Hacker News Reader

Comment by dfsegoat | Hacker News Reader

Meanwhile, USAF just fully designed and tested a 6th generation fighter [1] in record time [during 2020]:

https://www.defensenews.com/breaking-news/2020/09/15/the-us-...

They say the key to the record time was an 'all virtual' prototype design and test process. I found that pretty fascinating.

1 - https://en.wikipedia.org/wiki/Sixth-generation_fighter

View on HN

With modern combat aircraft the easy part is the working airframe actually. The YF-22 first flew in 1989 and finally entered service as the F-22 in 2005. F-35 had a similarly long development time and while its technically been operational for years, its software is like a modern EA release. A lot of the good stuff missing and available as later DLC. They are still patching in drivers for weapons that legacy aircraft already support.

While its definitely good that NGAD has produced a flying prototype so quickly, it isn't proof that they have achieved the goal of faster development.

The primary hinderance has been and still is the software. The defense industry has been slow to adopt modern coding practices. Sometimes that's a good thing. But on the balance its bad. F-35's software development has all of the hallmarks of a project saddled with a great deal of technical debt combined with outdated practices and overburdened with compliance.

What makes you think that Boom won't have a long time to go fixing their software after they have a flying prototype?

Stuff in civilian aviation is designed to be certified before its flown commercially, stuff in military aviation is designed to be adapted on an evolving battlefield.

Boom production models will use COTS avionics just like every other business jet manufacturer. They'll have to write custom flight control software but that will be much easier than on a military aircraft. No weapons, no external stores, no defensive systems, no tactical data link, no complex navigation modes, etc.

Except they'll need to write that flight control software themselves, for a plane so different from modern aircraft that there are literally no experts in software design for this class of aircraft. (Outside of anyone doing it for the military, and I can't imagine they'd be allowed to repurpose that work.) There will need to be a whole lot of new software development, along with the corresponding review process by the FAA. Simpler than a complex military fighter, but there's no COTS solution for the software and that's a huge part of this project.

Looking at the current state of open source software like ardupilot, and the fact that we've had supersonic jet fighters since the 1950s (F-100 Super Sabre) I don't think the control software is going to be a major bottleneck. If anything, not being tied to legacy control software may improve their velocity and testing. Navigation solutions should be drop in. Garmin, etc offer drop-in glass cockpit retrofit solutions for Cesnas from the 1960s.

If Boeing can't make the Max work without crashes, what makes you think a completely new supersonic aircraft can use a fork of Ardupilot with a few minor extras for supersonic flight?

Supersonic flight has much less overlap with subsonic flight than you might think. There are compression and twisting forces on the fuselage and flight surfaces which have no analog in conventional airliners.

And you can't just take the flight characteristics of one shape/size of aircraft and tweak them a little for your new design.

'If X can't do it...' is a bad argument that doesn't take any consideration of real life.

Just because they are a huge company with big spending power does not make it impossible for their codebases to be a pile of hot steaming garbage that even the best engineers struggle with.

Either way, your rebuttal about Boeing doesn't address the fundamental differences between control of a traditional aircraft-- which is all that any existing software can handle-- and control of a fundamentally different type of aircraft.

I wasn’t addressing that, though. If I was talking about your other point I’d have mentioned it in my comment.

> Supersonic flight has much less overlap with subsonic flight than you might think.

And every supersonic aircraft also has to fly subsonic. So you really need two sets of software in a supersonic aircraft. Or more accurately three, because the transonic regime is weird enough to be its own thing.

What about the airplane given in the original example, the F-100, whose design predated the integrated circuit by something like a decade? Presumably control software improves stability, but otherwise the aircraft behaves in a predictable manner above and below Mach 1 as supersonic control software is a relatively recent addition.

The F-35 has in-app purchases? Nice.

"It looks like you're gonna crash! Pick up a parachute for just 200 coins and you won't have to start a new pilot file."

You have 12 seconds until crash. Please enter credit card information.

Nah, first you get: "Please solve this captcha to prove that you are not a drone".

"Hyperion is offering a ninety percent off sale on all New-U respawns! This offer expires in three-two-one-zero aww, better luck next time."

> With modern combat aircraft the easy part is the working airframe actually.

Is there such a thing as a “working airframe” for modern combat aircraft? Reading https://en.wikipedia.org/wiki/Relaxed_stability, I have the understanding that, for modern combat aircraft, you can’t consider the airframe to be separate from the software.

Flight control software is a tiny fraction of the software on a combat aircraft.

As a side note: Early relaxed stability planes had no software. It was implemented with analog computers

Yes, most notably the F-16, which had a really good four-channel redundant analog flight control system. But is that needed on a transport aircraft? Fighters have it for extreme maneuverability, which a transport aircraft does not need or want.

Transport category aircraft are inherently stable by their aerodynamic design, it's a certification requirement.

If you have normal wings and the right size horizontal and vertical stabiliser the basic design of an aircraft is self stabilising. Especially with the wings sloping upwards (as all airliners have) because that wing shape also contributes to stability.

Per the Wikipedia article, MD-11 required active stabilization.

Never flown one, but as far as I know that's only moving the center of gravity by moving fuel. The CG is very important to inherent stability, if it moves behind the center of pressure you lose all pitch stability. And for efficiency you want it to be as far aft as possible because that reduces drag from the tailplane.

Dassault did something similar like 10 years ago with one of their business jets. In that case, it was also to showcase the capabilities of Catia.

Still amazes me, on the one hand you have the Air Forces one-year project. On the other hand you have the German Air Force that needs more than that just ginish the first draft of the requirements document for an existing plane.

Don't rule out the design/development of other planes in US arsenal. The F-35 is just in the news here this week (yeserday maybe) about it being a meh plane because of the bureaucratic process. The Air Force project seems to be an outlier and definitely not the norm.

That’s true, though given this success, hopefully they can name it the norm going forward.

I'd love that as well, but research projects are very very very different from actual huge contracts. Especially with how politically engineered supply chain has to be in USA. And changing that isn't technical challenge - it's political one, and no one will dare to do it.

The F35 is a primarily a sink for public cash that happens to fly. There is literally no incentive to make these projects cheap, fast, and economically streamlined.

The lifetime cost is projected to be $1.7tn. Someone is getting very rich off that.

For comparison, the entire Manhattan Project cost the equivalent of $30bn in modern money.

> The lifetime cost is projected to be $1.7tn. Someone is getting very rich off that.

Tons of people are getting rich off that, and tons of people are being elected for the office. Military spending is one of the biggest government welfare programs in USA.

I agree. Not only did they digitally prototype the fighter, but also its entire manufacturing process. That’s something new AFAIK.

Sounds like a step that Tesla skipped. In the VFX world of movies, this is known as PreViz. I remember when the 3D rendering first came to CAD. One of the projects my dad was working on discovered that if they built it exactly to the plan's specifications, there would have been plumbing pipes running through other pipes. Lots of value in these kinds of looking at things digitaly before doing it physically

Not new. See: Road Vehicles

This is fascinating. This reminds me of a book I read about John Boyd [1] who fought the Air Force to develop theorems about aircraft performance, namely the Energy-Manueverability Theory [2]. He went on to mathematically prove why certain aircraft would fail to live up to their assigned missions and help design the F-15, F-16, and other aircraft. Albeit, the mathematics of a "perfect aircraft" lost when faced with the bureaucracy of the government.

1 - https://www.goodreads.com/book/show/38840.Boyd

2 - https://en.wikipedia.org/wiki/Energy–maneuverability_theory

Thanks for the links, didn't know the USAF had done decided to go on house for their prototype. This caught my eye:

"the first Air Force aircraft to be built using the 'holy trinity' of digital engineering, agile software development and open architecture"

Building a flying airframe is relatively easy now. The slow expensive work tends to be in weight reduction, software development, and systems integration. But hopefully the Air Force has learned something from the problems in the B-2, F-22, and F-35 programs.

Sure, but how many people USAF has and how many billions did it cost? Compare that to Boom. Also, USAF does not have to comply with civil aviation regulations.

Military aircraft don't have to go through the FAA certification process but the actual requirements are generally even more rigorous. So that doesn't save anything. Most of the work is typically done by the manufacturer's employees.

im surprised we are still making manned aircraft. https://www.defenseone.com/technology/2020/08/ai-just-beat-h...

I was wondering what the timeline/records were but the article didn't say. Maybe it was kept secret?

I'm very hesitant to take anything "DefenseNews" says at face value.

Obviously heavily pro-military just by the title, as if the military is only used in a defensive manner.

Like it or not, the term "defense" to mean something military-related is all over the Department of Defense, the defense budget, and defense contracting. While you are welcome to write off all these sources as ideologically contaminated, I think there must surely be better reasons to do so than their use of the term "defense". And, of course, you'll deprive yourself of news about specific technological developments.

Well that’s good. How’s the F-35 coming along?

Apparently quite well if you look past the decade-old takes on it.

That’s the point that of the grandparent post - timelines have slipped a lot.

Reply on news.ycombinator.com