M777

From TekWiki
Jump to navigation Jump to search

The Tektronix M777 is an amplifier monolithic integrated circuit. As a raw die, it is used in the H2462 hybrid.

About the M777, John Addis said:

M777 was a number I asked IC Manufacturing to reserve because it was a rework of the M377 using SHPi technology. The process name SHPi indicated it was not a significant enough update of SH3 (Super High 3) to be SH4, so somewhere between SH3 and SH4 (which never happened to my knowledge).

I started designing the M777 when I moved from the Laboratory Oscilloscopes Division to Portables Oscilloscopes (“Portables”) in April 1988, just before Lab Scopes was to be abandoned. The reason for this abandonment was that plugin oscilloscopes were not deemed economically competitive with portables because of the added plugin-mainframe interface requirements and mechanical complications. I heard this explanation from Greg Rogers who managed the 11K disaster. Maybe it was his excuse for the 11K disaster, but it became the rationale for not doing another line of plugin scopes. The 7A22 plugin’s popularity was not enough justification for a version in the 11K line of instruments, let alone in a later plugin scope family. Spectrum analyzers and samplers were complex enough to justify separate stand-alone instruments that already existed, but not another lab scope plugin.

First a little history.

The reason the M377 was designed was that the 2465 parts would not fit into a 4 channel 11K plugin. These parts were almost 2” by 3” because they were hybrids.

Why were they hybrids? Because Portables had not figured out how to do a DC level shift or bandwidth limiting on a monolithic chip. Portables Division did need these capabilities although the requirement was not as essential as in the 11K series. 7K and 11K (which was intended to use most of the 7K plugin to mainframe interface for simplicity) signals came in at 0 V at the front panel and exited the plugins at 0 V (common mode) to the mainframe. With all NPN transistors, every stage ends up at a more positive voltage output than its input. Somehow you have to get back down to 0 V common mode.

7K and 11K plugins were required to enter the mainframe at 0 V. This was done in different ways in the 7K series.

The 7A11 (1969) used PNPs at every other stage to go back and forth between +7.1 V and –7.1 V with cascode amplifiers. The PNPs were 4 GHz, and this resulted in the fastest 1 MΩ input of any 7K plugin.

The 7A18 (1971) used common base PNP transistors to get back down to the 0 V common mode signal required by the main frame. The 7A24 and 7A26 later used the same technique. The 7A18 also used one of the earliest Tektronix ICs, the 155-0022-00 channel switch, made with the 50/450 IC process (referring to the two sheet resistivities in ohms per square used in fabricating the IC).

The 7A19 (1971) used a classic folded cascode, a fast NPN common emitter transistor followed by a PNP common base transistor. Here was only one stage. Trouble is, fast PNPs were only available in discrete devices, (and at that point fast NPNs were also only available in discrete devices too, at least at Tektronix), so the amplifier was a hybrid on ceramic with a nice heat sink. It was also a simple plugin, having no Variable Gain control or BWL filter.

The 485 (1972), being a portable, could afford to allow common mode signals to climb positive with each stage. There was no level shifting. It was the first use of the M84, later used by Tom Rousseau in the 7A26, and 7A24. One three pole BWL filter was switched in using the other output of an M84 (155-0078-xx), and some Ls and Cs on the ECB in the main vertical amplifier.

The 7A26 (1974) used M84s (155-0078-xx), simple (SH2 IC process) amplifiers whose outputs were about 3.2 V more positive than their inputs. At the last stage (where drift is less important, there was a common base PNP to get back down to 0 V. This is a variant of what is called a folded cascode where an NPN common emitter stage drives into a PNP common base stage. It’s a cascode except that the output is folded over in the middle of the cascode to make the common mode output voltage 0 V. The PNP and NPN stages both get their DC current through a resistor to a + supply. It’s a variant because the NPN part is actually an IC (the M84).

The 7A29 (1979) used Zener diodes mounted on transmission lines on the ECB between stages. The hybrids contained only metal patterns and thin film resistors, aside from the IC, so they were very simple. There was no bandwidth limit circuit, as had always been the case with 50 Ω inputs. (The 485 had a bandwidth limit circuit, but it also had a 1 MΩ input, necessitating the BWL circuit).

The 2465 (1984) used Zener diodes mounted on the hybrid which also mounted the amplifier IC. In other words, the IC did not have an on-chip level shift because no one had figured out how to do that in a monolithic process. Same for the bandwidth limit, no one had figured out how to do that on a monolithic chip either, so discrete capacitors and inductors were mounted on the hybrid. This and a number of thick film resistors made the hybrid quite large.

The 11K (1987) was not possible without a new IC in the plugins capable of fitting four channels on a single plugin ECB, so the M377 was designed specifically to be small enough that four of them would fit on a single plugin ECB. Whether it was a hybrid or monolithic did not matter, it just had to be small. The M377 had eight major and unique innovations that allowed additional performance.

  1. Six fixed gain settings allowing a 1 mV/div maximum sensitivity.
  2. Capability of being wire ORd to allow multiple channel operation, eliminating a channel switch IC
  3. On chip DC level shifting, eliminating external parts such as Zener diodes and keeping down the size
  4. On chip bandwidth limiter (two of them, both four pole filters), keeping down the size
  5. A really linear gain vs DC control voltage making automatic calibration and 1% calibrated gain steps possible
  6. Elimination of adjustments for thermals in the transient response which take calibration time and board space
  7. It was a monolithic self-contained IC, requiring no zeners, resistors, or capacitors on the ceramic
  8. Being self contained, the M377 was tested and trimmed in wafer form, all automatically

The 11K series was being abandoned starting in 1988, and along with it, the Laboratory Instruments Division.

Portables management (John Taggart, Rod Bristol) were aware of the M377, more so than I realized. They did need amplifiers for future portable oscilloscopes (the TDS series) because the 2465 parts (preamps designed by Jim Woo) were far too big and way too slow for higher speed portable scopes, which would be the only Tektronix offerings in the future.

They liked the M377 because it had DC level shift, bandwidth limit filters, was monolithic and could be put in a small package.

Taggart and Bristol were not about to embark on a new IC design when the M377 did everything needed for a portable scope. Even forgetting about the 2-3 year development time, there was nothing the M377 couldn’t do functionally and do well.

They hired me to help with implementing the M377 in the TDS540. The M377 was repackaged in a 44 pin J-lead package with just a big heat sink on top and soldered onto the ECB. ECB space was even tighter than the 11K plugins, and the M377 definitely ran hotter.

I also sold another Portable Scopes group on using the M377 as it was exactly what they needed too. I believe this became the 2245, 2245A, 2247, 2247A, and 2252 series. Thus the M377 is used in more than just the 11K and TDS500 series scopes.

My principal job in Portables was to redesign the M377 with the new SHPi process. It would be known as the M777. Bob Woolhiser left Tek’s excellent CAD group and worked with me to implement the M777.

I left Tektronix in April 1991 before the M777 was finished, but it was in capable hands. Woolhiser had responsibility for the M777. Woolhiser implemented the 250 MHz BWL filter in place of the 100 MHz BWL filter in the M377. He too moved on and Rich Huard finished the M777.

The M777 is bond pad compatible in every way with the M377, but faster. And because of the extra bandwidth, it was also a little noisier.

The M377 had a long life, still in production after Maxim Integrated bought the Tektronix IC facility (Building 59). Part of the purchase agreement was support for the Tektronix IC processes, including SH3, SHPi (that included p-channel JFETs), and CPi (included high speed complementary PNPs). I cannot speak with authority about these later years.

There was a chip that included the input IC for the TDS540 series that Art Metz originated and I adapted using the p-channel JFETs. Eventually, I believe, the two chips were combined into a new IC based on both these designs, although I am not sure of this. However, I believe the M377 and M777 continued in production until Maxim stopped.

There was some friction between Maxim and Tektronix as Maxim was bound by contract to supply legacy parts. Eventually (about 1997) the Tektronix processes went out of production.