KH-11 KENNEN

[Blackstar]

Thanks for th up Hoku. I remember Jim saying that HEXAGON didn't use CMGs, and I thought that I'd read in an NRO source that GAMBIT didn't but I'm now wondering if that's true, especially for the later G3 etc.

There is also an interesting comment from Bob O'Dell's oral history interview on HST:
O'Dell:
(...) We had things that we generally didn't expect, like this business about 2.4 meters being just about the largest size you could package with a low moment of inertia, a point I made in the review panel. That was really the driver, on 2.4 meters. That's what kept the costs down. That was the big break on the costs. The uncertainty of cost in making 3 meter optics was great, but very hard to define, whereas the certainty of the difference in cost between reaction wheels and control moment gyros was clear. That was black and white.
https://www.aip.org/history-programs/niels-bohr-library/oral-histories/4802-1

I've probably told this story before up-thread.

I'm never going to find it, but I remember back in the mid-1980s (probably around 1985-86) reading an article about Hubble that was possibly in Scientific American. The reporter visited Perkin-Elmer and was shown around by one of their top managers who showed off the big equipment they had built to construct and polish the Hubble mirror. The thing that stood out for me at the time was the manager's comment that now that they had built that equipment, it would "make us competitive for other programs." Even though I was but a wee lad, I remember thinking Ah ha! That means that P-E did not build the KH-11 mirrors! And that proved correct. Their HEXAGON mirrors were a lot smaller.

[LittleBird]

Thanks for th up Hoku. I remember Jim saying that HEXAGON didn't use CMGs, and I thought that I'd read in an NRO source that GAMBIT didn't but I'm now wondering if that's true, especially for the later G3 etc.

There is also an interesting comment from Bob O'Dell's oral history interview on HST:
O'Dell:
(...) We had things that we generally didn't expect, like this business about 2.4 meters being just about the largest size you could package with a low moment of inertia, a point I made in the review panel. That was really the driver, on 2.4 meters. That's what kept the costs down. That was the big break on the costs. The uncertainty of cost in making 3 meter optics was great, but very hard to define, whereas the certainty of the difference in cost between reaction wheels and control moment gyros was clear. That was black and white.
https://www.aip.org/history-programs/niels-bohr-library/oral-histories/4802-1

I've probably told this story before up-thread.

I'm never going to find it, but I remember back in the mid-1980s (probably around 1985-86) reading an article about Hubble that was possibly in Scientific American. The reporter visited Perkin-Elmer and was shown around by one of their top managers who showed off the big equipment they had built to construct and polish the Hubble mirror. The thing that stood out for me at the time was the manager's comment that now that they had built that equipment, it would "make us competitive for other programs." Even though I was but a wee lad, I remember thinking Ah ha! That means that P-E did not build the KH-11 mirrors! And that proved correct. Their HEXAGON mirrors were a lot smaller.

You did indeed tell that story   https://forum.nasaspaceflight.com/index.php?topic=29545.msg2295903#msg2295903, and it led Hoku to  an interesting article  https://forum.nasaspaceflight.com/index.php?topic=29545.msg2297210#msg2297210  , which he uploaded a scan from and is attached. However that was in Sky and Telescope, so it may be that the one you are recalling was indeed in Scientific American in mid 80s and may thus still be at large ...

PS I've never found searching within the site's own system very effective, but libra's tip about searching from *outside* works well e.g. "scientific american" "kh-11" site:forum.nasaspaceflight.com in DuckDuckGo worked well

[LittleBird]

Thanks for th up Hoku. I remember Jim saying that HEXAGON didn't use CMGs, and I thought that I'd read in an NRO source that GAMBIT didn't but I'm now wondering if that's true, especially for the later G3 etc.

There is also an interesting comment from Bob O'Dell's oral history interview on HST:
O'Dell:
(...) We had things that we generally didn't expect, like this business about 2.4 meters being just about the largest size you could package with a low moment of inertia, a point I made in the review panel. That was really the driver, on 2.4 meters. That's what kept the costs down. That was the big break on the costs. The uncertainty of cost in making 3 meter optics was great, but very hard to define, whereas the certainty of the difference in cost between reaction wheels and control moment gyros was clear. That was black and white.
https://www.aip.org/history-programs/niels-bohr-library/oral-histories/4802-1

Thanks, yes. Funnily enough when I saw you mention that quote before https://forum.nasaspaceflight.com/index.php?topic=29545.560 I wongly inferred that KH-11 wouldn't have used CMGs either. I couldn't really grok the simple reality that it was apparently OK to use them on about 20 odd KENNENs https://en.wikipedia.org/wiki/KH-11_KENNEN
 yet one for one Hubble it was too pricey ...

Some cost estimates for KENNEN:

Estimated unit costs, including launch and in 1990 dollars, range from US$1.25 to US$1.75 billion (inflation adjusted $2.8 to $3.92 billion in 2022).[36]

According to US Senator Kit Bond initial budget estimates for each of the two legacy KH-11 satellites ordered from Lockheed in 2005 were higher than for the latest Nimitz-class aircraft carrier (CVN-77)[21] with its projected procurement cost of $6.35 billion as of May 2005.[106] In 2011, after the launch of USA-224, DNRO Bruce Carlson announced that the procurement cost for the satellite had been $2 billion under the initial budget estimate, which would put it at about $4.4 billion (inflation adjusted $5.72 billion in 2022).[22]

In April 2014, the NRO assigned a "worth more than $5 billion" to the final two legacy KH-11 satellites.[107]

all from Wikipedia.

[hoku]

Thanks for th up Hoku. I remember Jim saying that HEXAGON didn't use CMGs, and I thought that I'd read in an NRO source that GAMBIT didn't but I'm now wondering if that's true, especially for the later G3 etc.

There is also an interesting comment from Bob O'Dell's oral history interview on HST:
O'Dell:
(...) We had things that we generally didn't expect, like this business about 2.4 meters being just about the largest size you could package with a low moment of inertia, a point I made in the review panel. That was really the driver, on 2.4 meters. That's what kept the costs down. That was the big break on the costs. The uncertainty of cost in making 3 meter optics was great, but very hard to define, whereas the certainty of the difference in cost between reaction wheels and control moment gyros was clear. That was black and white.
https://www.aip.org/history-programs/niels-bohr-library/oral-histories/4802-1

Thanks, yes. Funnily enough when I saw you mention that quote before https://forum.nasaspaceflight.com/index.php?topic=29545.560 I wongly inferred that KH-11 wouldn't have used CMGs either. I couldn't really grok the simple reality that it was apparently OK to use them on about 20 odd KENNENs https://en.wikipedia.org/wiki/KH-11_KENNEN
 yet one for one Hubble it was too pricey ...

Some cost estimates for KENNEN:

Estimated unit costs, including launch and in 1990 dollars, range from US$1.25 to US$1.75 billion (inflation adjusted $2.8 to $3.92 billion in 2022).[36]

According to US Senator Kit Bond initial budget estimates for each of the two legacy KH-11 satellites ordered from Lockheed in 2005 were higher than for the latest Nimitz-class aircraft carrier (CVN-77)[21] with its projected procurement cost of $6.35 billion as of May 2005.[106] In 2011, after the launch of USA-224, DNRO Bruce Carlson announced that the procurement cost for the satellite had been $2 billion under the initial budget estimate, which would put it at about $4.4 billion (inflation adjusted $5.72 billion in 2022).[22]

In April 2014, the NRO assigned a "worth more than $5 billion" to the final two legacy KH-11 satellites.[107]

all from Wikipedia.

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

[LittleBird]

Thanks for th up Hoku. I remember Jim saying that HEXAGON didn't use CMGs, and I thought that I'd read in an NRO source that GAMBIT didn't but I'm now wondering if that's true, especially for the later G3 etc.

There is also an interesting comment from Bob O'Dell's oral history interview on HST:
O'Dell:
(...) We had things that we generally didn't expect, like this business about 2.4 meters being just about the largest size you could package with a low moment of inertia, a point I made in the review panel. That was really the driver, on 2.4 meters. That's what kept the costs down. That was the big break on the costs. The uncertainty of cost in making 3 meter optics was great, but very hard to define, whereas the certainty of the difference in cost between reaction wheels and control moment gyros was clear. That was black and white.
https://www.aip.org/history-programs/niels-bohr-library/oral-histories/4802-1

Thanks, yes. Funnily enough when I saw you mention that quote before https://forum.nasaspaceflight.com/index.php?topic=29545.560 I wongly inferred that KH-11 wouldn't have used CMGs either. I couldn't really grok the simple reality that it was apparently OK to use them on about 20 odd KENNENs https://en.wikipedia.org/wiki/KH-11_KENNEN
 yet one for one Hubble it was too pricey ...

Some cost estimates for KENNEN:

<snip>

all from Wikipedia.

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

Thanks hoku, that's fascinating. One feels the author was instructed not to mention anything built by arch rival  Hughes while constructing that table ;-) (Oops: failed to see the Intelsat entry)

So I am now even more curious about the NRO and CMGs, Skylab, and the KENNEN.

Was Skylab thus probably the first LEO s/c to use CMGs, and if so was this experience directly useful to KH-11 designers ? If so they had more than just National prestige incentive to help NASA out by imaging it in its distressed state.

And did it actually use both the modes you describe-my guess is EREP would be high rate, whereas the solar telescope might or might not be ?

[Edit: also interesting to see gravity gradient described as experimental only-some Navy navsat designers would be surprised to hear this ?]

[Jim]

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

that MMS is not the 4 satellites that NASA launched in 2015.

[LittleBird]

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

that MMS is not the 4 satellites that NASA launched in 2015.

Indeed, and as source  is from  1999 it wouldn't be. Is it the Manned Maneuvering System, which I've seen in a s/c control textbook described as having small CMGs ? [Edit 2:No it isn't ... as that's called the Manned Maneuvering Unit.]

[Edit: Interestingly, on checking this, I see it was actually the AMU of the 70s that I'd seen described as using CMGs but I guess if that's right the later Suttle era system would do too ?]

[LittleBird]

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

that MMS is not the 4 satellites that NASA launched in 2015.

Indeed, and as source  is from  1999 it wouldn't be. <snip>

So in fact it's probably TRW referring to their own 3-axis stabilised GEO stage/spacecraft, the MultiMission Support stage, referred to in 1968 by Dirks in his Blue Book ... which I speculated https://forum.nasaspaceflight.com/index.php?topic=37169.msg2294863#msg2294863  was the basis for RHYOLITE ? Nice.

[LittleBird]

A few months ago, a Maxar satellite imaged the NASA/USGS Landsat 8:

https://landsat.gsfc.nasa.gov/article/satellite-sees-satellite-landsat-8-in-orbit/

On October 7, 2022, Maxar’s WorldView-3 satellite snapped this sequence of images showing Landsat 8 in orbit. The distance between the satellites ranges from between 91.4 to 129.9 km (56.8 to 80.7 miles).
...
The resolution of the Landsat 8 images in this sequence varies from 4.6 to 6.5 cm (SSD stands for Space Sample Distance).

At the time, news of the event was posted in the Landsat 8 tread in this forum. I thought I'd repost it in this thread because this is one of the tasks that the KH-11 KENNEN sats and their successors were designed to do.

The resolution of a picture taken at this distance is certainly impressive, but I imagine that a sat from the Evolved Enhanced CRYSTAL System-series could do even better.

Thanks. Blackstar added some discussion of recent progress in this general area to his nice GAMBIT and Skylab article in TSR:
https://thespacereview.com/article/4588/1 including the advertising snap by Worldview of ISS, see below.
 

[Star One]

Thanks for th up Hoku. I remember Jim saying that HEXAGON didn't use CMGs, and I thought that I'd read in an NRO source that GAMBIT didn't but I'm now wondering if that's true, especially for the later G3 etc.

There is also an interesting comment from Bob O'Dell's oral history interview on HST:
O'Dell:
(...) We had things that we generally didn't expect, like this business about 2.4 meters being just about the largest size you could package with a low moment of inertia, a point I made in the review panel. That was really the driver, on 2.4 meters. That's what kept the costs down. That was the big break on the costs. The uncertainty of cost in making 3 meter optics was great, but very hard to define, whereas the certainty of the difference in cost between reaction wheels and control moment gyros was clear. That was black and white.
https://www.aip.org/history-programs/niels-bohr-library/oral-histories/4802-1

Thanks, yes. Funnily enough when I saw you mention that quote before https://forum.nasaspaceflight.com/index.php?topic=29545.560 I wongly inferred that KH-11 wouldn't have used CMGs either. I couldn't really grok the simple reality that it was apparently OK to use them on about 20 odd KENNENs https://en.wikipedia.org/wiki/KH-11_KENNEN
 yet one for one Hubble it was too pricey ...

Some cost estimates for KENNEN:

Estimated unit costs, including launch and in 1990 dollars, range from US$1.25 to US$1.75 billion (inflation adjusted $2.8 to $3.92 billion in 2022).[36]

According to US Senator Kit Bond initial budget estimates for each of the two legacy KH-11 satellites ordered from Lockheed in 2005 were higher than for the latest Nimitz-class aircraft carrier (CVN-77)[21] with its projected procurement cost of $6.35 billion as of May 2005.[106] In 2011, after the launch of USA-224, DNRO Bruce Carlson announced that the procurement cost for the satellite had been $2 billion under the initial budget estimate, which would put it at about $4.4 billion (inflation adjusted $5.72 billion in 2022).[22]

In April 2014, the NRO assigned a "worth more than $5 billion" to the final two legacy KH-11 satellites.[107]

all from Wikipedia.

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

There was a study of the FIA optics and their suitability before they even moved forward with WFIRST.

[Jim]

Indeed, and as source  is from  1999 it wouldn't be. Is it the Manned Maneuvering System, which I've seen in a s/c control textbook described as having small CMGs ? [Edit 2:No it isn't ... as that's called the Manned Maneuvering Unit.]

No, the MMU did not have CMGs

[Jim]

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

that MMS is not the 4 satellites that NASA launched in 2015.

Indeed, and as source  is from  1999 it wouldn't be. <snip>

So in fact it's probably TRW referring to their own 3-axis stabilised GEO stage/spacecraft, the MultiMission Support stage, referred to in 1968 by Dirks in his Blue Book ... which I speculated https://forum.nasaspaceflight.com/index.php?topic=37169.msg2294863#msg2294863  was the basis for RHYOLITE ? Nice.

Exactly

[LittleBird]

KH-11 has/had different pointing and tracking constraints than HST. According to the attached table from the 1999 edition of TRW's "Space Data", CMGs are the preferred choice for "high" maneuvering rates (like frequent acquisition of new ground targets along the orbital track). HST on the other hand, is optimised for long exposures, and slow slew rates.

that MMS is not the 4 satellites that NASA launched in 2015.

Indeed, and as source  is from  1999 it wouldn't be. <snip>

So in fact it's probably TRW referring to their own 3-axis stabilised GEO stage/spacecraft, the MultiMission Support stage, referred to in 1968 by Dirks in his Blue Book ... which I speculated https://forum.nasaspaceflight.com/index.php?topic=37169.msg2294863#msg2294863  was the basis for RHYOLITE ? Nice.

Exactly

Thanks v much. I see that there is a  TRW brochure from late 60s that hints at what they hoped for with MMS, I'll move that to the RHYOLITE thread.

[Star One]

I wonder how much machine learn is used in the image pipeline from the KH-11, I’d bet that actual humans now only sit right at the end of the pipeline and that much of the initial sorting and interpretation of images is done by machine learning. You probably only need the same number of photo analysis as in the past even though something like the KH-11 must produce vastly more data.

[LittleBird]

I wonder how much machine learn is used in the image pipeline from the KH-11, I’d bet that actual humans now only sit right at the end of the pipeline and that much of the initial sorting and interpretation of images is done by machine learning. You probably only need the same number of photo analysis as in the past even though something like the KH-11 must produce vastly more data.

I have genuinely no idea, but this Verge piece was interesting: https://www.theverge.com/2019/7/31/20746926/sentient-national-reconnaissance-office-spy-satellites-artificial-intelligence-ai

[Blackstar]

https://thespacereview.com/article/4614/1

Spinning towards the future: crisis response from space
by Dwayne A. Day
Monday, July 3, 2023

In the early morning of June 5, 1967, hundreds of Israeli aircraft took off from their bases and headed out over the Mediterranean and the Red Sea before turning toward Egypt. They attacked multiple Egyptian airbases, and soon more than 300 Egyptian aircraft were smoking wrecks with their airfields torn to shreds. Shortly thereafter, the Six-Day War was over.

In August 1968, Soviet tanks rolled across the border into Czechoslovakia, suppressing a movement that threatened Soviet control over the country. The Americans learned a lot about how Soviet forces operated, but had no clear warning that the attack was about to take place, and thus no ability to take diplomatic action to try and stop it.

In both cases, American intelligence agencies were caught flatfooted. American reconnaissance satellites only returned images of the events on the ground long after they were over. These two incidents prompted American intelligence officials to consider development of a new “crisis reconnaissance” capability in the late 1960s. A leading candidate was a satellite proposal named SPIN SCAN, which would have used a mature, but limited technology to return images to the ground. SPIN SCAN and several other satellite proposals of the late 1960s were concurrent, competitive, and complementary to the slow-burning American effort to develop a near-real-time reconnaissance satellite capability that culminated in the 1971 decision to build the KH-11 KENNEN reconnaissance satellite. SPIN SCAN came closer to development than any of the other crisis reconnaissance satellites evaluated during this period. But the two significant intelligence failures of the late 1960s did not lead to its approval. SPIN SCAN’s existence, and just how close it came to becoming reality, have never been public before now.

[edzieba]

I wonder how much machine learn is used in the image pipeline from the KH-11, I’d bet that actual humans now only sit right at the end of the pipeline and that much of the initial sorting and interpretation of images is done by machine learning. You probably only need the same number of photo analysis as in the past even though something like the KH-11 must produce vastly more data.

Use of modern deep-learning techniques is near certain, as computerised image analysis and classification long predates the recent resurgence of SLNNs. Not a new technique, but making existing tools faster, more effective, and easier to create.

I'd be more interested in how far back assistive pattern matching techniques were used. Prior to computerisation, optical Fourier correlation allowed identification and location of a pattern within an image to be performed entirely within the optical domain - effectively a laser source plus a series of mirrors/lenses and a fourier transformed filter of the pattern you wished to search for would allow taking a large strip of film, and projecting through it via the correlator to produce a projected image where intensity correlates to the closeness of match to the desired pattern. e.g. take a fourier transform of a SAM site footprint, run a Hexagon film strip through the system, and watch for the bright spots to appear and mark them for manual inspection (or even just count the spots to rapidly guesstimate sites per few thousand square km). Attached is a 70's era paper on modifying that sort of setup for scale and rotation invariance, so tools of the era were up to this task. I can't recall this being mentioned in any of the released NRO documentation, but that often seems sparse or non-existent when it comes to what happened to the film after recovery.

[LittleBird]

I'd be more interested in how far back assistive pattern matching techniques were used. Prior to computerisation, optical Fourier correlation allowed identification and location of a pattern within an image to be performed entirely within the optical domain - effectively a laser source plus a series of mirrors/lenses and a fourier transformed filter of the pattern you wished to search for would allow taking a large strip of film, and projecting through it via the correlator to produce a projected image where intensity correlates to the closeness of match to the desired pattern. e.g. take a fourier transform of a SAM site footprint, run a Hexagon film strip through the system, and watch for the bright spots to appear and mark them for manual inspection (or even just count the spots to rapidly guesstimate sites per few thousand square km). Attached is a 70's era paper on modifying that sort of setup for scale and rotation invariance, so tools of the era were up to this task. I can't recall this being mentioned in any of the released NRO documentation, but that often seems sparse or non-existent when it comes to what happened to the film after recovery.

Agreed, fascinating topic to me as well.

Hoku posted about an optical correlator a while ago, but I can't remember whether in SAR or imaging context. I'll dig it out.

[hoku]

I'd be more interested in how far back assistive pattern matching techniques were used. Prior to computerisation, optical Fourier correlation allowed identification and location of a pattern within an image to be performed entirely within the optical domain - effectively a laser source plus a series of mirrors/lenses and a fourier transformed filter of the pattern you wished to search for would allow taking a large strip of film, and projecting through it via the correlator to produce a projected image where intensity correlates to the closeness of match to the desired pattern. e.g. take a fourier transform of a SAM site footprint, run a Hexagon film strip through the system, and watch for the bright spots to appear and mark them for manual inspection (or even just count the spots to rapidly guesstimate sites per few thousand square km). Attached is a 70's era paper on modifying that sort of setup for scale and rotation invariance, so tools of the era were up to this task. I can't recall this being mentioned in any of the released NRO documentation, but that often seems sparse or non-existent when it comes to what happened to the film after recovery.

Agreed, fascinating topic to me as well.

Hoku posted about an optical correlator a while ago, but I can't remember whether in SAR or imaging context. I'll dig it out.

This was for SAR - the radar signal being recorded on film, and then fed through the correlator to reconstruct the image (1D scan + distance) information.
https://forum.nasaspaceflight.com/index.php?topic=29545.msg2303311#msg2303311

There was a proposal (by LMSC?) around the mid 1960s to use computed-aided pattern matching for image interpretation. I'll try to dig up the source.

Edit: not exactly what I've been looking for, but attached is Chapter 4 on "Automatic Target Recognition Systems" from Volume 8 of the 1993 edition of the Infrared & Electro-Optical Handbook. It starts with a brief history, and then gives quite some details on pattern recognition techniques. The claim is that a paper by Minsky & Papert in the late 1960s convinced funding agencies to withhold support for research in this field for the following decade.

[LittleBird]

https://thespacereview.com/article/4614/1
 
In both cases, American intelligence agencies were caught flatfooted. American reconnaissance satellites only returned images of the events on the ground long after they were over. These two incidents prompted American intelligence officials to consider development of a new “crisis reconnaissance” capability in the late 1960s. A leading candidate was a satellite proposal named SPIN SCAN, which would have used a mature, but limited technology to return images to the ground. SPIN SCAN and several other satellite proposals of the late 1960s were concurrent, competitive, and complementary to the slow-burning American effort to develop a near-real-time reconnaissance satellite capability that culminated in the 1971 decision to build the KH-11 KENNEN reconnaissance satellite. SPIN SCAN came closer to development than any of the other crisis reconnaissance satellites evaluated during this period. But the two significant intelligence failures of the late 1960s did not lead to its approval. SPIN SCAN’s existence, and just how close it came to becoming reality, have never been public before now.

Many many interesting threads spin off this, more than usual even for your pieces, Blackstar.

One that intrigues me  is based on Ray Cline, quoted in the 1971 Excom meeting, your ref [35] and attached. See grab, and statement that

Dr Cline thinks the classification level of the photographs is too high. He would like an unclassified photo he could take to the UN for example.

This echos Merton Davies' pov to my mind, were there others with a similar view at that time ? Cline is an interesting figure, as I'm some of you will know: https://www.nytimes.com/1996/03/16/us/ray-s-cline-chief-cia-analyst-is-dead-at-77.html