DDS limits and perspectives

29
DDS limits and perspectives Alessandro D’Elia on behalf of UMAN Collaboration 1

description

DDS limits and perspectives. Alessandro D’Elia on behalf of UMAN Collaboration. Damped and detuned design. Detuning: A smooth variation in the iris radii spreads the dipole frequencies. This spread does not allow wake to add in phase - PowerPoint PPT Presentation

Transcript of DDS limits and perspectives

Page 1: DDS limits and perspectives

1

DDS limits and perspectives

Alessandro D’Elia on behalf of UMAN Collaboration

Page 2: DDS limits and perspectives

2

Damped and detuned design

• Detuning: A smooth variation in the iris radii spreads the dipole frequencies. This spread does not allow wake to add in phase

• Error function distribution to the iris radii variation results in a rapid decay of wakefield.

• Due to limited number of cells in a structure wakefield recoheres.

• Damping: The recoherence of the wakefield is suppressed by means of a damping waveguide like structure (manifold).

• Interleaving neighbouring structure frequencies help enhance the wake suppression

Page 3: DDS limits and perspectives

3

VDL

Page 4: DDS limits and perspectives

Why a Detuned Damped Structure (DDS) for CLIC

4

• Huge reduction of the absorbing loads: just 4x2 loads per structure

• Inbuilt Wakefield Monitors, Beam Position Monitors that can be used as remote measurements of cell alignments

• Huge reduction of the outer diameter of the machined disks

Page 5: DDS limits and perspectives

5

CLIC_DDS_A: regular cell optimizationThe choice of the cell geometry is crucial to meet at the same time:1. Wakefield suppression2. Surface fields in the specs

Consequences on wake function

Cell shape optimization for fields

DDS1_C DDS2_E

Page 6: DDS limits and perspectives

6

RF Properties of CLIC_DDS_A in comparison with CLIC_G

Parameters Units CLIC_DDS_A 8 x DDS_A 8 x DDS (Circular cells) CLIC_G

Fc (Amplitude) - 1.29 x 1024 * 3.4 x 105 * 6573 * 1.06 **

Frms (Amplitude) - 1.25 x 1027 * 2.8 x 107 * 5 x 106 * 5.9 **

Fworst (Amplitude) - 1.32 x 1028 * 7.5 x 108 * 1.55 x 108 * 25.3 **

Pulse length ns 276.5 - - 240.8

Peak input power (Pin) MW 70.8 - - 63.8

No. of bunches - 312 - - 312

Bunch population 109 4.2 - - 3.72

Max Esurf MV/m 220 - - 245

T K 51 - - 53, 47

SC W/m2 6.75 - - 5.4

bXm-2 1.36 x 1034 - - 1.22 x 1034

RF-to-beam efficiency % 23.5 - - 27.7

RF cycles - 8 - - 6

Cost - -

* 312 bunches, only first dipole band** 120 bunches, quarter structure GdfidL wake

Page 7: DDS limits and perspectives

7

A new approach: a Hybrid Structure for CLIC_DDS_B

WGD_Structure

+DDS_Structure

=

Hybrid Structure

Page 8: DDS limits and perspectives

8

Study of the wake functionThe problem

15 16 17 18 19 200

0.2

0.4

0.6

0.8

1

Freq (GHz)

Am

plitu

de (a

. u.)

RectangleGaussian

Product Re(Z)

F

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-1

100

101

102

time (ns)A

mpl

itude

(%)

Convolution WakefieldFFT(Rectangle)FFT(Gaussian)

571MHz; F=2GHZ

Question: How big must be F in order to have acceptable wake damping starting from 0.5ns?

Page 9: DDS limits and perspectives

9

Study of the wake function

Wt16-7V/[pC mm m], considering that W(0)170-180V/[pC mm m], the maximum acceptable bump must be 4%

F2.9GHz and 0.830GHz

0 1 2 3 4 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)FFT(Gaussian)

F=2GHZ

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)FFT(Gaussian)F=2.5GHZ

0 1 2 3 4 510

-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)FFT(Gaussian)

F=2.9GHZ

Page 10: DDS limits and perspectives

10

0 0.5 1 1.5 210

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

What about a “Sinc” wake?

0 0.5 1 1.5 210

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

Wake uncoupledWake coupled

This is the wakefield considering only the first dipole band

2Kdn/dfReal(Zx)

Page 11: DDS limits and perspectives

11

0 0.5 1 1.5 210

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

GdfidL “Full Wake”

1st Dipole wake from GdfidL

The presence of the higher order bands makes the scenario even less comfortable

Conclusion: It is not possible to control the position of the zeros along the wake, a smooth function of the

impedance is needed

What about a “Sinc” wake?

Page 12: DDS limits and perspectives

12

Can other types of distributions improve the wake decay?

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution Wakefield

FFT(Exp[-(x2/22)2)]FFT(Rectangle)

15 16 17 18 19

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Freq (GHz)

Am

plitu

de (a

. u.)

Rectangle

Exp[-(x2/22)2]

Product Re(Z)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)FFT(Gaussian)

906MHz F=2.9GHZ

830MHz

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Wakefield for Gaussian

Wakefield for Exp[-(x2/22)2]

Page 13: DDS limits and perspectives

13

Can other types of distributions improve the wake decay?

967MHz F=2.9GHZ

1.036GHz

0 1 2 3 4 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)

FFT(sech2[x2/2])

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)

FFT(sech1.5[x2/2])

14 15 16 17 18 19 200

0.2

0.4

0.6

0.8

1

Freq (GHz)

Am

plitu

de (a

. u.)

Rectangle

sech1.5[x2/2]

Product Re(Z)

14 15 16 17 18 19 200

0.2

0.4

0.6

0.8

1

Freq (GHz)

Am

plitu

de (a

. u.)

Rectangle

sech2[x2/2]

Product Re(Z)

Page 14: DDS limits and perspectives

14

Can other types of distributions improve the wake decay?

=1GHz

926MHz

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)

FFT(sech1.5[x2/2])

14 15 16 17 18 19 200

0.2

0.4

0.6

0.8

1

Freq (GHz)

Am

plitu

de (a

. u.)

Rectangle

sech1.5[x2/2]

Product Re(Z)

14 15 16 17 18 190

0.2

0.4

0.6

0.8

1

Freq (GHz)

Am

plitu

de (a

. u.)

Rectangle

Exp[-(x2/22)2]

Product Re(Z)

F=2.5GHZ

0 1 2 3 4 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)

FFT(Exp[-(x2/22)2])

Page 15: DDS limits and perspectives

15

What about 0.67ns?

F=2GHZ

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)

FFT(sech1.5[x2/2])

14 15 16 17 18 190

0.2

0.4

0.6

0.8

1

Freq (GHz)

Am

plitu

de (a

. u.)

Rectangle

sech1.5[x2/2]

Product Re(Z)

0 1 2 3 4 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)FFT(Gaussian)

0 1 2 3 4 510

-2

10-1

100

101

102

time (ns)

Am

plitu

de (%

)

Convolution WakefieldFFT(Rectangle)

FFT(Exp[-x4/(24)])

Page 16: DDS limits and perspectives

16

How big is the bandwidth we may achieve?

2 2.5 3 3.52.3

2.4

2.5

2.6

2.7

2.8

SlotW (mm)

Ban

dwid

th (G

Hz)

Assuming SlotW constant throughout the full structure

1.5 2 2.5 3 3.50

200

400

600

800

1000

1200

1400

SlotW (mm)

Avo

ided

cro

ssin

g (M

Hz)

CLIC_DDS_ACLIC_P

We must consider that 400-500<Av. Cross.<800-900 in order to get Qs in the order of 500-600 which will preserve the fsyn distribution

NB: The BW has been evaluated considering the difference between 1st Reg. Cell and Last Reg. Cell, i.e. Cell#27, but the total number of the cells is 26 (26 cells 27 irises); then the real BW will slightly decrease in the real structure

Geometric Parameters

a (mm) 4.04-1.94

L (mm) 8.3316

t (mm) 4-0.7

eps 2

WGH (mm) 5

WGW (mm) 6

Page 17: DDS limits and perspectives

17

Bandwidth coupled and uncoupled

5 10 15 20 2515.5

16

16.5

17

17.5

18

18.5

# of cell

Fsyn

(GH

z)

Coupled (from GdfidL)Uncoupled - Uncoupled 27 cells: F= 2.685GHz

- Uncoupled 26 cells (not shown): F= 2.47GHz- Coupled (GdfidL): F= 2.363GHz

2 2.5 3 3.52.3

2.4

2.5

2.6

2.7

2.8

SlotW (mm)

Ban

dwid

th (G

Hz)

From theoretical distribution to real structure one must take into account a reduction of ~200MHz in the BW

Av. Cross~600MHz

Page 18: DDS limits and perspectives

18

0 0.2 0.4 0.6 0.8 110

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

What is the bandwidth of the real coupled structure?

GdfidL

Reconstructed wake (only 1st Dipole band)

Uncoupled wake with 25 peaks (F=2.314GHz)

0 1 2 3 4

100

101

102

s (m)

wak

e

The uncoupled wake with 25 frequencies (black dashed curve, F=2.314GHz) falls faster than the 1st dipole band reconstructed wake from GdfidL (red dashed curve): is there any strange effect from uncoupled to coupled that further reduce the bandwidth?

Page 19: DDS limits and perspectives

19

Non Linear Fit to improve wake reconstruction

The procedure:• I take GdfidL wake as “objective” function of my

non linear regression• I use reconstruction formula as my fitting function • Fsyn are considered as given from Lorentzian fit of

the impedance peaks while Qdip and Kicks are the parameters to be optimized

• Initial guess for Qdip and kicks are from Lorentzian fit

Page 20: DDS limits and perspectives

20

Results (1)The agreement with GdfidL is quite good and, as expected, the new procedure produces a major correction at the beginning of the curve while for the rest there are no appreciable variation with the wake reconstructed using the data from Lorentzian fit.

0 0.2 0.4 0.6 0.8 110

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLFrom LorentzianNon Linear Fit

0 5 10 15 20 25-40

-20

0

20

40

60

N

Kic

ks (V

[pC

mm

m])

Non Linear FitLorentzian

0 5 10 15 20 250

500

1000

1500

2000

2500

N

Qdi

p

Non Linear FitLorentzian <Qdip>=312

<Qdip>=512

=94=67

It is clear that the wake is reconstructed from unphysical values of kicks and Qdip. Constraints on the parameters are needed.

0 1 2 3 4 510

-2

100

102

104

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLLorentzianNon Linear Fit

Page 21: DDS limits and perspectives

21

Results (2)

<Qdip>=312<Qdip>=337

=94=67

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLFrom LorentzianNon Linear Fit

0 5 10 15 20 250

1

2

3

4

5

6

N

Kic

ks (V

/[pC

mm

m])

LorentzianNon Linear Fit

0 5 10 15 20 250

200

400

600

800

N

Qdi

p

LorentzianNon Linear Fit

With same constraints and an appropriate length of the wake, kicks and Qdip starts to converge.

Page 22: DDS limits and perspectives

22

First results for sech1.5

15.5 16 16.5 17 17.5 18 18.510

20

30

40

50

60

70

Freq (GHz)

2Kdn

/df (

V/[

pC m

m m

GH

z])

0 0.2 0.4 0.6 0.8 110

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

Uncoupled Wake for 26 CellsGdfidLBunch position (6 RF cycles)

2Kdn/df Very sharp deep, before 0.15m

Need to finalize the simulation to finalize the analysis

Very preliminary

Page 23: DDS limits and perspectives

23

Conclusions• With conventional DDS (DDS_A) it seems very difficult to meet beam

dynamics criteria• With hybrid DDS, using Gaussian distribution, it seems non realistic to

get damping within 6 RF cycles • With different distribution (in particular sech1.5) it is possible to relax the

constraint on the BW and this could allow to stay in the 0.5ns bunch spacing

• Play with Kdn/df would be interesting to see what happen and especially whether it is possible to increase the bandwidth by distributing differently the frequencies

• However the requirement of 0.5ns is quite tricky and we have not yet considered surface fields…

• I would not close totally the door to 8 RF cycles

Page 24: DDS limits and perspectives

24THANKS Igor

Page 25: DDS limits and perspectives

25

Additional slides

Page 26: DDS limits and perspectives

26

Physical interpretation of the resultConstraints:• First and last three peaks in the impedance are well separated then their Qdip and kicks are considered fixed• The rest of the kicks must be positive and spanning in a range from zero to roughly 10• The rest of the Qdip can span from zero to a maximum of 1500

<Qdip>=312<Qdip>=576

=94=67

Wake is still well approximated but kicks and especially Qdip do not seem correct. The constraints I gave are still not enough.

0 0.2 0.4 0.6 0.8 110

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLFrom LorentzianNon Linear Fit

0 5 10 15 20 250

1

2

3

4

5

6

N

Kic

ks (V

/[pC

mm

m])

Non Linear FitLorentzian

0 5 10 15 20 250

200

400

600

800

1000

1200

N

Qdi

p

LorentzianNon Linear Fit

Page 27: DDS limits and perspectives

27

Extrapolation for longer wakeIf I extrapolate for a longer wake it is clear that Qdip and kicks evaluated from Non Linear Fit are not correct.

0 5 10 15 2010

-4

10-2

100

102

104

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLLorentzianNon Linear Fit

0 1 2 3 4 510

-2

100

102

104

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLLorentzianNon Linear Fit

0 0.2 0.4 0.6 0.8 110

-1

100

101

102

103

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLLorentzianNon Linear Fit

I need more wake to improve Qdip calculation

Page 28: DDS limits and perspectives

28

Increasing the length of the wake: 10m

<Qdip>=315<Qdip>=312

=67=67

This makes me much more confident on the wake reconstruction

0 2 4 6 8 1010

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

GdfidLFrom LorentzianNon Linear Fit

0 5 10 15 20 250

1

2

3

4

5

N

Kic

ks (V

/[pC

mm

m])

10m5m

0 5 10 15 20 25100

200

300

400

500

600

700

N

Qdi

p

10m5m

Page 29: DDS limits and perspectives

29

Going back to the beginningQuestion was: can I evaluate the bandwidth reduction from uncoupled?

0 0.1 0.2 0.3 0.4 0.510

-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

Uncoupled 27 CellsCoupled from Non Linear FitUncoupled 26 CellsUncoupled 25 CellsCoupled from Lorentzian 0.1 0.12 0.14 0.16 0.18 0.2

10-1

100

101

102

s (m)

Wak

e (V

/[pC

mm

m])

Uncoupled 27 CellsCoupled from Non Linear FitUncoupled 26 CellsUncoupled 25 CellsCoupled from Lorentzian

From GdfidL

Uncoupled 25 Cells

Uncoupled 27 Cells

Uncoupled 25 Cells

Uncoupled 26 Cells

2Kdn/df

Answer: It seems Yes, with some minor approximation. In particular in this case it is clear that the major reduction comes from one peak which is missed. Then I estimate a reduction of ~230MHz and not of 322MHz if I choose ~2.75GHz, I should stay around 2.5GHz which is the minimum required for sech1.5 distribution.