Twinax A Twinax is a dual coaxial cable

Twinax • A Twinax is a dual coaxial cable with a common shield (“Extension” of TWP). • 3 custom prototypes were produced for R&D purposes. • Currently Cu wire (later CCAl), low density polyethylene dielectric, Al foil shield, PU or polyester jacket. AWG 28 30 34 Martin Kocian (SLAC) Width/mm 2. 39 1. 75 1. 12 Height/mm 1. 32 0. 97 0. 64 Twinax Mass (kg/km) 5. 2 2. 55 1. 1 14 September 2016 1/9

Bit Error Rate Tests • Used a Virtex 4 bit error rate tester with a number of fixed rates to determine the maximum rates at which the transmission is error-free. • The table shows the max rates in Mbps. • The eye pattern is for 6 m of AWG 28 at 5 Gbps. AWG 28 30 34 length(m) 6 6 4 Martin Kocian (SLAC) Raw 3110 1555 DC bal. /preem. 8000 6220 Twinax 14 September 2016 2/9

New Bit Error Rate Tester • • • Kintex 7 instead of Virtex 4. Xilinx-supplied ibert core. Max rate is about 10 Gbps. PRBS-31 pattern is random non-DC-balanced. No 8 b 10 b capability but PRBS-7 pattern is equivalent to 8 b 10 b. 2 types of adaptive equalizers: LPM and DFE (high performance). AWG/l (m) 28/6 30/6 34/4 PRBS 31 6. 4 5 5 PRBS 31/pre 6. 4 PRBS 31/DFE 8 10 8 PRBS 7 8 6. 4 PRBS 7/pre 8 8 8 PRBS 7/DFE 10 10 10 =⇒ Higher rates with the new board. • Much improvement on transceivers between Virtex 4 and Kintex 7. • The equalizer on the Kintex 7 makes a big difference. Martin Kocian (SLAC) Twinax 14 September 2016 3/9

New Prototype • We plan to order a new Twinax prototype in about a month: • CCAl wires instead of copper. • Lower impedance (70 Ω? ) for smaller cable size. • AWG 30? • 1000 ft. Any input on specs or requests for cables is welcome. Martin Kocian (SLAC) Twinax 14 September 2016 4/9

Connections Interposer • • The twinax cable has to be connected on both ends (opto/EOS). Fuzz buttons provide a technology to do this. Interposer with two PC boards attached on boths sides. Twinax soldered to PC board. + Reliable and reasonably high density. - Expensive (about 4 Fr. per button). • An alternative could be a ZIF connector with kapton flex soldered to the Twinax. Martin Kocian (SLAC) Twinax 14 September 2016 5/9

Twisted Wire Pair Solution • TWP is low-mass solution high-density electrical data transmission solution for inner pixel layers • Achieves 5 Gbps bandwidth for short lengths [O(1 m)], but bandwidth drops quickly with increased length – Most likely final solution continues twisted wire pair with longer impedance-matched twin-axial cable – We expect the total module-to-optobox distance to be 1. 2 m + 6 m (more input on optobox location needed) • 100 Ω twisted-pair requires large conductor separation 120 ✓ z(Ohms) = p ln ◆ 2 s ✏ d • Signal propagation in skin-effect region (multistrand conductor) – Large number of strands should approximate solid-core performance J. Nielsen (UCSC/SCIPP) ITk Week -- 2016/09/14 6

Twisted Wire Pair Solution J. Nielsen (UCSC/SCIPP) ITk Week -- 2016/09/14 7

Current Models Under Test • • TWP 36 AWG multistrand w/ PEEK dielectric and shield TWP 36 AWG multistrand w/ PEEK dielectric, no shield TWP 40 AWG MS Rad-Soft (used for comparison) Hybrid [6 m 28 AWG twinax + 1 m 36 AWG MS TWP]: soldered connection between twinax and TWP • Hybrid [6 m 30 AWG twinax + 1 m 36 AWG MS TWP]: soldered connection between twinax and TWP • We are working on detailed material calculations for comparisons of current and future designs J. Nielsen (UCSC/SCIPP) ITk Week -- 2016/09/14 8

Summary of Bit-Error Rate Results • Comparison of solid core and multi-strand, 36 vs 40 AWG • We also see large variations in results for different PRBS – With PRBS-7, BER<10 -14 , even at 8. 25 Gbps! Need a standard for testing J. Nielsen (UCSC/SCIPP) ITk Week -- 2016/09/14 9

Design Adjustments • Even if the TWP is low-mass solution, non-negligible volume complicates stave loading – Current cables have outside diameter of 1100 microns (area 1 mm 2) – PEEK insulation contributes 660 microns of the total – Shield contributes 150 -200 microns • Can we reduce the volume and number of X 0? – Reduce conductor diameter d from 150 microns to 100 microns (36 to 38 AWG), maintaining multistrand conductor for strength – PEEK reduced by corresponding amount ✓ 120 – Total diameter 1100 -> 900 microns 2 s z(Ohms) = p ln • Reduce impedance from 100 Ω to 70 Ω? ✏ – Allows larger reduction of PEEK – Reduce outside diameter to 700 microns (0. 4 mm 2)? – Power considerations for long low-impedance cables • Drain wire for ease in connecting shields (no effect at high freq. ) J. Nielsen (UCSC/SCIPP) ITk Week -- 2016/09/14 ◆ d 1 0

CABLES TESTS Cables tested: TWISTED PAIR • • • 36 AWG (American Wire Gauge) ≈ 0. 127 mm diameter Multi-stranded Copper : 7 strands/ cable Each conductor is insulated with PEEK (is a colourless organic thermoplastic polymer, Poly-Ether-Ketone) - Different thicknesses has been measured for this insulator • • 390 µm for “transparent” PEEK 307 µm for “blue” PEEK Both conductors are shielded with 10 μm aluminium foil Backed with kapton insulator, 25 um thick Kapton insulator (25 µm) Signa l Peek Insulat or 0. 7 mm 14 September Leyre Flores Foil Shield (10 µm) 3 9 0 µ 1. 10 3 0 7 µ 11

CABLES TESTS Cables tested: TWINAXIAL (3 different versions) Twin. Ax Type/ parts Cu 28 Cu 30 Cu 34 Signal Wire Cu Cu Cu Signal wire OD (µm) 321 255 160 Drain wire OD (µm) 255 202 127 Dielectric material Polyethylene Dielectric diameter (µm) 1060 840 480 Al shield thickness (µm) 25 25 25 Jacket material Polyurathane Polyester Jacket thickness (µm) 100 100 Cable width (mm) 2. 39 1. 75 1. 12 Cable height (mm) 1. 32 0. 97 0. 635 Weight (kg/km) 5. 2 2. 55 1. 1 14 September Leyre Flores 12

CABLE TESTS Cables tested: SPACEWIRE • It was given by a cable company AXON for us to test it • 28 AWG • Conductor is aluminium coated with silver • Silver is not possible to use in the experiment!!! • Multistranded wire • Insulator is expanded PTFE 14 September Leyre Flores 13

CABLE TESTS RESULTS TABLE Type of cable AWG Length S 21 (d. B loss) @ 2. 5 Gbps Twisted pair 36 1 meter -3. 36 d. B Twisted pair 36 3 meters -8. 66 d. B Twisted pair 36 5 meters -16. 43 d. B Twinaxial 28 8 meters -10. 50 d. B Twinaxial 30 8 meters - 12. 37 d. B Twinaxial 34 8 meters -20. 74 d. B Spacewire 28 3 meters -4. 77 d. B Spacewire 28 7 meters -9. 79 d. B ***We will need to calculate the spacewire mass • Band-width completely dependent on the distance and AWG • We consider any result up to -12 d. B (thumb rule) 14 September Leyre Flores 14