This is a write up of various measurements of the MC2 violin modes done from 7/29 to 9/12.
Much of the procedure was based on Keiko's LLO alog 5097. All excitations were done at H1:SUS-MC2_M2_DRIVEALIGN_L2L_EXC, and readback at H1:IMC-F_OUT (later H1:IMC-F_OUT_DQ).
For initial mode ID (7/29) I used awggui for excitation, with amplitude of 1000000, and a cheby1("BandPass",1,1,620,680) filter, and searched with DTT in the same range (620-680 Hz). I tried measurement bandwidths of 0.01 and 0.001 Hz but 0.01 Hz turned out to be plenty. For the higher harmonics I scaled the range by 2, 3 and 4. With this approach I was easily able to find all n=1 and n=2 modes, and 3 of n=3, but no n=4.
Later (9/5) I went back and did narrow band searches where I had previously found peaks, and where I suspected n=3 and n=4 peaks. I used DTT for excitation, with Gaussian noise of amplitude 5e6 over a range of 1 Hz for n=1, 2 Hz for n=2, etc, with filters like cheby1(BandPass",1,1,644,645). I labelled the four wires "a", "b", "c" and "d" in increasing order of mode frequency. Because the main focus was to find the n=4 peaks, I did 1a, 1b, 1c, 1d, 2a, and then skipped straight to the ones I hadn't found before:3a, 4a, 4b, 4c, and 4d, and this time I found them easily.
On 9/6 I started trying ringdown measurements. I adapted the DTT files from 9/5 as monitors, switching off the excitations (e.g., . To excite the modes I reverted to awggui, with a sine excitation with amplitude 30000 at the previously identified frequencies. I then did a parallel set of Triggered Time Response measurements, with 1 average of 500 s.
On 9/8 I went back and dide mode ID for the ones I skipped on 9/5, 2b, 2c, 2d, 3b, 3c, 3d.
On 9/12 I did ringdown measurements for all 16 modes, except that to work with Keiko's analysis script I used 10 averages of 1/10 the length. Since the Q's for the higher harmonics were expected to be similar but the frequencies were higher, I used 10*24 s for n=2, 10*16 s for n=3 and 10*12 s for n=4. This was nearly a disaster because I didn't export as I went and DTT has a bug that it doesn't save the separate segments in the .xml file for a triggered time response with averaging. Fortunately the start times were store in the files so I was able to rerun them retrospectively and export them as .txt files.
I adapted Keiko's scripts to process all the files as a batch and save the plots as a PDF (see first attachment) in the manner of the many other SUS scripts such as plotquad_dtttfs.m. The adapted scripts are in ^/trunk/Common/MatlabTools/ViolinModes and the primary script is plot_vm.m. It writes a summary PDF to ^/trunk/HSTS/H1/MC2/SAGM2/Results (or the like, depending on the settings at the start of the script). It also generates a summary table like the following (except that the notes in parentheses have been added by hand):
1a: f = 644.641 Hz, Q = 227793.765
1b: f = 648.734 Hz, Q = 223441.7436
1c: f = 654.758 Hz, Q = 225393.2642
1d: f = 661.492 Hz, Q = 220474.6201
2a: f = 1289.38 Hz, Q = 158588.9576
2b: f = 1297.58 Hz, Q = 160877.9141
2c: f = 1309.7 Hz, Q = 145061.1334
2d: f = 1323.02 Hz, Q = 155196.8117
3a: f = 1934.4 Hz, Q = 169442.0117 (poor S/N)
3b: f = 1946.62 Hz, Q = 142432.5364
3c: f = 1964.75 Hz, Q = 131593.7049
3d: f = 1984.85 Hz, Q = 131407.1051
4a: f = 2579.77 Hz, Q = 1454265.3967 (very poor S/N)
4b: f = 2596.1 Hz, Q = 2302393.4224 (very poor S/N)
4c: f = 2620.29 Hz, Q = 255698.7597 (poor S/N)
4d: f = 2647.04 Hz, Q = 161984.0682 (poor S/N)
All the n=1 and n=2 ringdowns had excellent S/N, as did 3 of the n=3. The n=4 ringdowns ranged from poor to very poor. I tried running the data analysis from T1200418 on the data through n=3 (see second attachment), but the fit is poor and results in unphysical negative values of structural phi.
When the modecleaner is next available, I will redo the n=3 and n=4 modes, experimenting with different actuation frequencies to get higher levels of excitation of the modes.
