May xx, 2011
<soon to be updated>
September 9th, 2008
In August of 2004 I purchased a Pmag from Emagair to replace one of my magnetos. During operation over the next months, I had numerous issues with intermittent missing. Brad and Tom at Emagair were extremely good at customer service and replaced my units numerous times whenever I had an issue. Obviously, I was not happy about the problems, but I truly wanted them to succeed - the idea for the units is a great one. By the summer of 2005, after having problems with a unit during my Instrument Instruction, and further problems with replacement units, I threw in the towel, sent the unit back, and got a refund. While replacing the unit with a rebuilt magneto with Ken Miller, we discovered some anomalous wear on the Emagair soft drive gear - Ken was very concerned about this, and I'll get to that issue later.
Fast forward to February, 2007. I figured they had had more than enough time (1.5 years) to fix any issues that they had, so I wanted to give Emagair another try. I know that many of you thought that I was insane, and given the rest of this story, you may have very well been correct. But I wanted them to succeed and put Slick/Bendix out of business... At any rate, I got a new generation 3 unit (version 113) and installed it. Within the first month, I had issues with mis-timing, and worked with Emagair to replace the unit. After problems with a second unit, I had been able to run for over a year and 120 hours with no issues whatsoever on the third version 113 unit that I had installed in the spring of 2007. I finally felt like they had the issues licked.
After having a magneto failure (reported on list) back in early June of this year (2008), I decided that since I had been having good luck with the Pmag, that I'd get an Emag to go along with it to replace the magneto that had failed. Emagair offered to upgrade my existing Pmag from the version 113 that I had to a version 114 at the same time that they were sending me a new version 114 Emag. Hey - something for nothing - take it, right?
I installed both units and flew locally for about 10 hours, taking friends for rides, going places for lunch, etc. No problems at all - engine was running like a top.
In mid July, 2008, I headed east at the crack of dawn for a two week trip that would take me to Provincetown, MA for a week vacation, a few days in NJ with my mother and sisters for my father's footstone dedication, and then to OSHKOSH with my wife for the obvious reasons and to give the COZY and Canard fora. I took off at 5:30 AM, just as the sun was rising, and headed east. After about an hour and a half, near Kingman, AZ, the engine started running a TINY bit differently - a mag check seemed clean - it was running OK on either EI, but the CHT's had gone up about 10-20 degrees and the power was down a bit. I continued the flight for another 4.5 hours with no issues, landing in Clinton, OK for gas, oil and a bathroom break.
Upon runup, I found that one EI was NOT working at all, but I was in the middle of nowhere and had places to be. I'll be writing another story about all of the poor decisions I made on this 36.5 hour trip, and won't address them here, but I'll just describe the events because that's what's germane to the warning.
I decided to take off on one EI to get to Ohio, where I'd be staying overnight with Bill Kastenholz. I took off on one EI, climbed over the airport to cruising level, and headed off. About 10 minutes after takeoff, the second EI kicked in and the engine continued running as it had for the previous 4.5 hours - CHT's a little elevated (but not dangerous), power a bit down from normal. 5.5 hours later I landed in Ohio.
The next morning, after refueling, I took off (again on one EI - the bad one seemed to not work when cold and work when hot) and flew towards MA. On the way I stopped at Westerly, RI to visit Don Ponciroli but we didn't connect and I called Emagair to talk to them about the issues I was having with the 114 units. I then flew to Norwood, MA to visit Jose Velez, and after a couple of hours on to Provincetown, MA. Each time, I took off on one EI, and each time, the second one kicked in after 10 minutes of warming up. It was apparent that there was a problem with timing, but since it didn't run on the ground when cold, I couldn't adjust it.
Emagair sent me a replacement unit in P'town and I replaced it. A test runup indicated that it was running fine, along with the other one. On Friday of that week, I flew down to Brookhaven airport on Long-Island with my son and nephew to bring my nephew home after the vacation week. There was no issue on that flight, and no issue on runup. However, about 7 minutes after taking off from Brookhaven, just about over Orient Point, the engine suddenly started running like crap - RPM's dropped off, CHT's shot through the roof on two cylinders, and power decreased substantially. I brought the power back and made a precautionary landing with my son at Easthampton.
On the ground, I called Ken Miller (whom I had spoken to for a few minutes on the ground at Brookhaven), described the symptoms, and he gave me a few things to look for. It turned out that one of the EI's had failed catastrophically, with the timing shifting by a huge amount (no wonder the CHT's were through the roof - it was firing on the intake stroke). Retiming it twice did nothing - it wouldn't hold a timing set. The one that had failed was NOT the replacement unit - it was the second unit that had worked OK on the whole flight east. The replacement unit was working OK. I had to completely unplug the power from the unit in order to get it to stop firing, but once I did, the engine ran OK on the one EI (the replacement unit) that was left.
My son and I took off from Easthampton on the one EI and flew uneventfully back to Provincetown. I called Emagair to let them know, and they agreed to send another replacement to NJ, where I would be the next day. I sent my wife to NJ via car instead of having her fly with me, and we made arrangements for her to fly commercial to OSH rather than with me in the plane - I would pick her up when I got there in Appleton.
In NJ, I received another replacement unit, went out to Caldwell airport and installed it. Checks of both EI's indicated that they were working.
The next day I took off and headed to OSH. About 1.5 hours out, over Pennsylvania, one of the EI's started acting up with the same symptoms as the first bad EI - slightly elevated CHT's, timing off a bit, slightly lower power. I navigated from airport to airport, staying as high as was necessary for glide distance, and made it to OSH in 6.5 hours - about 5 after the EI failure. At OSH, I removed the bad EI and took it, along with the other two failed units, to the Emagair booth at OSH. We disassembled the units and found that all three had had catastrophic mechanical failures inside that caused them to not be able to hold timing to any extent at all.
Emagair stated that they had not had any failures other than mine of this type, but agreed that this was a design failure. They attributed the failure to excessive vibration caused in some way by my engine, and this plays back to the wear on the removed EI drive gears that Ken Miller and I had seen two years previously. Emagair believed that there was some strange thing going on in my engine that was causing excessive vibration, which beat up the gears and which caused the catastrophic failure of the sensor magnet holder that I had occur three times.
I told them that I just needed to fly 11.5 hours more to get home to Tehachapi, and that the plane would be grounded (which it is) for an engine teardown. They did a temporary repair on two of the EI's, which, along with the one in the plane that was still working, gave me three to use for the 11.5 hour flight(s). When I left OSH a couple of days later, I determined that I would also change my flight characteristics to see if I could lessen the stress on the EI's - I never let the RPM's exceed 2500, either in climb or cruise, and I climbed at 140 mph to ensure excessive cooling to the EI. Since the examination of the failed units had pointed to heat and vibration as being the cause of the failures (mechanical - NOT electrical), I wanted to change the heat and vibration signatures of my flying.
I was able to fly from OSH to TSP, with one stop in Tucumcari, NM, without any further failures. The plane is grounded and is coming apart; the EI's have been removed and returned, and I've got my refund.
Next, I'll describe the issues, the warning, and my recommendations.
So what's happening here? As Mr. Cottner and Mr. Read have stated, the Pmag/Emag units lose timing information, firing at arbitrary and sometimes random times. Clearly, this screws up the workings of the engine, sometimes to the point of having no power whatsoever. Mr. Cottner had four failures and lost his airplane because of this failure mode - Mr. Read had to make a precautionary landing, and I had three failures and one precautionary landing.
Emagair has issued a SB on both the 113 and 114 units. If you recall, I had timing issues with my 113 units as well. They have instituted a "fix" for the issue, notwithstanding that the 114 mechanical redesign was in part already a "fix" for the problems with the 113 versions.
The Emagair units use a small magnet and a hall effect sensor to determine crankshaft position. It's an ingenious solution, because it uses a sensor that can determine crank position within 1 degree, not just when at TDC or within 10 - 30 degrees, as some other EI's do. However, the mounting of the sensor magnet has been flawed from version 113 onward. In version 113, the magnet was epoxied into a metal cup at the end of a shaft. The other end of the shaft had the magneto gear on it and was in intimate contact with the accessory case gears inside the engine, along with the engine oil. This guaranteed that the shaft would get just about as hot as the oil in the accessory case - about 200 F, if not more.
Due to the very thin bond line of the epoxy holding the magnet in the cup and the differential CTE's (coefficients of thermal expansion) of the magnet and cup, the thermal stresses in the epoxy can be very large, cracking the epoxy. This would allow the magnet to rotate, thereby losing positional accuracy and timing.
After having this failure mode pointed out to them (by me) a year and a half ago, and after having refused my offers to assist them in redesigning the mechanical portion of the units gratis (I am a mechanical engineer with 27 years of experience), Emagair, with an EE but no ME on staff, redesigned the magnet holding portion of the units for the version 114's. They soldered the magnet into a large brass holder, and then glued the holder onto the same shaft that the cup had previously been attached to. They then added two locktited set screws to the holder which applied force to the shaft. There was no flat on the shaft where the set screws touched it.
This "fix" was far worse than the disease it was attempting to solve, and is the root cause of the horrible performance of the version 114's under heat and vibration loading. Now that the mass of the brass holder has been made much larger than the mass of the magnet alone, the stresses in the glued bond-line are far higher than previously, and under heat and vibration is guaranteed to fail eventually. The set screws do absolutely nothing, since the CTE of brass is much higher than the CTE of the steel shaft, so as the system gets hot, the set screws don't even touch the shaft. Failure of the glue bond line is sufficient to cause the holder to be able to rotate, with the same mis-timing issues as with a magnet disbond in the version 113 units.
Although Emagair has issued an SB, describing a "fix", I do not have ANY confidence in this "fix"'s actually working in the long term. As with Mr. Cotner's warning, I don't believe that the keyway and roll-pin are anything resembling an adequate solution to the magnet mounting problem, which is obviously NOT specific to my engine and/or installation. As more hours are put on these units (I fly 120 hours/year - more than twice the homebuilt average), more will fail, and I have no confidence that this fix will substantially change this situation.
My recommendations, based on my opinions and my examination of MANY Emagair unit disassemblies are these:
- If you are flying with a version 113 or 114 product from Emagair, ground your plane immediately and remove the units. Do NOT fly with them, or with the "fix" described in the SB. It will work for a while, but for how long?
- If you have a unit that you have purchased but haven't used yet, return it for a refund - do NOT put it on your aircraft and fly it.
- If you were considering Emagair products for your engine, consider something else - do NOT purchase one and put in on your aircraft.
When Emagair have hired a competent mechanical engineering firm to redesign their systems, have FULLY tested the units under O-360 type vibration loads for thousands of hours, and publicly published the results, THEN in may be appropriate to consider these units. Until then, it is not.
If Mr. Cotner, Mr. Read, and my experiences with these units put Emagair out of business, and it costs some of you the opportunity to get a refund on your units, I'm sorry for that, but Mr. Cotner was lucky, and relying on luck to keep people alive is not acceptable.
Everything that I have written here is either my personal experience, my opinion, or my recommendation based on my opinion.
I know that many of you are saying (to yourselves or to others) "I told you so", and you're right - you told me so, but my desire for Emagair to succeed overrode what should have been enough evidence to the contrary. Feel free to write me with "I told you so" messages, if it makes you feel better.
I'm happy to address any issues, comments, or questions that anyone may have. Since Emagair never responded to my entreaties to sign an non-disclosure agreement with them, I do not have any responsibility not to explain the inner workings of the units or the problems therein.
PLEASE, PLEASE, PLEASE - distribute this message, or the link to this web page (and Mr. Cotner's and Mr. Read's messages) to any and all aviation related mailing lists, fora, and printed newsletters.
After posting this warning, I received a number of emails regarding Emagair product failures. Two of these were from Daniel Tracy, who graciously agreed to allow me to post our exchange here:
MZ,
I just read your diatribe re E-mag air and their pretend engineering. I could have signed that set of letters myself. In fact, some of the very same phrases are identical to mine in letters that I have written. The thoughts are exactly the same anyway. E-mag Ignition people are pretty good at software engineering but absolutely lousy at system and mechanical engineering. My RV-7 with an IO-360 had ignition failure on takeoff - @ 100 ft. alt.. It was the 5th P-mag failure in 160 hours on the, since new, meter - an awful record!
They've sent me a set of #114s to replace the cobbed up set of #113s (which went back and forth between Seattle and Texas many times over the last year and a half). I currently have one #114 on the right side and a standard Slick mag on the left - a hybrid. I need to know more about the 'fix' E-mag has dreamed up for the #114s. Is it the roll pin setup? Why don't they make a housing for the magnets and the shaft out of one piece of metal? - EDM a snug encapsulating place for the little magnets and use a soft rivet to hold them there? Forget the goddamn glue... This stuff is not rocket science. I happen to know about rocket science. (41 years as a mech design engineer at The Boeing Company).
I've cancelled a trip from SEA to OAK in my little plane today because of your informative letter. It really has me spooked and it's really, really nice weather today.
dt
You can post my note to you re the P-mag affair. I talked with Brad today. I'm gonna send him one of my #114s and have him do his thing with it. I'll replace the one in the right side of my engine when I get it back. I'll try it out with the somewhat dubious comfort of being able to instantly switch it off and fly home on the Slick in the left side. It's a hellova way to fly - always hoping to get to 500ft on takeoff and once I'm up, always looking for a place to land when the engine decides to quit. Was not my original vision...
BTW Brad sez the shaft material has to be steel and the sensor holder cannot be because the sensor magnets need to exhibit a certain flux to trip the position sensor and the ferrous material, if touching the magnets, would smother the flux. I'd sure like to do some geometry investigations with a one piece shaft and some carefully shaped sensor magnets. We don't like to use roll pins at Boeing - at least on anything that counts. Press fits with small diameters don't work so well and neither do keyways. They're sloppy. Maybe with all three schemes Brad will get a unit that doesn't slip. I wonder if he's doing any close tolerance electric discharge machining on these parts.
Then there is the circuit board smack on the back of the hot-cycling and white-noise vibrating engine - couldn't be a worse environment for a CB. That will be the next item to need the "treatment". I've asked E-Mag what their proof testing and acceptance testing criteria might be (how far beyond expected temps and vib amplitudes, etc.). Apparently they do a single thermal cycle exam on each delivered CB. Yikes! They need to explore some vast testing regimes with a reasonable test environment - Vibration simultaneous with Thermal cycling - and go thousands of cycles and hundreds of hours for the Proof tests and then some significant exam for acceptance tests for the production units. How in the hell can they expect to get the damn things certified otherwise? Maybe we are (or have been) the test program....
MZ,
Okay! publish if you like. There's a lot of folks in the 2 EAA clubs to which I belong up here near Seattle that are watching my struggle. I cannot yet provide them any incentive, at all, to get acquainted with E-Mag Air. The troubling thing about the whole experience is E-Mag's seeming sincerity and dedication. I like that type of passion. Eventually, I hope they will get it, and possibly without any burning bodies at the end of various runways. I, too, am contemplating the Lightspeed/Slick hybrid. My Bride of 40 years has as well....{"Get those goddamn P-Mags out of there!"}
Thanks to a forwarded message from the Aero-Electric Connection mailing list from Dave Vigliercho, I became aware of a posting from Bob Nuckolls, whom I respect very highly, regarding this issue that tended to dismiss the "alarmist" tone of my warning. I started an email exchange, and post portions of my responses to him here, as it addresses the questions that I've been getting from many folks regarding the efficacy of the Emagair SB "fix".
Since I had my three failures of the 114's within a 25 hour period on a flight east, I've done some digging within the canard and RV communities, and have come up with over 30 failures, most catastrophic loss of timing failures. I've had 8 failures over the years, with the three 114 failures in the last 25 hours of flight. I've found other people that have had 5 failures over 160 hours, 4 failures within a few hours, and many with two or more multiple failures.
<Emagair> is NOT being totally forthcoming when <they> say that this is a "rare event" - it is NOT. With approximately 1000 delivered units, maybe 500-700 of which are flying, with most of those having less than 200 hours on them, this many catastrophic failures of late generation (113, 114) units is not rare, and not acceptable...
That, Bob, is just pure crap. I am a mechanical engineer, and am in charge of the Rocket Motor development at Scaled Composites. I've been an engineer for 28 years. Daniel Tracy, another ME from Boeing who's been an engineer for 41 years, concurs with everything I wrote on the previously mentioned web page. When I was at OSH this year with my three failed units, I did a post-mortem on them WITH <Emagair> in the back of their booth, and it was OBVIOUS what the design flaws were that led to the failures, and are leading to the increasing incidence of failures that they're seeing as more and more version 114 units start to actually fly and get some hours on them. Having disassembled numerous 113 units as well, the issues with magnet disbonding due to thermal stress and poor bond line design is also clear.
There is nothing unique about the problems - it's related to heat and vibration, and poor design of bond lines in the 113's and high stress levels, poor bond lines, too much mass and a complete lack of understanding of CTE issues in the 114's, which are far worse with respect to magnet loosening than the 113's were. <Emagair's> "fix" of the 113's to the 114's, at my instigation (but NOT using my design) made the cure far worse than the disease...
No, Emagair CLAIMS they have been addressed, but given the history, there is no reason to believe that's the case. If nothing else, nowhere near enough testing could possibly have been done in the given time period to prove that on the hottest, worst vibration level O-360 engine that could exist in the field that the new design will not fail...
It's not a mystery at all. The failures in the 114 units tend to occur soon after the units start being subjected to the heat and vibration, especially in the O-360 line of engines where the torque pulses are much higher than in other engines...
See above. That's just wrong. The causes are know, and fixes are relatively simple, and Emagair has rejected useful fixes to come up with questionable ones of their own...
But the fact that Emagair has not retained ANYONE to help with the mechanical design until just now IS evidence of irresponsibility, when the problems have been known and pointed out for at least a year and a half (when I first had 113 failures of magnet mounting). Me or someone else - don't care - just get someone that knows <something>...
They still have a lot of mass at the end of the shaft. They obviously had no idea what the natural frequency of the system was previously, and the claim that they've reduced the mass is indicative of the fact that they still don't, else they'd have claimed that they had removed any torsional vibratory effects.
The press fit is still CTE dependent, just like the setscrews were, because the CTE of the brass magnet holder is much higher than the CTE of the steel shaft.
Keys/keyways do absolutely nothing in this case, because there's no axial retention from the key, and if the press fit holds, there's no reason for the key, and if the press fit doesn't hold, the keyway would get the crap beat out of it in a couple of seconds even if it DID retain axially. Same for the key/rollpin relationship - the keyway is not redundant, not a secondary retention scheme, not a backup, and indicates a total lack of understanding of mechanical design principals.
The rollpin is not being used in an optimal manner - they're generally not used in oscillatory, stress resisting situations. Yes, it's better than the glue, but is it adequate? Don't know, and only many hundreds/thousands of hours of testing AT TEMPERATURE with NO press fit and WITH the vibratory spectrum of a poorly balanced O-360 will indicate the effectiveness.
They have done little if any testing, and certainly have not PROVED the "fix" in a manner that supports the claim that this is a production ready design.
The problem is that I do not trust Emagair to do any mechanical design - they have no history of successfully doing so. Because of this, I have far more suspicion of anything they come up with than I would if they did have a successful history. Why would <Emagair> all of a sudden be able to come up with a great long-term fix when the last fix <they> implemented reduced the MTBF of the system by two orders of magnitude?
Let a damn ME design the thing...
If you were told that these were isolated, very few instances, then you were misled. Whether due to lying, self-deception, or something else, I cannot say. All I know is that you haven't been told the truth. When 3% - 6% of their units have failed, and many of those catastrophically, and essentially all of the failures on low time units, it is disingenuous at best to claim that the failures are "isolated". This is an epidemic. If 3% of B&C's starters were failing, what would you say to Bill Bainbridge about that? No one will die if a starter fails - they just won't fly. EI's are a different story.
They have offered something different - whether a fix or not is yet to be seen by the field alpha testers that they call customers. "Offering" an untested remedy for a known catastrophic failure mode is not an acceptable route....
That's exactly what I propose. A REAL fix, not one of <Emagair's> band-aid shot in the dark imitation engineering fixes must be designed and tested. Even without the real engineering, at LEAST hundreds of hours of testing as described above (IN A VIBRATORY, HOT ENVIRONMENT) must be completed before even considering offering a fix and/or telling folks it's safe to fly with these units. It is not...
Save folks from dying. Mr. Cotner is alive by pure luck. 3 seconds earlier on the failures, and he doesn't make it back to the airport. If my statements keep people from using an unsafe device on their airplane and dying because of it, I will have achieved my goal.
I have spent four years buying, testing, using, and failing Emagair products, touting them to others, supporting them in the face of all evidence to the contrary, and generally ignoring my engineering judgment when all of these facts have been pointed out to me by others who have given up on them long ago. I did this because the IDEA for the units is brilliant - it's a great architecture, a great idea, and I wanted them to put Slick out of business. But the oscillations around a safe system, without any convergence toward one, have gone on too long, and the safety issue is too big, for me to keep the blinders on any longer...
First and foremost, I want people to be safe, and they're not if they're flying with 113's and especially 114's. Second, I want Emagair to get a REAL design done, and then perform REAL testing to prove it out. Third, I want them to continue selling safe products and put Slick and LASAR out of business.
But the safety comes first...
The issue isn't the static grip. With no vibration, the installation would be fine. In many cases, however, with no retention of the pin, vibration can cause the pin to "walk" out of the hole in one direction or the other. Should this happen, since there's no retention mechanism, and since the press fit of the brass holder onto the steel shaft may come loose due to CTE, and since the keyway does nothing, there's no positive guarantee that the magnet can't come loose...
FEA may be useful in some situations, but it will do nothing to determine whether the pin will walk out under vibration. Without a Monte-Carlo statistical analysis using the worst case tolerance levels for the holder and shaft, they also don't know the likelyhood of the CTE issue causing loosening of the press fit.
Once again, it's not my responsibility to do their work for them for free. I offered that in the past and was turned down when I was a customer. I'm no longer a customer and other than keeping other folks alive, have no interest in helping Emagair...No, thousands of hours in that ENVIRONMENT. Put the units on a stepper motor drive, in a 220F environment, and feed it the worst case vibration through the gear train that could reasonably be expected.
You could test 10 units for 2000 hours in a few months. Imagine the publicity from showing pics of THAT testing...Yes. there's no retention mechanism. Have they stated that they're staking the ends of the hole? Both ends? Personally, if I were them, I'd want a solution that was repairable, and allowed for the disassembly of the system. See? I can come up with simple solutions for their design. But I'm tired of it, after doing it for years and being rebuffed. As I say, I don't care about <Emagairs>'s education anymore - just the folks that might use their products...
I have performed vibration testing on many medical instruments in my 17 years of working for HP/Agilent/Philips Medical Systems. Under random and/or sinusoidal vibration excitation, I have seen MANY unconstrained items, theoretically under very high frictional force levels, migrate unidirectionally and allow parts to disassemble themselves...
See above. Vibration can substantially change the effective frictional force coefficients. I'm not suggeesting that the roll-pin WILL come out - I'm only suggesting that it wouldn't surprise me a bit if it did, since I've seen similar things occur on products in development in the past during actual vibration testing.
This was on defibrillators - I was mechanical engineering manager for the last two generations of defibs that Philips released in the market, in 2001 and 2004. Since these devices can also kill people, a higher standard of development and testing was assumed - we never just designed something and set it out in the field, assuming that the customers would test to see if we had done it right...
The fix may or may not work - I've explained what I think the potential pitfalls of it are. The architecture of the design indicates a lack of understanding of basic mechanical engineering principals, therefore I prima-facie don't trust it. Only redesign or extensive testing PRIOR to flying with it could make me believe in its efficacy...
Realize, also, that NONE of this addresses the cause of Jeff Mallia's two failures, which I was told by Emagair were NOT magnet holder fixturing failures. If that is true, then there is at least one more catastrophic failure mode that is clearly not addressed by the SB that speaks only to the magnet holder issue.
I've recently read on one of the RV aircraft fora that a rumor has been going around (I have no idea where it started) that my engine had some "significant" engine vibration that MIGHT have been a cause of the Pmag failures I experienced. First, let me state that while Brad at Emagair suggested that a vibration issue might be causing the failures, there was no vibration issue with the engine. It was running extremely smoothly, as it always had, as long as the Pmags were firing correctly. Recall that others had exactly the same type of failures which were traced to the poor mounting of the magnets. No "significant vibration" was necessary to cause the failures, and none was extant in my engine.
After grounding the aircraft, I had the engine rebuilt by Performance Engines in Corona, CA. They found no significant damage to the bottom end or accessory case gears, and did not indicate that there was anything that might have caused any vibration that could cause damage to a magneto or an electronic ignition run by the accessory case gear train. I had specifically asked them to carefully inspect the accessory case for anything that might be an issue.
Just to set the record straight.
Return to: Cozy MKIV Information
Copyright © 2008, 2011 All Rights Reserved, Marc
J. Zeitlin
e-mail: marc_zeitlin@alum.mit.edu