Talk:Telebit

2006
I don't recall the exact timing, but the FastBlazer was introduced AFTER Telebit acquired Octocom. It had been developed entirely by Octocom, rather than resulting from anything Telebit had been developing. In fact, Telebit cancelled its internal V.fast development plans and pursued the Octocom acquisition specifically because Octocom claimed to be very close to completion of the FastBlazer, a claim which was in hindsight obviously false.

During the ITU development of the V.fast standard (which eventually became V.34), Telebit proposed using an enhanced version of PEP with echo cancellation for full-duplex operation, but this was rejected in favor of a conventional single-carrier approach. Full-duplex PEP would have had better performance on poor quality telephone connections, but this was apparently less of a concern to the committee than the peak data rate. --Brouhaha 21:49, 10 April 2006 (UTC)

Lingering questions...

 * 1) Did the T1500/1600 support PEP, or just V.32?
 * 2) Why did they do the T1600? There was no way to beat the price of the Rockwell stuff.
 * 3) Why didn't the T3000 support PEP? Was it a Rockwell chipset again? Or was the "rumor" I mentioned in the article correct?
 * 4) Why didn't the FastBlazer support PEP or TubroPEP? It was all custom hardware, right?

Maury 13:49, 22 April 2006 (UTC)


 * 1) The T1500 and T1600 did not support PEP.
 * 2) The T1600 was developed because the Rockwell V.32 module actually was more expensive than Telebit's own bit pump design, and because Telebit's implementation had better performance (e.g., tolerance of line impairments).  It was the first Telebit modem using the second generation hardware platform which was intended to support future development of PEP, TurboPEP, full-duplex echo-cancelled PEP, V.32bis, V.fast, etc.  However, the T1600 hardware was not fast enough to support V.32bis or TurboPEP).
 * 3) The T3000 hardware was capable of PEP, but the software wasn't ready.  The Worldblazer used the same hardware and added PEP support.  An upgrade from the T3000 to Worldblazer was offered.  Part of the reason was product differentiation, as was done with the T1500 vs. the T2500.  Telebit never used the Rockwell V.32 module after the T1500/T2500.
 * 4) The Fastblazer was developed by Octocom's modem engineering staff, and the Telebit modem engineering staff had little or no involvement.  The hardware probably could have supported PEP, but the Octocom people probably saw no value in it.

I just recently donated one of the early prototype PEP modems to the Computer History Museum. It was a huge circuit board (more than 24" on a side) designed and built by Paul Baran, David Caulkins, and Eugene Wilson circa 1980-1982, a decade before I joined the company. Single-chip DSP processors didn't yet exist, so it used TRW 16-bit multipliers and hundreds of logic chips.  Recently Paul and Eugene told me of their experiences with the prototype; two were made, and tested on a link between Silicon Valley and LA.  They got very poor performance, and thought something was wrong with them, then discovered that the phone connection was so bad it couldn't be used for speech.

I was given the prototype by the manager of the modem engineering group around 1993 or 1994, because the company didn't want it but he thought it should be preserved for posterity, and he knew that I was interested in computer history.

--Brouhaha 01:17, 23 April 2006 (UTC)


 * So did the T2500 support V.32 and PEP, as it was stated in the original article? If so, was it basically offered as the "upscale" T1500 to customers with PEP installs? Also, if the T1600 could do PEP, from a hardware perspective, was there a corresponding T2600 (or similar) based on the same internals? Maury 20:49, 24 April 2006 (UTC)


 * No, the T1500 was introduced as the "downscale" version of the T2500.


 * There was no T2600. The T3000 and Worldblazer used identical hardware, which was very similar to the T1600, but both the DSP and control microprocessors ran faster in order to handle the higher data rates.  It is unclear whether the slower T1600 hardware might have been adequate for PEP. --Brouhaha 07:37, 25 April 2006 (UTC)


 * I had a T2500, and I clearly recall it supporting 19.2k PEP, which I often attempted from my home in an Alaskan village to PSI in San Diego, but because of satellite echo, rarely achieved. It usually dropped back to v.32 if I recall correctly. (Bryan Lockwood) 71.231.183.123 (talk) 05:47, 4 July 2010 (UTC)

Notes from Jim Norrod
''No, the FastBlazer was designed and being sold by Telebit before the Octocom merger. It was a very inexpensive product completely outsourced. Problem was that there were so many commodity modems at that time. It did not do PEP to keep the cost down so it could compete against other commodity modems. PEP would only provide benefits if it worked with other PEP modems. Actually V.34 provided many of the same benefits that PEP provided. Hope this helps.''

It's diffult to know what exactly he's saying "no" to in the context above, but the question was worded such that he's actually saying "No, Octcom did not design the FastBlazer." Comments?

Maury 20:44, 24 April 2006 (UTC)

Mr. Norrod must have the FastBlazer confused with something else; the FastBlazer 8840 was definitely from the Octocom side of things, and was in fact the main reason for the Octocom acquisition, at least according to what the Telebit executive staff told the engineering team. Telebit did NOT have a V.34 solution (in house or otherwise), and Octocom convinced the Telebit executive staff that their V.34 devlopment was very nearly completed. Presumably it was not called the FastBlazer prior to the acquisition talks, but was introduced under the name FastBlazer 8840 after the acquisition.

The modem that Telebit outsourced before the merger was the QBlazer, which did V.32, not V.34. The product worked well enough, but it was a fiasco from a business perspective, as was the followon QBlazer+ which added v.32bis support.

It is possible that the FastBlazer 8820 and/or TeleBlazer might have been outsourced designs, but that wasn't introduced until long after the merger, when the former Octocom staff were firmly in control of modem development.

As far as the relative benefits of V.34 and PEP go, it was still the case that PEP worked on phone lines with much more degradation than V.34 could. V.34 did offer more graceful degradation of throughput in the face of moderate impairments than the earlier standards such as V.32bis could, but PEP had even finer-grained fallback increments. The big advantage of V.34 was that it was full-duplex, while PEP and TurboPEP were half-duplex. The Telebit modem engineering staff did some research into full-duplex PEP, and believed that it was technically feasible. It was proposed to the CCITT early in the V.fast standards process as a candidate, but when CCITT rejected it the Telebit chose not to develop it, since a standards-compliant solution was expected to bave a larger market opportunity. Unfortunately Telebit was never really able to compete effectively in the V.34 market, because the prices declined so rapidly.

--Brouhaha 07:34, 25 April 2006 (UTC)


 * ...but so was the case for everyone but USR basically. Hayes died the same year as Telebit, and by that point the majority of the "incoming brands" like Supra had already given up. There's only so much one can do when faced with the twin problems of rapid commoditization and a rapidly shrinking market. Hats off to USR, really, for a finely managed transition. Maury 12:44, 25 April 2006 (UTC)

Both the 8840 and 8820 were 100 % Octocom in-house designs, the 8840 being used in the merger negotiations. The fastblazer 8840 was far too expensive to build with two large PCBs and a complex internal construction, hence the introduction of the cost reduced 8820.

The Fastblazer 8840 and 8820 could have easily supported PEP but there was no affection on the Octocom side for this modulation. As a result, Telebit lost for instance Deutsche Telekom as a customer because of the lack of backward compatibility with the installed base of PEP modems in the core network of Deutsche Telekom.

Accuracy in copying material from outside sources
It appears that some of the contents of this article, particularly with regard to Telebit's accounts payable handling after the merger, may have been copied from my web page. (Possibly even by me; I can't seem to get at the history of the old Trailblazer page to check.) Anyhow, some possible inaccuracies may have crept in. My web page states that Telebit stopped paying one of it's semiconductor suppliers in a timely manner. The article says that they were unable to pay the bills on time. I am not aware of enough details of the financials to know whether they were *able* to pay the bills on time; I only know that they didn't do so for at least one particular major supplier, who dropped Telebit like a hot potato as a result.

Also my web page states that the west coast office was closed after the VP of Engineering quit; the article states that it happened after the "head of NetBlazer development" quit. I'm not sure who that would be, but it was definitely the VP quitting that triggered the closure. He had a golden parachute, and the executives undoubtedly didn't want to pay it out, so they waited for him to leave.

Unless there is some objection to it, I intend to edit the "End of the company" section of the article soon in order to correct/clarify these matters.

--Brouhaha 08:13, 25 April 2006 (UTC)


 * By all means, go for it! I am going to re-work the fastblazer/octocom part based on the info above, so note that there will be a larger intro to octocom earlier in the article. Maury 12:22, 25 April 2006 (UTC)


 * Actually I see you've completed most of it already. I did minor cleanup only. Maury 12:41, 25 April 2006 (UTC)

=How does TurboPEP work?= There needs to be a description that explains how TurboPEP works. Right now it says, "WorldBlazer model; essentially a T3000 with the new 23,000 bit/s TurboPEP mode," which doesn't explain to the reader how they increased the speed from ~18000 to ~23000. - Theaveng 14:27, 9 October 2007 (UTC)

Important lessons
I'm not sure the lessons learned section is complete without a slightly more thorough look at the technology platforms and some key decisions. Both Octocom and Telebit were developing a v.34+ modem on (2) TI DSP chips, C25s if I am not mistaken (TMS320C25). US Robotics had decided to develop this generation on a single TI chip - and had begun to price more aggressively to reflect better future pricing. So the chatter about whether the Sunnyvale team or the Chelmsford team would be faster to market was never going to be relevant. This 1994 modem was not economical in the world's major markets at the time of launch. Neilhenry 22:28, 20 October 2007 (UTC)

Memory of early Telebit
The Trailblazer was not really a modem, not in the sense we not think of modems. Instead, it was a very powerful 68010-based computer that acted like a modem. It used a couple of co-processors to do fast D-to-A and A-to-D conversions and digital signal processing.

I remember being in the office I shared with my boss in, I believe 1984, when the chief HW design engineer came in and gave us a set of EEPROMS for the Trailblazer saying that this was the final release of FW for the Trailblazer and we could now start production of the first modems. My boss said that we would do that as soon as I had a chance to do some simple tests on the FW. The engineer literally blew up, shouting that they had been fully tested and we were wasting time. My boss calmly told him that the tests would not take long and we just wanted a comfort level with it. The engineer threw up his hands and stormed away.

I duplicated the EEPROMS and installed them onto two Trailblazers. Then I had those trailblazers connect to each other on a simulated phone connection. I then put a DB-25 connector on one modem looping back RTS/CTS and Tx/Rx, and I pumped randomized data into the other modem. The data went to the other modem, out and back in the loopback, and returned to the first modem, where I counted the bytes and check them for accuracy. I found that I was losing data. Please recall that one of the main Telebit claims was data integrity, a LOT rested on that.

After running the test several times I discovered that I was losing data in multiples of 247 bytes. We checked and rechecked and retested several times. Then we took our findings back to the HW engineer. He exploded again, obviously we were mistaken. We showed him the test and the data and he then took us to the FW engineering group so that they could correct us. The chief FW engineer listened to us, and then the HW engineer spoke up starting to say that obviously we were idiots. But the FW engineer shushed him and said that they were transmitting packets of 256 bytes, but with overhead each packet actually held 247 bytes of real data. Obviously the modems were dropping packets. The HW engineer was aghast!

The source of the lost packets was found and corrected and the problem never made it to the customer, I verified the fix myself. That one stupid little bug might have sunk the company coming at that time, making a lie of data integrity. I continued working at Telebit for another 6 years doing verification testing and creating all the production tests for the modems. I was there when they went public. 2603:8000:2:78C7:E1CF:5520:34BF:E108 (talk) 22:05, 19 June 2024 (UTC)