Tankers. A new price category?

[pippin]

I don't think "big" comm birds right now can be seriously used for internet. GEO sats have way too much latency to be useful for that, for internet you need LEO or cable.
Latency these days is more of an issue than bandwidth.

[oldAtlas_Eguy]

I don't think "big" comm birds right now can be seriously used for internet. GEO sats have way too much latency to be useful for that, for internet you need LEO or cable.
Latency these days is more of an issue than bandwidth.

The IP (Internet Protocol) can handle latency better than other forms of communications since there is no actual time specification for how long a packet is to take point to point. There is only a timeout which is purely based on the local software which is usually on the order of 60s to 3min.

[oldAtlas_Eguy]

Sorry for the off topic. Back to topic.

I was curious about the sensitivity of per flight cost of RLV and ELV to observed reliability (not design) to see if there were possible local minimums that showed that a general level of reliability would produce better costs than a lower one or higher one.

Results:
For an RLV there indeed is a minimum (for the successful flight costs: a failure must be charged to the other successful flights increasing their costs) and in my cost model it occurred around a 92% observed reliability. The reason for this is that the vehicle designed for use 10 times was averaging only 6 uses because of a failure cutting the vehicle life short. Lower reliability such as 80% made it nearly an ELV.

For an ELV there is also a minimum (for the successful flight costs: a failure must be charged to the other successful flights increasing their costs) occurring at 50% observed reliability.

Using the same relationship between reliability and original hardware cost and operation costs the resulting best RLV cost per successful flight and best ELV successful flight is nearly the same, in my model $26M RLV and $21M ELV which is loosely based on FH costs. The only significant item here is that the ELV not being designed for return would deliver nearly twice the payload to orbit than the RLV:  ~$1,000/kg for RLV to LEO and ~$500/kg for ELV. A very high reuse RLV that has a high reliability might be able to compete with the ELV price, I will see what happens when the model uses higher design reuse numbers stay tuned. There is also the fact that schedules would be easier to keep with the higher reliability RLV vehicle than the very low reliability ELV. The last item due to the payload capacityies and the total number of flights to succesfully deliver a set amount of prop the flight rate for the higher reliable RLV and the low reliable ELV would be the same, you need 2 ELV for each 1 RLV flight to succesfully deliver prop.

[pippin]

IP can handle the latency, but it makes the whole communication look laggy and some application layer stuff, e.g. games, will not work at all.
TCP, to begin with, doesn't really like it, too.

It doesn't help if the protocol supports it well if it isn't suited for the application. I have used a high-latency network in the early days of 3G and I can assure you it doesn't make for a good user experience so it' something you can only sell in places where you don't have any alternatives for internet access.

But OK, 'nuff OT....

[Xplor]

I was curious about the sensitivity of per flight cost of RLV and ELV to observed reliability (not design) to see if there were possible local minimums that showed that a general level of reliability would produce better costs than a lower one or higher one.

Very interesting topic.  Are you aware of any operational data shows the payload cost vs. reliability?  It seems like there are so few vehicles and so many differences that it would be very difficult to decipher a trend.

[baldusi]

I will only state that you are discussing a supply solution to what's fundamentally a demand problem. For a fuel tanker there must be a fuel depot necessity. And willingness to pay the cost of transport. There's nothing of that. The comm market is very small. Even Iridium NEXT will need no more than eight flights. How many LEO constellation are really needed? Four? With an average life of ten years that would mean 3.2 launches/year. Now what else do you have?

I was curious about the sensitivity of per flight cost of RLV and ELV to observed reliability (not design) to see if there were possible local minimums that showed that a general level of reliability would produce better costs than a lower one or higher one.

I would love to see your model. But have you taken in consideration the cost of the payload?

[savuporo]

I will only state that you are discussing a supply solution to what's fundamentally a demand problem.

Hear hear.

Or, to put it more accurately, the world launch market is so far up the left hand peak of the price / demand curve, that no valleys are seen.

[baldusi]

I will only state that you are discussing a supply solution to what's fundamentally a demand problem.

Hear hear.

Or, to put it more accurately, the world launch market is so far up the left hand peak of the price / demand curve, that no valleys are seen.

One of the ways to find out if it's a problem of demand or supply, is to assume one is infinity quantity at zero (for supply)/infinity(for demand) price. So let's make the mental exercise of assuming free launches.
For any of current satellite systems, the cost of a satellite is so high that it wouldn't make more than a 30% price difference.
Why, because you have to use space rated elements, because everything has to be meticulously integrated and tested. Because every design has to be validated. And because you can't really have much commonality. I mean, a GSO sat can be very standardized. But even for SSO or Polar orbital satellites you have lots of different requirements.
But even if you could, and you assumed free launches and free satellites, you'd still rather have a couple of huge comm sat than thousands of small size. Because operations cost is almost lineal to the amount of sats that you put up.
And even if you had that for free, you have 720 orbital slots. If the satellites have a half life of 15 years, that would be 48 launches per year. Now a days you have around 20 or so. And that's not accounting that satellites might extend their lives to 20 years. But putting everything for free you'd still just double the number of launches.
There's only so much observation sats, and ditto for LEO constellations. If you make your numbers and everything save operation was free, you'd double, may be triple the amount of launches. And you're not considering the fact that every wannabe power likes to have their own launch vehicle and satellites.
What I'm trying to say here, is that you can't grow very much the LV market with current payloads (i.e. comm sats/earth observation sats). You've gotta create a whole new category of market. A need. A tanker is a solution, is an expansion of the supply. It doesn't creates demand.
Now, tell me that you have a business idea as ridiculous as making a golf course on the moon and charging 100M of affiliation fee, plus 100M for a two week stay, and the cheapest way is to do it with tankers, then that's getting close to a solution.
Build huge ships and make a "cruise" on a free lunar return trajectory? That's another idea. But whatever you propose has to make the need for extra launches. Current needs are less than current supply.
And let me state very clearly that the only reason that NASA and DoD are paying more each year has to be with their restriction to use US LV. Let them use Russian, European, Japanese, Indian or even Chinese launches, and your launch costs would be much lower.

[savuporo]

...What I'm trying to say here, is that you can't grow very much the LV market with current payloads (i.e. comm sats/earth observation sats). You've gotta create a whole new category of market. A need. A tanker is a solution, is an expansion of the supply. It doesn't creates demand.
Now, tell me that you have a business idea as ridiculous as making a golf course on the moon and charging 100M of affiliation fee, plus 100M for a two week stay, and the cheapest way is to do it with tankers, then that's getting close to a solution...

Very much right. There is an angle here though : current payload builders are NOT thinking in terms of cheap bulk payloads at all, just because such launchers do not exist. Hence, nobody comes up with these required ridiculous-but-may-actually-work business plans that would utilize such launchers.

Say that one would run numbers on a theorethical orbital 3D printing/manufacturing facility that requires tons and tons of bulk supplies of metal and plastic powders for the printers and SLS sintering machines, water and other bulk supplies. And it would actually have customers.
Right now, the business plan would have to include the design, development, tooling, manufacturing setup of the theoretical bulk launcher.
Its two leaps of faith in one business plan, that will never get funded by angels, VCs or anyone really.

The only remote possibility is that someone funds the launcher side with a "build it and they will come" mentality, and starts flying dummy or almost worthless payloads on it. And then has years of funding to keep it ticking like that without any real customers, allowing business plans to come forward.

[savuporo]

To put it very clearly, there is no current supply OR demand for bulk, maybe unreliable materials launches, including propellant.

The are no such launchers, and no applications demanding such launchers.

Orbital RLVs are better off in that regard : they always have humans as payloads as the market to rely on. That demand and market undoubtedly does exist, and business plans hedge on estimated the size and price elasticity curves of that market.

For "bulk" or "tanker" launches, both sides do not exist. Very hard to break that without either public funding on one side, or really really massive business plan trying to break both sides at once.

[baldusi]

That was a bit what I was trying to make. Exploration is not a viable market now a day (unless the 10B+ billionaires of the world suddenly decide to spend money on it for the sake of it). Communication and Earth Observation are probably at 25% of their maximum even if launches and satellites were free. And in any case we are are seeing the market moving towards fewer and more capable birds rather than smaller and cheaper. And many of the increases can't be launched by the same vehicle given national interests.
So if you want cheap LV you have to find or invent an economically viable necessity. Try tourism, energy, manufacturing or whatever you can think of. Just remember, a 40' container cost 2.5USD/kg to move across the globe. So don't even start to talk about mining other objects unless you find a source of monopoles or a clump of antimatter. Anything else is too expensive. And don't do the circular fallacy of saying that you make fuel/food on the Moon or Asteroids for exploration, since the exploration itself isn't economically viable. And if you are thinking about space tourism, let me remind you that if a ticket was 100k/passenger, and SpaceX was transporting 5000 passengers per year (ten times the total passengers ever), they would still have less revenue than four CRS flights.
I would love there to be a market. But I think even something as sophisticated as crystals for semiconductors can be done more cheaply on the Earth. I would love to be able to pay just 100k for a Zond like flight to the Moon. But there just isn't a market for that yet.

[oldAtlas_Eguy]

There are cheaper US LV's but they are not certified to carry NASA and DOD payloads yet. That's the rub for entry LV's, getting certified in order to get government payloads.

BTW COTS being a contract for succesful delivery of, relative to launch costs, low cost supplies is structured such that attempts are not paid for just success. No certification neccesary for the launch of one of a kind costly payloads, since the supplies are neither one of a kind or costly.

[oldAtlas_Eguy]

The first results I reported were on a log/log relationship between reliability and manufacturing/launch processing & ops.

I checked to see what effect the basic relationship with reliability has with the conclusions:

For RLV

Case 1: log/log – this model has a minimum per successful flight cost in the low 90% observed reliability.

Case 2: linear/log – this model has a minimum per successful flight cost in the mid 90% observed reliability close to 95% and probably will be below 95% for most specific situations.

Case 3: log/linear – this model has a minimum per successful flight cost in the mid 90% observed reliability close to 95% and probably will be below 95% for most specific situations. Although this looks to be the same as case 2 this model produces a minimum at reliability consistently higher than case 2.

Case 4: linear/linear – this model does not have a local minimum. It shows that for an RLV the highest obtainable reliability will be the lowest cost, which sort of goes against first impressions.

The RLV cost model definition:

Manufacturing cost:
=((1-.95)/(1-Reliability))*$95M – logarithmic relationship;
=($95M/.95)*Reliability – Linear relationship
[Note: both models are pined at 95% observed reliability to a cost of $95M. Observed Reliability is the number of successful flights to the total number of flights, not the design reliability.]

Launch Processing & Ops cost:
=((1-.95)/(1-Reliability))*$20M – logarithmic relationship;
=($20M/.95)*Reliability – Linear relationship
[Note: both models are pined at 95% observed reliability to a cost of $20M or ~20% of the total launch costs at the 95% case which the model is pinned to. Observed Reliability is the number of successful flights to the total number of flights, not the design reliability.]

Average number of successful flights per vehicle:
[Design life]*Reliabilty^[Design Life]
[Note [Design life] in the model is set at 10. Also if you have a better more accurate mathematical model of this it would be appreciated.]

Additional loss of use of vehicle due to early loss prior to design life retirement:
=(1-Reliability)*(SUM[x from 1 to [Design life] ]([ Launch Processing & Ops cost]+[Manufacturing cost]*([Design life] – x)/[Design life])/[Average number of successful flights per vehicle]
[Note: For a service that only gets paid when the payload is delivered (the pure commercial model) in this case prop at the depot, the additional costs of loss must be added to the cost of the successful flights.]

Total cost per successful flight:
=([Manufacturing cost]/[Design life])+[Launch Processing & Ops cost]+[Additional loss of use of vehicle due to early loss prior to design life retirement]


For ELV:

Case 1: log/log – this model has a minimum per successful flight cost at 50% observed reliability.

Case 2: linear/log – this model has a minimum per successful flight cost at 50% observed reliability.

Case 3: log/linear – this model has a minimum per successful flight cost at 50% observed reliability.

Case 4: linear/linear – this model does not have a local minimum. It shows that for an ELV the no matter the reliability the costs are the same. This is an improbability and shows that the linear/linear model does not reflect reality invalidating this case for both the ELV and RLV scenarios.

The ELV cost model definition:

Manufacturing cost:
=((1-.95)/(1-Reliability))*$95M – logarithmic relationship;
=($95M/.95)*Reliability – Linear relationship
[Note: both models are pined at 95% observed reliability to a cost of $95M. Observed Reliability is the number of successful flights to the total number of flights, not the design reliability.]

Launch Processing & Ops cost:
=((1-.95)/(1-Reliability))*$20M – logarithmic relationship;
=($20M/.95)*Reliability – Linear relationship
[Note: both models are pined at 95% observed reliability to a cost of $20M or ~20% of the total launch costs at the 95% case which the model is pinned to. Observed Reliability is the number of successful flights to the total number of flights, not the design reliability.]

Additional loss requiring replacement flight:
=(1-Reliability)*([Launch Processing & Ops cost]+[Manufacturing cost])
[Note: For a service that only gets paid when the payload is delivered (the pure commercial model) in this case prop at the depot, the additional costs of loss must be added to the cost of the successful flights.]

Total cost per successful flight:
=[Manufacturing cost]+[Launch Processing & Ops cost]+[Additional loss requiring replacement flight]

[baldusi]

Just to start, when you want to know how many launches you can expect form a launcher given:

p: Success Probability
Dl: Design Life

Then it is
Expected Life = Dl x p ^ Dl + [(1-p) x SUM (n=0 to Dl-1){n x p ^ n}]

Here is where I would ask you to start a new thread on Policy or somewhere like that, starting with your previous post. Because if we are going to develop an economic model, it will get long.

First thing to define here is if you are going to do an integrated analysis, where you make both payloads and launcher (i.e. govt style) or are you trying to fit a launcher in the market.
In the first case, you'd want to minimize the total present value cost, but that means adding the average cost of the payloads to each failure.
In the second case, you'd need a demand curve that's dependent on the reliability and service price, but then again, you'd have to derive a demand curve that includes payload cost, operation, insurance and lost revenue.
Generally when you start an economic analysis you try to make a very simple model, and then start adding things. Yours is a good point to start a discussion.
One of the things that I would think on first approach, is that you have to take into account the development cost, as a function of the desired reliability, expected launch rate, design life and desired launch cost, at the very least.
The difficulty is in defining the model. Then finding the optimum is basic calculus (basic if you are used to bordered hessian and dynamic problems, that's it). How would you like to go forward?

[oldAtlas_Eguy]

p: Success Probability
Dl: Design Life

Then it is
Expected Life = Dl x p ^ Dl + [(1-p) x SUM (n=0 to Dl-1){n x p ^ n}]

Thanks for the equation correction the RLV cases only showed movement to lower reliability values for the local minimum cost in case 1,2,&3 (81%,89%,89%) but no basic differences for the conclusion that for RLV’s designing with good reliability is best in the long run economically.

My goal for the model was to see if the basic premise of lower reliability is better from a cost per successful flight is correct or not. For RLV not really, for ELV probably yes but it will probably still be more expensive than a cost efficient RLV or end up breaking even on a $/kg to LEO.

[baldusi]

My point was that you have to add the payload cost. This is not only the hardware, but the whole operation and lost profit. Which again will run the operations towards high reliability.
Now, this doesn't means that having a fault tolerant architecture where the individual parts are less reliable but the overall mission is highly reliable might not be a possible solution.

[oldAtlas_Eguy]

My point was that you have to add the payload cost. This is not only the hardware, but the whole operation and lost profit. Which again will run the operations towards high reliability.
Now, this doesn't means that having a fault tolerant architecture where the individual parts are less reliable but the overall mission is highly reliable might not be a possible solution.

For the cost model the payload was propelant. A low cost item that would be contained probably in a stretcehed US to eliminate the need for a seperate container and set of systems. So there is no cost of payload per say just the booster. Unfortunately this would also be a single use LV, something that at this time is definitly not an economic possibility.

Yes in the end I believe higher reliability will in general be better economically especially for RLV's.

I think SpaceX has persued the redundancy method both electronic and engines as about as far as the current tech alows while still keeping costs low. Although there may still be more that could be done.