PDA

View Full Version : Two options for Primary Scrambling Code in planning, which one is better?



leteo
2010-07-10, 02:09 PM
Hi experts,
I have a question needing help from all of you.
We have two options in planning Primary Scrambling Code (PSC) for WCDMA system:
1, All cells (or sectors) in ONE site have consecutive PSCs. For instance, if PSC of cell 1 is 100, then PSC of cell 2 is 101 and PSC of cell 3 is 102.
2, All cells (or sectors) in ONE site have discrete PSCs. For instance, if PSC of cell 1 is 100, PSC of cell 2 is NOT 102 (maybe it is 105, 109, or 95....), if PSC of cell 2 is 102, then PSC of cell 3 is NOT 103 (may be it is 200, 45, or...)
In theory, if I choose option 1, I am always true, because all PSC is orthogonal with each others and they do not make interference to each others, right? And my site and UE are still working normally.
In practical, I see that all vendors recommend to operators that they should choose option 2. Tracking this recommendation, I found in a Nokia's document, they explain that because of radio environment, if we choose option 1, fading (or something else) can make all PSC lost its orthogonality, and stop here. :confused: :confused: :confused:
By the way, anyone on this 4fum can make more details on this issue???
Appreciate if you attach relating docs.

Thank you.

cococrunch
2010-07-10, 02:50 PM
Can you post the document where you read that? Thanks. :)

Mctest
2010-07-10, 03:08 PM
All 512 primary SC are divided into 64 code groups with 8 codes per group.I suppose that second option is used for having codes from different code groups for the reason of better performance on cell search and cell sync procedure.

boring
2010-07-10, 03:17 PM
Hi experts,
I have a question needing help from all of you.
We have two options in planning Primary Scrambling Code (PSC) for WCDMA system:
1, All cells (or sectors) in ONE site have consecutive PSCs. For instance, if PSC of cell 1 is 100, then PSC of cell 2 is 101 and PSC of cell 3 is 102.
2, All cells (or sectors) in ONE site have discrete PSCs. For instance, if PSC of cell 1 is 100, PSC of cell 2 is NOT 102 (maybe it is 105, 109, or 95....), if PSC of cell 2 is 102, then PSC of cell 3 is NOT 103 (may be it is 200, 45, or...)
In theory, if I choose option 1, I am always true, because all PSC is orthogonal with each others and they do not make interference to each others, right? And my site and UE are still working normally.
In practical, I see that all vendors recommend to operators that they should choose option 2. Tracking this recommendation, I found in a Nokia's document, they explain that because of radio environment, if we choose option 1, fading (or something else) can make all PSC lost its orthogonality, and stop here. :confused: :confused: :confused:
By the way, anyone on this 4fum can make more details on this issue???
Appreciate if you attach relating docs.

Thank you.

From theoretical point of view there is absolutely no difference between the two options. Primary scrambling codes are very long and have extremely good cross-correlation properties which means that even in fading or more significantly in high multi-path conditions they do not lose their cross-correlation properties.

(Note: Becareful on the terminology: PSCs do not have perfect orthogonality but they have not been designed for this, like the OVSF spreading codes. They have been designed to have extremely good auto- and cross-correlation properties)

Now from experience and practise (cross-feeders, datafil errors etc), when it comes to drive testing it is far better to know that consecutive codes e.g. 100,101,102 come from the same site.

Thanks

PS: If you liked my response please add to my reputation (need to become member quickly!!!)

Mctest
2010-07-10, 03:46 PM
From theoretical point of view there is absolutely no difference between the two options. Primary scrambling codes are very long and have extremely good cross-correlation properties which means that even in fading or more significantly in high multi-path conditions they do not lose their cross-correlation properties.

(Note: Becareful on the terminology: PSCs do not have perfect orthogonality but they have not been designed for this, like the OVSF spreading codes. They have been designed to have extremely good auto- and cross-correlation properties)

Now from experience and practise (cross-feeders, datafil errors etc), when it comes to drive testing it is far better to know that consecutive codes e.g. 100,101,102 come from the same site.

Thanks

PS: If you liked my response please add to my reputation (need to become member quickly!!!)

Thanks,good reply,but you haven't mentioned code groups and belonging codes to the same or different code groups and how it influences on the cell search procedure.

s52d
2010-07-10, 03:52 PM
Hello!

There are 64 groups of 8 codes. According to The book (I think even Harry Holma mentioned it),
terminals decode a group first, then they find which of 8 codes it is.
In order to make decoding simpler/faster, there is rule 3:

"No two neighbors can be in the same group of codes."

Thus, when terminals is decoding pilots, it is enough to decode group and just verify
proper scrambling code.
Question: is this urban myth? Has anyone confirmed any benefit of terminal side?
With todays chips, not some 9 years old NEC?

Now, how to fullful this request?
We generate marix out of neighbor list and assign one group to any cell.
Thus, no two cosited cells are in the same group.
For each cell we need up to 31 groups. Some 50 groups are enough to find proper sollution.

In next step, we assign one out of 8 subcodes: this time we try to minimise fasle decoding
while searching for undefined neighbors. Distance is combination of distance in kilometers
and number of hops in neighbor relation matrix.

There are several networks using this principle: you can easily verify by TEMS pocket
whil roaming. From your post, step 2 is close.

For any engineer with some experience with heuristics, this is simple software.
I modified my BCCH planning software to assign groups, and BSIC planning to do subcodes.
SImple, fast - there is 64 gorups and 512 codes avaialble!
(ok, we use few more tricks just for fun).

One more question: how about multicarrier?
As scrambling code is used to distinguish cells on same frequency, we just use same
scrambling code for all carriers in one sector.
We heard some rumors of some terminals getting confused years ago - but never verified it.
You can see this approach in several networks.


So, to summarise:
If you do not understand step 3, then use step 2.
Just make a rule: at least 8 distance between codes in same site, for example:
80,88,96. For me, 87,83,99 is fine as well.

In fact, while traveling, I saw rule 1 on the air only years ago with young untuned networks.


BR
s52d





Hi experts,
I have a question needing help from all of you.
We have two options in planning Primary Scrambling Code (PSC) for WCDMA system:
1, All cells (or sectors) in ONE site have consecutive PSCs. For instance, if PSC of cell 1 is 100, then PSC of cell 2 is 101 and PSC of cell 3 is 102.
2, All cells (or sectors) in ONE site have discrete PSCs. For instance, if PSC of cell 1 is 100, PSC of cell 2 is NOT 102 (maybe it is 105, 109, or 95....), if PSC of cell 2 is 102, then PSC of cell 3 is NOT 103 (may be it is 200, 45, or...)
In theory, if I choose option 1, I am always true, because all PSC is orthogonal with each others and they do not make interference to each others, right? And my site and UE are still working normally.
In practical, I see that all vendors recommend to operators that they should choose option 2. Tracking this recommendation, I found in a Nokia's document, they explain that because of radio environment, if we choose option 1, fading (or something else) can make all PSC lost its orthogonality, and stop here. :confused: :confused: :confused:
By the way, anyone on this 4fum can make more details on this issue???
Appreciate if you attach relating docs.

Thank you.

firstmaxim
2010-07-10, 04:07 PM
Hi experts,
I have a question needing help from all of you.
We have two options in planning Primary Scrambling Code (PSC) for WCDMA system:
1, All cells (or sectors) in ONE site have consecutive PSCs. For instance, if PSC of cell 1 is 100, then PSC of cell 2 is 101 and PSC of cell 3 is 102.
2, All cells (or sectors) in ONE site have discrete PSCs. For instance, if PSC of cell 1 is 100, PSC of cell 2 is NOT 102 (maybe it is 105, 109, or 95....), if PSC of cell 2 is 102, then PSC of cell 3 is NOT 103 (may be it is 200, 45, or...)
In theory, if I choose option 1, I am always true, because all PSC is orthogonal with each others and they do not make interference to each others, right? And my site and UE are still working normally.
In practical, I see that all vendors recommend to operators that they should choose option 2. Tracking this recommendation, I found in a Nokia's document, they explain that because of radio environment, if we choose option 1, fading (or something else) can make all PSC lost its orthogonality, and stop here. :confused: :confused: :confused:
By the way, anyone on this 4fum can make more details on this issue???
Appreciate if you attach relating docs.

Thank you.
If you assign adjacent PSC so that they belong to the same PSC group (there are 64 of these), then cell acquisition can be faster.

Also, if the cells from site belong to same code group, the S-SCH from the different cells of the site can be superimposed to create stronger S-SCH signal as measured by the UE.

s52d
2010-07-10, 04:14 PM
Ups, this is in direct contradict to my post above.

I assume you have all cosited cells TS in sync (tcell parameter in E///)? We try to move
TS a bit to make life easier for terminals.

Have you actually verified it?
Any hint, how we can compare two approaches?
Years ago I did some experimetns with no statistically significant results, so I guess it
is more theoretical as practical issue?



If you assign adjacent PSC so that they belong to the same PSC group (there are 64 of these), then cell acquisition can be faster.

Also, if the cells from site belong to same code group, the S-SCH from the different cells of the site can be superimposed to create stronger S-SCH signal as measured by the UE.

Mctest
2010-07-10, 04:14 PM
Thanks,good post!But you didn't explain how the rule 3 makes decoding easier and faster.If neighbouring cells have SC under the same group - UE should process all SC under this group.If SC belongs to different groups it should detect code group and also process all codes under this group. In both cases the same algorithm.


Hello!

There are 64 groups of 8 codes. According to The book (I think even Harry Holma mentioned it),
terminals decode a group first, then they find which of 8 codes it is.
In order to make decoding simpler/faster, there is rule 3:

"No two neighbors can be in the same group of codes."

Thus, when terminals is decoding pilots, it is enough to decode group and just verify
proper scrambling code.
Question: is this urban myth? Has anyone confirmed any benefit of terminal side?
With todays chips, not some 9 years old NEC?

Now, how to fullful this request?
We generate marix out of neighbor list and assign one group to any cell.
Thus, no two cosited cells are in the same group.
For each cell we need up to 31 groups. Some 50 groups are enough to find proper sollution.

In next step, we assign one out of 8 subcodes: this time we try to minimise fasle decoding
while searching for undefined neighbors. Distance is combination of distance in kilometers
and number of hops in neighbor relation matrix.

There are several networks using this principle: you can easily verify by TEMS pocket
whil roaming. From your post, step 2 is close.

For any engineer with some experience with heuristics, this is simple software.
I modified my BCCH planning software to assign groups, and BSIC planning to do subcodes.
SImple, fast - there is 64 gorups and 512 codes avaialble!
(ok, we use few more tricks just for fun).

One more question: how about multicarrier?
As scrambling code is used to distinguish cells on same frequency, we just use same
scrambling code for all carriers in one sector.
We heard some rumors of some terminals getting confused years ago - but never verified it.
You can see this approach in several networks.


So, to summarise:
If you do not understand step 3, then use step 2.
Just make a rule: at least 8 distance between codes in same site, for example:
80,88,96. For me, 87,83,99 is fine as well.

In fact, while traveling, I saw rule 1 on the air only years ago with young untuned networks.


BR
s52d

s52d
2010-07-10, 04:25 PM
As I reacall (To lazy to start browsing books):
8 codes in within subgroups are distingushed by different pattern how some information is sperad over 15 TS. Thus, once in the group, it is faster to check one in the group (YES/NO) as to check two SC in the same group.

I found this referred few times in the literature years ago. Based on Moore observation, HSPA etc
it is quite likely chips today can do it much easier as those from 10 years ago, so it might be nonissue. I asked around and newer got decent answer.

firstmaxim pointed out S-SCH superimposing, and it smells good!

BR
s52d




Thanks,good post!But you didn't explain how the rule 3 makes decoding easier and faster.If neighbouring cells have SC under the same group - UE should process all SC under this group.If SC belongs to different groups it should detect code group and also process all codes under this group. In both cases the same algorithm.

cococrunch
2010-07-10, 07:13 PM
If you assign adjacent PSC so that they belong to the same PSC group (there are 64 of these), then cell acquisition can be faster.

Also, if the cells from site belong to same code group, the S-SCH from the different cells of the site can be superimposed to create stronger S-SCH signal as measured by the UE.

How does S-SCH is superimposed from different cells with same code group? Thanks.. :)

cococrunch
2010-07-10, 07:22 PM
As I reacall (To lazy to start browsing books):
8 codes in within subgroups are distingushed by different pattern how some information is sperad over 15 TS. Thus, once in the group, it is faster to check one in the group (YES/NO) as to check two SC in the same group.

I found this referred few times in the literature years ago. Based on Moore observation, HSPA etc
it is quite likely chips today can do it much easier as those from 10 years ago, so it might be nonissue. I asked around and newer got decent answer.

firstmaxim pointed out S-SCH superimposing, and it smells good!

BR
s52d

PSC detection is only done during cell synchronization, is this correct? Kindly explain if not for the benefit of us all.. Thanks :)

Mctest
2010-07-10, 08:38 PM
How does S-SCH is superimposed from different cells with same code group? Thanks.. :)

If cells belong to the same group they have the same S-SCH pattern of 16 S-SCH codes.But if I'm not mistaken for S-SCH superimpose avoidance tcell parameter is used.It is some kind of timeshift of S-SCH channel between co-sited cells,because all co-sited cells have the same timing.

boring
2010-07-10, 09:15 PM
Hello!

There are 64 groups of 8 codes. According to The book (I think even Harry Holma mentioned it),
terminals decode a group first, then they find which of 8 codes it is.
In order to make decoding simpler/faster, there is rule 3:

"No two neighbors can be in the same group of codes."

Thus, when terminals is decoding pilots, it is enough to decode group and just verify
proper scrambling code.
Question: is this urban myth? Has anyone confirmed any benefit of terminal side?
With todays chips, not some 9 years old NEC?

Now, how to fullful this request?
We generate marix out of neighbor list and assign one group to any cell.
Thus, no two cosited cells are in the same group.
For each cell we need up to 31 groups. Some 50 groups are enough to find proper sollution.

In next step, we assign one out of 8 subcodes: this time we try to minimise fasle decoding
while searching for undefined neighbors. Distance is combination of distance in kilometers
and number of hops in neighbor relation matrix.

There are several networks using this principle: you can easily verify by TEMS pocket
whil roaming. From your post, step 2 is close.

For any engineer with some experience with heuristics, this is simple software.
I modified my BCCH planning software to assign groups, and BSIC planning to do subcodes.
SImple, fast - there is 64 gorups and 512 codes avaialble!
(ok, we use few more tricks just for fun).

One more question: how about multicarrier?
As scrambling code is used to distinguish cells on same frequency, we just use same
scrambling code for all carriers in one sector.
We heard some rumors of some terminals getting confused years ago - but never verified it.
You can see this approach in several networks.


So, to summarise:
If you do not understand step 3, then use step 2.
Just make a rule: at least 8 distance between codes in same site, for example:
80,88,96. For me, 87,83,99 is fine as well.

In fact, while traveling, I saw rule 1 on the air only years ago with young untuned networks.


BR
s52d


some answers here:

1. the tcell parameter is used to offset the start of the radio frame for cells controlled by the same Node-B is VERY important for the cell search procedure during the 1st stage where the UE monitors the primay synch channel (PSH). if you don't use this number and the UE is located exactly in the mid distance betwwen the antennas of two different cells then it cannot differentiate the transmissions of these cells, hence the difficulties (default setting of 256 chips is sufficient)

2. from speed point of view, it does not make any difference to the UE whatsoever if neighbours belong to the same or different code groups. in fact the 3rd stage is always very fast because the pilot is trasmitted always at relatively higher power than the secondary SCH. from some past work i did while working for chip manufacterer, i came up to the conclusion that for the UE it makes better sense to have all neighbours from the same group. but the gains in all cases are marginal and because the hassle for the planner is more, this is why i don't really follow any rules when assigning codes to neighbour (which effectivelly is the point that you make) but it helps a lot (during optimisation) to have consecutive codes assigned to cells of the same site as i said earlier.

3. when multiple frequencies apply, simply re-use exactly the same code plan. And one note, we should always keep asside 100 codes for special usage (i.e. indoor sites) which are not used elsewhere.

firstmaxim
2010-07-11, 02:52 AM
Ups, this is in direct contradict to my post above.

I assume you have all cosited cells TS in sync (tcell parameter in E///)? We try to move
TS a bit to make life easier for terminals.

Have you actually verified it?
Any hint, how we can compare two approaches?
Years ago I did some experimetns with no statistically significant results, so I guess it
is more theoretical as practical issue?

There is a paper by: S. Kourtis, "Code Planning Strategy for UMTS - FDD networks".VTC 2000, Tokio, Spring 2000.


According to Kourtis, the distance between primary scrambling codes assigned to neighboring network cells impacts the performance of the UE, directly affecting the computational cost of the cell search procedure, and the synchronization time of the UE, at the frame level. In particular, with the increase of the number of scrambling codes groups assigned to adjacent cells, within the pool of 512 scrambling codes, the time needed for achieving synchronization rises, whereas the computational complexity is reduced. On the contrary, with the decrease of the number of scrambling codes groups assigned to adjacent cells, within the pool of 512 primary scrambling codes, the time needed for achieving synchronization decreases, but the computational complexity increases.


The choice of which strategy to adopt is not univocal, and depends on the considered scenario. According to Kourtis, in a case of a micro-cell network , typical of a urban environment, it is important to minimize the time required by the cell search procedure, particularly if the UEs exhibit a medium/high mobility: a poor synchronization performance means that the number of measurements taken by the UE per measuring period is small, thus the UE may not have the necessary information required to perform the soft-handover
efficiently.


In the case of a macro-cell network (with network cells of relatively wide area coverage), more typical of suburban and rural environments, the greater cells’ dimension allows tolerating higher delays in the cell reselection procedure.


Then, it is preferable to privilege strategies that minimize the computational cost of the mobile terminal, thus increasing the battery charge life, also in view of the fact that, on average, the transmitted power of the terminals is higher in macrocell environments than in micro-cell ones.


Regards.

firstmaxim
2010-07-11, 02:58 AM
PSC detection is only done during cell synchronization, is this correct? Kindly explain if not for the benefit of us all.. Thanks :)
PSC detection comes in the third stage, once the UE has acquired the WCDMA cell (after detecting Primary Sync Code) and the PSC code group (thanks to the Secondary Sync Code). The cell's PSC is arrived by doing a symbol to symbol correlation of the CPICH with the 8 PSCs (narrowed from the 512 codes).

Mctest
2010-07-11, 03:27 AM
It would be great if somebody explain us why in case of many groups computational complexity increases and in case of one code group per adjacent cells it reduces.

boring
2010-07-11, 04:01 AM
It would be great if somebody explain us why in case of many groups computational complexity increases and in case of one code group per adjacent cells it reduces.

Actually, it is the opposite:

case #1: Many groups --> low computational complexity (but long time)

case #2: Few groups --> high computational complexity (but short time)

now case #1 relies on heavy secondary SCH processing (this is one 256 code correlation per slot) whereas case #2 on heavy PSC processing (this on the other hand is one 256 code correlation per symbol, i.e. 10 per slot)

So, for macro (roof-top antennas so even for urban environements) deployment, according to paper (and to what i pointed out earlier), case #2 is preferable despite the complexity. if it helps imagine that a UE under case #2 does "small bursts" of calculations and most of the time nothing (the opposite for case #1)

BR

Mctest
2010-07-11, 04:35 AM
Why did you mention that case #1 relies only one SCH processing?After heavy stage 2 UE should process codes.why does it have low computational complexity?


Actually, it is the opposite:

case #1: Many groups --> low computational complexity (but long time)

case #2: Few groups --> high computational complexity (but short time)

now case #1 relies on heavy secondary SCH processing (this is one 256 code correlation per slot) whereas case #2 on heavy PSC processing (this on the other hand is one 256 code correlation per symbol, i.e. 10 per slot)

So, for macro (roof-top antennas so even for urban environements) deployment, according to paper (and to what i pointed out earlier), case #2 is preferable despite the complexity. if it helps imagine that a UE under case #2 does "small bursts" of calculations and most of the time nothing (the opposite for case #1)

BR

arefo1o
2010-07-11, 04:45 AM
Hi experts,
I have a question needing help from all of you.
We have two options in planning Primary Scrambling Code (PSC) for WCDMA system:
1, All cells (or sectors) in ONE site have consecutive PSCs. For instance, if PSC of cell 1 is 100, then PSC of cell 2 is 101 and PSC of cell 3 is 102.
2, All cells (or sectors) in ONE site have discrete PSCs. For instance, if PSC of cell 1 is 100, PSC of cell 2 is NOT 102 (maybe it is 105, 109, or 95....), if PSC of cell 2 is 102, then PSC of cell 3 is NOT 103 (may be it is 200, 45, or...)
In theory, if I choose option 1, I am always true, because all PSC is orthogonal with each others and they do not make interference to each others, right? And my site and UE are still working normally.
In practical, I see that all vendors recommend to operators that they should choose option 2. Tracking this recommendation, I found in a Nokia's document, they explain that because of radio environment, if we choose option 1, fading (or something else) can make all PSC lost its orthogonality, and stop here. :confused: :confused: :confused:
By the way, anyone on this 4fum can make more details on this issue???
Appreciate if you attach relating docs.

Thank you.


can you please share the document "I found in a Nokia's document, they explain that because of radio environment"
thanks in advance :)

cococrunch
2010-07-11, 05:08 AM
PSC detection comes in the third stage, once the UE has acquired the WCDMA cell (after detecting Primary Sync Code) and the PSC code group (thanks to the Secondary Sync Code). The cell's PSC is arrived by doing a symbol to symbol correlation of the CPICH with the 8 PSCs (narrowed from the 512 codes).

Thanks for your reply. So you confirm that PSC detection only occurs during cell synchronization?

cococrunch
2010-07-11, 05:18 AM
If cells belong to the same group they have the same S-SCH pattern of 16 S-SCH codes.But if I'm not mistaken for S-SCH superimpose avoidance tcell parameter is used.It is some kind of timeshift of S-SCH channel between co-sited cells,because all co-sited cells have the same timing.

Thanks a lot for your reply.. :)

Just one more thing. What do you mean by avoidance tcell parameter? Do you mean to be able to superimpose S-SCH we have to standardized the same value of Tcell for co-sited cells?

Mctest
2010-07-11, 05:53 AM
It was my guess,because I also try to find a clear explanation.But I know that tcell parameter is used for making time shift for SCH channels relatively to the CPICH timing in the co-sited cells.

Thanks a lot for your reply.. :)

Just one more thing. What do you mean by avoidance tcell parameter? Do you mean to be able to superimpose S-SCH we have to standardized the same value of Tcell for co-sited cells?

boring
2010-07-11, 02:29 PM
Why did you mention that case #1 relies only one SCH processing?After heavy stage 2 UE should process codes.why does it have low computational complexity?

UE does not have to go through all codes. in order to detect the PSC it uses the pilot which importantly carrieis a known spread bit sequence. so after de-scrambling/de-spreading the UE looks simply for this bit sequence, and if detected the search stops.

Mctest
2010-07-11, 03:36 PM
Now it's getting clearly and clearly.But if you want to skip high PSC processing you should assign on the adjacent cells 1st codes from different groups.For example cell1 belongs to group #1 and has SC1 then cell2 belongs to the group #2 and has SC9.But if you assign to cell2 SC 16 - it will process all codes in group2 until it find SC16.Am I right?


UE does not have to go through all codes. in order to detect the PSC it uses the pilot which importantly carrieis a known spread bit sequence. so after de-scrambling/de-spreading the UE looks simply for this bit sequence, and if detected the search stops.

boring
2010-07-11, 05:00 PM
Now it's getting clearly and clearly.But if you want to skip high PSC processing you should assign on the adjacent cells 1st codes from different groups.For example cell1 belongs to group #1 and has SC1 then cell2 belongs to the group #2 and has SC9.But if you assign to cell2 SC 16 - it will process all codes in group2 until it find SC16.Am I right?

if you refer to my previous e-mail, i say that this approach takes more time than the other so if you combine time and complexity it turns out that the overall processing needed is lower when assigning codes from the same group.

boring
2010-07-11, 05:02 PM
It was my guess,because I also try to find a clear explanation.But I know that tcell parameter is used for making time shift for SCH channels relatively to the CPICH timing in the co-sited cells.

a small correction. tcell param shifts all channels e.g. pilos, SCHs etc (actually the start of the radio frame) with respect to Node-B's common clock

Mctest
2010-07-11, 05:15 PM
a small correction. tcell param shifts all channels e.g. pilos, SCHs etc (actually the start of the radio frame) with respect to Node-B's common clock

Yes,you are right!

firstmaxim
2010-07-11, 05:38 PM
Thanks for your reply. So you confirm that PSC detection only occurs during cell synchronization?

I want to add a paper on 'Cell Search in CDMA'. Hope it helps.

Regards

d0000h
2010-07-14, 06:00 AM
Ahh, the timeless PSC question arises again (s52d’s post).
Our original design was with 8 groups of 64 though I was not involved in this decision. Later the national optimization group said the theory of quicker scanning via “same group planning” was not valid. Sorry I did not hear details at the time but this is what happened. After that, PSC planning was “random” with only strict emphasis on reuse (which can be brutally unforgiving).

Consider a design of 8x64 clusters. A phone camping in the middle of cluster “A” powers off, drives to cluster “B” and power on. ”Quick” initial scan happens on group “A” PSCs. Let’s say you can decode a group “A” cell though it is the 4th best server since you are now in group “B”. Does the UE not have to scan ALL the PSC in order to determine the strongest cell? You can’t set up on the 4th strongest...

The attached paper tries to claim an improvement in SSCH Ec/Io with same group planning vs different group planning. Is also states the UE camp-on time is unaffected by both frame timing offset and group planning. My opinion is that the only real issue is over all reuse as it can and will cause big problems even if sites are not neighbors and are not particularly close together. Whatever makes planning and managing of PSC the easiest is best.

I agree with boring’s comment of possible positive of consecutive PSCs is it can help with field testing.

We used same PSC for both frequencies and never had any reported issues.

Excerpt from part of the conclusion section:

“UE Camp-On Time”
As a final verification, the time from UE power-up until cell acquisition (reception of the first broadcast control channel [BCCH] message) was measured for all combinations (same and different code groups, 0- and 256-chip frame timing offsets). These measurements were repeated for three different types of UE. The results show no distinct pattern to indicate that the performance of any given scenario is better or worse than another. Thus, it was concluded that the use of the same code groups and 0-chip cell frame timing offset does not affect UE camp-on time.”

(No password, please give rep [I need it])

-Deuce-

leteo
2010-07-14, 02:03 PM
Ahh, the timeless PSC question arises again (s52d’s post).
Our original design was with 8 groups of 64 though I was not involved in this decision. Later the national optimization group said the theory of quicker scanning via “same group planning” was not valid. Sorry I did not hear details at the time but this is what happened. After that, PSC planning was “random” with only strict emphasis on reuse (which can be brutally unforgiving).

Consider a design of 8x64 clusters. A phone camping in the middle of cluster “A” powers off, drives to cluster “B” and power on. ”Quick” initial scan happens on group “A” PSCs. Let’s say you can decode a group “A” cell though it is the 4th best server since you are now in group “B”. Does the UE not have to scan ALL the PSC in order to determine the strongest cell? You can’t set up on the 4th strongest...

The attached paper tries to claim an improvement in SSCH Ec/Io with same group planning vs different group planning. Is also states the UE camp-on time is unaffected by both frame timing offset and group planning. My opinion is that the only real issue is over all reuse as it can and will cause big problems even if sites are not neighbors and are not particularly close together. Whatever makes planning and managing of PSC the easiest is best.

I agree with boring’s comment of possible positive of consecutive PSCs is it can help with field testing.

We used same PSC for both frequencies and never had any reported issues.

Excerpt from part of the conclusion section:

“UE Camp-On Time”
As a final verification, the time from UE power-up until cell acquisition (reception of the first broadcast control channel [BCCH] message) was measured for all combinations (same and different code groups, 0- and 256-chip frame timing offsets). These measurements were repeated for three different types of UE. The results show no distinct pattern to indicate that the performance of any given scenario is better or worse than another. Thus, it was concluded that the use of the same code groups and 0-chip cell frame timing offset does not affect UE camp-on time.”

(No password, please give rep [I need it])

-Deuce-
This paper is so good :p. But it does not still explain pros and cons between two options. Any one can do ?

d0000h
2010-07-15, 02:12 AM
This paper is so good :p. But it does not still explain pros and cons between two options. Any one can do ?

Seems to me that several people have already presented their opinion on the pros/cons of each method in this thread.

firstmaxim
2010-07-15, 02:55 AM
Ahh, the timeless PSC question arises again (s52d’s post).
Our original design was with 8 groups of 64 though I was not involved in this decision. Later the national optimization group said the theory of quicker scanning via “same group planning” was not valid. Sorry I did not hear details at the time but this is what happened. After that, PSC planning was “random” with only strict emphasis on reuse (which can be brutally unforgiving).

Consider a design of 8x64 clusters. A phone camping in the middle of cluster “A” powers off, drives to cluster “B” and power on. ”Quick” initial scan happens on group “A” PSCs. Let’s say you can decode a group “A” cell though it is the 4th best server since you are now in group “B”. Does the UE not have to scan ALL the PSC in order to determine the strongest cell? You can’t set up on the 4th strongest...

The attached paper tries to claim an improvement in SSCH Ec/Io with same group planning vs different group planning. Is also states the UE camp-on time is unaffected by both frame timing offset and group planning. My opinion is that the only real issue is over all reuse as it can and will cause big problems even if sites are not neighbors and are not particularly close together. Whatever makes planning and managing of PSC the easiest is best.

I agree with boring’s comment of possible positive of consecutive PSCs is it can help with field testing.

We used same PSC for both frequencies and never had any reported issues.

Excerpt from part of the conclusion section:

“UE Camp-On Time”
As a final verification, the time from UE power-up until cell acquisition (reception of the first broadcast control channel [BCCH] message) was measured for all combinations (same and different code groups, 0- and 256-chip frame timing offsets). These measurements were repeated for three different types of UE. The results show no distinct pattern to indicate that the performance of any given scenario is better or worse than another. Thus, it was concluded that the use of the same code groups and 0-chip cell frame timing offset does not affect UE camp-on time.”

(No password, please give rep [I need it])

-Deuce-
Thanks for the post Doooh. There seems to be a disconnect between theoretical reasoning and actual emperical measurements when it comes to PSC planning. I have not done any field measurements. So, I am tilting towards your reasoning.

Can someone post some results of PSC acquistion time for the two different types of PSC assignments?

Regards