that is able to adapt to filter and tank differences.

Anyway, welcome aboard.

Dennis

 

[OzIA]

I changed out the course media a couple of days ago and restarted the system. While I believe the media I chose was a contributing factor, I am pretty sure that the primary cause of the blockage was the use of the drain sock. The fine mesh essentially became a nearly impenetrable barrier, once the bacteria slime formed on it. Don't use that to separate your media. I will go back and cut a piece of 1/8" acrylic and drill some 1/8 or 3/16 holes before I put the Dingo Down version on.

The wave generated by the new media is much more saw-toothed. I am flushing roughly 3 gallons through, on the hour, followed by .55 ml of carbon, three minutes later. The ORP will top out (or bottom if you are looking at my upside down graph) almost exactly 10 minutes later, falling around 75 mV. It then takes about 50 minutes to rebound and the cycle is started again, time based. I have made and continue to make adjustments in the amount of carbon and length of the flush cycle. As you can see, it doesn't seem to change the swings too much but it does affect the slew rate. ORP was increasing pretty steadily until I changed the amount of flush by 5 seconds yesterday at 18:14 and you can see things flattened out after that. This morning, I have reduced the carbon from .55 to .45 ml to see what effect that will have. I am logging the data in a spreadsheet. I thought that might help with the algorithm later.

Aaron

 

[OzIA]

Sorry, wrong graph. See attached for correct one.

 

[dartier]

Sorry, wrong graph. See attached for correct one.

That looks nice and regular. I assume you are forcing a flush once per hour or when the ORP get below xxx. I would leave it like this for awhile and see if the cycle length starts to shorten by hitting the low point and starting a new cycle.

Also how is your nitrate in this tank doing? That should be the primary indicator of the filter actually performing as intended. The DyMiCo version typically takes 2 to 3 weeks to actually become fully operational (though they mention it could take up to 6 weeks).

Dennis

 

[OzIA]

No, actually I am not doing anything with the ORP, other than watching it. It's the simplest program I have run so far. It doesn't check anything. It just doses three minutes after the flush, which comes every hour. This is the result. The swings are getting larger, now about 100 mV, even though I lowered the carbon dose earlier. I guess the big question is, "what is the correct / ideal range?

Unfortunately, nitrate was already low, almost unreadable. I was already dosing before. I rarely get anything above .25 ppm. That trend has continued. I am hoping for more pod activity to indicate the filter is working.

 

[dartier]

Ah, OK, so what you are observing is like the pattern that the DyMiCo filter operates in after just being setup, but before it is fully cycled. It flushes once an hour regardless of ORP and then eventually as the bacteria bed becomes mature the ORP drops faster and faster and the ORP starts to control the length of the cycle rather than the hard limit of a minimum of 1 flush per hour.

If you can add another condition that causes a flush even if the 1 hour limit has not been exceeded, but the ORP has reached -250mv, then that should give you more head room because the 1 hour limit will start to be too short of a time to keep the ORP in range. I think we should be shooting for -100mv on average. For example a 0mv to -200mv swing would be -100mv on average.

The fact that your nitrate is low and staying low, shows that the filter is having an effect. You don't have a protein skimmer on this tank?

Dennis

 

[OzIA]

At this point, it doesn't appear anything is going below -250 mV any time soon.

This morning's graph attached.

Avg peak = -146.6 mV to - 45.8 mV, so about a 100 point swing.

There is a bit of a sinusoidal wave in there that may be due to the BRS doser. They only have two pinch rollers and there seems to be a dead spot, right before the next roller enters the track. I noticed this when I was purging the line. The carbon moves steadily except for this point, where it almost stops for a few seconds. Depending on the timing of the cycle, I am sure that my carbon dose is varying and this may be the result.

The protein skimmer has been out or off since starting this project, since late Sept.

Aaron

 

[dartier]

The protein skimmer has been out or off since starting this project, since late Sept.

That is pretty good then. You are able to maintain such a low nitrate reading with no protein skimmer export. How is your phosphate?

Dennis

 

[OzIA]

Phosphate is OK but not great, 0.08. This is also pretty consistent with what I got with just the carbon dosing. I am running my CaOH dose straight into the overflow drain. I was hoping that the pH might be high enough before it reached the roller mat that some of the phosphate might be precipitated out and picked up by the roller mat. I am also running a small amount of GFO in a reactor.

Keep in mind, this system was using something similar to a deep sand bed before. I had stacked trays of sand (12) in those compartments (1&2) that were probably performing similarly. I was not cycling the water through them like now, it was running through them all the time and of course the carbon was not being dose and circulated through either.

No water changes since I chanced things out in late Sept, except for a small 5 gallon change / refill when I switched the media out the other day. I had to remove the water out of that compartment to get the media out. The new is the Two Little Fishes Reborn, the same thing you will have.

Aaron

 

[maltes]

From what I could gather from the small DyMiCo filter system (by the pictures), the processing pump will push water into the coarse sand bed at elevated height in the middle, while at the bottom of one side, water returns to the processing chamber through a standpipe.

On the other side at the bottom is a pipe for the return pump. The standpipe of the processing return will prevent the return pump to draw water from the processing chamber.

Pictures of the commercial version:
http://uploads.tapatalk-cdn.com/20150925/996b1b2a30413be665e4f87a24428e9d.jpg

Pictures of a prototype:
http://uploads.tapatalk-cdn.com/20150804/ed740435772b159c5ae63d2164434d2b.jpg

More prototype pictures (very helpful) that require a registered account at zeewaterforum:
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422891&d=1425221926
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422890&d=1425221892
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422893&d=1425221972

I wonder about two topics:
1. Why would they connect the return pump directly to the sand bed pipe and not just let it pump the water from the return chamber? Why is there water in the return chamber then anyhow?

2. The way it is built, the processing pump will only circulate water through half of the sand bed, as the second pipe is only used by the return pump. Is that by design or in order to save space/money?

I was interested in setting up a reef tank to replace my freshwater tank- now I am really tempted to do so with a DIY Nutrient Pulse Reactor
(the filter itself will be super easy to build, we have online services in germany where you can order custom cut plexiglass/acryl with holes and other cutouts)

 

[OzIA]

Just got another message in my inbox that doesn't show up on the forum. I am guessing that it has something to do with all the links. Salty had the same issue.

From Maltes:

From what I could gather from the small DyMiCo filter system (by the pictures), the processing pump will push water into the coarse sand bed at elevated height in the middle, while at the bottom of one side, water returns to the processing chamber through a standpipe.

On the other side at the bottom is a pipe for the return pump. The standpipe of the processing return will prevent the return pump to draw water from the processing chamber.

Pictures of the commercial version:
http://uploads.tapatalk-cdn.com/20150925/996b1b2a30413be665e4f87a24428e9d.jpg

Pictures of a prototype:
http://uploads.tapatalk-cdn.com/20150804/ed740435772b159c5ae63d2164434d2b.jpg

More prototype pictures (very helpful) that require a registered account at zeewaterforum:
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422891&d=1425221926
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422890&d=1425221892
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422893&d=1425221972

I wonder about two topics:
1. Why would they connect the return pump directly to the sand bed pipe and not just let it pump the water from the return chamber? Why is there water in the return chamber then anyhow?

2. The way it is built, the processing pump will only circulate water through half of the sand bed, as the second pipe is only used by the return pump. Is that by design or in order to save space/money?

I was interested in setting up a reef tank to replace my freshwater tank- now I am really tempted to do so with a DIY Nutrient Pulse Reactor
(the filter itself will be super easy to build, we have online services in germany where you can order custom cut plexiglass/acryl with holes and other cutouts)
***************

 

[maltes]

I guess it has to do with me being a first time poster and all the (external) links inside. Thanks for reposting

And OzIA - thank you so much for keeping us updated on your setup and results. I love it

 

[OzIA]

No problem on the repost.

To answer your questions, with my limited understanding...

I have no idea why they would connect that return directly to the sand bed instead of just letting it pump water from the chamber. I am not connected directly to the sand bed so I don't think it matters.

Regarding the process pump... I believe they only want to circulate the dosed (CO2 and carbon) water through the course sand bed. If you have been following, you have seen the issues I had with the finer media and drain sock I used to separate the two levels of media. The bacteria builds up a slime that would probably block the fine sand. Also, you probably need a level of media (fine sand) to perform the first step of nitrification process. This then gets fed into the next level to be denitrified by the bacteria fed by the carbon in suboxic conditions.

At least that is the way i understand it.

Aaron

 

[maltes]

Ah regarding the process pump I think I have not made myself clear enough.

Top view of the prototype looks like this:

-----------------
1=========

0000000000000

2=========
-----------------

1=== : Return pipe connected to the processing chamber (at the bottom)
000000: Output pipe of the processing pump (elevated)
2=== : Return pipe connected to the return pump (at the bottom)

So it seems to me that with this setup, water only circulates between the middle output and the top return pipe, so only through half of the course sand bed.

 

[OzIA]

I can't see the view/s you sent but from what I have seen with Dennis's version and the stuff from the manual, the process pump (left) takes water from the treated side and pushes it in the upper middle of the course media. From there it moved down toward the bottom pipes and is returned to the left chamber.

The right most chamber (return side) has a single pipe, near the bottom of the course media that allows water, that has been treated in the course media section, to be returned to the tank. Water also flows from the tank to the top of the fine sand and pushes water down into the layers of the sand bed.

There may be some course media just above the top process pipe / pipes but the majority of it lies at or below the level of those pipes.

Is that the way you see it?

Aaron

 

[dartier]

Just got another message in my inbox that doesn't show up on the forum. I am guessing that it has something to do with all the links. Salty had the same issue.

From Maltes:

From what I could gather from the small DyMiCo filter system (by the pictures), the processing pump will push water into the coarse sand bed at elevated height in the middle, while at the bottom of one side, water returns to the processing chamber through a standpipe.

On the other side at the bottom is a pipe for the return pump. The standpipe of the processing return will prevent the return pump to draw water from the processing chamber.

Pictures of the commercial version:
http://uploads.tapatalk-cdn.com/20150925/996b1b2a30413be665e4f87a24428e9d.jpg

Pictures of a prototype:
http://uploads.tapatalk-cdn.com/20150804/ed740435772b159c5ae63d2164434d2b.jpg

More prototype pictures (very helpful) that require a registered account at zeewaterforum:
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422891&d=1425221926
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422890&d=1425221892
http://www.zeewaterforum.info/forums/attachment.php?attachmentid=422893&d=1425221972

I wonder about two topics:
1. Why would they connect the return pump directly to the sand bed pipe and not just let it pump the water from the return chamber? Why is there water in the return chamber then anyhow?

2. The way it is built, the processing pump will only circulate water through half of the sand bed, as the second pipe is only used by the return pump. Is that by design or in order to save space/money?

I was interested in setting up a reef tank to replace my freshwater tank- now I am really tempted to do so with a DIY Nutrient Pulse Reactor
(the filter itself will be super easy to build, we have online services in germany where you can order custom cut plexiglass/acryl with holes and other cutouts)
***************

Welcome Maltes.

I can answer your 2 questions as I have been following this filter for a few years before thinking about building one.

1. The return chamber is hooked directly to the sand bed to keep that chamber dry. They use it as an emergency overflow. The baffle on that end has a weir on it. As an option, they also suggest using as a resevoir for top off by filling it with RODI. You are probably seeing RODI in it in the photos.
2. The standpipe's are actually probe holders used in the first BETA units. The production ones forego those and use a probe holder on the side of the chamber. The smaller filter, model 700, uses 3 pipes. That is the model you linked to in the photo. The larger model 2000 uses 5 pipes. Yours and Aaron's analysis of water flowing from the central pipe, to one of the bottom pipes, while the other bottom pipe is for the return, is correct (for the 700 model).

Dennis

 

[saltyclaus]

Since my last invisible post I've further read studies on this subject, more specifically into SBR's for wastewater. As far as I can tell Dymico is based on the principles employed here. I have come to a few speculatory conclusions that only raise a few new questions.

Let me first go into SBR or sequencing batch reactor. I read into this after going through Dennis's post about the nitrate knee. Waste water plants employ among other techniques biological nitrification and denitrification through this kind of reactors. The principle is very similar to the Dymico. Waste water is put into a tank where it is aerated to induce nitrification. Once the nitrification is complete the air is turned off to let the tank go into an anoxic state which induces denitrification. The entirety of the process is a bit more complex than that but those two phases are the ones of interest to us. The detection of the end of nitrification and denitrification is done through ORP monitoring and/or ph monitoring, the trigger points are either fixed thresholds or are detected through the ammonium valley (ph based) and/or the nitrate knee (ORP based).
The easiest implementation is based on fixed thresholds (ORP, ph or time), similar to the experiment Aaron is running currently (time based). Higher efficiency can be obtained by basing it on ORP value (Dymico). Even better throughput is achieved by basing the cycle on ph and of course, combining ph and ORP yields maximum result.
In waste water, time based and fixed threshold are rarely used apparently. The reason is simple: they have fluctuating loads and a non adaptive system such as this either yields effluent with too much N or P in it or augments the idle time (nitrification and denitrification all done and the tank is sitting there doing nothing) which lowers the efficiency. I'll include a screenshot from a typical SBR cycle.

Dymico uses the ORP adaptive method. Dennis mentioned he'll try to incorporate ph based control. It would allow for a perfectly tuned system but I believe the Dymico doesn't incorporate ph control for a few reasons:
- first, simplicity: driving the system on one parameter less
- the ammonium valley is less noticeable than the nitrate knee and is hence difficult to detect
- it requires more precise probes and a higher sampling rate to detect it (might be a non-issue)
- I read the mention of not easily detectable in two separate SBT studies, this concerns waste water and hence fresh water. I presume (speculative) that the ph fluctuation would be even less in seawater with the higher buffering capacity
- lastly, and most important for me personally, an NPR doubles as a Ca reactor. As soon as you add CO² to the mix the ph swings are related to that and will most likely mess up any 'ammonium valley detection algorithm'

I'll post the links to the studies I base the above on in a follow-up post (to prevent this post from disappearing yet again).

ORP on the other hand has important drawbacks as well. The readings are not empiric. Quote from one of the studies:

A major problem with the absolute value (either ORP or pH) based control strategy is the determination of the limits or set-point. It is commonly known that ORP as measured is not a true thermodynamic parameter. It is merely an indication of the overall oxidative-reductive state of the system, the absolute ORP value per se does not impart any process significance (Al-Ghusain et al., 1994).

In short ORP depends not only on the probe, it also depends on the system it is used on (the entirety of all molecules present). Furthermore, ORP isn't set in stone, the same potential in the same system can fluctuate as ORP probes tend to drift over time.
Hence the reaction of the system is based on a relative change in ORP: detecting highs and lows and changes in rise or decline. I've pondered a bit more on detecting these as well as how they would influence the cycle which will be for a follow up post as well.

There are some important differences as well between an NPR and a SBR. An SBR goes from totally aerobic to totally anaerobic and the cycles tend to be at least 24h. The bio mass of aerobic and anaerobic bacteria coexist in the reactor, they are active depending on the state of the reactor. A NPR on the other hand is like a DSB with the controllability of an SBR. In a NPR we want a zone of nitrification (fine layer) and a zone of denitrification (coarse layer). The system does not swing from aerobic to anaerobic. In conclusion a fine grained balance is necessary to keep the top layer aerobic and the bottom anaerobic. I believe that this renders the required algorithm even more complex than the ones used for SBR's. I might be wrong about this though, maybe the slightest control is enough to support the bio load of a reef tank, after all we're not trying to denitrify a varying load from a couple of thousand of households ;-) Time, experiments and the resulting data will tell.

Nevertheless, some questions popped up:
- will the nitrate knee be detectable as the system pumps a full tank's worth of volume through the sand bed? It might well be that the coarse layer is not anoxic enough to consume enough N and clearly generate a nitrate knee. A NPR might be continuously balancing on the line between aerobic and anaerobic (which still allows for denitrification as apparently both can occur at the same time at an ORP in the vicinity of 20mV).
- what will adjustment to the different parameters yield regarding the ORP swings? Think of all the possible fluctuations where each one will have an influence on the slope, extremes and periodicy of the ORP evolution through a cycle: amount of carbon dosed, volume of tank water pumped, frequency of tank water pumped (all at once or at intervals), frequency and volume through the process pump, amount of N in the influent and desired NO3 for the tank (0 versus 1-5ppm). Now imagine the amount of possible combinations with all of these. Some derivations can be made from the Dymico system but a lot of data will be required from different setups to end up with a bullet-proof system that readily adapts to any bio-load and filter combination. But hey, who doesn't like a good challenge

@Aaron: could you post a blown up screenshot from two or three of your last cycles to see if there are any nuances in the values measured? Do you know what the sampling frequency is and/or do you have the individual data points in your Excel?