February 2009

Process Automation

Simulated gas

Researchers automate atmospheric simulation chamber for gas introduction control


  • Emission monitoring simplified with automated atmospheric simulation chamber.
  • Users need only set parameters, avoid human errors.
  • Chamber gives precise measurements in atmospheric pressure.
By Aubri C. Buchanan, Randall Dannemann, and Christopher B. Winstead

The threat of toxic emission is nothing to sneeze at. When industrial users provide data for modeling concentration and transport of greenhouse gases or monitoring continuous emissions, they need reliable measurement technologies. That is why researchers are looking more into ultra-sensitive sensors that measure trace atmospheric gases. Users need a flexible, yet accurate, system to develop and test such sensors. In response, our team in the physics department at the University of Southern Mississippi developed an automated atmospheric simulation chamber that facilitates computer control over all components including gas introduction.


To simplify the process from a user's perspective, we decided to completely automate the process so the user only needs to set parameters. The system was able to simulate atmospheres with resolution in the parts-per-million range through a software-controlled automated process. Automation of this process resulted in reliable and repeatable atmospheric compositions.

The intention of this project was to build a vacuum chamber to simulate any atmosphere in composition with accuracies in the part-per-million range and with theoretical minimum concentrations into the parts-per-billion range. Automating the process facilitates more accurate and reproducible results and allows the recording of much more data. Any error is repeatable because there is no human manually operating the system. And with proper implementation of safety features, it will be harder for an operator to damage the system.

Vacuum chambers

To simulate an atmosphere effectively, we needed to recreate the main parameters of the atmosphere. First, we established the desired composition of the bulk atmosphere with gases in the largest percents.  We selected higher flowrate valves to allow timely introduction of the main gases, while fine metering valves control introduction of trace gases. Another requirement included controlling final total pressure. The most important elements to achieve accurately are the relative molecular concentrations, followed by the total pressure.  In later system upgrades, we will implement temperature control of the final gas mixture.

For this system, we used an 18-inch spherical stainless steel chamber with six 8-inch ConFlat ports, four 4.5-inch ConFlat ports, one 6-inch port, and eight 2.75-inch ports as the containment vessel. We constructed a three-gas manifold with stainless steel tubing, metal gasket face seal fittings, and ¼-inch bellows seal-vacuum valves to allow for gas introduction into the chamber. The chamber has the ability to bake out at a temperature higher than operating temperature in order to remove contaminants from the chamber wall through out-gassing. This helps bring the chamber to a lower ultimate pressure, after the bake-out cycle, and it is inherently cleaner as a result.


We designed the system with the intent to use it primarily for spectroscopic experiments. These include cavity ring-down spectroscopy and integrated cavity output spectroscopy, as well as emission measurements using spectrometers. To attain this versatility, we left ports available for later use. We also left the gas introduction system general in design, from having a multi-valve manifold with quick-connects for changing gases, to high and low gas introduction rates facilitating small gas concentrations and quick pressure rises.


Operation and control of the chamber is completely automated, as are pumps, valves, displays, and data acquisition. The only way to bypass this automated control is to manually pull a relay, disconnect pneumatic pressure from a valve, or direct-wire a voltmeter to the sensors. The user has simulated manual control by use of pushbuttons on the front panel located above the real-time controller.

The first level of automation is in the field programmable gate array (FPGA) and input/output (I/O) modules. At this level, Boolean logic is the only option, as floating point arithmetic and most other higher order programming features are not available. However, for simple logic, the FPGA allows for fast response. We did not fully use it in this project because of the length of time it takes to compile a program and time constraints on the project. (These compile times have decreased, and the functionality of the FPGA has increased in subsequent versions of the software.) Data acquisition and transfer are the only operations at this level; even calibration occurs at a higher level. Yet data sees use to calibrate the I/O modules at this level.

On the real-time controller, users can implement the most basic functionality, calibrate inputs and outputs, implement safety locks and alarms, and interface with the system at this level. By having all the basics at this level, you can disconnect the chamber from the host computer and it will still operate. Although at this level of operation, manual control only simulates with safety soft-wired into it.

Automated atmospheric simulation chamber

The highest level of operation is on the host computer, where you store and use automatic sequences. This is also where you log, graph, and display data in formats not limited to a simple numerical display. Most of the human-machine interface takes place at this level, so we use carefully designed operating instructions to minimize errors.  Software checks are also implemented to ensure variables entered are within reasonable bounds.

Pump-down procedure

The procedure used to evacuate the chamber must be versatile because there are several ways to evacuate the chamber, depending on the current pressure and final pressure setpoint. Also, the status of some of the pumps can greatly alter the procedure. At first glance, the procedure might seem unnecessarily complicated, but such detail is required to reproduce the decisions a human operator would make under manual operation conditions. This will also reduce wear on pumps, total pumping times, and risk of contamination and hazards.

We considered these main criteria to determine the exact procedure:

  • Is the diffusion pump on?
  • Is the chamber pressure setpoint low enough to require the diffusion pump?
  • Is the chamber pressure low enough to open the gate valve and use the diffusion pump?

If you know system evacuation will be necessary more than once in a short time, leave the system pumps on, and close the valves they pump through when you reach the desired pressure. 

Backfill procedure

The procedure for backfilling the chamber with known gas compositions is designed to make precise atmospheres in percentages down to .0013%, or 13 parts per million, using existing measurement hardware. It includes every step going from atmospheric pressure and composition all the way through to a completed atmosphere of the user's chosen composition and pressure.

First, the user must initialize all set points on the graphical user interface (GUI) or the program will return an error and not run. Initializing ensures the program does not become entangled in a logical trap and helps to ensure the user knows what they want before randomly pushing buttons. Because of the latter possibility, errors will list in a descriptive manner to help instruct the user.

Before making the atmosphere, clean the chamber, or at least evacuate it. To accomplish this, call and instruct the auto sequence vacuum virtual instrument (VI) on how to evacuate the chamber. When this is complete, refill the chamber with a cleaning gas such as ultra-high purity nitrogen. You may use other gases depending on the application. The user will be able to set how many times this cycle will repeat. At the end, the chamber will be evacuated, and the manifold will be cleaned to prepare for introducing the first component of the simulated atmosphere.

To make the requested atmosphere, the process will introduce gas until it meets the partial pressure of the first gas (p1); it will then switch gasses and introduce the second gas (p2) until meeting the condition p1+p2. Again, this repeats for the third gas (p3) so total pressure = p1+p2+p3, at which point the VI will go idle. Between each gas, clean the manifold to minimize error in gas introduction. Also, proportional integral derivative (PID) type control tends to ensure a minimum of overshoot when properly tuned and to avoid introducing unnecessary errors into the gas composition because of the difficulties in reducing pressure in accurate and precise amounts.


The hardware used to control gas introduction and pumping includes a gas manifold, two roughing pumps, a diffusion pump, a gate valve, a bypass valve, and a metering valve. With this configuration, you can isolate the high-vacuum pump from the chamber by the gate valve and the chamber pressure, and concentration may range from low milli-torr up to atmospheric pressure, thus allowing the high-vacuum pump to continue operating even though the pressure is high. This type of operation makes it possible to quickly pump down and backfill the chamber, and then pump again because you have eliminated the need to shut down and cool the diffusion pump between cycles. You may also open the manifold to either the diffusion pump or chamber so it can evacuate without contaminating or pumping on the chamber. Evacuating the manifold is useful in creating accurate and precise atmospheres because it means you can introduce one gas, then remove the small leftovers from the manifold, then introduce another gas, leaving the system with little to no hysteresis.



The program to evacuate the chamber is the Vacuum Sequence VI. You can call the vacuum sequence concurrently from two programs: Master Conrol and AtmospherePID. The code for the backfilling sequence is Atmosphere, implemented in a graphical programming language. This is a dataflow type language as opposed to a sequential programming language. The graphical user interface includes an error dialogue box, set points for operation, and general program interface buttons.

The functions of the buttons and numeric controls and indicators on the GUI include:

  • Number of cycles: Determines how many times the system will go down to evacuation pressure, then back up to backfill pressure before the final pump down to evacuation pressure.
  • Back press F: The final backfill pressure the system will attain before stopping.
  • Primary backfill gas: Calibrates the ion gauge to the main gas concentration.
  • Cleaning gas: User can select the gas to use when backfilling during the cycles part.
  • Percent partial pressure (gas 1-3): Allows user to determine the concentration of each gas by percentages. Note this number must add up to 1 as a floating-point double.
  • G1-(1-3): Determines the introduction order of gas 1-first, second, or third, depending on the second number.
  • G2-(1-3): Determines introduction order of gas 2-first, second, or third, depending on the second number.
  • G3-(1-3): Determines the introduction order of gas 3-first, second, or third, depending on the second number.
  • Start/Stop backfill: When the light is on, the sequence will start and continue until finished. When the light is off, the sequence will not start. If the light is on and the sequence has started, the user can press this button, turning off the light, and the sequence will abort.
  • Stop VI: Halts execution of the program. This is not an abort or pause; it is a shutdown.


At present, we are still optimizing control algorithms.  We obtained partial pressures as low as 325.5 +/-.5 mTorr, with the system in its initial control configuration. Also, we obtained a pressure of 9.7 +/- .51 Torr when attempting to reach a set point of 10 Torr.  The first pumping speed test conducted with the system, where the units for the graph are Torr vs seconds. We lost software control at around 100 seconds, but it returned at the first spike at around 3,700 seconds. The program ended at .84 +/- .001 Torr.

The ramp and plateau of the curve at approximately 200 seconds is associated with a cleaning cycle where the program setpoint for the cleaning pressure was .35 Torr. It reached .3255+/- .0005 Torr with the control system as initially tuned.

We have not yet reached our initial goal of introducing a minimum of ~10 mTorr, but the total procedure for evacuating and refilling the chamber with a desired gas combination was successful. With the current error in the measurement of pressure, you can greatly lower the minimum concentration and possibly make it more accurate through careful tuning of the PID control, as well as by tuning the program itself. In fact, the program in general is in need of tuning, as well as the associated other programs for seamless operation. However, there is no obvious impediment to achieving the design goals. 

Future work

We need more work to achieve the initial goals of this project. Most of the software needs some slight adjustment and fine-tuning. We made a hardware change to install a bypass valve to allow the manifold gas to go around the metering valve, therefore, letting the chamber increase in pressure much faster. This made concentrations larger than 50 Torr obtainable much more quickly and allows for the chamber to be ready for service on short order. We should also place more redundancy in the two programs. This means, at critical areas in the software, the program needs to either make sure something is in the proper state or (if it is in a loop or close to one) constantly publish the correct state to the Master Control VI through the corresponding shared variables. We also need to more finely adjust the PID control of the valve. Currently, the top of the error will stop the loop instead of the bottom. It might be advantageous to work on smoothing the chamber pressure measurement; although we need to think about this carefully because too much smoothing could introduce a lag in the response signal, and this could introduce more errors at small pressures where the pressure rise is high compared to the current pressure.


Christopher B. Winstead ([email protected]) is a professor in the department of Physics and Astronomy at University of Southern Mississippi in Hattiesburg, and he is the director of the Signal Research Center. Aubri C.Buchanan ([email protected]) is a Physics graduate from USM and currently works as a research engineer at Radiance Technologies. Randall Dannemann is a former undergraduate student in the department of Physics at USM, a current graduate student at the University of Alabama at Huntsville, and an employee of Miltec in Huntsville.