Cable Filter Design

N. Kitamura

03/26/2004

Purpose of the Cable Filter

• The cable filter is to be inserted between a DOR channel and a DOM.
• The cable filter is to be used when the cable between the DOR channel and the DOM is approximately 20* m (TBR), the length of the cables to be used for wiring the DFL.
• The cable filter is to introduce a high-frequency roll-off as well as a DC attenuation mimicking a 3.5* km-long In-Ice cable.
• The cable filter is to allow a RAP cal when the cable between the DOR card and the DOM is short (length specified above).
• The cable filter is to allow a smooth** switch-over from the short-cable testing environment to the long-cable deployment environment.

*TBR--To Be Reviewed.

**What is considered smooth must be discussed and agreed upon.

This webpage focuses on the design and implementation of the filter.  For information of the RAPcal measurements at UW-Madison, see Mark Krasberg's webpage.

Assumptions

In-Ice Cable	Nominal length = 3.5 km Characteristic impedance = 145 Ohm Loop resistance = 44 Ohm / km
RAP cal input pulse characteristics

COMM ADC

Sampling rate = 20 MHz Resolution = 10 bit Scale = 400mV* per 1024 counts.  *ADC input range (2V) divided by the differential preamp gain (5).

Filter Design Information

Summary: CableFltDoc.pdf

Andy Laundrie's design info for various cable length with datasheets: http://docushare.icecube.wisc.edu/docushare/dsweb/View/Collection-994 

Gerber files: CableFlt2.zip

Parts and cost information

Simulated response (3.5km filter)

Ouptut  DFLFILT3K5MA.pdf

Netlist DFLFILT3K5MA_NET.pdf

Schematic DFLFILT3K5MA_sch.pdf

Measured response with an arbitrary waveform generator source

Setup diagram: labsetup.jpg Numerical data: Input (Trace B): scope_in.txt Output (Trace A): scope_out.txt

Comparison: Simulation vs Actual Response

(The curves have been manually shifted in x-direction to overlap with one another.)

Comparison: Filtered Response vs Long Cable's Response 

(The curves have been manually shifted in x-direction to overlap with one another.)

Why is the red curve smoother? ==> Because Kalle's FPGA code performs a running average over every four sampled points.  See Kalle's presentation (page 17): running_ave/DOR_DOM_Comm.pdf

How the running average was applied to the scope data is shown in the mathcad document: running_ave/running_ave.pdf 

If you want to run the MathCad document yourself, use the following:

• running_ave/running_ave.mcd
• running_ave/Filtered.txt
• running_ave/LBNL3k5m.txt

The comparison of the two curves, after applying the running average is shown below:

The scope measurement with and without the running average, as well as the LBNL cable data, are compared in the below (magnified view):

Finally, the red and the green are shifted relative to each other and are compared (now in physical units):

The response of Andy Laundrie's filter looks very much like the response of the 3.5km-long cable.

==================================================================================

Andy's Revised Filter

Andy Laudrie announced a revised component values on 2/19/2004 (read your email), saying that the previous design gave too fast a rise time in the response.  As seen in the above plots, that doesn't seem to be the case.  The actual response measured using the revised filter is shown below.  The previous design seems to match the actual 3.5km cable response more closely than this revised version.

The input signal is distorted and has the amplitude smaller than what is programmed into the arbitrary waveform generator.  The symptom is consistent with an impedance mismatch.  The measurement was done in the identical setup as in the previous scope shot.

The following is a comparison that after applying the running average to the scope data.

See more detail in the following files:

• MathCad calculation of the running average:revised_filter/running_ave2.mcd
• Data for MathCad document: revised_filter/revised_filt.txt, revised_filter/LBNL3k5m.txt
• Excel sheet: revised_filter/comparison3.xls

===================================================================================

By the way, the following is still an unresolved mystery:

Comparison: RAPcal with Long Cable vs RAPcal with Filter

(The curves have been manually shifted in x-direction to overlap with one another.)

The large discrepancy in amplitude is a mystery !? at this time (2/13/2004).

Excel datasheet: comparison2.xls