Starship prototypes assembly timeline

[fael097]

Alright, here's an attempt to establish an assembly timeline for Starship serial numbers so far. Let me know if it makes any sense.

Most of those "they're building it much faster than the previous SN" comments sound baseless, so I figured I'd try to determine how much faster, if at all. (Of course I'm not comparing them with Mk1)

Some notes:
-First of all, this BB code table isn't optimal. Lots of compromises so everything fits. I wanted to keep it forum friendly instead of attaching a spreadsheet though. I might have to revisit this in the future;
-Space is limited so I'm calling bulkheads "domes" and common bulkhead "mid dome";
-Prototypes don't follow the same steps in the exact same order;
-They changed stacking order after SN1 so one of the steps is exclusive for SN1;
-"Since #1" means days elapsed since step #1 happened;
-Between brackets is the number of days since previous step. I didn't want to make yet another set of rows for that;
-SN3 reused SN2's common dome section so I didn't include that as it was made way before SN3 started;
-The date a SN was first spotted doesn't mean much, since it's impossible to know the exact day they started that SN, so I'm considering first dome being sleeved as step #1.

I'll eventually need to swap rows and columns if I'm going to keep adding SNs, but I suspect we won't be able to keep track of assembly like this for long, and coding tables in bbcode is tiresome.

Starship SN =>

SN1

-

-

SN3

-

-

SN4

-

-

SN5

Assembly step

Step #

Date

Since #1

Step #

Date

Since #1

Step #

Date

Since #1

Step #

Date

Since #1

First spotted

0

Jan 31

-5

0

Mar 4

-4

0

Mar 23

-10

0

Apr 11

-6

Fore dome sleeved

2

Feb 8

3 (+3)

1

Mar 8

0

1

Apr 2

0

1

Apr 17

0

Mid dome sleeved

1

Feb 5

0

-

-

-

2

Apr 4

2 (+2)

2

Apr 22

5 (+5)

Mid dome flipped

3

Feb 9

4 (+1)

-

-

-

3

Apr 5

3 (+1)

3

Apr 25

8 (+3)

Mid dome onto 4stack

N/A

N/A

N/A

3

Mar 17

9 (+2)

4

Apr 9

7 (+4)

5

Apr 30

13 (+2)

Aft dome sleeved

4

Feb 12

7 (+3)

2

Mar 15

7 (+7)

5

Apr 10

8 (+1)

4

Apr 28

11 (+3)

Fore dome onto mid

5

Feb 13

8 (+1)

5

Mar 19

11 (+2)

6

Apr 11

9 (+1)

7

May 4

17 (+3)

Aft dome flipped

6

Feb 15

10 (+2)

4

Mar 17

9 (+0)

7

Apr 12

10 (+1)

6

May 1

14 (+1)

Skirt top onto bottom

7

Feb 16

11 (+1)

6

Mar 21

13 (+2)

8

Apr 15

13 (+3)

8

May 6

19 (+2)

CH4 tank onto 4stack

8

Feb 17

12 (+1)

N/A

N/A

N/A

N/A

N/A

N/A

N/A

N/A

N/A

Tanks onto skirt

9

Feb 22

17 (+5)

7

Mar 26

18 (+5)

9

Apr 17

15 (+2)

9

May 12

25 (+6)

Moved to the pad

10

Feb 25

20 (+3)

8

Mar 29

21 (+3)

10

Apr 23

21 (+6)

First test

11

Feb 28

23 (+3)

9

Apr 2

25 (+4)

11

Apr 26

24 (+3)

Starship SN =>	SN6	-	-	SN8	-	-
Assembly step	Step #	Date	Since #1	Step #	Date	Since #1
First spotted	0	Apr 28	-8	0	Jun 5	-?
Fore dome sleeved	2	May 10	4 (+4)	-	-	-
Mid dome sleeved	1	May 6	0
Mid dome flipped	4	May 28?	22 (+2)
Mid dome onto 4stack	5	May 30	24 (+2)
Aft dome sleeved	3	May 26	20 (+16)
Fore dome onto mid	6	Jun 1	26 (+2)
Aft dome flipped	7	Jun 3	28 (+2)
Skirt top onto bottom	8	Jun 8	31 (+5)
Tanks onto skirt	-	-	-
Moved to the pad	-	-	-
First test	-	-	-

And some observations based on the data:
-It took them 20 days to move SN1 to the pad since its first dome was sleeved;
-21 days for both SN3 and SN4.
-SN5 is a week or more behind previous SNs.
-SN6 also delayed due to SN4 testing.

[CrazyHorse80]

I think that if you want to make it much more readable, you should use a spreadsheet and attach both the SS file itself (optional) and a picture of the table. Bbcode tables without borders and color formatting are really hard to read (for me at least) and rapidly become hard to manage as the dataset grows...

Edit: typo.

[envy887]

I tried doing something similar in Excel. It looks a lot better with a visual, reading dates is hard

[fael097]

yeah bbcode tables are suboptimal. I have it on excel but wanted a forum friendly format. I'll try something else eventually.

I tried doing something similar in Excel. It looks a lot better with a visual, reading dates is hard

The reason I made it is for comparing SN timelines so what matters to me are the actual dates and numbers, so we can infere how much ahead/behind schedule a SN is, and estimate how many days until it moves to the pad, begins testing, etc.

[georgebower]

@fael
@envy

Good data.
not sure if envy used fael's data or not for SN's but totally useful.

IMO what it shows :

1) The latest set of starships (labeled as SN) are showing significantly shorter assembly times than earlier MK

2) All SN SS seem to be taking approximately the same times

If I wrote article's I'd write it up:

"SpaceX demonstrates significant reduction in assembly time"

Note I purposely mis used "article's" just to irritate ChrisC

[fael097]

@fael
@envy

Good data.
not sure if envy used fael's data or not for SN's but totally useful.

IMO what it shows :

1) The latest set of starships (labeled as SN) are showing significantly shorter assembly times than earlier MK

2) All SN SS seem to be taking approximately the same times

If I wrote article's I'd write it up:

"SpaceX demonstrates significant reduction in assembly time"

Note I purposely mis used "article's" just to irritate ChrisC

My headline for the same article would be "SpaceX doesn't demonstrate any reduction in assembly time"

Of course I omitted Mk1, it was a completely different thing, they only had a fraction of their workforce and facilities, it was a full prototype and it took them 9 months since they started building it until it rolled to the pad, so it goes without saying that current SNs are being built faster than Mk1.

But I don't see a significant assembly time difference between SN1 and SN3 or even SN4 so far.

[georgebower]

@fael
@envy

Good data.
not sure if envy used fael's data or not for SN's but totally useful.

IMO what it shows :

1) The latest set of starships (labeled as SN) are showing significantly shorter assembly times than earlier MK

2) All SN SS seem to be taking approximately the same times

If I wrote article's I'd write it up:

"SpaceX demonstrates significant reduction in assembly time"

Note I purposely mis used "article's" just to irritate ChrisC

My headline for the same article would be "SpaceX doesn't demonstrate any reduction in assembly time"

Of course I omitted Mk1, it was a completely different thing, they only had a fraction of their workforce and facilities, it was a full prototype and it took them 9 months since they started building it until it rolled to the pad, so it goes without saying that current SNs are being built faster than Mk1.

But I don't see a significant assembly time difference between SN1 and SN3 or even SN4 so far.

I didn't mean to minimize your efforts.
I think both points are good ones

[fael097]

@fael
@envy

Good data.
not sure if envy used fael's data or not for SN's but totally useful.

IMO what it shows :

1) The latest set of starships (labeled as SN) are showing significantly shorter assembly times than earlier MK

2) All SN SS seem to be taking approximately the same times

If I wrote article's I'd write it up:

"SpaceX demonstrates significant reduction in assembly time"

Note I purposely mis used "article's" just to irritate ChrisC

My headline for the same article would be "SpaceX doesn't demonstrate any reduction in assembly time"

Of course I omitted Mk1, it was a completely different thing, they only had a fraction of their workforce and facilities, it was a full prototype and it took them 9 months since they started building it until it rolled to the pad, so it goes without saying that current SNs are being built faster than Mk1.

But I don't see a significant assembly time difference between SN1 and SN3 or even SN4 so far.

I didn't mean to minimize your efforts.
I think both points are good ones

No, I didn't take it that way, I just came to the opposite conclusion is what I mean 

[waveney]

-First of all, this BB code table isn't optimal. and coding tables in bbcode is tiresome.

There are some free bbcode table generators out there - I have used them in the past, they make tables a LOT easier.

I can't remember which one I used.

[OTV Booster]

@fael
@envy

Good data.
not sure if envy used fael's data or not for SN's but totally useful.

IMO what it shows :

1) The latest set of starships (labeled as SN) are showing significantly shorter assembly times than earlier MK

2) All SN SS seem to be taking approximately the same times

If I wrote article's I'd write it up:

"SpaceX demonstrates significant reduction in assembly time"

Note I purposely mis used "article's" just to irritate ChrisC

So, if appropriate in this discussion, why are “The latest set of starships ... showing significantly shorter assembly times...” while “All SN SS seem to be taking approximately the same times”? IIUC, you’re saying that from start to finish takes the same time but once the pieces are all together the stacking and close out is much faster.

Did I get that right? I’ve got problems reading the chart too.

Phil

[envy887]

yeah bbcode tables are suboptimal. I have it on excel but wanted a forum friendly format. I'll try something else eventually.

I tried doing something similar in Excel. It looks a lot better with a visual, reading dates is hard

The reason I made it is for comparing SN timelines so what matters to me are the actual dates and numbers, so we can infere how much ahead/behind schedule a SN is, and estimate how many days until it moves to the pad, begins testing, etc.

Yeah, mine was only for comparing Hopper/Mk1 to SN build speed, since some people hadn't notices the change in rate.

Here's a graph with your data plus a couple gaps interpolated. Darker lines are newer vehicles, lower slopes are faster builds. SN4 was pretty average until the last few milestones.

[Eka]

The complexity is also increasing from SN to SN.

[fael097]

yeah bbcode tables are suboptimal. I have it on excel but wanted a forum friendly format. I'll try something else eventually.

I tried doing something similar in Excel. It looks a lot better with a visual, reading dates is hard

The reason I made it is for comparing SN timelines so what matters to me are the actual dates and numbers, so we can infere how much ahead/behind schedule a SN is, and estimate how many days until it moves to the pad, begins testing, etc.

Yeah, mine was only for comparing Hopper/Mk1 to SN build speed, since some people hadn't notices the change in rate.

Here's a graph with your data plus a couple gaps interpolated. Darker lines are newer vehicles, lower slopes are faster builds. SN4 was pretty average until the last few milestones.

now this is absolutely fantastic. thank you! I just think SN3 has 4 days inverted as the steps happened in a different order

I just updated SN4 stacking and SN5 bulkhead sleeving

[Eylrid]

The other relevant metric is how long it takes them to go from one SN to the next.

Looking at the time between when each one's first dome was sleeved*:
SN1-SN3: 32 days**
SN3-SN4: 25 days
SN4-SN5: 15 days

As they refine their process they aren't waiting as long between the testing of one and the start of the next:
SN1 first test to SN3 start: 9 days**
SN3 first test to SN4 start: 0 days
SN4 first test to SN5 start: -? days

The time it takes them to build each one may stay the same, but they seem to be getting faster in terms of how often they pump them out.

*Ideally we would compare the first test dates, but we only have that for two SN's so far

**In fairness there was the SN2 test tank inbetween SN1 and SN3. So time from one test article to the next was less than that, but that's the time between full scale articles.

[steveleach]

Here's another visualisation of the same data.

I'd prefer to flip the X/Y axis, but I'm struggling a bit with Google Sheets charting.

[livingjw]

Here's another visualisation of the same data.

I'd prefer to flip the X/Y axis, but I'm struggling a bit with Google Sheets charting.

Totally agree, flip the axes!

John

[wes_wilson]

First, huge thanks to Rafael for this thread and data!  Here's another visualization variation mapping elapsed days from same step to same step over the builds. 

[meberbs]

First of all, thanks for this data, it can be hard to get these things in perspective.

The other relevant metric is how long it takes them to go from one SN to the next.

Yeah, this was one of the first things I checked with this data. SN 2 test article messes with the data a bit, but with SN 5 it now seems clear the overlap has increased, increasing the build rate even if start to finish time is unchanged.

Also, one source of improvement expected in start to finish time comes from the bulkhead assembly, which we don't have a good way to see directly. The reduced time between SNs implies some improvement has been made (if I remember right Musk had said it had been a week per bulkhead before.)

[wes_wilson]

Here's another visualisation of the same data.

I'd prefer to flip the X/Y axis, but I'm struggling a bit with Google Sheets charting.

I thinking flipping the axis is difficult because with a date axis, the dates are sequential, so the alignment and scale break.   Instead of dates, we can do a timeline.  Take the sighting of each SN as day 0 and baseline each step from there; then the axis flip works out.
 

[Shrike DeCil]

I'm seeing this as a much more "parallel process", which means fundamentally weird things.

Maybe I'm looking at it weird, but everything fits to my thinking better if you simply imagine separate "teams", each running their tent (or part of a tent) process. The high bay final assembly team is a different team (or two), no?

There's the ring team. They had a lot of trouble with early rings. But we have glimpses of the scrapyard, and after fiddling with the ringmaker, they've been drowning in "rings". To the point they aren't even tracked here . And the ring-making tent was dismantled and replaced with a new building (unfinished! ) ... without appearing to ever run out of rings.

There's the nose team. None of the noses have ever done anything but appear for a Cameo with Elon. My pet theory is the nose is the whole reason for the "high tent" - and I haven't seen photos inside since it was finished. How many nosecones are there? Now ... inside?

When SN5 gets the nosecone originally intended for SN2, does its construction time shift to "8 mo"?
And ... up to Raptor SN#26 ... grand total of one used in Boca Chica. When were they made?

Penalizing "Construction time" when the legos are already sitting there. Done. Seems odd.

"First sighting" is affected by how much space they have inside. Some of the early stuff was shoved outside for lack of room and sat untouched out there for days.
We've already reached a new place where they (seemingly) mostly come out for 'sleeving', 'flipping', or 'waiting for high-bay.'

The fact that one piece of SN5 is outside implies "Because we needed the darn space to work on..."


Hard to explain, but each lego needs its own stopwatch. Somehow. Until the point the actual production is moving through all the channels all the time and it can be judged on "Number per month".