[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Ext-GDE-87] comment from LLRF team (CCR#20)

Note that for the proposal we submitted, a 10% power overhead is included (see page 35 if the Main Linac BCD - an earlier version of the proposal with 27 cavities had only 6% power overhead) - the rf power distribution math is as follows

33.5 MV/m * 9.5 mA * 1.038 m = 330.3 kW  (Cavity Input Power)
* 26 Cavities
* 1 / 0.95 (Distribution Losses)
* 1 / 0.90 (Tuning Overhead)
= 10.0 MW

Note also that the 33.5 MV/m maximum gradient is 6.3% above the design gradient of 31.5 MV/m, so there is some additional overhead to allow for the variation in sustainable gradients among the cavities. Finally, when running at IP energies < 500 GeV, there will be more overhead.

The current BCD has 11% power overhead, which is close to that originally proposed in the TESLA TDR:

23.4 MV/m * 9.5 mA * 1.038 m = 230.7 kW  (Cavity Input Power)
* 36 Cavities
* 1 / 0.94 (Distribution Losses)
* 1 / 0.88 (Tuning Overhead)
=  10.0 MW

The cavities are assumed to have piezo-electric tuners, which should make the overhead required for Lorentz force detuning compensation small. Partial-quenches should be rare (if not, the gradient will be lowered) and the FB system should probably not try to compensate for them (remember that even with the full loss of one cavity, the IP energy changes by only ~ 1/8000). Beam loading may be more of a problem with the large variation in cavity gradients - we are working on a way to adjust power to individual cavities without dumping power (a prototype is being built) - with control of the cavity input power and Qext, one can compensate beam loading at gradient G by choosing

External Q = Qe = Qeo * ln(2) / ln (1 + G/Go * Qeo/Qe)

Input Power     = PI = PIo * (1/4) * (1 + G/Go * Qeo/Qe)^2 * (Qe/Qeo).

where Go and Qeo are the nominal values (from which the cavity fill time is determined). In this case, the energy gain along the bunch train is uniform but there is non-zero reflected power,

Reflected Power = PR = PIo * (1/4) * (1 - G/Go * Qeo/Qe)^2 * (Qe/Qeo).

which is relatively small.

SNS has a large overhead (times 2-3 in power) because it is a proton machine that includes low beta cavities, and because they want to eventually increase the beam current. TTF generally has a large overhead because they run relatively few cavities per klystron and do not have piezo-tuners in most of the cavities. No doubt they find the extra overhead useful, but it is also a short linac with bunch compression, so fine local control is needed.

XFEL will have a fairly large overhead (40-60% in power), but they also have a higher Qext (times 2), and with a 'smaller' machine, they can afford to be safe. As the linac BCD says, ' ... an overhead based on operation experience with ILC-like cryomodules should eventually be used'. 

-----Original Message-----
From: Shinichiro Michizono [mailto:shinichiro.michizono@xxxxxx] 
Sent: Thursday, November 02, 2006 10:23 AM
To: Nobu Toge; Warren Funk; GDE CCB; ml-ext-gde@xxxxxxxxxxxx
Cc: Brian Chase; Stefan Simrock; Shinichiro.michizono@xxxxxx
Subject: [Ext-GDE-87] comment from LLRF team (CCR#20)

Dear Toge-san,

This is the comment about CCR#20 from ilc-llrf team.

1.  6% overhead (3% amplitude overhead) is too small for feedback 
compensation of Lorentz force detuning, beam loading, exception 
handling (such as quench).
2. Operationa experiences at FLASH, SNS indicate >10% (in amplitude) 
will be necessary for stable operation.

Best wishes.

Shinichiro MICHIZONO
Accelerator Laboratory
High Energy Accelerator Research Organization(KEK)
1-1 Oho, Tsukuba, Ibaraki 305-0801 Japan

  e-mail: shinichiro.michizono@xxxxxx
      + + 4641 (PHS)