Airflow reactor manifolds (Part 10)

Background

This is part of a series of project pages, describing an attempt to build cheap controlled temperature reactors, capable of incubating microtubes with different biological reactions used for DNA engineering and diagnostics. Commercial instruments cost thousands of dollars, and even self-built devices can cost $100's. One constraint is the use of metal heating blocks, with attendant machining costs and heating requirements. Some early polymerase chain reaction (PCR) instruments employed air heating, albeit in relatively expensive instruments from Roche and Corbett.

This series of projects is summarised at https://www.hackster.io/jim-haseloff/summary-of-airflow-reactor-development-327330. The aim is to explore the use of low cost air heating devices, computer controlled fans for switched airflows and heat resistant plastics for 3D printing to create very low cost and open platform for education, screening in health and agriculture and globally accessible research.

At the time of writing (Aug 2020), and after a number of trials, the general form of the reactor has been established. Different aspects of the design are being tested using a programmable test-rig (https://www.hackster.io/jim-haseloff/programmable-test-rig-part-1-d7df62) and running the heating/cooling elements through different regimes with digital temperature sensors installed in the vessel. A closer description of the design and assembly of the device can be found at: https://www.hackster.io/jim-haseloff/refactoring-the-airflow-reactor-design-part-9-b64e2a. That project also describes improved designs for the mixing chamber of the reactor, and the cooling fan assembly.

Airflow manifolds

This project tackles the design of airflow manifolds, designed to support microtubes, and direct airflow across their surfaces for efficient heating and cooling. In future designs, optical sensors will be integrated into the manifolds in order to provide real-time monitoring of reactions.

Manifolds were 3D printed in heat resistant material, and designed to drop into the reactor vessel as a modular unit, sitting over the exhaust outlet for the fan heater component. Heated air was directed through the manifold, and over the surface of microtubes, or other samples, mounted on the top of the manifold. The short distance allowed for efficiency of heat transfer, but even, precise and predictable heating was also required.

Several different manifolds were designed and tested in the Mk II airflow reactor - to get started, several design constraints were adopted. (i) 8-tube (200µL) strips were assumed for the sample format. (ii) The geometry of the fan heater exhaust and the tube strip dimensions were different (55mm diameter circle vs. 80mm x 10mm strip) - so air channels needed to be fashioned for even heating. The initial approach was to fit the sample tubes into a column of heated air, experiment with different size and arrangement of channels, and test the prototypes.

Manifold 1

Single tube strip with a bifurcating arrangement of merging channels created by lofting elipse and circular cross sections - corresponding to adjacent paired tube or track outlines in the manifold block.

Manifold 2

Single tube strip with channels created by lofting alternate tube outlines and tracks and progressively merging.

Manifold 3

Manifold 4

Manifold 5

Comparison of sample heating profiles

The Mk II reactor vessel included the Mk II lower cooling fan housing and was operated with a cycle of 95ºC 10 min; 40ºC 10 min; 60ºC 10 min. Yellow traces corespond to heater activity. Blue traces indicate the set temperature. Red and green traces show the temperatures of sensors placed in a middle and outer tube of an 8-tube strip in the particular manifold that was being tested. Details of the five different manifolds are shown above.

Observations:

(i) All of the manifolds show some temperature differences between the inner (red trace) and outer (green trace) sample tubes, particularly with fast heating and higher temperatures. All performed better at 40ºC and 60ºC. Some of the temperature differences were unacceptable, several degrees or more at some points. More efficient mixing or even distribution of air is required. Manifolds 3 and 4 performed best - and these may be sufficient for initial biological testing.

(ii) Thermal oscillations were evident, particularly at 95ºC - again manifolds 3 and 4 perform better. In the simple control system being used, the fan for the heater was always energised. A DS18B20 digital sensor was used to read the temperature (every 100msec) inside a middle tube in the sample rack. When the target temperature was reached, the power to the heater was turned off, and the cooling fans were energised. It was noticeable that this would sometimes be accompanied by a small initial increase in temperature - as if activation of the intake and exhaust fans transiently slowed or shifted airflow past the sample space, or pushed a new reservoir of warmed air to the samples. This effect was most noticable at 95ºC. Previous attempts to minimise oscillations through adjustment of the PID controller had made little difference to the system, while adjustments to the geometry of the mixing chamber had led to clear improvements (https://www.hackster.io/jim-haseloff/refactoring-the-airflow-reactor-design-part-9-b64e2a). Some complex pneumatic effects are possible due to the recirculating and dual intake geometry of the reactor vessel, and the blower type cooling fans have a higher flow rate than the heater fan. Two possible ways forward are:

1. Reduce the resistance to air flow in the manifolds by increasing the diameter of passages for air flow. Rethink the arrangement of baffles to minimise air resistance, and create evenly balanced airflow for the sample tubes.

2. Experiment with modified fan heaters with higher powered fans. I have already experimented with CeraDyna high air flow fans from ARX (6, 500 and 7, 000 rpm with ceramic bearings, £15), which showed marked improved performance in early heatsink-based prototypes. These are 60mm x 25mm fans, but can be adapted to replace the 60mm x 15mm fan on the car heater component. This might help drag air through the recycling loop in the vessel, and better match the power of the cooling fans. This will require a small change in the design of the 3D printed vessel.

Revised manifold designs

Manifolds 3 and 4 share a similar basic design - with a conical inflow duct, split arrangement of passages for airflow to two rows of tube-strips, and a wider (11mm diameter) mounting holes, to allow less obstructed flow of air past the sample tubes. The two manifolds differ in the presence of an inset region around the sample tubes for manifold 4, to facilitate air flow around the mounted tube strips for heating/cooling. In order to more systematically and empirically explore manifold properties the observations shown above were used to make some educated guesses about potentially useful design adjustments, and a series of manifolds were printed for comparison.

Several elements were incorporated into a base design. (i) The manifold supported a single 8-tube strip for simplicity (with the aim of incorporating a detection system later). (ii) The outlets that ran past the sample tubes comprised a series of fused tubes that were larger (11mm x 20mm), elliptical in cross-section and reduced resistance to air flow through the system. (iii) The inlet port adjacent to the fan heater was made conical in cross-section, similar to Manifolds 3 & 4 above. It was reasoned that the directed inward flow of air might aid mixing. (iv) A smaller curved baffle was printed, suspended on ribs in the centre of the pathway for the entry port, and shielding all of the exit ports behind it. The aim was to minimise air resistance, while promoting mixing and even, indirect airflow through the manifold.

This was printed as a test manifold, to check its performance, and then to systematically remove or modify features to empirically determine the relative contribution of the different features.

Standardised manifold for testing

The manifold was designed in Fusion 360 as a drop in module, and variants were printed using Extrudr GreenTEC Pro filament. The heat resistant material is proving very useful for exploring this type of application, with enzymic reactions that can reach over 100ºC. Each manifold requires approximately 8 hour printing time.

The test manifold was run through the same set of temperature steps described above (using p=1 setting for the PID controller). The revised manifold performed especially well in terms of even heating across the different sample tubes - with the best performance of any manifold so far tested. The temperatures of the inner and outer sample tubes only diverge under rapid heating - the inner tube readings extend to 2-3ºC higher than the outer tube, and converge as the chamber reaches the set point. The manifold has a internal baffle, which may help with mixing and even distribution of airflows past the sample tubes. The reduced size of the baffle may help reduce overall resistance to air flow (compared to the full-length baffle used in Manifold 5).

The amplitude of the temperature oscillations around the setpoints was higher that Manifolds 3/4 (shown above), similar to the performance of Manifold 5, which was also printed with an internal baffle, so there is no direct air path through the manifold. A series of modified versions of this manifold are being printed to test the contributions (positive and negative!) to the performance of the test manifold.

Even heating during a mock PCR test:

The new manifold showed even heating of the sample tubes during 10 min temperature shift experiments. In order to test performance during more rapid temperature cycling, the manifold was subjected to a mock PCR reactions - cycling between 95ºC 1 min; 55ºC 1 min; 72ºC 2 min. The gratical markings are set 10ºC apart on the Y-axis, and 100 secs on the X-axis.

The temperatures of inner and outer sample tubes remained close throughout the cycling conditions of the PCR reaction. On rapid heating, temperature differences of 2-3ºC appeared on approaching 95ºC, but otherwise temperatures remaining within ~1ºC.

Systematic testing of variant manifolds

Taking a biologically inspired empirical approach to better understanding manifold behaviour - I have taken the "standardised" manifold and created mutant version for testing, Design variants were printed to better test the effects of particular manifold features. The "control" manifold shared (ii) internal baffle, (ii) conical inlet port, and (iii) 11x20mm elliptical outlets around the sample tubes. The curved surface of the internal baffle has two possible benefits - first, it is more feasible to print without supports, and second, airflow across the surface might benefit from the Coanda effect (https://en.wikipedia.org/wiki/Coandă_effect), which would aid flow into the body of the manifold.

Testing the mutant manifolds

The manifold designs were printed using Extrudr GreenTEC Pro filament, and dropped in as replacements inside the Mk II Airflow Reactor. The device was operated with a stepped temperature cycle of 95ºC 10 min; 40ºC 10 min; 60ºC 10 min (as above). The sample heating/cooling profiles were collected and captured for comparison.

Observations:

(1) The control manifold performed consistently, and the trace of sample temperature looks similar to previously seen traces (shown above). Inner and outer tube positions show very similar temperatures through the different heating and cooling steps, which is an excellent feature for the reaction vessel.

(2) The removal of the internal baffle resulted in the appearance of marked temperature difference between inner and outer tube positions. especially under rapid and high heating, and cooling - which were not evident when the baffle was present (1 & 4). This confirms the important role of the baffle in avoiding uneven heating due to direct airflow from the heater through the central tubes of the sample holder. The baffle would be expected to break up the airflow and direct it in a turbulent fashion around the manifold. This might be at the expense of increased air resistance for flow through the manifold. It was noticeable that the sample temperatures were maintained with lower amplitude oscillations at the setpoint temperatures, compared to those using manifolds with baffles. This supports the growing conviction that these temperature oscillations may result from periodic activation of the relatively powerful cooling fans pushing against the resistance offered by the lower flow rate fan heater with the manifold-sample stack, and triggering a transient backflow through the recirculating system - accounting for a small delay in cooling that is a function of air resistance in the normal direction of flow. The oscillations are relatively minor but might be further improved by replacement of the heater fan with a higher power fan. Plans are underway to test this.

(3) The removal of the narrowing conical inlet port, and replacement with a wide cylindrical inlet results in degraded perfomance of the manifold. Temperature differences emerge between the inner and outer tubes of the sample rack - indicating uneven distribution of heated air through the manifold. It is likely that the positive effect of the conical inlet is due to the way it is positioned to direct incoming heated air onto the baffle, and to facilitate even mixing and airflow inside the manifold.

(4) Expansion of the outlet ports adjacent to the sample tubes made little difference to the performace of the unmodified manifold - suggesting that the aperture of the outlet ports is not rate limiting.

Future work:

I am building a revised vessel design to house more powerful 60mm x 25mm fans (replacing the 60mm x 15mm standard fan). This will allow testing of ideas for the balance of air pressures in the vessel, and to see whether it is possible to attenuate setpoint temperature oscillations.

The standardised control manifold would potentially benefit from aerodynamic modelling, and further experimentation, to see whether the performance could benefit from better exploitation of aerodynamic properties like the Coanda effect.