Before the Dawn of IP Telephony - Part 17"BAKUON GA GINSEKAI NO KOUGEN NI HIROGARU" (December 1999 ~ January 2000)
These contents translated a serialization article carried by ITPro IP telephony ONLINE published by Nikkei Business Publications, Inc. Jump to the original (Japanese).
Shinji Usuba
General Manager
eSound Venture Unit
Oki Electric Industry Co., Ltd
We were at the final stage of line unit (LU) development. It was then discovered that the LSI, which had already entered mass production, did not perform sufficiently under certain conditions during software assessment. The situation seemed hopeless with no measures to be found.
In December 1999, LU development of OKI's IP-PBX seemed to be at a dead-end with the fatal problem of LSI signal processing capacity. Just when I was feeling like abandoning the project from absolute despair, I was stricken with an idea. I had remembered that members handling signal processing of cellular phones for the U.S. market were developing the core technology of VoIP gateway. I visited Department Manager Hasumi who was in charge of the team as if I were grasping for straws.
Photo 1 Hiromi Aoyagi, master of signal processing
After explaining the situation, Department Manager Hasumi agreed to let Hiromi Aoyagi, the core engineer in this field and his associate come to the site. Upon their arrival, they immediately picked up the receiver of the IP telephone and started to howl into it.
I'll never forget the scene. As the IP-PBX development members watched breathlessly, they shouted, "BAKUON GA GINSEKAI NO KOUGEN NI HIROGARU (*)"
It was a bizarre scene to say the least. They repeated the words over and over again in a volume far above what one would expect in a normal phone call-they were shouting. Although this is now common during verification in an anechoic room for VoIP related personnel, the bizarreness was beyond words for those watching the scene at the time. But at the same time, there was a strange expectation that they might just be the ones to save us from this potentially tragic situation.
"Let's fiddle with the parameters"
After 10 minutes of repeating what seemed like a spell of some sort, Hiromi and his associate placed the receiver with an expression that they had sensed something. At that moment, the eyes of everyone in the room who had been waiting breathlessly had shifted to the lips of the two to hear what they had to say.
Let's fiddle with the parameters. The words that came out from Hiromi and his associate　were calm and quiet, unlike the shouts they were making moments before. But there was a sense of confidence in their words. And the words felt like a ray of light shining through the darkness.
Although the details are omitted here due to the complexity, the procedure for supplementing the shortage of LSI processing capacity by combining the four technologies of echo canceling, PAD, mute and level diagram was taken.
I was deeply impressed with the accurate analysis and flexible imagination of Hiromi. It was a skill that was possessed only by those who had complete understanding of sound. Some may regard the measure for supplementing the LSI capacity as a deception; however this later became the foundation for the technical advancement of IP telephony toward wide-spread deployment.
In retrospect, many communication engineers visited this anechoic room, buried their heads in their hands and gave us negative comments or words that pressured us. And many times, our expectations had turned into betrayal.
I thought of the solution presented by Hiromi as a true technical integration. This meeting with Hiromi was a stroke of faith that lead to sharing of the same vision even later on.
Remaining issue
January 2000 came. The first customer was already decided, and we had less than a month for the shipment judgment. Although LU development was in its final stage, there was a crucial technical issue remaining that had been giving us a headache ever since the beginning. It was a problem concerning detection error of DTMF (Dial Tone Multi Frequency) signals. It was a phenomenon that co-workers referred to as "missing or garbled numbers."
DTMF is the push-button signals or the series of beeps that are sounded to notify the telephone number to the switching equipment. Since asynchronous, narrow-band data communication was the basic environment of early VoIP, there was a concept of handling voice separately using protocol such as TCP/IP considering the amount of data and making sure voice was transmitted successfully. In the case of VoIP gateway, there is the need to distinguish with certainty PB signals from voice by monitoring the communication status to accurately transmit PB signals over an IP network. In other words, there was a mechanism for identifying PB signals and voice so that wrong PB signals are not sent to the other party when they are accidentally mixed.
In the case of an IP-PBX, however, there are cases when PB signals are received during transmission of voice guidance. Hence, there was the need to pursue contradictory issues of not recognizing PB signals in voice as PB signals, and detecting PB signals even with the presence of voice guidance when such signals are being passed through.
Photo 2 Hachioji region of OKI
At this period just prior to shipment, I also played a role in the cycle of problem solving by being part of the development flow from a standpoint of development management. I acted as a technical liaison among offices including the Shibaura region handling algorithms, Yokohama region handling firmware and Hachioji regional handling LSI development (photo 2) and sent information and program files concerning measures of problems to system evaluation team members by email.
Unification of communication engineers, performing system design and signal processing engineers capable of tweaking with the inside of the DSP as the core of voice signal processing, played a major role in the successful development of IP stage, the first full IP-PBX in Japan. It was a project that consisted of developing the system while toying with the inside of devices. It is a good example of our company's policy of promoting the integration of technologies bearing fruit.
In other words, signal processing engineers in Shibaura did not know the events that could occur, but they would be able to find a mechanism as complete measure if such events can be accurately grasped. On the other hand, communication engineers understand communication and thus can inform the possible events to the signal processing engineers. And LSI engineers in Hachioji understood how such mechanisms can be accumulated. Integration of the technologies of these three parties resulted in the commercialization of the first full IP-PBX in Japan.
Signal processing tuning of VoIP-LSI took approximately two weeks. By this time, many of the developers starting with Koji the Sub-Team Leader were confident in that the product would be complete for shipment. Although there are many developments that never make it into the market, I, too, was confident that this product would be released soon from experience.
Two days before the shipment judgment. And something had occurred.
Sharp pain
That morning, I woke to a sharp pain racing through my back. The pain was so intense that I broke out in a nervous sweat. Just then, I remember the change in my health condition five months ago. It's probably that. Fighting the dreadful pain, I thought about the shipment judgment meeting that was to be held in two days.
- *Japanese for "A sound roars through the highland of a snowy world."
... To be continued
