This project page describes part of a series of experiments to design and cheap instruments for DNA engineering and molecular diagnostic reactions - based on air heating and switch airflow.
The airflow reactors are based on use of PTC resistive heaters, fan-driven convective heating, recirculating and switched airflows for heating and cooling of sample tubes. A fan heater device sits at the centre of the reactor, with the fan driving heated air through a manifold directly onto the sample tubes.
In a previous set of experiments, a range of parameters for manifold design were tested, and produced a design that gave even heating with the default fan heater component. This was a component intended for use in car heating systems and is supplied with an integrated 60x60x15mm axial fan. This type of axial fan is widely used for air cooling of electronic equipment, something as exhaust fans, other times as inlet or heatsink/radiator fans. In the first case, fans are optimised for high rates of air flow, and the second, to deal with resistance due to high static air pressure. The design of the fan heater allows relatively simple direct replacement of the standard fan with different 60mm fans.
This project describes the effects of dropping in more powerful fans which could better cope with the high static pressures found in a sealed reaction vessel, and increase the efficiency and speed of air heating in the reactor.
Axial fansA range of 60x60mm axial fans are available, with different aerodynamic properties. Fans with higher air flow rates and better tolerance of higher static pressures can be obtained, which can have different sizes - with depths of 25mm and 38mm. In addition, specialised (and more expensive) 78mm deep fans with a pair of stacked, counter-rotating fans are made for exceptionally high flow rates and high static pressures found in devices like computer servers.
Airflow reactors were redesigned to accommodate different size fans, which involved extending the height of the main vessel in Autodesk Fusion 360, and reprinting using Extrudr GreenTEC Pro filament.
An ARX 6025V CeraDyna Fan (FD1260-C2451E2AL) was fitted as a replacement for the fan of a 12V 150W car heater. The original fan was a 15mm deep 60x60mm axial fan, and this was replaced by the 25mm deep ARX fan. When fitted to the standard Mk II Reactor, the fan-heater-manifold assembly protruded from the top deck of the vessel, but could still be run if the lid was not fully pressed down - so some tests were made with the standard vessel while modified reactor vessels were printed. The "control" manifold with internal baffle and conical inlet, described in Part 10 of the series, was used.
The fitted ARX fan rotates at 7000 rpm and produces a noticeably higher airflow (37.1 cubic feet per minute, CFM) and noise (44 dB). Use of the higher power fan produced faster temperature shifts - both for heating and cooling steps. There was also some evident overshoot on heating, and extra noise or temperature variation at the programmed set points - seen with the higher powered fans.
Mock PCR reactions with ARX 60x60x25mm fanIn order to better evaluate the effect of higher fan speeds on temperature cycling times, the different fan-powered vessels were tested with mock PCR conditions, cycling between 95ºC, 1 min; 55ºC, 1 min and 72ºC, 2 min. A comparison between the standard Mk II setup, and the use of higher powered fan is shown below - again with duplicate runs for the ARX 25mm fan. Generally, repeated runs gave very consistent profiles for all of these experiments.
The use of the control manifold ensured maintenance of closely controlled temperatures across the sample rack. In each case, the heater activity and setpoint target temperatures were plotted for over an hour, and red and green traces, corresponding to inner and outer sample tubes maintained very similar temperatures through the cycling regime.
It should be noted that the Noctua NF-A6x25 fan did not tolerate higher temperature - it ceased working after a few minutes above 95ºC - to be fair, the fan's working temperature is only rated to around 60ºC. It is evidently composed of different types of plastic than the other fans used here. The ARX fans are tested at 80ºC, and not recommended for use at higher temperatures - but have so far worked reliably - where sometimes the temperatures can rise transiently above 110ºC - longer term reliability is something to explore in the future.
The use of the higher powered 25mm fan appeared to promote better heat exchange and substantially faster heating and cooling steps - very helpful for improved performance. The next step was to fully adapt the reactor vessel for the larger sized fans.
Testing higher powered fan with different test manifoldsThe design of the reactor was "stretched" by 10mm to allow proper fitting of the larger 25mm fan (shown in black in the diagram below). The 150W PTC heaters are shown in red, the manifolds in yellow, and the cooling fan assemblies in green.
New versions of the reactor were printed in heat resistant Extrudr GreenTEC Pro filament. The full reactors were assembled as described at: https://www.hackster.io/jim-haseloff/refactoring-the-airflow-reactor-design-part-9-b64e2a
The reactors and test assemblies were hooked up to a custom test-rig with thermal sensors (https://www.hackster.io/jim-haseloff/programmable-test-rig-part-1-d7df62 - and subjected to a standard series of temperature steps (95ºC 10min; 40ºC 10 min; 60ºC 10 min) - intended to examine heating and cooling capacity and stability of temperature control - where different airflow manifolds were used.
The tests confirmed that the control manifold performed best. The higher powered ARX fan produced faster heating and cooling cycles at the expense of some overshoot during heating and oscillations/noise at the setpoint temperatures.
Temperature controlOne objective is to reduce the amount of oscillation and noise in the control of temperatures throughout cycling and step changes. One concern was that there were continued small increases in temperature after the cooling fans were switched on - say, due to thermal inertia, or perhaps because of aerodynamic flushing or blocking of airflows, that could trigger elevated temperatures. In order to get a better view of this, the Mk II Plus25mm vessel was run normally, or with the top and bottom cooling fans switched off - to get a better idea of their contribution to temperature regulation.
It seemed that the top cooling fan plays a relatively more important role in cooling, providing an exhaust for heated air, forcing it's expulsion from the vessel. The switching off of either fan made little difference to the oscillations seen at setpoint temperatures.
The benefits of higher airflow for faster cycling of temperatures led to exploration of more powerful fans. 60mm axial fans are commonly used in computer servers, and are often replaced as part of preventative maintenance. This means than specialised high power fans are often available cheaply - new, as spare parts, or on sources like eBay, as resold components. There are several suppliers of specialised high power axial fans like these - notably Sanyo Denki, Delta, Nidec and Panaflo. These also include dual blade counter-rotating fans with high rpm and the highest flow rates and resistance to static pressure. To start with, I tested ex-server fans with single blades. These high power fans tend to be deeper, and include fans that are 25mm and 38mm deep. 60mm counter-rotating fans can be up to 76mm in depth. Ultimately, three versions of the airflow reactor were printed - designed for 15mm, 25mm and 38mm fans, respectively, used to drive air flow through the heater.
The performance of the 25mm ARX (6025V CeraDyna Fan FD1260-C2451E2AL) was compared to the San Ace 60 (Sanyo Denki 60 109R0612J401) and 38mm San Ace 60 (Sanyo Denki 9G0612G1031) fans.
The three fans were installed in the modified housings, and run through a standard series of temperature shifts (95º 2.5 min; 40º 2.5 min; 60º 2.5 min), using a simple on-off control routine. The relevant XOD nodes are shown above. Temperature sensors were placed in sample tubes in the middle and edge of the rack, and when the middle sensor reached the particular set point, cooling or heating was stopped, then switched on as the target was passed again - "bang-bang" type control.
Both 25mm fans showed improved performance, in terms of speed of temperature cycling, compared to the stock 15mm fan that is incorporated in the car fan heater device. The 38mm fan was considerably more powerful than the others in terms of airflow and resistance to static pressure - however, it performed relatively poorly, especially noticeable on the heating part of the cycle. It is likely that the higher internal air pressures provided by the larger fan result in "leakage" by pushing air through the unactivated cooling fans, and could interfere with return airflow through the system.
Optimising the PID temperature controllerFurther characterisation work was done with the 25mm Sanyo Denki San Ace 60 fan. This was done with particular focus on reducing thermal oscillations at the setpoint temperatures - presumed due to hysteresis in the system, related to heating/cooling effects and/or air flow dynamics. A PID controller was employed in the XOD sketch, and relevant parameters were explored using the standard temperature shift regime.
Proportional, Integral, and Derivative (PID) controllers are widely used in control systems to manage feedback and lagging responses. There are a number of online explanations for how these work - for example: https://www.machinedesign.com/automation-iiot/sensors/article/21831887/introduction-to-pid-control. In particular, the controllers incorporate 3 parameters that need to be tuned. From the above website:
The XOD PID node accepts these three parameters, but the values generally need to be determined, or at least confirmed, experimentally. There is an excellent practical (and graphic) description of this for the XOD node and robot control, from Max Danilin at https://medium.com/xodlang/how-to-program-mbot-with-xod-pid-controller-c3e310f8eceb
Varying the Kp parameter between values of 0.1 to 10 made little difference to the oscillations around the setpoint temperatures - which was unexpected. Lowered Kp values caused mixed heating/cooling cycles as setpoints were approached. These can be seen on the charts above for Kp <1.
Even small variations in Ki made large differences to the temperature cycling profiles, in (at this stage) unpredictable ways.
Varying Kd had some beneficial effects on lowering the amplitude of thermal oscillations at setpoints.
Throughout this and previous testing, I have been concerned to better isolate the effect of the dual switched cooling fans that have been used in the airflow reactor - with control effected by (i) maintaining the fan heater on at all times, (ii) switching off the heater current when required, and (iii) coordinately turning on the cooling fans for intake of ambient temperature air, and exhaust of heated air from the vessel. The fan switching could have major impact on air flow within the vessel, and may not be helpful for the fine adjustments of temperature required at the the setpoints.
Modification of fan coolingTherefore, I adjusted the XOD controller code to only employ the dual cooling fans during major shifts in temperature, and not to maintain equilibria. The cycling temperature shifts are regulated by a simple state machine implemented in XOD, and these were converted to provide outputs that could be used to control switching of the cooling fans. In this way, the cooling fans were only used in the single main cooling step, when the vessel was cooled from 95ºC to 40ºC - and cooling to maintain temperatures at setpoints solely relied on switching off the heater current.
I started to experiment with different PID parameters - where a suitable balance of P and D parameters were required for faster cooling times (to avoid heating pulses as the vessel cooled from 95ºC, and approached the 40ºC setpoint.
Likely PID parameter sets were tested for mock PCR cycling (95ºC 30 sec; 55ºC 30 sec; 72ºC 2 min). PID (1, 0, 5) values showed less overshoot on heating to 95ºC and 72ºC, with cycle times of about 8.4 minutes.
Comments
Please log in or sign up to comment.