User:Pranav Rathi/Notebook/OT/2011/06/21/Noise issue with the optical tweezers
This page is dedicated to the different noise issues we experience during the data acquisition. I might decorate and elaborate this page very well but for now I am concentrating on the more important issue being faced right now. I am experiencing a weird oscillation in the data plane while acquiring the data. I mean by data plane is the plane in which the data is acquired. Some preliminary information about the oscillations:
- Oscillations are more prominent during the DOG scan over the DOG profile.
- Oscillations in X-signal are not continuous, they come and go.
- Oscillations in y-signal are also like that.
- But oscillations in the sum signal look continuous.
- According to my observations two or more oscillations exists; 60~70 Hz and 120 Hz.
- I do not know the amplitude yet.
- The prime suspects are x-Piezo, QPD(x and y channel interference); environmental acoustic noise/unidentified noise object (UNO, DAQ cross talk. This list might increase.
- 1 June/21/2011
- 2 June/22/2011
- 3 June/23/2011
- 4 June/24/2011
- 5 June/27/2011
- 6 Aug/16/2011
- 7 Aug/20/2011: Investigation of 120Hz oscillation problem
- 8 Sep/01/2011: Search for the source and remedy
- 8.1 Search for the sources
- 8.2 10 Oct/01/2011: Remedy: Design and Solution
- 8.3 Design & solution
- 8.3.1 New Redesigned Parts (stage 1)
- 8.3.2 Isolation of parts by redesigning (stage 2)
- 8.3.3 Deal with active and passive noise sources and build an Enclosure (stage3)
The investigation is under progress.
- There are two different oscillations exist. One is around 60Hz, and these are due to on-track (amplifier and controller of QPD). This oscillation is on-track gain dependent and never goes away even when the QPD is not connected to on-track. These oscillations can be controlled by manipulating the ground of the on-track power supply. I designed a weird circuit to do that and it is successfully controlling the noise.
- 120 Hz is the oscillation exists. This oscillation is not continuous like 60Hz. It comes and goes. I am not surprised that this oscillation is piezo damping dependent. I compared sets of data, between piezo is damped and not damped. I found some difference and looking more into it.........
The 60Hz periodic oscillations are due to the power supply came with on-track. The power supply does not have ground enforced polarity and maybe it is a half-wave rectifier (I am not positive on that). I tried different power supplies with the same result. The ground manipulation on on-track power supply worked but I was not satisfied. So I used a different power supply model: TR2V1000N00 with regulated +12V DC at 300mA. This power supply gives a regulated voltage with ground. If I use this power supply without the cancellation circuit, than it is no different. So I used a cancellation circuit in which I have a 300uF 200V capacitor in parallel to the output of the power supply and a ground which is directly connected to the common. The capacitor absorbs any fluctuation in the current and keeps the modulation at minimum, and any unwanted noise is grounded. Another possible noise source might be AC power-lines and shutter-powersupply situated close to the signal carrying wires. A comparison is shown below.
<html><iframe height="400px" width=850px" src="http://www.evernote.com/pub/pranavrathi1/noiseissues#b=4dabf319-691b-42f5-ac6c-8fc5dcf7bbe7&n=38476a58-734b-488b-9815-4bb5d30f074d"></iframe></html>
I have also tested 12 V battery option. This option is little better overall noise wise. On-track needs 12V DC at 300mAmps. Any rechargeable-battery over 5 Ah can give 15 hours of constant power which is more than enough. Another exercise done today is isolation of the data acquisition hardware from various AC power lines. Data acquisition hardware includes, on-track, BNC wires, DAQ and QPD. I revised the whole electrical setup for tweezers, so now all the AC lines are far away from the acquisition hardware. This will definitely help in reducing the overall noise. On Monday I will verify it. Next Task on Monday: investigation of 120 Hz oscillations with nano drive piezo-alternate.
60Hz noise issue is fully resolved now. The were two causes behind this noise. First x,y and sum signal BNC-wires were running close to the AC line and shutter-Powersupply. Power-supplies with transforms have 60Hz electromagnetic filed which might induce current into a conductor running close to it through electromagnetic induction. AC lines can do the same. Second the power supply for on-track was missing the ground.
I rearranged everything so the signal wires are isolated from any power-lines and power-supplies in vicinity. I changed the power-source for on-track to battery ( am using RYOBI cordless power drill's battery, which can supply power for hours to on-track, and we have two of them. We can use one when the spare is on charge). This solves the problem.
The results are presented in slide 2 to 7. Slide 2 and 3 shows the noise through power-spectrum. I wrote a lab view program in lab view v7.1 for noise detection see slide 29. This program can take the data produced by the feedback program as an input and find-out any existing frequencies. The most prominent frequencies can be clearly identified on x-axis which is frequency axis. Y-axis is power-axis in DB. We uses quadrant photo diode as a detector, which gives three separate responses; x-signal, y-signal and sum signal of both. Sum signal is most appropriate to identify the noise frequencies.
Stiffness Vs time presents the sum signal Vs time so it is more appropriate to find the frequencies in comparison to x-signal OR y-signal Vs time. The data in slide 2 was taken on may/18/2011. 60Hz noise with its 3rd and 5th order multiples (180 and 300) can be clearly identified in power-spectrum graph.
Data in slide 4 was taken on June/02/2011. Similar result can be seen here too. The 60Hz noise oscillations in a form of beating can be seen by zooming into the data as shown in the satellite figure.
Slide 5 shows the solution with ontrack and its power-supply. Slide 6 and 7 data was taken on june/30 and july/14. There are no prominent oscillations, the relative strength of the noise is in the order of other background noise which is great!!
In case of 120Hz, i found that piezo might be the problem. I compared few sets of data between piezo dumped and not dumped. I found that when piezo is dumber the oscillation are weaker in magnitude. I am looking more into it, but if this is the case than i doubt it that this can be solved without changing the piezo stage.
Designing a new microscope stage.
Aug/20/2011: Investigation of 120Hz oscillation problem
In the beginning there were two major noise-traces in the data. One was 60 Hz and other one was 120Hz. 60 Hz was due to two main causes; AC power line running close to the signal-wires and ontrack-power-source was not being grounded. This problem was solved by rearranging all the wires and using DC power-source (battery) for ontrack QPD-driver.
120 Hz noise was not easy to investigate; It is an airborne noise, the bandwidth of this noise is roughly 10 to 15 Hz, which means it goes from 110 to 125 Hz, and this might not the only frequency. So I had to device a way to investigate this noise in the optical tweezers: The entire device I made including the techniques and results are presented in the slide-presentation. So I will refer to the corresponding slide-number while discussing the relative.
In slides 9 and 10 a data comparison is shown: Ideal Vs noisy. The 120 Hz noise oscillations can be clearly seen. An optical tweezers can resolve the DNA-unzipping to the DNA base-pairs which are .338nm apart. This information is vital when DNA is unzipped in the presence of site-specific proteins. This noise definitely interferes with that.
To solve this problem the first step is to investigate the structure (setup) dependence (OR find the resonance of the setup) on the airborne noise frequency and vibration-frequency. The optical-tweezers setup consists of variety of materials like Aluminum, Steel, Glass, plastic and rubber. Each material has its unique response to the acoustics called its mechanical-response to its resonant frequency. This response is also 3-D design dependent; for an example a plane thick square sheet (12"X12") of aluminum can propagate a particular frequency with different attenuation in comparison to the same sheet with a grid of wholes (1/4" of hole diameter at every square inch). But mechanical-resonance will somewhat stay the same. This affects both vibration-frequency and airborne noise frequency. So first I did an experiment to test for the resonant frequency (most sensitive frequency band) for the whole setup all together which includes aluminum stages and breadboard, glass slide and steel posts.
Test for frequency response of the setup
I am not going into the whole chapter of transmissibility Vs frequency and airborne acoustic noise Vs material frequency response to that noise by the materials. I have chosen an easier way. I wrote a lab view V9 program to produce an acoustic wave from 0Hz to 22 KHz. It can generate sin, triangle, square and sawtooth waves of any amplitude with any offset with a power-spectrum display DB Vs frequency see slide 12. I used computer built-in sound-card and an external low power sound amplifier of unknown power-output. Since the computer sound-card, external amplifier and speakers all have their own range of frequency response, so no matter what frequency I generate at the program, it will only show-up at the speakers if it falls in that range, which is 100Hz to 20KHz. But this range is enough for the test; I am more interested in the range of 100 to 1000Hz, because I think this is what I am seeing in my data. The frequencies above 1.5KHz are useless for direct observation because we use a low-pass filter at 1.5kHz in data acquisition.
The test setup is relay simple see slide 11. I have a speaker connected to the output of the external amplifier which is connected to the sound-card output at the ear-phone jack. First I placed the speaker on the optical table about 1 meter away from the tweezers and second placed the speaker on a chair so no physical contact between the tweezers and the speaker. The sound and vibration produced by the speaker can affect the tweezers in two ways; by direct acoustic airborne noise and vibration transmissibility through the structure in first and only through acoustic airborne noise in second. I will start at 100Hz and take data (DOG; I made a stuck bead sample with water) every 5Hz increment.
There are 3 things to consider in advance in this test; first it is not necessary that the vibration frequency is same as the sound frequency produced by the speaker at all frequencies, second it is possible that the structure (setup) can absorb the energy from one frequency and resonate at another. But this behavior should not affect the results much. In other words we can still find the resonant frequency or frequencies of the structure by producing different frequencies at the speaker. Third, (because of the first and second) in the data we can see a result due to beating between structure vibration frequency, speaker generated frequency and other frequencies present. So counting oscillations at the time axis does not always work to find the frequencies present, because the oscillations do not have to belong to a single frequency. So the right way is to do a FFT-power-spectrum of the data, so all prominent frequencies can be found simultaneously.
My expectation from this test is to find: Which frequency or frequencies are more transmissive through structure, which frequency or frequencies are more intervening through airborne noise, which is more dominating between vibration transmissibility and airbornen and at what frequency range? This will help me with redesigning the setup with particular choice of the materials.
The frequencies are found out of a huge background of airborne noise, it is like finding the noise out of noise in the data. The data I receive through our Feedback program is a 2D array with number of options at x and y axis. So I wrote a noise detection program in lab view V7.1 see slide 30, which takes this and identify any prominent frequencies. We use a quadrant photo diode as a detector, which gives three separate responses; x-signal, y-signal and sum signal of both. Sum signal is most appropriate to identify the week noise frequencies. And x-signal is more appropriate in finding the noise in relative magnitude in DB. The program finds the frequencies through FFT.
The test results are presented in the slide 13 to 24 and power-spectrum from 25 to 29. The results are generated through a DOG scan right across a 1μm stuck bead with water as buffer. Each time a scan is done with different frequency generated at the speaker starting from 0Hz when no frequency is generated (OFF). Similar test is also done when the speaker was not in physical contact with the optical table but was at the same distance from the trap.
First the slides 13 and 25: The data in these slides is taken when the speaker was off (0Hz). A DOG profile in slide 13 is given with a lower section zoomed in the inset: The oscillations can be counted to 12 over .1 second periods (this is 120Hz noise), but the oscillation profile looks of beating. Why is that (It means there might be more than 1 frequency)? In slide 25 the power-spectrum shows two prominent bands; around 120Hz, 180Hz and 295Hz. 120 Hz band is wider and much stronger than other two, maybe this is why it is more prominent in the beating. So it is sure now that there are more than 1 frequency bands exist in the data (similar is shown by the data collected when speaker was on the chair slide 25 right portion).
From slide 13 to 16 the DOG-profiles are presented over a frequency span of 150Hz. As we reach 120 Hz the profile gets thicker and gets thickest at 120 Hz (I tried to keep the amplitude same as I changed the frequency every time). I remember that I saw the bead vibrating in the camera at this frequency, which suggests that the present setup is most sensitive (vibration transmissibility and noise absorption wise) to this frequency. In slide 14 the 120Hz DOG is presented; the oscillation looks perfect (sine wave) no beating in that section. In slide 26 the power spectrum is presented; 120 Hz band can be clear seen with its higher-order bands of 240, 360, 480 and 600 Hz bands from left to right. The multiple bands are due to the constructive and destructive interference between the original and reflected waves. The waves are reflected at the ends of the optical table and interfere like the water waves in a pound are if generated at the center travel to the ends reflected back at the ends and interfere with the original. Since 120Hz is the only frequency band around, so all the generated bands are integer multiple of it with power descending with ascending order. The same thing also happens with other frequencies close to 120 Hz like 115 and 125Hz but it is strongest at 120Hz, so 120Hz is definitely more favorable, than other frequencies.
For next few frequencies the profile keeps going thinner until for 180 Hz in slide 17. This is another frequency we see in the 0Hz profile. In slide 27 the power-spectrum is presented in comparison with when speaker was not in physical contact. Profile on left (physical contact) is much stronger then on right (no-contact), which shows that this frequency is also definitely vibration transmissibility favorable. On right the 120 Hz frequency is dominating on 180Hz when the speaker is not in contact, but on left 180Hz is dominating when it is in contact.
After 180Hz 300Hz is next on to hit. In slide 28 power-spectrum shows that in-contact profile is much stronger than non-contact profile; suggests that this frequency is vibration transmissibility favorable. After 300Hz, no single frequency is dominating so there is effective interference between the frequencies. As can be seen in the slide 29; 600Hz is as strong as 120 Hz in both left and right (in-contact and no-contact).
The data is briefly analyzed in the slides 31 to 35. In the slide 32 a plot is presented with three strongest frequency Vs scan frequency. The blue line shows the strongest frequency band received. At 0Hz the first strongest frequency is 120Hz after that the received strongest frequency is the scan frequency up to 400Hz, after that the strongest frequency discontinued to be scan frequency and becomes 120Hz for all the higher scan frequencies. This suggests that the structure (setup) supports (less attenuation comparable to other higher frequencies) more to the vibration frequencies less than 400Hz. In slide 32 the plot shows the relative magnitude (DB) of 120Hz Vs scan frequency. At 0Hz the blue line starts with 15DB reaches 44DB at scan frequency of 120Hz, after that it remains below 15DB. This plot also shows the strongest frequency (120Hz) magnitude after 400Hz which still remains below 15DB; this suggests that there was not any power coupling to this frequency, otherwise it must be higher than 15DB. This suggests three things: The structure is more sensitive to the vibration-frequencies below 400Hz (the transmissibility is higher for the frequencies below 400Hz), power-coupling to 120Hz is void and there is not much airborne noise intervening above 400Hz. In slide 32 second and third strongest frequencies behave non-monotonically. Second strongest starts at 180Hz, between 105 and 150Hz it is the second harmonics of the scan frequency after 150 it drops down to 120Hz. After 400Hz it stays at 295Hz for more than once. Third strongest shows same behavior, it starts at 295hz becomes 3rd harmonics for frequencies from 100 to 130 Hz and drops down to 295 stays there for more than once. This suggests that the structure is definitely very-very sensitive to vibration-frequency band of 100-150Hz and sensitive to frequencies less than 400Hz. On average the 295Hz stay over the 0Hz magnitude for scan frequencies over 150Hz, which might be due to power coupling or noise absorption to this band (I am not sure).
Slides 33 and 34 presents when the speaker was not in physical contact with tweezers but placed at the same place and distance to avoid any transmissive-vibrations generated at the speaker. Slide 33 shows the first strongest(blue) frequency of 120Hz at 0Hz. From 105 to 140 the strongest is similar to the scan and then it drops to 120 and stays there for all the higher scan frequencies. Slide 34 presents the relative magnitude of 120Hz Vs scan frequency. At 0Hz it starts at 18.7DB and goes to 28DB at 120Hz scan frequency. After 140 Hz this plot presents the values for the first strongest frequency which still remain less than 18.7DB. This again suggests two things: First structure is more sensitive to the airborne noise frequencies less than 140Hz and there is no sufficient airborne noise power coupling (out of higher frequencies) to 120Hz. Second strongest frequency starts at 180Hz (-1.3DB) between 145 and 180Hz it is same as scan frequency, from 200Hz it stays on 295Hz which is on average of 1.5DB. This is above 0Hz DB which suggests that for the scan frequencies above 200Hz there might be power coupling to this frequency via airborne noise. Slide 35 compares the in-contact and no-contact data. All three profiles of no-contact (120,180 and 295Hz) are higher than the in-contact. This is due to the vibration dominance and interference created by the strong vibration scan frequency at any scan frequency. The resonance for all three frequencies can be clearly seen in the data.
The test completes with the following summary:
- Vibration transmissivity wise structure is most sensitive to 100 to 150Hz and more sensitive to the frequencies less than 400Hz.
- Power coupling from frequencies higher than 150Hz wise 295Hz is the strongest candidate.
- Airborne noise wise structure is most sensitive to frequencies less than 140Hz.
- 120 Hz is the mechanical resonant frequency of the structure.
- 180Hz and 295Hz are the secondary resonance.
The data files can be downloaded from the links:
Sep/01/2011: Search for the source and remedy
A painful task of searching for the source is over with some surprising and not so surprising results. In this section I will discuss my struggle with the search for the sources and better design for the setup. I must say I learned a lot about the airborne noise and mechanical vibrations; this gave me completely new insight into the matter which I used to design better setup and parts. I found that it is literally impossible to design a setup (structure) without an intrinsic resonance/resonances based on mechanical and structural properties. And it does not stop there; even different parts of a setup (structure) have their own resonance which they respond to. And little changes in the setup can shift it in frequency but can never eliminate it. So how do we solve this problem? There are three ways to deal with it in enclosed environment like in my lab. First find the sources of noise and remove them. Second if they cannot be removed than isolate the setup from them by building enclosures and isolators. Third design an appropriate setup or some parts of the setup (by using different structural-geometry and material) to drift the resonance spectrum away from the noise spectrum generated around the setup. In addition dispersers can also be very affective in breaking and absorbing the airborne noise and vibrations. I used all three methods to solve this problem.
Search for the sources
I have few sources in the lab, which I can divide into two categories: Active and passive. In active-source category I have two CPU (Steve-office and steve-daq2), AOM driver, x-piezo driver, z-piezo driver, Luca CCD camera and laser power-supply cooling fans. In passive I have the hood and may be the lab wall which separates us from the chase. I will start with the active sources. But before I start the process first.
So my goal is to analyze the noise power-spectrum generated by each source and find if it generates any of the frequencies I see in the data and what is the magnitude of the generated frequency. To achieve it I invented an Acoustic and Mechanical Noise Reader device (AMNR) with lab view V9 program. This device gives me a power-spectrum DB Vs frequency of the desired source. It can also help in designing the setup by helping me choose the right material and structural-geometry. In few words I can say that this device does all what I need to solve the problem.
I start with the two CPU, because they are the closest to the trap. First I scanned the Steve-office which performs the oscilloscope and live-feed functions. I scanted it at few different places as results show in the slide 2. BINGO!! In the graph 1 it shows the power-spectrum of Steve-office with 120Hz the major frequency generated with -10.78DB magnitude. The other frequency peaks are at 150, 225, 945 and 1170Hz. The data shown in three graphs were taken at three different places upon the CPU shows the same that 120Hz is the major frequency generated by this CPU with magnitude above -11DB. Other frequencies it generates are at 150, 225, 945 and 1170 Hz but there relative magnitude is pretty low below -21 DB. So is this CPU the source of 120Hz? May be!!!
Now I turn off the Steve-office and move over to Steve-daq2 slide 3. Steve-daq2 shows the peaks all over from 50 to 1250Hz. But the strongest generated frequencies are 120Hz at -12.9DB and 455Hz at -13.03 DB in graph 1. All three graphs are taken at different places upon CPU. Graph 2 shows the major frequencies at 210, 140 and from 340 to 460Hzs with magnitude above -15 DB. In graph 3 again 120 Hz is the major frequency with -3 DB magnitudes. Other frequencies are 140 and 200Hz around 10DB. It is definite that the two CPU are the major source of the noise especially 120Hz. If I do not find another source which generates 120Hz stronger then these, then these two are the source of 120Hz. My device has the capability that it can let me listen to the specific band. I listened to 120 Hz and it does not sound much even at higher magnitudes which mean that this frequency can be strongly present around the tweezers without bothering me much, because my ear response (human ear response) is not very good at this frequency makes it not very noticeable.
I do not have the AOM driver present, so I am skipping it to x-piezo driver. I had already checked the x and z piezo stages and there is no noise there. Slide 4 presents the x and z piezo drivers power-spectrum. It generates all kinds of frequencies from 50 to 3000Hz graph1. Frequencies above 1500Hz are useless because we use a low-pass filter at 1.5 KHz. The major frequency the driver generates is at 180 Hz at -8.97 DB another BINGO!!. The story of z-piezo driver is same graph2. It generates two major frequencies at 130 and 2155Hz. These frequencies might not have any direct impact because 2155 will be filter out and 130 Hz is at pretty low magnitude of -20 DB. All these noise can affect the trap in two ways; via airborne noise and mechanical vibrations propagating through wires. So removing the drivers and CPUs from the lab will help with airborne noise and using metal dampers on wires before they reach to the tweezers will help with vibrations.
Slide 5 present the power-spectrum for Luca CCD camera. It generates somewhat a frequency comb from 50 Hz to 1445 Hz. The major frequency it generates is at 390Hz at -2.4 DB, other are at 130, 260, 520, 650, 780, 915 and 1050 which are all multiples of 130Hz. This camera is the closest to the trap and it is bolted next to the microscope on the optical table directly. So it can affect the trap most through airborne and mechanical vibrations. Graph 2 cooling fan look very quite they generate a major frequency at 150Hz.
So out all the active sources it looks like the CPU are the major sources of 120Hz and x-piezo driver is the major source of 180Hz.
In search for the passive sources of noise I started with getting power-spectrum of airborne noise around the optical table. I just hanged the AMNR in open air by the tripod and collect the data. This data gives me an insight into the airborne noise spectrum in open air in the lab. I collected this data at three different places in the lab; around the optical table and near the hood. Slide 6 & 7 presents the airborne noise spectrum in the lab.
In all power-spectrum 20 and 50 Hz is common and around the same magnitude. In slide 6 midlab1 graph shows the power-spectrum of airborne noise in the middle of the lab. The major frequency it reports is 150Hz with a band 160Hz wide. This band is from 105 to 165Hz peak at 150Hz at -31 DB. Other major frequency present in the lab is at 270Hz this spectrum repeats itself every time with little difference. Midlab2 shows another power-spectrum I took near the computer desk. It reports the same band but peaking at 120Hz at -25.62 DB which makes sense since it is close to the desk (150Hz has the same magnitude of -31DB). I took another data set on the other side of the optical table near hood wall. This data set reports the same band peaking at 150Hz at -27.51 DB. Here the magnitude is little higher which probably means that the source of this noise is the hood.
Slide 7 presents the power-spectrum taken near the exhaust wall. It reports the same band peaking at 150Hz at -20 DB. The other peaks are at 225 and 270 Hz. To make it sure further I took the next data set inside the hood, which confirms that the hood is the source of 150Hz (-8.41 DB) band with other bands at 270 Hz (-21DB) and 325 Hz (-21 DB). I know that I did not see this specific frequency of 150Hz in the data but it is very possible that this band is providing some of the inputs to 120 and 180 Hz. So I cannot ignore it. It also produces a band at 270 and 325 which might be providing inputs to 300Hz.
The hood fan is off so the noise coming out of the hood exhaust is coming through the air tunnel from the AC unit hanging off the roof in the chase. We have a thin wall separating our lab from the chase, so it is also important to collect a power-spectrum of the wall to make sure that it is quiet enough not to work as airborne/mechanical noise source. Slide 8 presents the data. In first graph I took the data close to the hood and 150Hz (-27 DB) mechanical vibrations can be clearly seen with other peaks at 220 and 270Hz. Next data set I took away from the hood shows a wide disperse band of frequencies at relatively much higher DB peaked at 165Hz (-16DB). This band is a combination of different frequencies from 80 to 400Hz, Which suggest that the wall is definitely not quiet. It is propagating and dispersing some of the noise generated on the other side of the wall.
So I went outside in the chaseto prove it to myself that it has the source for noise in the lab. I did many airborne power-spectrum tests at different places there. The results are presented in the slide9. BOY!! This place is very noisy it has some huge laser power-supplies on the floor and a giant AC-unit hanging off the roof. The major generated frequencies I found are 150, 125,135, 270, 435 and 535 Hz.
A summary of the results is presented in the slide 10 and in the picture above. The major source of the 120Hz is the two CPU which might affect the trap via airborne and mechanical vibrations through wires. Luca-camera and z-piezo driver also generates a band around 120 and 300Hz so these two can also contribute in addition. The major source for 180 Hz is x-piezo driver which can affect the trap via airborne and mechanical vibrations through wires. I did not find any other source for 180Hz but the airborne noise at 150Hz through the hood might affect the trap and contribute to 120 and 180Hz. I did not find any source for 300Hz but airborne noise of 270Hz can definitely contribute to this. So to fix the noise problem I will have to fix the CPU, drivers and hood. I cannot do anything about the Chase which is the primary source for 150 and 270Hz. But I am happy that I have more knowledge of what is going on.
10 Oct/01/2011: Remedy: Design and Solution
Design & solution
So now how do I save the trap from the noise? I planned and executed a three-stage plan. First stage; determine the sensitive frequency/frequencies of the current setup and parts. Find the right material and redesign some of the parts of the trap-setup (the setup which holds and moves the sample around) to move the resonance/sensitivity away from the most prominent frequencies generated in the lab. Second stage; isolate microscope and trap-setup from the optical table and surrounding. Third stage; move all the active noise sources out, chock the hood-vent and build an enclosure around the tweezers OR change the lab. Changing the lab is also an option; we might shift our tweezers to other much quieter lab!!Maybe I should do that!!
New Redesigned Parts (stage 1)
I started with the trap-setup. Slide 12 shows the old set-up which consists of 2 motor stages, 1 piezo holder stage, 1 piezo stage and 1 sample holder stage. Slide 13 shows the new setup in which a redesigned piezo and sample holder stages are used. I redesigned the setup for two things; to attenuate any mechanical vibrations coming from below the piezo holder stage by choosing the plastic made piezo holder and sample holder stages and to have less absorption and propagation of airborne noise through sample holder stage (plate) by choosing plastic made advance designed sample holder stage.
New sample holder plate
The sample holder plate is very important part of the setup because it holds the sample directly at the trap center, so any mechanical vibrations in the plate can be directly seen in the data. It is also very sensitive to the airborne noise because it is bolted at one end and all its weigh is at other end, so any small pressure change in the surrounding makes it work like a tuning fork slide 12. So I keep this in the mind for the new design. Now I dealt with mechanical vibration in two ways; first I choose acrylic plastic over aluminum because it is softer and hence have higher attenuation for higher frequencies, second the design is such that it absorbs less mechanical vibrations. Slide 14 shows all three slides. Comparisons between the old and new plates are such: new plates are made of acrylic plastic which is softer and lighter than aluminum and hence higher attenuation at higher frequencies. New plates have less weight at the sample-end because of having chopped edges. New plates have less surface area so less exposure to airborne noise. There are two versions of new plates; regular and advance. Advance plate is advance in the sense of having less surface area and more reflection edges; less surface in contact due to the circle-gape with piezo stage so less material in contact for lower vibration absorption from below. Less surface area in middle due to the triangle-gape means less material to carry the vibrations (just like a church bell; the heavier the bell the longer it resonates) and less surface area to absorb the airborne noise. This also decreases the weight further. Geometry is very important because inside the material-medium noise propagates and reflects back at the edges and interferes. So by manipulating the geometry we can manipulate the interference and hence the overall amplitude at particular band of frequencies. I choose the circle, triangle and chopped edges so I can encourage the reflection and interference inside the plate. I used Google sketch up for the drawing slide 14. Will this work I test next.
The test setup is shown in the slide 16. The basic idea in the test is to generate and read the vibrations at a distance. I used a 5.5 inch audio speaker to generate the vibrations. I detached the diaphragm from the frame to suppress the airborne noise generation (and it actually worked pretty well) and used a tone generator program written in lab view v9 for frequency generation. I read the vibrations through AMNR sensor. To avoid any airborne noise contamination through speaker I detached the diaphragm and kept the speaker power such that it generates just enough vibration-amplitude to do the test. I also used an acoustic-foam curtain between the speaker and the sensor. I used a frequency range of 100Hz to 800Hz because it is best suited for the subwoofer speaker I used. I designed this test specifically for mechanical noise (vibrations) transmissibility through the plates. All this setup is built on a black aluminum plate which is isolated from optical table through rubber padding. To get a transmissibility (%) Vs frequency graph I start at 100Hz and moved on to 800Hz in small increments. To estimate the transmissibility I produced unknown amplitude at the speaker at particular frequency, measure the magnitude through sensor in DB at the place where the plate is to be bolted, then bolt the plate and measure the magnitude at the middle of the plate again at that frequency without changing anything. This gives me an input magnitude in DB below the plate and an output magnitude in DB at the desired spot on the plate. I choose the middle of the plate because that is the source-point of noise for the sample slide 18 & 19.
Lab View does a FFT of the input signal and gives a power-spectrum magnitude DB Vs frequency. For reference it uses 0 DB for amplitude of 1 Volt of input signal. Which means if an input signal has 1 Volt of amplitude, Lab View will refer it as 0 DB.
In transmissibility Vs frequency on x-axis, we usually use frequency-Lorentzian which has a ratio of input frequency over natural frequency (resonance) of the structure. Since I do not know the natural resonance of the structure and I do not bother because it does not affect the results, I just use the input frequency at the x-axis. Since vibrations are longitudinal waves so this whole argument can be done in terms of forces but that would not change the overall behavior of the plates to the generated vibrations (noise), so I keep it simple. But this test and all the other tests are half a picture because they are related to mechanical noise only. I did not have enough time and resources to test them on airborne noise basis (to do that I would need a soundproof enclosure an upgraded sensor and better sound generator with appropriate sound absorber units).
Since the test is conducted on and through black aluminum platform so before starting the test I run a small test to know the behavior of this plate. Slide 16 & 17 shows the setup and result of the test. I used a frequency from 100 to 800 Hz at 50 Hz increment; result suggests that the system (plate + speaker) has highest transmissibility at 450Hz. So for the plate at this configuration, 450Hs is the most sensitive transmissible frequency. In the entire test I do not bother to mathematically verify this, because in principle the experimental analysis is enough to design a plate and verify a design. The test setup and the results for the plates are shown in the slide 18, 19 & 20. Aluminum plate: Slide 20 shows the result, aluminum plate graph is shown with triangle spots. Graph shows the response of the plate frequency Vs transmissibility. The plate has maximum transmissibility of 100% at 300Hz and 8% at 700Hz. On average it transmits 40% of the input magnitude. Most sensitive frequencies are 300, 350 and 650Hz. Blue curve shows the response of the new plastic plate regular. It is much better than the aluminum plate with average transmissibility of 18%. The most sensitive frequencies are 200 (47%) and 700 Hz (25%). It is pretty calm in between 250 and 650Hz. The red curve shows the response of the new plastic plate advance. It has the best response in three with 12% transmissibility which is 3 times as better as aluminum and 50% as better as new plastic regular. The most sensitive frequency is 100Hz at 77%. This plate has pretty low transmissibility of 7% above 100HZ. My goal was to design a plate with very low transmissibility for the frequencies generated in the lab. Lab generates the frequencies from 120 to 900 Hz and this plate has an average transmissibility of 10% over this range which is pretty impressive. So I am going to use new acrylic plate advance replacing aluminum plate.
New piezo holder stage
Now the next task is to design the piezo holder stage. Piezo holder stage is another important part of the setup; it attaches the x-y linear stages to the x-piezo stage. So if there are any vibrations in the linear stages they get transferred to x-piezo stage through this. By redesign it some of the vibrations can be dumped and attenuated, even though it is hard to achieve. I designed two version of the stage; regular and advance slide21. Both stages are the same size, but little thicker. Stages are made of polypropylene plastic. Regular stage is like aluminum stage with two squares carved out on each side; this is to reduce the surface of contact. And also less material is available for noise propagation. This stage directly sits on the linear stage with piezo stage on the top. Advance stage has the same size with two circular cups carved out in the middle. These cups will hold two isolators which will isolate the stage form holding screws. In the bottom I carved out four circular gaps to fit rubber cushions to isolate the stage from the bottom completely. This stage sits on four rubber cushions with screw isolators in hope of less vibration absorption and propagation. I used Google sketch to design the stages slides 22 & 23.
The test setup is shown in the slide 24, 25& 26; it is same as previous setup with acoustic foam curtain in the middle to cancel any air borne noise contamination from the speaker. Results are somewhat surprising; regular plastic stage and aluminum stage response very same, advance plastic stage seems noisier. Aluminum stage has the lowest transmissibility of 18.6% in comparison to regular plastic stage at 22% and advance plastic stage at 24%. Aluminum stage has the most sensitive frequency of 1000Hz at 60%, between 100 and below 1000Hz it has an average transmissibility of 12%. Regular plastic stage has most sensitive frequency of 950Hz at 55%, between 100Hz and below 950 it has an average transmissibility of 21%. The worse is advance plastic stage which has the most sensitive frequency of 150Hz at 93% and of 950Hz at 55%. So the result shows that aluminum plate is most appropriate but why is that? This is due to the strange behavior-combination of noise and structure. Aluminum and piezo stage above it, are full in contact with each other and with the surface below so the vibrations couple into the aluminum stage, transfer through piezo stage, and reflect back from the top surface of it and couple back into black aluminum platform. So in this structure all the vibrations are dumped back to the aluminum platform. The same happens with regular plastic plate and that’s why these two structures are very close to each other in response. Advance aluminum plate is on rubber cushions which is good when no screws are attached because in this case the vibration propagation will depend on the rubber natural response to frequency. This changes when the stage is tightly clamped to the platform in two ways; the force upon the rubber cushions due to clamping deviates their natural frequency and increases overall transmissibility at the tested frequency range, second the vibrations start propagating through the screws. These vibration transfer through the plastic stage into the piezo stage and reflect back, but could not couple back effectively through the screws and cushions to the aluminum platform. So vibrations enter into the structure couldn’t be dumped back effectively, so this is why this stage is nosiest in the test. This is valuable information to keep in mind when design the isolators/dampers. I used this information in second part when I design a steadier dichroic holder platform and an isolator platform for the microscope, will discuss next.
Isolation of parts by redesigning (stage 2)
Microscope breadboard platform
Trap setup is directly built on and around microscope, and both are directly bolted on the optical table. So isolating microscope and trap-setup from the optical table may be helpful in reducing mechanical vibrations in the setup. The microscope and trap-setup will be directly built on this platform. I use 18”X18” aluminum breadboard platform shown in the slide 28. To reduce vibration-propagation through screws a cup on each side is milled out. The cups will hold a rubber-isolator which will hold a screw clamped on the optical table. I have two versions to hold the breadboard on the optical table; shown in slide 29; top view shows that the breadboard can be hold down by screws or by plates on the sides. The plates are isolated from the breadboard by two rubber padding. Later will have less noise propagation then former because when clamped down,the screws may propagate more noise (this is just what I saw in the piezo holder stage). Slide 30 shows the bottom view; the breadboard does not directly stay on the optical table, it is put on four rubber feet of 1.4 cm diameter. Since the vibrations are strongest in the corners the feet are 3” away from either side into the plate. I used two different feet types; regular and holy. Holy feet have a circular gap of 8mm diameter, so 33% less surface and material as shown in the slide 30. I designed the platform such that it has no direct contact with optical table; all the contacting surfaces are rubber with least surface and material. This is done, keeping in mind the results of advance piezo holder stage. So I have two setups to test; breadboard with screws and regular feet Vs breadboard with plates and holy feet.
The test setup is shown in the slide 31; test is conducted on a small optical table. Speaker is kept directly on the optical table wrapped with acoustic foam enclosed in a box. Sensor is put in the corner where the vibrations are strongest. I first recorded the vibration magnitude for the table for frequency range of 100 to 900Hz. Then repeat the same test for breadboard with screws and regular feet and breadboard with plates and holy feet slide 32 and 33. The result is presented in the slide 34; red curve shows the response of the table at that particular place. Green curve is the response of the breadboard which is lower than the red curve. The difference between the two setup is minimal and later is little better but both the setup are far better than the floor. Average magnitude of noise vibration on floor is -27, on breadboard with screw is -39 and on breadboard with plate is -42.
The slide 35 shows the transmissibility curves for the two setups. There is not much difference between the two transmissibilities but later is better between 100 and 500Hz. The most sensitive frequency for holy feet setup is 500Hz. The average transmissibility of the breadboard with screw is 29% and breadboard with plate is 22%. So breadboard with plate and holy feet is definitely better then the two (breadboard with screw and regular feet and floor). But there is not much difference between the two breadboard setups, so I think I will choose the one easier. But remember, the behavior is not going to be the same once I put the microscope and the trap setup on it. The weight and the structure of the microscope and trap set up will change the behavior of the breadboard itself, so this test should be repeated once everything is built on the breadboard to know the final behavioral difference.
Dichroic holder stage platform
In this section I discuss the problems with dichroic holder stage which came with microscope and its new design. Dichroic holder stage is the platform which holds the dichroic below the objective to bounce the laser into the objective. This stage is clamped into a holder which is clamped directly on the side of the microscope as shown in the slide36; I am calling it regular stage. There are two problems with this stage (design); it is clamped on its side, so it is hanging by its side with all its weight which makes it very venerable for low frequency mechanical and airborne vibrations (<150Hz) and this stage is also directly in contact with microscope which is in direct contact with trap set up and other structure, so any vibration in any of these can directly make their way to dichroic and eventually show up in the data. So I need to design something steadier (very well damped), isolated from the rest of the structure (microscope, trap setup and other structure), less responsive to airborne noise and mechanical vibrations at the cost of risking mechanical vibration propagation directly from optical table. Slide36, 37 & 38 shows the regular stage at three different angles.
Slide 39-45 show the designed dichroic holder stage platform. I kept it simple; I used 2 four by four inches breadboard plates attached together with another aluminum plate from below. The dichroic stage sits directly on the breadboard plate attached to very well machined holder slide41 ( I designed this holder to accommodate the fitting of the stage in x and y directions which are transverse to the optical axis; for alignment point of view it is very important that dichroic sits directly under the objective). The dichroic holder stage is designed so the dichroic can be moved in and out of the laser light; I wanted to keep this feature. So in present design dichroic can still move on the plate while keeping itself in full contact with the plate. This design also offers a lens holder (in which a lens can move in 3D) directly in front of the dichroic slide42. The lens has to be twice its focal length away from the back aperture of the objective, right in front of the dichroic. The lens holder design offers an exact solution of this problem. In the design a 2d lens tube holder is directly clamped on the breadboard slide43; this 2d holder will hold a lens tube. This 2d holder is attached to breadboard by 2 machined clamps slide44. I designed the clamps to accommodate the holder fitting in x and z direction (where z is the direction of laser light). The whole set up will sit on 3 aluminum posts which are directly bolted on to the optical table slide45. But this design has a problem; it has too much material. I call it bell-effect; the heavier the bell the longer it resonates. The dichroic holder stage platform sits on the aluminum posts and there is no easy way I can isolate it completely from the optical table without risking its stability which is absolutely important. So if vibrations reach to the platform, then it is going to resonate for longer time because of having more material. So I tried to solve this problem by redesign it slide46-49. I did few things; milled out a square and a rectangle from the breadboard plates and attached the two plates internally. In this design the platform is lighter, fewer vibrations would travel from one plate to another because of having fewer mediums, and more reflection surfaces would induce more reflection and hence higher interference. Fewer surface would be affected less by airborne noise. I am calling this dichroic stage holder platform an advance stage.
To test the new design I performed the similar experiments (setup is shown in slide 38 and 50). I tested four different combinations (I had to test the microscope breadboard platform performance again in presence of the microscope and other structure, so I combined it here); Regular stage and microscope is on the optical table, advance stage and microscope is on the optical table, regular stage and microscope is on the breadboard (breadboard is on the optical table on 4 holyfeet with plates on the sides; a setup just like in test slide 33) and advance stage and microscope is on the breadboard. I conclude the results in slide51 to 55.
Slide51 shows a comparison between the results achieved when regular stage was attached to the microscope and microscope was in direct contact with optical table Vs microscope on the breadboard. Regular stage on breadboard was little better between 140 and 550Hz, but at 140 it was noisier than the optical table and after 550 it is again noisier then the optical table. Slide 52 presents a comparison between advance stage and microscope are on the optical table Vs advance stage and microscope on the breadboard. Result shows that the setup on breadboard is much noisier than the optical table. In next two slides I test and compare the stages separately on table and breadboard. Slide53 shows that regular stage is less noisy. Regular stage has peak at 140Hz advance stage has peak at 180Hz. In rest regular stage is less noisy but very close to advance stage. Slide54 shows the same test but on breadboard. The result is rather surprising; advance stage has more vibrations propagating than the regular stage on every frequency with a peak on 400. Regular stage has a peak on 140Hz as usual. Slide55 shows all four results; advance stage on breadboard is the noisiest, advance stage on optical table is after it, regular stage on optical table is better than both advance stage setup but worse than regular stage on the breadboard, so regular stage on breadboard is the best in these tests. But unfortunately this particular setup does not qualify on the bases of stability and dichroic isolation. Breadboard platform gives me a good isolation from optical table but it might not be very stable. Rubber feet and plates give good isolation but reduce stability of the structure over time, which might be a problem to tweezers alignment. Regular stage on other hand is directly attached to microscope and other structure so any vibrations in the structure can directly reach to the dichroic and I do not want that to happen. So beside their outstanding performance I am discarding the microscope breadboard platform and regular stage. I am going to use advance stage and microscope directly on the optical table setup, because it provides me better stability. But I will definitely keep that option in mind, if in future this does not work well then I will definitely give a chance to breadboard.
Next is to design an appropriate enclosure for the microscope and trap setup, but before this I have to align the tweezers and start taking data.