4 Profound Lessons from the First Flight of PSLV-D1
PSLV, the workhorse of ISRO, launched for the first time on 20 September 1993 as the PSLV-D1 mission. Though the mission failed, a lot of lessons were learnt.
On 15th February 2017, ISRO wrote history. A world record for most satellites in a single flight was made. India's Polar Satellite Launch Vehicle, PSLV, launched the Cartosat-2 satellite and 103 co-passenger satellites into a 505 km polar Sun Synchronous Orbit. This is only one of the numerous feats of the workhorse of ISRO, PSLV. The first flight of PSLV, PSLV-D1, was unsuccessful. Let us celebrate the anniversary of this flight with a lessons learnt article.
Conceptualisation of PSLV
PSLV, the third generation launch vehicle, is the successor of India's SLV-3 and ASLV launch vehicles. ISRO's orbital flight journey started with the SLV-3 rocket on 10th August 1979. Even before this SLV launch, Dr Satish Dhawan, then Chairman of ISRO, formed a study group responsible for defining a configuration for a vehicle capable of placing a 600 kg satellite in Sun-synchronous orbit (500–1,000 km). S. Srinivasan was the head of this group targeting the first experimental launch in 1984–85.
The study group initially considered three configurations for PSLV, to be launched from the Sriharikota Range. At one time, ISRO even toyed with the idea of establishing another range down south from where PSLV could be launched due south. Such a sight will enable straight launch into polar orbit instead of azimuth constraints of the SHAR launch site.
2 × S80 + S80 + L30 + PAS (Perigee–Apogee System)
9 × S10 + L45 + L15 + PAS
4 × S80 + S80+ S20 + PAS
In the meantime, the French Space Agency, CNES, and ISRO came together into an agreement where CNES would transfer the knowledge for their liquid engine, and India would supply 10,000 space-qualified transducers. This agreement resulted in a breakthrough for ISRO. ISRO then developed the Vikas engine, the indigenous version of Viking (French counterpart). Now, a new configuration was considered for PSLV:
6 X S9 + S125 + L33 + S7 + S2
In simpler words, six solid strap-on boosters with 9 tonnes propellant loading for PS0 with a single solid core of 125 tonnes for the first stage, PS1, followed by a 33 tonnes propellant loaded liquid second stage, PS2. And finally, 7 tonnes and 2 tonnes of solid cores for the third and fourth stages, respectively. ISRO received the sanction for this configuration in 1982.
Lesson-1: Liquid over Solid for Forth Stage
Over the next few years of the development phase, many new technologies evolved, but a few critical changes in the configuration were done quite early. The final solid stage would have resulted in an inaccurate injection of the satellite. The liquid stage's ability to be switched off or on mid-operation and change the flow rates makes wins here. This ability can help in the design of a closed-loop control system as compared to the solid stage. Therefore, the final and fourth stage became liquid as well. The propellant loading of PS2 was also increased to 37.5 to augment payload capacity. Thus the configuration of the first PSLV flight, PSLV-D1, was 6 X S9 + S125 + L37.5 + S7 + L2.
PSLV-D1 Mission Objectives
Realisation of the complex integrated mission involving many new technologies related to vehicle, range, and ground stations.
Flight testing the integrated flight system and to inject IRS-1E satellite into a polar sun-synchronous orbit.
Evaluation of the in-orbit performance of IRS-1E carrying multispectral LISS and MEOSS payloads.
PSLV-D1 Vehicle details
Height: 44 m
Diameter: 2.8 m
Lift-Off Weight: 280 tonnes
Number of Stages: 4
Payload Weight: 846 Kgs (IRS-1E)
Launched on 20th September 1993
Lesson 2: Change designs mid-course | Lessons Learnt from ASLV flights
While the development phase of PSLV had started after the sanction from GOI in 1982, the ASLV launch vehicle project was still in the testing phase. Therefore, the first two failures of ASLV necessitated ISRO to review and rework the PSLV Mission Design. Without going into the details of the ASLV flights, I would list a few significant guidelines incorporated. While all these guidelines are lessons learnt, the more important lesson for me is the change of Mission Design during the development phase.
Aim for peak dynamic pressure as low as possible.
Provide for the seasons' worst-case peak wind angle of attack, wind shear and gust values.
Ensure availability of control force during the entire atmospheric phase of the flight.
Introduce real-time decision making during critical regimes of flight.
Aim for positive or neutral static stability margins.
Take into account reliable estimates of dynamic pressure, tail-off thrust, wind and angle of attack while setting the jettisoning energy for stage separation.
In case you have doubts regarding the reasoning behind any of these guidelines, please comment below, and I would revert.
The Bold Decision to Work with Digital Systems
Today, the world around us revolves around digital systems. We work on embedded systems right from our undergraduate levels, and binary coding is taught even in high-schools. Three decades ago, the scenario was different. Space agencies in the USA, Russia and Europe were using analog techniques for control loop design. ISRO decided to go with digital implementation, and that too with a processor with severe performance limitations. They were doing everything in binary arithmetic and programming in machine language. This bold decision became of the reasons for the failure of PSLV-D1.
I should add that the decision itself was not wrong, but the implementation required something extra. This will be indeed clear by the end of this article, so read on.
There were many more design decisions, trade-offs and discussions. Some of these are listed in the article, 'PSLV: The Workhouse of ISRO', by N. Narayanamoorthy in the book, From Fishing Hamlet to Red Planet. I have inspired most of this blog post from the article only.
PSLV-D1 Launch | The Maiden Flight of PSLV
The launch campaign for PSLV-D1 took 115 days as opposed to the expected 90 days. This delay was okay being the first attempt. Finally, the countdown for the PSLV-D1 mission commenced on 16th September 1993. A minor snag on the launch day, 19th September, pushed the launch to the next day 20th September. Overall, the countdown performing for the first time was satisfactory.
The PSLV-D1 lifted off majestically on 20th September 1993 through thunderous applause and a sense of relief. The strap-on boosters, first stage and second stage, performed nominally. Though the mission control room was anxious for the stage separation events, they also went quite satisfactorily. But the mission was not over yet. Immediately after the ignition of the third stage, though control force being available, the pitch error went beyond limits, and the vehicle started tumbling. The result was a mission failure, even though the third and fourth stage propulsion system performed normally.
The post-flight analysis of the telemetry identified three issues, namely the following:
During the second stage separation, two of the retro rockets has not functioned as the initiators were not commanded, indicating a break in the pyro chain electrically.
During the second stage regime, yaw error build-up was excessive and was not being corrected due to specific autopilot problems.
The pitch error was increasing due to an error in software implementation during the third stage regime.
Lesson 3: The Bane of Digital Systems
Further detailed investigations pointed towards an overlook during the in the software implementation. To understand this error better, we need it to understand the mounting of the electromechanical actuators of the third stage (PS3) flex nozzle control system. These actuators were mounted in such a way that the pitch command signal needed polarity inversion before being fed to the control electronics.
Now, the error computation by the Digital Autopilot (DAP) as implemented in the GCP (Guidance and Control Processor) was done in 32-bit fixed-point arithmetic using 2's complement representation and truncated 16 bits and then to effect the inversion. The following article can help in understanding the fixed-point binary arithmetic: https://inst.eecs.berkeley.edu/~cs61c/sp06/handout/fixedpt.html
The overlook was that no one foresaw the possible danger in this scheme during design and simulation. The software, as implemented, had no provision to limit the signal to a predetermined value or to recognise overflow condition. This meant that in case the error became large enough, the bits will not be enough and there will be an overflow. This overflow will then result in no inversion, which in turns lead to the flight going beyond control. All this happened during the PSLV-D1 mission, and it ended in tragedy.
Lesson 4: Quality control and assurance
ISRO realised the above error and corrected it. But such as error should have been detected and solved on the ground before the flight itself. The space agency acknowledged that the concepts of quality control, quality assurance, configuration control are to be applied as much to software as to hardware. This realisation resulted in setting up of an exclusive mission software testing and validation group which would evaluate the entire flight software on different simulation testbeds to eliminate any errors due to singularity, scale factor overflow and other parameters. Also, the documentation process and code walkthrough reviews were made mandatory before the commencement of any software testing.
Learning from the Failure: PSLV-D2
The next flight of PSLV after the PSLV-D1, PSLV-D2, performed flawlessly. Narayanamorrthy Ji describes this achievement as "We can also say that for the first time we successfully orbited an application satellite of the IRS class with a precision of which any advanced spacefaring nation could be proud." The learnings from SLV-3, ASLV, and PSLV-D1 mission, were all responsible for this success and all 47 subsequent successful PSLV flights. Even our interplanetary missions, Chandrayaan-1 and Mars Orbiter Mission, were launched using PSLV-XL. It has been nine months since the last PSLV launch. Hopefully, we will see one before the year ends.
Writing this article gave me the idea to write an article comparing the PSLV-D1 configuration with the current PSLV launch vehicle and list what all has been improved. Let me know if you want me to write such a blog post.