I first worked for Charles Rhodes in January of 1959. At that time, he, Ron Olson, and Walt Lowy were working to bring the 526 Vectorscope to the market. The 526 enabled precise measurements of that part of the new NTSC signal carrying the color information. Later, network executives claimed that the 526 probably hastened acceptance of color television by 3 or 4 years.
Charlie continued to contribute to Tektronix in many ways, leading projects, providing insights and revolutionary ideas, but perhaps most importantly, identifying and encouraging individuals having exceptional talent and drive. He finally left Tektronix in 1982, ending up as Chief Scientist at the Advanced Television Test Center (ATTC) in Washington, D.C.
In the summer of 1989 I was visiting Washington to attend a conference with my wife, Irene, as part of our volunteer work and decided to give Charlie a phone call. He told me of his plan to convert the various proposed analog HDTV signals to a common digital format that could be digitally recorded on a Sony HDD-1000 tape recorder, to enable recording, duplication, and playback of source material and test results so that the various proposed standards could be compared and analyzed.
The proponents included:
- Sarnoff Labs, evolved from RCA Labs, fielding two systems, 525/59.94p (progressive) and 1050/29.97i (interlaced)
- Zenith/ATT, with 787/59.94p and innovative proposals for over-the-air transmission.
- NHK/Sony, with 1125/60i, as used in Japan.
- General Instrument – A late arrival supporting the Sarnoff proposals, but using a digital implementation.
I continued with the project work that I was doing at the present, but early in 1990, I got drawn into a proposal whereby Tektronix would manufacture a limited production instrument to be called the "Format Converter". At that time, instrumentation-grade A/D and D/A conversion at the requisite sample rate of nearly 75MHz. was difficult. Additionally, sampling jitter had to be held to less than a few tens of picoseconds, a technically challenging objective, but seemed possible.
I shared design responsibility with Micheal Cranford, a talented and versatile fellow engineer. We developed an easy way of dividing responsibilities and ensuring that all the bases were covered. Paul Barton and Jan Kuderna supported us in aspects of circuit board design and prototype construction. Fortunately, the Tektronix TV Test and Measurement group had built some HDTV-compatible test signal generators, betting that HDTV, in some form or another, was going to become a reality.
Tektronix had also recently developed a series of VXI (VME Extensions for Instrumentation) card cages that had the power, cooling capability, and module size (VX1505) enabling a five module implementation of the Format Converter in a 5 1/4" rack space.
The modules were to be:
- A/D Analog Inputs: Y, Pr, Pb, Composite Sync, H Drive, V. Drive inputs. Functions: Y, Pr, Pb, signal amplification, anti-alias filtering, Analog to digital conversion, sync separation, sample clock generation, frame pulse generation, self-test signal generation and multiplexing. 2*8 bit differential ECL signal, clock and frame pulse outputs to FIFO module.
- FIFO (First In, First Out) Functions: ECL to TTL conversion, demultiplexing, FIFO buffering, multiplexing, TTL/ECL conversion.
- I/O Functions: Microcontroller and user interface, RS-232 interface, line and frame format code generation, data extraction from FIFO, source format ID code insertion (record)/detection (playback).
- FIFO Same as input FIFO.
- D/A Data extraction from FIFO, D/A conversion and reconstruction (anti-aliasing) of Y, Pb, Pr signals, H, V drive and composite sync signal reconstruction.
Micheal suggested using FIFO chips, avoiding the task of calculating addresses for the data. Although that involved demuxing (input) and muxing (output) to TTL levels in order to implement. At that time, chips of that type operating at speed (~75MB/Sec (Y) and ~37.5MB/Sec (Pb, Pr) were unavailable.
Micheal also suggested the use of PALs (Programmable Array Logic) to implement the state machines that would be needed to recognize signal events and generate the signals necessary to operate the phase lock loops (PLL) needed for source and destination timing.
One of the first issues concerned the choice of microcontroller and programming language. Mike and Paul had earlier worked on a project using an Intel 8051 and a compact implementation of the FORTH programming language, with which I was also familiar. Those choices greatly sped up hardware development because, for example, we could know to expect another H sync edge in 2200 clocks if we gave the PAL that data through the microcontroller bus.
So, the state machines on the A/D, I/O and D/A boards could be loaded with the appropriate data, the PLLs would lock, and we could go on to discover the next problem. And there were quite a few, dumb ones like how I reversed the pinout on a 50 pin I/O connector to the more serious problems described later.
The self-test signal generator employed a Tek-made I.C. developed by the TV T&M group, integrating the functionality of a "zone plate generator", a term borrowed from camera designers to test resolution. In essence, it is a mathematical engine generating pointers to a sine look-up read-only memory (ROM). The result, when D/A converted, is a sine function of X(Horizontal), Y(Vertical), and T(frame-to-frame) values in response to user-supplied parameters Kx. Ky. Kxy. Kxt and so on. Using that instead of the A/D to generate the output stream of the A/D module, and bypassing the HDD-1000, produces an output from the D/A module to display on an oscilloscope or feed to a second Format Converter, greatly simplifying test and calibration.
By mid-October, the first prototype staggered into life. The most serious problem was the step response of the A-A filters - they displayed serious overshoot and ringing due to inappropriate ATTC specifications for the group delay of the purchased A-A filters and the inability of the filter designers to design filters meeting time domain parameters. It was clear that the time domain behavior of the filters would need to be specified to produce a result that would be acceptable to the users, the testing agencies and proponents.
The A-A filters were custom, from an English outfit called Faraday Technology Ltd. This presented a problem in that the filters were being designed in England and design iterations would be slow and painful over the 10,000 mile feedback loop. To solve that problem, using a 2T/30 test pulse from one of our Tektronix generators as an example, I took an oscilloscope photograph of the signal from the generator and of the signal after it had passed through the filter. Then using a mathematically generated 2T/30 pulse and a discrete Fourier transform having the magnitude and phase parameters derived from a network analyzer measurement of the filter, I was able to mathematically generate the response, duplicating the overshoot and ringing of the "real" filter. That convinced the designer at Faraday that we could work together using the magnitude and group delay values from his simulations with me doing the discrete Fourier transform calculations (on my Amiga 2500!) at my end to converge on a satisfactory solution. It took about four weeks.
The HDD-1000 came with two different modules - one having a field frequency of 59.94Hz., compatible with the U.S. analog NTSC standard, and 60Hz, compatible with the new Japanese HD standard. It was popularly believed that a low-jitter sampling clock could not be regenerated from the 33.75KHz line sync (or H drive) pulses without using a crystal-controlled voltage controlled oscillator (VCO). I had patented a sampling phase detector a decade earlier, but there was something that had eluded me - I felt it could be improved upon. It turns out that it could be and I used that improvement in the PLL design for the sampling clock so that an LC oscillator having a wider tuning range than a crystal oscillator could be used in the PLL. That enabled us to lock on either field rate and to handle the samples/line variations needed to handle the proponent standards.
The parameters to accommodate each standard resided in the programmable read-only memory of the FORTH operating system and the microcontroller would load the appropriate parameters into the PALs when the user pushed the button. A standards change could occur after only a 250mS delay. In fact, from power-up the Format Converter could do the self-test and start up in the desired standard before the cooling fan got up to speed.
Early in the project I managed to equip our lab with a Tek 2467B "Brighteye" oscilloscope in order to see things like setup and hold time for signals that were related to the frame pulse. It meant that I was trying to measure time intervals of less than 100 picoseconds that were occurring at a 30Hz repetition rate. Although such measurements were much easier later with advances in digital oscilloscopes, the 2467B greatly facilitated making these measurements.
While using the built-in zone plate generator to look at a 20MHz sine wave, I noticed that if I positioned the waveform so that the sine wave packet was just above the top of the display area a trace would infrequently flick below the bottom of the packet. I had uncovered a defect in the A/D converter chips that we were buying from the vendor. Infrequently, a comparator would not produce the binary one-zero determination, but instead would hang up in an in-between state, a condition called metastability. We found that these parts had never been tested by the manufacturer at the conditions under which they were specified to operate. Component selection and careful adjustment of the clock duty cycle turned out to minimize (but not completely cure) the problem. It turns out that this kind of problem exists with tiny but non-zero probability in many products.
While all of this was going on, Charlie Rhodes had kept me apprised of the correspondence between ATTC, NHK and Sony. It had been ATTC's intent to use a multi-standard camera developed by the German firm Bosch-Fernseh. Their camera would supply the V and H drive signals that ATTC was expecting the Format Converter to lock to. However, in reviewing the correspondence, I never saw a clear statement that Sony and NHK were going to generate or use those H and V drive signals. But since some of the HTDV test signal generators we were using to test the Format Converter supplied composite tri-level sync, the standard used in Japan, the Format Converter could respond to and generate either synchronizing signal format.
We were about to fire up the Format Converter in an outside broadcast (OB) van on 42nd St. in New York City in three weeks (no stress!). But about a week after New Year's Day of 1991, I received a call from an engineer at ATTC, telling me that Sony had decided not to use the Bosch-Fernseh camera, instead deciding to use a Sony camera instead. It looked like the ATTC test plan had been derailed. I had been criticized for including the tri-level sync functionality, but now it turned out that it was essential in keeping the test plan on track. My response to the concerned engineer was, "Just keep on doing what you are doing, but officially, I will say we're screwed until we get a revised statement of work including tri-level sync capability".
We had the equipment in the basement of the Ed Sullivan theater, but couldn't really fire it all up at the same time because IBM had rented the stage to do an HDTV extravaganza including a model of the Sistine Chapel and the word, "EXCELLENCE" in eight foot high characters to be dollied up to by 3 HDTV cameras, creating a 48x9 HDTV image.
We finally got things hooked up together in the OB van and everything worked! While we were setting up, two technicians from Sony arrived with their suit-wearing supervisor to perform some modifications on the HDD-1000s. The supervisor seemed very interested in the setup, noticing the Format Converter sitting between the two HDD-1000s.
Him: "Does the Format Converter operate on V and H drive or tri-level sync?"
Me: "Yes, you can select either."
Him: "Does the Format Converter lock to 59.94 or 60Hz field rate?"
Me: "Yes, it doesn't care."
I had suspected that Sony and NHK would take the position of "Why look for something else? We have everything you need." But my belief then and now is that that would never have happened. The Sony/NHK plan involved a transmission bandwidth of nearly 30MHz, compared to the 6MHz we used then, and still use, due to these advances in digital technology. I'm proud of our role in helping to make HDTV a reality and grateful for the chance for us to contribute to the technology.
Since it was a limited production instrument and was to be used only over the evaluation period that was expected to be less than two years, we were responsible for service support and had more freedom over nomenclature, so we called it the FC (Format Converter) 511 (after the model number of the first Tektronix oscilloscope). Jim Geddes came on board to assist with the later phases of the project and he, Micheal, and Jan were mostly responsible for keeping things going as the project was winding down.
We also developed a prototype to support development of the European HDTV efforts (yes, 50Hz field rate with a 60Hz recorder. It worked!), but changes in management prevented that from happening.