Wow, that is impressive. I on the other hand have been using the Betty Crocker method. I have a nice set of measuring spoons and cups.
Cool thread
Cool thread
cypho's 36" DIY reef (1 Viewer)
- Thread starter cypho
- Start date
Still working out the kinks in the auto-titrator. The biggest problem is that I have not found a way to verify the accuracy of the results. Here are the test results I got on Saturday - all from the same sample of water.
Auto-titrator: 2.4 meq/l
Auto-titrator: 2.4 meq/l
Auto-titrator: 2.4 meq/l
Auto-titrator: 2.4 meq/l
Hanna Checker: 3.2 meq/l
Hanna Checker: 2.8 meq/l
New Wave: Off the scale HIGH
New Wave: 4.1 meq/l
So depending on who you ask my alk was either way too low or way to high and no one is close to agreeing with anyone else!
On saturday I had no idea who to trust because I knew my pH probe was poorly calibrated so even though the auto-titrator was consistent, I suspected it may be consistently wrong. New Wave was out of calibration fluid and said even on a good day they never stock pH 4 calibration fluid which is what I really needed. So I will have to order pH calibration fluid from BRS before I can find out how far off the pH probe is.
In the meantime, I checked in with the R2R chemistry forum and JimWelsh pointed me towards some interesting info about the endpoint of the titration. Instead of stopping the titration at 4.5 and calling that one datapoint the alkalinity, I can go past the endpoint and look at all of the data to find the inflection point, the point where the change in pH is the greatest, and use that as the endpoint. The beauty of this is that it eliminates the need for a precisely calibrated pH probe.
The downside is that it takes longer because I have to titrate further and I have to go slow over a wider range. But this is totally worth it since I know my pH probe is poorly calibrated. Using the inflection point confirmed my calibration was off - but the correction was in the opposite direction of I expected. The new readings were even lower than before.
I really wish I could verify the results. But now that the pH probe's calibration is no longer relevant, there are only two potential sources of error. Bad reagent and miscalibrated dosing pumps. Everything else is just math. The the reagent is brand new and I have checked the calibration on the pumps multiple times by weighing what the pump thinks is 10ml of RO water and I get 10 gm +/- 0.03, so the results really should be accurate.
Even though I have not gotten it to agree with a commercial test kit, I decided to trust the auto-titrator and raise my alkalinity. For now just manual adjustments but once I see a few more days worth of consistent readings I'll let it automatically dose 2-part.
In addition to improving the accuracy, the new testing method creates these cool graphs.
An interesting thing you can see in this graph is that even though 4.7 feels pretty far from 4.5. The endpoint correction only changes the result by around 0.02 meq/l (0.05 dKh). This shows that even with a fixed endpoint, pH calibration errors do not matter very much.
Auto-titrator: 2.4 meq/l
Auto-titrator: 2.4 meq/l
Auto-titrator: 2.4 meq/l
Auto-titrator: 2.4 meq/l
Hanna Checker: 3.2 meq/l
Hanna Checker: 2.8 meq/l
New Wave: Off the scale HIGH
New Wave: 4.1 meq/l
So depending on who you ask my alk was either way too low or way to high and no one is close to agreeing with anyone else!
On saturday I had no idea who to trust because I knew my pH probe was poorly calibrated so even though the auto-titrator was consistent, I suspected it may be consistently wrong. New Wave was out of calibration fluid and said even on a good day they never stock pH 4 calibration fluid which is what I really needed. So I will have to order pH calibration fluid from BRS before I can find out how far off the pH probe is.
In the meantime, I checked in with the R2R chemistry forum and JimWelsh pointed me towards some interesting info about the endpoint of the titration. Instead of stopping the titration at 4.5 and calling that one datapoint the alkalinity, I can go past the endpoint and look at all of the data to find the inflection point, the point where the change in pH is the greatest, and use that as the endpoint. The beauty of this is that it eliminates the need for a precisely calibrated pH probe.
The downside is that it takes longer because I have to titrate further and I have to go slow over a wider range. But this is totally worth it since I know my pH probe is poorly calibrated. Using the inflection point confirmed my calibration was off - but the correction was in the opposite direction of I expected. The new readings were even lower than before.
I really wish I could verify the results. But now that the pH probe's calibration is no longer relevant, there are only two potential sources of error. Bad reagent and miscalibrated dosing pumps. Everything else is just math. The the reagent is brand new and I have checked the calibration on the pumps multiple times by weighing what the pump thinks is 10ml of RO water and I get 10 gm +/- 0.03, so the results really should be accurate.
Even though I have not gotten it to agree with a commercial test kit, I decided to trust the auto-titrator and raise my alkalinity. For now just manual adjustments but once I see a few more days worth of consistent readings I'll let it automatically dose 2-part.
In addition to improving the accuracy, the new testing method creates these cool graphs.
An interesting thing you can see in this graph is that even though 4.7 feels pretty far from 4.5. The endpoint correction only changes the result by around 0.02 meq/l (0.05 dKh). This shows that even with a fixed endpoint, pH calibration errors do not matter very much.
Last edited:
If you want to get serious, you could always make up your own standard 
They discuss it here.
https://www.google.com/url?sa=t&sou...AA_cQFggoMAE&usg=AOvVaw0pu52XHaeUgiknyUDZmQJ3
Sent from my Pixel using Tapatalk
They discuss it here.
https://www.google.com/url?sa=t&sou...AA_cQFggoMAE&usg=AOvVaw0pu52XHaeUgiknyUDZmQJ3
Sent from my Pixel using Tapatalk
If I could buy a (reasonably priced) standard, I probably would. But I don't think I would trust a DIY standard any more than I trust the values now.
I thought about this more and decided to see how difficult it would be to make an alk standard. Turns out my gut instinct was right. It is not easy to DIY an alk standard.
If understand the math correctly to make a 2.50meq/l standard I would mix 100 liters of RO water with enough sodium chloride to be roughly seawater like and then add 13.2 grams of sodium carbonate. If making a 100L batch was not bad enough, even reagent grade sodium chloride is insufficiently pure (the article discusses how you can measure the impurities in the reagent grade sodium chloride and correct for those impurities) so using regular table salt is definitely out of the question.
I have been knocking my head against the wall trying to turn my new alkalinity graph into a perfectly flat line. When I ran back-to-back tests the results were +/- 0.003 meq/l. But when I have it pull samples from my tank every two hours I am seeing the alk bounce around way more than that. Occasionally changes as big 0.1. That is a small change in the grand scheme of things, but it is 2 orders of magnitude worse than I was expecting based on the previous testing.
Eventually I figured out the problem. If I take a sample shortly after dosing kalk (or other buffer), I will get a result that is too high because the water in the sump has not yet had time to fully mix with the water in the tank. Similarly if I take a sample shortly after topping off with RO water, I will get a result that is too low because the water in the sump is diluted. Even if I do wait for the top off water to be fully distributed, the salinity drop of an average topoff event will reduce the alkalinity by a detectable amount. I disabled all top-off and dosing for a while and all of the biggest jumps disappeared, confirming my theory.
Moral of the story is that I built this thing too precise. It is not testing errors, the alk really does bounce around a bit. I just need to zoom out a bit on the graph and show the results with fewer decimal points.
I have also tweaked the testing routine to speed things up. Pre-dose more reagent, and use fewer data points over a narrower pH range.
Since I can use the calculated curve to estimate between data-points, I don't think I have lost much precision. And I have cut the test time it takes for the test to run in half.
Eventually I figured out the problem. If I take a sample shortly after dosing kalk (or other buffer), I will get a result that is too high because the water in the sump has not yet had time to fully mix with the water in the tank. Similarly if I take a sample shortly after topping off with RO water, I will get a result that is too low because the water in the sump is diluted. Even if I do wait for the top off water to be fully distributed, the salinity drop of an average topoff event will reduce the alkalinity by a detectable amount. I disabled all top-off and dosing for a while and all of the biggest jumps disappeared, confirming my theory.
Moral of the story is that I built this thing too precise. It is not testing errors, the alk really does bounce around a bit. I just need to zoom out a bit on the graph and show the results with fewer decimal points.
I have also tweaked the testing routine to speed things up. Pre-dose more reagent, and use fewer data points over a narrower pH range.
Since I can use the calculated curve to estimate between data-points, I don't think I have lost much precision. And I have cut the test time it takes for the test to run in half.
Last edited:
If your readings were sampled one sump chamber upstream from where top off and kalk are added, would that make a difference in stabilizing your readings since the water wouldn't be concentrated or diluted, or not really?
That would probably reduced the fluctuations somewhat. But that is not possible in my sump. I don't have multiple chambers and even if I did, both the tank water and the in-sump filters drain directly into the same filter sock so there is no location in the sump I can pull a sample of water from that has not been contaminated by sump-water.
I have programmed my controller to stop dosing and top-off for 10 minutes prior to each test. That did greatly reduce the fluctuations.
It would only require changing one line of code to have it slowly draw the sample over the course of an hour, then each test would represent the average alk over that hour. That should eliminate evidence of any short term alk fluctuations. But I don't really love that idea. Now that I have convinced myself that the results are accurate, it does not bother me too much to see the fluctuations.
If you are still considering doing a slow sample draw, I would make sure the pH isn't dropping substantially in your sample chamber. It might be overly cautious, but something worth looking at.
That would not be a problem. Changes in pH due to changes in CO2 will not have any impact on the result. Evaporation in the sample chamber would not matter either. The only thing that would be a problem is calcification - and I could pre-dose a bit of HCl to ensure the pH is low enough that there will be no calcification.
However I realized there is a big problem with my slow-draw idea. The pH probe would be out of the water for an hour and that would damage the probe.
There was an interesting debate on that on RC a while ago, but a moot point nonetheless in your case. http://www.reefcentral.com/forums/showthread.php?t=2130613
nate_mcnasty
Senior Member
I could have missed this while reading through this but I had a couple questions regarding your auto-traitor
How is the sample chamber being mixed during the titration? Is there a stiring mechanism? it seems like your titration curves look good so my guess is you are.
Do you ever notice bubbles trapped under by the probe? or considered a smaller probe so you could use a smaller sample size? like these probes, they are an additional cost though.
Have you considered using a Gran Function plot instead of the inflection point? I believe it accounts for sulfates giving a more accurate reading.
How is the sample chamber being mixed during the titration? Is there a stiring mechanism? it seems like your titration curves look good so my guess is you are.
Do you ever notice bubbles trapped under by the probe? or considered a smaller probe so you could use a smaller sample size? like these probes, they are an additional cost though.
Have you considered using a Gran Function plot instead of the inflection point? I believe it accounts for sulfates giving a more accurate reading.
In that discussion they say that the initial CO2 concentration can move the endpoint. That would cause problem if you are relying on a die to change color at a particular pH to signal the endpoint.
But since I am titrating past the expected endpoint and then finding the endpoint by calculating the inflection point of the least squares poly fit, the endpoint can move around all it wants and it will not mess with the test. This also lets me get an accurate value no matter how poorly calibrated the pH probe is.
Eventually I am planning to post the whole project in sufficient detail that anyone could recreate it.
I have a small motor with magnets attached under the reaction chamber and a magnetic stir bar in the chamber.
I have not thought to look for bubbles. Not sure if it would matter though, I seem to get OK results when the probe's junction is just barely touching the sample - and in that case the whole room is a bubble.
I don't feel any need for a smaller sample size. With my current vial (a shotglass) the minimum sample size is 10 ml. With a 10 ml sample it just barely touches the junction so I have been using a 12 ml sample just to be safe(although I get the same results using the 10ml sample). From a tank water consumption side of things, that is plenty small. And if I want to reduce reagent usage I can just use a more potent reagent. Plus, that probe looks really fragile.
I am considering switching to a liquid filled probe instead of a gel filled probe so I can speed up the titration. The main thing holding me back is I can't find any good stats on how often you have to refill the liquid in a liquid probe. Everything I read just says check and refill the probe if needed each time you use it which is not possible since the probe is constantly in use.
I'm embarrassed to say I have no idea what a "Gran Function" is. But you can bet I'll be googling that this evening.
I wouldn't expect that you would need to refill super often since the water has a fair bit of ionic strength. I could be way off base however.
Sent from my SM-T800 using Tapatalk
Fair point. I didn't think about it in terms of inflection as end point
Sent from my SM-T800 using Tapatalk
Sent from my SM-T800 using Tapatalk
nate_mcnasty
Senior Member
I'm not sure what you mean by this. Having a good connection with the fluid and junction will stabilize the reading more quickly. The plastic pH probes that protect the bulb often times trap a lot of bubbles in the plastic guards with vigorous stirring. I prefer the all glass pH probes for analysis because of this. It also makes it easier to see if the fluids in the probe look ok too. The plastic ones make sense in a aquarium, because they are a lot more durable. That being said, I agree with you, your titrations look good and it's probably not a major contributor to error.,
This is because it is dependent on how often the probe is used. The fill hole needs to be opened when it is in use. This allows the fluid to flow through the junction. If the fill hole is closed it can take longer for the reading to stabilize. I'm not sure how long it would take to deplete but I'm guessing if it siting in higher ionic strength solutions, like seawater, it would deplete more quickly. The storage solution for these probes is a KCl solution that is similar to the concentration of the internal probe solution, which prevents fluid transfer between the junction. I think it would be something to check weekly maybe refilling monthly. If you could keep the probe stored in a storage solution it would be even less frequent. Don't store it in RODI.
I think this is code that uses a modified gran plot function that doesn't require a calibration. I'm inept when i comes to programming so I don't know if it's useful
I've been tweaking things constantly both with the testing and with the dosing so I'm not sure what I changed that did the trick, but I got a perfectly flat line last night & this morning.
2017-12-08 01:20:34,292 INFO The alkalinity is 2.83795800402 meq/L
2017-12-08 03:19:40,212 INFO The alkalinity is 2.80019799011 meq/L
2017-12-08 05:44:44,120 INFO The alkalinity is 2.84770074243 meq/L
2017-12-08 07:19:44,748 INFO The alkalinity is 2.8408669927 meq/L
2017-12-08 09:19:05,919 INFO The alkalinity is 2.83794258621 meq/L
2017-12-08 11:19:51,419 INFO The alkalinity is 2.83605644593 meq/L
2017-12-08 13:20:02,344 INFO The alkalinity is 2.83562270995 meq/L
2017-12-08 15:19:47,253 INFO The alkalinity is 2.84093273951 meq/L
For some reason I find the sudden stability just as troubling as the previous fluctuations. Was my alk really stable +/- 0.01 over 12 hours?
2017-12-08 01:20:34,292 INFO The alkalinity is 2.83795800402 meq/L
2017-12-08 03:19:40,212 INFO The alkalinity is 2.80019799011 meq/L
2017-12-08 05:44:44,120 INFO The alkalinity is 2.84770074243 meq/L
2017-12-08 07:19:44,748 INFO The alkalinity is 2.8408669927 meq/L
2017-12-08 09:19:05,919 INFO The alkalinity is 2.83794258621 meq/L
2017-12-08 11:19:51,419 INFO The alkalinity is 2.83605644593 meq/L
2017-12-08 13:20:02,344 INFO The alkalinity is 2.83562270995 meq/L
2017-12-08 15:19:47,253 INFO The alkalinity is 2.84093273951 meq/L
For some reason I find the sudden stability just as troubling as the previous fluctuations. Was my alk really stable +/- 0.01 over 12 hours?
Last edited:
That's the thing. Everything I read assumes you are closing the hole and putting the probe into a storage solution between uses. In that scenario it is not too big of a deal how often you have to refill, which is probably why no one talks about it.
But in my case, there is no storage. Unless I build a tiny robot arm capable of plugging the hole between tests, the hole would be open 24/7. When it comes to routine maintenance, I am very lazy. Checking/filling the probe once a month would be on the frequent side of acceptable.
I should probably stick with the gel filled probe. Although, a tiny robot arm does sound like a fun challenge.
I spent several hours reading about gran plots and although I can't say I fully understood it all, I have not read anything that convinced me a gran plot would be better than what I am doing now. It seems to be more of a backup when you are unable to collect data at the inflection point or when it is otherwise difficult to find the exact inflection point than an inherently better method.
Since I don't have those problems, (and since I have not yet figured out how to calculate the gran endpoint) I think I will stick to the inflection point method.
Also, this USGS manual suggests that the inflection point method is more appropriate here. ( alk > 0.4, high conductivity, and primarily carbonates )
https://water.usgs.gov/owq/FieldManual/Chapter6/section6.6/pdf/6.6.pdf said:
Last edited:
I would like to vote for a little servo powered robot arm or rubber lined clamp. Bonus if you use tiny doll hands like in an snl skit along with a tiny tambourine monkey head that turns and looks at you each time it does it (o;