Below are the raw results for two of the largest ballheads I have at my disposal, the RRS BH-55 and the Feisol CB-70D. I don’t normally post the raw results, since, they aren’t that interesting, but in this case they are critical to showing why this test hasn’t worked well.

The leftmost column shows ten trials of measuring the resonance frequency of the tripod and test rig without the head. The other two columns show five trials for each of the two heads I tested. Note how little change there is. The standard deviations of each data set would overlap with the means of the other data sets. In theory we could just build up enough statistics to get a reasonable standard error on the measurement, but in practice this incredibly time consuming and highly dependent on the assumption of Gaussian statistics.

Compared to the camera and tripod MOIs that we have measured so far, the MOI of the heads measured here are insignificant. Despite being heavy, the heads are compact with their mass located close to the axis of rotation. For example, lets look at the CB-70D and calculate its MOI by approximating it as a solid cylinder. The MOI for such an object is given by:

From the CB-70D review page, we can see that the mass is 0.971 kg and the radius as measured from the base diameter is 4.2 cm. That gives us an MOI of 0.00086 kgm^2. This is in the ballpark of what we measured above. Given the error in our measurement and the crudeness of this approximation, these results are in enough of agreement to form a reasonable sanity check. Because I am barely able to measure the MOI for the largest heads I have, any of the smaller ones are going to read effectively zero. Going forward, if we need to use the MOI of a head for any reason, we will simply make a rough approximate calculation.

I also attempted to measure the pitch MOI, with similarly poor results. But lets talk about pitch MOI briefly. For yaw, the axis of rotation extends up through the middle of the tripod and so the head will always be centered on that axis. For pitch, the axis of rotation is horizontal and extends roughly through the point in space where the imaginary lines representing the center of each leg intersect. Typically this is somewhere in the middle of the head, depending on the width of the top plate and apex. For tripods with a center column though, the MOI of the head and camera in the pitch direction will increase dramatically as the center column raises them up and away from the center of rotation. There is much more to investigate on this, and will be a theme in our future discussions on center columns.

**Takeaways**

- Ballhead MOI in the yaw direction is basically negligible.
- Ballhead MOI in the pitch direction depends mostly on how far above the center of rotation the center column takes it.

Next up: Camera and Lens MOI

]]>Moment of inertia is the angular equivalent of weight, or mass. The movements of a tripod are inherently rotational, so it is important that we analyze the damping problem in the correct way. Similarly to how it is more difficult to slow down a heavier object, it is more difficult to damp out a rotational vibration in a system in that has a higher MOI. Unlike mass, MOI is different based on the axis that the object is rotating. Here, I am only measuring yaw MOI as it is the only axis about which I can have confidence in my measurements. Also, because the camera is typically placed high above the center of rotation, the dynamics are very different and the weight of the camera will be much more important. More on this later. For now, we are only talking about yaw vibrations.

As it takes me roughly 20 minutes to test the MOI of a single tripod, I was only able to test a small number, which I chose to be a somewhat representative sample across different weights. The results are below:

Its no great surprise that heavier tripods have higher moments. The ratio column is simply the Inertia/Weight. We can immediately see that there isn’t a direct linear relationship between weight and Inertia. This is simply because heavier tripods tend to also be taller, and the taller the tripod is, the further the legs splay out, resulting in more MOI. We can also see this as a result of leg angle. The TFC-14 has a greater MOI despite weighing less than the very similarly constructed LS-284C. The narrow leg angle on the 284C reduces its MOI (but also its stiffness). When looking at the ratio, there are two significant outliers, the 3 Legged Thing Leo and the Feisol CT3472. These tripods are quite heavy for their height, and very light for the height, respectively. Again, a short tripod’s legs don’t extend out as far from the center of rotation, and for MOI, radius from the center tends to matter more than weight.

For context, I measured the MOI of the Fuji GFX 50S and 45mm lens at about 0.004 Kg*m^2. So the MOI of the tripod will almost always exceed that of most normal sized camera and lens combinations. This will invalidate our damping time calculations, as those assume that most of the MOI comes from the camera and ball head placed atop the tripod. If the damping is occurring within the legs, then our calculations should still work. However, many tripods get a significant portion of their damping from the rubber pad on the top plate. This rubber pad will do nothing to damp the energy contained in the legs themselves. Its efficacy will be grow as the MOI on the tripod increases.

Because the tripod MOI will be so important for damping, we will want some way to estimate the tripod MOI from things easier to measure, such as the height and weight. To do this I am going to fit a couple quantities to the formula:

First, lets use weight as our tripod metric, and we get:

Here, I have also plotted a straight up linear fit (exponent = 1) and we can see that it doesn’t describe the data well at all. The exponential fit is reasonable, but we can still see the significant outliers of the 3LT Leo below the red exponential fit line and the Feisol CT3472 above the line. An exponent around two is consistent with what we expect. If the tripods were identical aside from their weight, we would expect the exponent to be straight up linear. But as we said before, weight correlates with height, and height has a much stronger relationship to MOI. So next, here is MOI vs Height:

We see a much larger exponent, but now we aren’t taking into account weight. Also, the range in heights is pretty low compared to the range in weights, making this fit much more questionable. This plot is less useful.

In theory, we could try a fit placing an exponent on both height and weight, but frankly, we don’t have enough data to get meaningful results out of such an approach. Instead (after some experimentation) lets fit the exponential function to MOI as a function of the tripod’s weight*height^3.

Not too bad! I have thrown the linear fit back in here to show that it is functionally the same. This demonstrates that we are likely using the correct exponents on our height and weight factors. So we have reduced this to a single parameter fit, which is the slope of the line. So, moving forward, we will use the following formula for tripod MOI:

where the height and weight are given in meters and kilograms. I’m sure we could get more accuracy by bringing leg angle into this, but that won’t be necessary. This is plenty good enough for our purposes.

**Takeaways:**

- Tripod MOI is larger than expected, and will be important to incorporate into our damping calculations
- Tripod MOI can reasonably be calculated as a function of height and weight.

**Next up:**

- Head MOI
- Camera / Lens MOI
- Damping times

For my stiffness tests, I use a large enough angular mass to do the test such that the MOI of the tripod is negligible. In real world conditions though, the MOI of the tripod is important and may very well be larger than that of the camera. To measure the tripod’s MOI, I used the typical equation for the resonance frequency of a system and broke out the MOI term into and .

Now by varying and observing the resulting change in we can fit the data to the function above and extract the MOI of the tripod. This method is highly dependent on having a very accurate estimate of the MOI of the test mass, which in our case here is questionable. Using simple MOI formulas for basic objects, I suspect that we can get within 1% or so. Again, this is plenty accurate for our stiffness tests as it would throw off the test by that percentage. Here though, the error would be added to our result, which can throw it off wildly.

To test this method, I picked out one of the smallest, lightest tripods I have at my disposal, the Leofoto LS-224C. If any tripod should display a MOI close to zero, it is this one. I used a much lighter MOI for the test mass than I typically do to maximize the effect that the tripod’s MOI should have on the data.

The raw results are shown in the table to the right. The test MOI is shown in the leftmost column, then the resulting measured frequency. The implied stiffness is the yaw stiffness of the tripod calculated from the MOI and frequency of that row. I typically use an MOI of about 0.4 kg*m^2 for tripod stiffness testing, and at that mass, we measured a yaw stiffness for the LS-224C of 230 Nm/rad. As we reduce the MOI of the test mass, we see the implied stiffness drop indicating that the MOI of the tripod is becoming relevant.

To actually calculate the tripod MOI, I fitted the data to the resonant frequency equation and allowed python’s curve fitting function to calculate the tripod stiffness and the tripod’s MOI. The results are displayed below:

First off, the fitted tripod yaw stiffness is at 230 Nm/rad, exactly what we had calculated previously with the larger MOI test mass. This is a great sanity check on our methods. Second, the tripod MOI is fairly small at roughly 0.003 kg*m^2. This is a fairly plausible value. It is small but positive, exactly what we expect to see for this tripod. It means that we are successfully calculating the test moment to less than this value in error. We should not take the 0.003 kg*m^2 value for the tripod as particularly accurate. It is probably within about 50%. What it does mean though, is that we can take our future measurements for the MOI of larger tripods to be reasonably accurate.

For reference, in the previous post I measured the GFX 50S with the smallest lens at 0.004 kg*m^2, so in this case the tripod MOI is less than that. Given how small the LS-224C is, it is reasonable to expect most tripods to have significantly higher MOI, and thus dominate the total MOI of the system (camera + head + legs). For the purposes of damping calculations, it will thus be imperative to get reasonable measurements of tripod MOI. It remains to be seen whether or not I need to calculate the MOI for each tripod during testing. I certainly hope not as this is a rather time consuming process.

**Takeaways**

- Fitting the test MOI vs frequency data yields reasonable results, and will result in reasonably accurate MOI measurements for larger tripods.
- The tripod’s MOI is large enough to be a significant factor in calculating damping times.

**Questions**

- So what do the MOI’s for other tripods look like?
- What about MOI in the pitch direction?

]]>

Calculating the moment of inertia of an object is straightforward. Take every bit of the object, weigh it, and multiply by the distance from the center of rotation (radius) squared. Or, in math terms, integrate the mass times radius squared. This is great and all, but we don’t know what the mass distribution is for cameras and lenses, so we will have to measure it instead, bringing us here.

We know that tripod are near perfect harmonic oscillators, and thus adhere to the following equation regarding their oscillation frequency:

where is the oscillation frequency, is the stiffness of the tripod, and is the moment of inertia. So if we add additional weight in the form of a camera to a known system, we would get:

The stiffness remains the same, so we can combine these equations to get:

So, if we know the initial moment of inertia of the system (which we do), we can calculate the inertia of the camera and lens simply by measuring the frequency of oscillation before and after adding the camera.

The accuracy of this measurement will be mostly dependent on lowering the amount of noise and variance in our frequency measurements. We want to minimize the amount of initial inertia placed on the tripod such that adding the camera will result in the largest possible difference in frequencies. However, we want to keep enough initial inertia to make the oscillation last a long time. We therefore also want to use a tripod with as little damping as possible. Fortunately, the perfect tripod for this recently came across my test bench in the form of the Jobu Design Killarney.

Typically I measure the frequency of a tripod by fitting a sinusoidal function to the time domain oscillation. Coupled with an exponential decay, this is the typical method I use for measuring stiffness and damping. For measuring the frequency only, I wanted to compare this with measurement directly in the frequency, or Fourier domain. Here is an example of the tripod oscillation in the time domain:

And now in the frequency domain:

A perfect oscillator in the time domain looks like a delta function in the frequency domain. In practice, no oscillator is perfect and there will always be some width to the frequency distribution. Here though, we only really care about fitting the center peak. The Lorentzian function was a slightly better fit to the data than a simple Gaussian, but they both fit to the same center frequency of 11.26 hz. This is functionally equivalent to the 11.27 hz fitted in the time domain. In practice I found neither method to result in particularly more accurate measurements than the other.

Now as an example, here is what happens when I place a large lens, the Pentax 67 400mm F/4, onto the tripod:

The frequency of oscillation has dropped considerably, down to 10.31 hz. Even though this is the largest, heaviest lens I have, its moment of inertia is still clearly small compared to that of the test bar, despite the lens weighing more. This is directly a result of the deliberate weight distribution on the test bar. The weights are placed at the ends, 33 cm away from center, while the weight of the lens is mostly at the center of rotation. The r^2 term in the moment of inertia tends to be more important here than the actual mass.

I briefly measured the inertia for several lenses mounted on the GFX 50S, and the results with the implied damping time on the Jobu are shown below.

Even with the terrible damping on the Jobu, the damping times are reasonable for the smaller lenses. This requires a more in depth analysis, but likely means that damping is not going to be that important except for when using large telephoto lenses. In the future we are going to want to invert this analysis and ask “how much stiffness and damping are necessary to achieve reasonable damping times?”

In the GFX with 45mm lens is a fairly normal size for a camera and the inertia is a factor of a 100 smaller than what we typically use for testing tripods. It is going to be somewhat challenging to measure even smaller camera and lens combinations. We may have to reduce the test inertia further.

For stiffness testing, we use enough inertial mass in the test setup that the inertial mass of the tripod itself doesn’t matter. When using a small camera though, the inertia of the tripod should be in the same range as the camera, so we will have to take that into account when calculating damping times. This can also result in more complicated behavior that we are modeling here, as the damping elements within the tripod are not placed to damp out the energy in the tripod itself. Fortunately for us, when the load on the tripod is that small, the damping times are likely going to also be so small as that this more complicated behavior will be irrelevant to a practical guide on tripod use.

Stay tuned for a more comprehensive set of inertia testing.

**Takeaways**

- We can measure the inertia of camera and lens combinations by looking at the change in resonance frequency of a tripod after placing the camera upon it.
- For most normal sized camera and lens combinations, the inertia is quite small, and therefore damping times will be as well.

**Questions**

- How much inertia does the tripod itself exhibit?
- What kind of inertia can we expect for the most typical camera and lens combinations?
- What kind of damping times can we expect from typical systems?
- When is damping most important?

Our focus on stiffness in understand the effects of wind is well founded. Wind varies on a much longer time scale than that of the resonance frequency of the camera-tripod system. This means that damping will do nothing to help keep the camera still and the image sharp during longer exposure times. It will of course still help the camera return to a neutral position after a push from the wind more quickly. This is helpful and can reduce the total amount of motion blur in the image, but cannot eliminate it in the way that stiffness can. This dynamic motion is interesting and will be the subject of future research. For now though, I want to start simply and develop a framework for much more basic problems in damping.

Many informal tripod tests floating around the internet revolve around the settling time for a tripod after an impulse is applied. The test will usually be conducted as such: Place a large lens on the tripod and then give it a little push. Then measure the time before the vibration in the viewfinder settles. This method does correlate with tripod performance, but not in a way that clearly distinguishes between the effects of stiffness and damping or the value of each. It is neither repeatable nor well defined in what it means for the image in the viewfinder to be ‘still’. Here, we are going to formalize this style test and relate it back to real world tripod use.

When handling the controls on a tripod mounted camera, we exert more torque onto the camera than it will typically receive from wind. We cannot reasonably expect the tripod to hold the camera still to within the pixel level under such forces. This is mostly true for telephoto lenses. We all know it is possible to obtain sharp images with a physical shutter press for shorter focal lengths and shutter times. We have also all had to wait for the camera to settle after handling the controls with a longer lens. It is this settling time that we want to ensure is not a hindrance towards workflow. Exactly how much settling time is acceptable will of course vary between photographers, but we will need to accept some standard amount of time. I will propose 1 second, as cameras typically have a 2 second timer and we need to leave some small amount of buffer time for the finger to come off of the shutter after pressing it. If a different amount of time seems more reasonable to you, please say so in the comments. I will adjust the standard to consensus.

Drawing from the test methodology post, after an momentary impulse of torque such as the touching of the camera, the angular position of the camera will be given by:

where

and is the initial amplitude of the vibration.

Because we are only focused on the decay of the vibration, we will throw out the cosine oscillation term and are left with:

We are going to assume that the initial displacement (or initial amplitude of vibration) is caused by a torque that is slow compared to the resonance frequency. A good example of this would be touching or nudging the camera. A bad example would be the mechanical shutter, which moves very quickly. So, we get:

where is the torque from the impulse and is the stiffness of the tripod. Following our previous standard for pixel level sharpness, we want the amplitude of the vibration to reduce below:

where is the focal length of the lens and is the pixel width. Putting everything together we get:

where is the damping constant and is the moment of inertia. Solving for we find that the time required for the tripod to damp vibrations down below the pixel level is:

This is far less neat than our calculations of required stiffness. We find that the amount of damping time depends on a lot of factors. Even if you ignore the pixel level requirement and just look at how long it takes for the viewfinder to get still, the damping time is still dependent on the tripod stiffness, torque, and inertia on the load on the tripod.

This formula can be interpreted more intuitively in graphical form:

In red we have the initial displacement caused by whatever torque happens to be acting on the camera. In this case it may be operating the camera’s controls or pressing the shutter button. The green line shows when the camera has settled to a point where any motion is too small to cause blurring in the image. The difference between these two lines is what is given inside the natural log term in the damping time formula above, . The stiffer the tripod used, the initial displacement (red line) will be smaller. With a sufficiently stiff tripod, the red line can drop below the green line and therefore zero damping will actually be required. This is also the scenario we require when we want the camera to remain still, for example, during a long exposure with light winds.

The rate at which the vibration is damping out is shown by the black arrow and given by the ratio of the moment of inertia to the damping constant . The higher the damping constant, the fast the damping occurs, and the more inertial mass placed on the tripod, the slower it occurs. This has an interesting consequence. Minimizing inertial mass is just as important as increasing the damping of the tripod when it comes to settling time. Hold that thought for a future post.

We have slowly been assembling all of the data to solve this damping time equation. Pixel pitch and focal length are just pulled from the camera’s specifications. Tripod stiffness is well understood and measured at this point, and we have some data regarding how much torque is placed on the system. What we don’t have is any measure of the moment of inertia of the cameras and lenses we are asking the tripod to hold. This will be our final necessary piece of information. So, when you see a post in the near future measuring the inertial mass of various equipment, don’t think that this blog has gone completely off the rails into the most boring possible subject. There is a good reason we need to know.

**Takeaways:**

- The amount of time it takes for a tripod to damp out vibrations from an impulse can be calculated in a straightforward manner.
- This time depends on a combination of many factors such as inertial load, tripod stiffness, and focal length in addition to the amount of damping in the tripod.
- Therefore, damping time is a poor way to measure tripod performance in a general way.

**Questions:**

- What are the moments of inertia for typical camera and lens combinations?
- Can we minimize the inertial mass of our existing equipment for better performance?
- What are the calculated damping times for typical shooting scenarios?
- Do these calculated times match up with reality?
- How much tripod damping is necessary?

This rubber cap fits somewhat loosely. So to make sure that I was getting the best possible performance out of the tripod, I removed the rubber caps and tested the tripod with the spikes on the concrete floor. The results was a shockingly low amount of damping in the tripod.

Here is the damping and oscillation curve for the tripod with the rubber caps:

The damping here is poor, but not outside what I have seen in other tripods. Now with the spiked feet:

Note the difference in time scale. The damping is about half, all the way down at 0.082 Js/rad. This is beyond the point of being bad and actually becomes really cool. Nearly half a minute later after the initial excitation, it is still vibrating. Often, designers want as little energy loss in an oscillating system as possible, such as when building a clock. You want the clock to run for a long time without needing to be wound up again. For tripods though, we want the opposite, for those vibrations to damp as quickly as possible.

By removing the rubber feet, we have removed a major source of energy loss from the system. The small amount of rubbing between the spike and the rubber, and the rubber and the floor, was accounting for about half of the damping in the system. The sharp point of the spike directly on the hard concrete floor of my garage removed all of this friction and thus damping. Note thought that the stiffness of the tripod increased slightly from 1308 to 1353 Nm/rad. This increase in stiffness is likely not worth the massive loss in damping observed though.

I don’t mean to imply here that spikes are worse than rubber. The story would probably be very different on a softer surface where they spike could actually embed itself somewhat into the ground. Testing this further is on my to do list, but currently my test setup is confined to the lab, so it won’t happen soon. We can probably safely conclude that spikes are not the best solution for use on firm surfaces.

This also tells us that the feet of a tripod can make a big difference in the damping performance. I didn’t expect this as I have typically seen only marginal differences in the performance of the system based on foot type. Here though, the Jobu Killarney has so little inherent damping to begin with that the damping effect of the feet is amplified. This piques my curiosity as to whether a foot with exceptional damping performance is possible.

**Takeaways:**

- Apparently the choice of foot on a tripod can have a big impact on its damping. This could be an interesting avenue for further research.
- Probably don’t use spikes on hard floors.

**Questions:**

- How much does damping even matter anyways? Working on it, stay tuned
- What are the best feet for different types of surfaces? Concrete? Dirt? Grass? Sand?
- Is it possible to make a foot that adds a lot of damping without any loss in stiffness?

For this test, I placed the Fuji GFX and 120mm GF lens on a couple different tripods. The first was a very stable RRS TVC-33 and BH-55 ballhead. This is the standard for what people generally consider to be a very high quality tripod setup. The second was a Feisol CT-3342 Tournament and Sirui FD-01 pan-tilt head. While no slouch, this setup is notably weaker than the RRS one. I am not listing exact stiffness here because the connection between the RRS L-bracket and the camera appeared to be a significant, and unknown contributor to the stiffness of the system as a whole, and the point I am presently trying to make isn’t exactly dependent on overall tripod stiffness.

Previously, I measured the amount of torque placed on the camera when I was pressing the shutter button. I observed large enough torques to indicate that only the very stiffest tripods ( much more so than the RRS-TVC33 ) could hold the camera steady enough for razor sharp pics when pressing the shutter manually. Those torques occurred on a fairly slow time scale though, and so it was an open question as to what shutter speeds would be most detrimentally affected.

Here, I use MTFmapper to measure the sharpness of images in cycles/pixel (higher values = sharpe images) for a range of shutter speeds. At each shutter speed I take 3 images and report the average MTF observed from the set. As always with camera shake, there is variance between the shake observed in each image so the data is somewhat noisy. Below is the plotted loss in sharpness observed when not using a cable release for the two tripods referenced above as well as a control tripod.

The results largely speak for themselves. At faster shutter speeds around 1/60s and above, we observe virtually no loss in sharpness from pressing the shutter button by hand. Slower than 1/60s, and we begin to see a steady decrease in the average sharpness of the images that doesn’t appear to stop even at the slowest exposure time tested of 2 seconds. The stiffer TVC-33 did a better job at holding the camera steady by approximately 2 stops, but there will still plenty of soft images. The relatively small and lightweight Peak Design Travel tripod had no trouble holding the camera steady while a cable release was used to trigger the shutter. Use a cable release.

If a cable release is not available, using the camera’s built in 2 second timer can also work. The plot below shows the same data as above except that RRS data was removed and replaced by another series taken with the Feisol and a 2 second timer.

When the 2 second timer is used, I observed no loss in sharpness. This is of course dependent on the damping of the tripod being sufficient to steady the camera in under 2 seconds. Cameras also typically have a 10 second timer, but this isn’t practical for real world shooting. For analysis on exact damping times, stay tuned.

Note that the loss in sharpness observed in these plots is not huge. Above 0.30 cycles/pixel, the images are stunningly tack sharp in a way that only the very best lenses can achieve. At 0.25 and above, the images still look pretty sharp to the eye, though in a head to head comparison it is clear some fine detail is being blurred. On its own though, you would still label this image as sharp. 0.20 is still pretty good and probably usable. 0.15 shows some noticeable blurring when viewed at 100%, but could still be used for smaller prints. See the previous post for some examples. So we aren’t talking about massive losses in sharpness here. But if you are spending the big bucks on the best glass, get a $5 cable release too.

**Takeaways:**

- Use a cable release or 2 second timer at mid to slow shutter speeds.

*Edit*

I have received some flak for using MTF 50 for this analysis vs. MTF 20. Here is the data set for hand pressing the shutter with the CT3342 tripod using both MTF 50 and MTF 20:

The results are essentially the same, but as expected, MTF20 occurs at higher spatial frequencies. MTF20 is a much noisier measurement as the MTF is not particularly well behaved in general for looking at the loss of sharpness due to image shake. For example, here is a particularly poorly behaved MTF curve for an edge showing significant image shake:

A properly behaved edge showing only a little bit of image shake is as such:

]]>

In short this MTF test looks at the sharpness of an white to black edge transition. White to black to white would be considered one full ‘cycle’, and so the maximum observable MTF from one pixel to the next is half of one cycle, or 0.5. In practice we never actually observe MTF values that high due to optical imperfections, diffraction, bayer filters, etc. If you want to go down a rabbit hole of MTF testing for handheld images, I will refer you to Jim Kasson’s excellent blog. Here though, suffice it to say that MTF is simply a measure of image sharpness.

For this test I used the Fuji GF 120mm lens as it is the sharpest lens that I have available with image stabilization. I varied the shutter speed using a variable ND filter while keeping the aperture and ISO the same. At each shutter speed, three images were taken and the MTF values from each averaged to give the data point shown. Each image was independently focused by the contrast detection auto focus of the GFX 50S so that any focusing errors could be averaged out. The lens was stopped down to F/5.6, the sharpest aperture for the 120mm and one with relatively little focus shift. The MTF values are given in cycles/pixel and taken from camera generated JPEGs with the sharpness set at minimum (-4 for the GFX 50S). Because the MTF is not taken from raw files, these values should only be taken as relative to each other, and not absolute.

The graph below shows the average MTF recorded at each shutter speed for handheld shots taken with the built in OIS both on (red) and off (blue). Using identical methodology, the black dots show the MTF for images shot on a tripod with OIS off.

The images with OIS on are sharper, not exactly a great shock. We do see the image sharpness of the shots lacking OIS die off pretty rapidly around 1/100s, which is consistent with the general wisdom of using 1/focal length shutter speed while handheld. Note though that we see a small but steady loss of sharpness before that. With OIS on, we are able to consistently achieve reasonably sharp (>0.20 cycles/pixel) images until about 1/8 second shutter. This represents a roughly 4-stop improvement in image sharpness, less than Fuji’s advertised 5 stops.

Surprisingly to me, the images with OIS displayed better sharpness and consistency even above the standard 1/f shutter speed. Even with OIS though, the images didn’t get to quite the same sharpness as those shot off of the tripod. This is reasonable, as we can’t expect the OIS system to perfectly cancel out camera shake. Though, as we will discuss below, this difference in sharpness would be imperceptible in real life photography. I couldn’t shoot fast enough to get perfectly sharp images with this lens, so we will have to choose something with a shorter focal length and easier to hand hold.

Next lets look at the same data but taken with a shorter, lighter lens, the Fuji GF 45mm. This is also an excellent lens and while it displays more significant focus shift, it is well corrected by the camera when using autofocus. This is a smaller and lighter lens than the 120, so I wanted to see if I could match the tripod sharpness while handheld:

At the faster shutter speeds, I saw no difference between the handheld and tripod images. At about 1/100, we start to see a fast and steady decline in the average image sharpness. This is about 1 / 2 x focal length in shutter speed. At 1/ focal length, the images were acceptably generally sharp, but displayed more variance. Dropping below 1/30 I got perfectly sharp images, though one was occasionally usable. Of course, if you have particularly stable hands, your results may be different.

For reference, here is a perfectly sharp image from the 120mm GF:

The numbers are MTFmapper’s annotation for the sharpness of that edge in cycles/pixel. Now for a less sharp image:

While we have lost some of the very fine detail that can be seen in the perfectly sharp image, this is a usable level of sharpness and you can’t really notice any image shake.

This is now noticeably soft but could be acceptable for smaller prints. The point I am trying to make with this images is that the MTF test is very demanding. Images that look sharp when viewed in Lightroom may not be perfectly sharp. So, please don’t send me a message saying “I can get sharp images handheld at much slower shutter speed X” without quantifying your image sharpness. Everyone’s hands are different though, so results will indeed vary. Jim Kasson has done similar tests and achieved similar results.

While not the primary subject of this post, I also tested a the 120mm GF on the tripod with the OIS on:

I see now difference between the images that is statistically significant. Note that the sharpness appears to dip slightly around 1/10. I’m not sure exactly why this is, but I suspect that it is from the variable ND filter causing a slight loss in optical quality as it is rotated. It is also possible there is some slight vibration from the shutter or focusing mechanism that I am not accounting for. If the later were the case though, I would have expected more of a difference between having OIS on and off. Anyways, I still have yet to see any evidence that leaving OIS on while the camera is tripod mounted is harmful.

Takeaways:

- At fast enough shutter speeds, perfectly sharp images can be obtained without a tripod.
- Image stabilization helps a lot, but doesn’t perfectly correct for image shake.
- I received roughly 4 stops of benefit from OIS. It may be different for you.
- For handhold use, I would want to use 1 / 2 x focal length to get super sharp images. This applies to all camera formats with similarly sized pixels. Your results may vary depending on how stable your hands are.

These results roughly follow the conventional wisdom for shooting in today’s world of high resolution digital sensors. So, while there aren’t any dramatic conclusions here, it is important data for us to establish and will be a reference point moving forward.

]]>For this set of tests, I mounted the test camera onto the torque meter and recorded 20 second of torque data while taking images without a cable release. I also recorded the torque involve for changing a variety of settings and generally handling the camera. Unfortunately, the torque meter only measure one axis, so we only have data for the yaw torque applied to the tripod. The yaw stiffness is generally weakest on tripods, so this approach is at least somewhat reasonable. I also measured the yaw torque with the camera mounted in landscape orientation with a telephoto lens with lens foot. This should at least tell us if we are getting drastically different amounts of torque in the pitch direction under normal usage.

Before diving into the data below, please take these results with a grain of salt. Most of the tests on this site are precise and should be repeatable by a third party. This one, not so much. Everyone’s hands are a little different, and will handle the camera differently. I tried to handle the camera and press the shutter button as gently as possible, but certainly there are people out there with more precise control. From this data, I just want to extract a reasonable estimate of how much torque is placed onto the system in normal usage. As we will see in posts later on regarding damping, the exact amount of torque is going to be less important than pulling a consistent number out that we can apply across all tripods for comparison purposes.

Here is the torque observed over four exposures using the GFX 50S and 45mm GF lens:

Its a mess. But we see that while taking the exposures, I am putting an average of about 15 N*cm of yaw torque on the camera. Because my hand is on the camera, and hands are fantastic dampers, we don’t see the same kind of high speed oscillation observed with the wind. This has interesting implications for what kinds of mid-range shutter speeds can be used in each situation, but we won’t get into that now. I also tested things with the 120mm GF lens attached, but the results were pretty much the same in terms of torque.

Here is the torque for GFX 50S and 250mm GF. This is of course a little bit different as the camera is mounted away from the center of rotation:

It is much more obvious and consistent when my hand is placed on the camera. Surprisingly, the magnitude of the torque is roughly the same. As advertised, here is the torque with the same lens but the camera rotated into portrait orientation to simulate pitch torque:

Again, similar, but actually a little bit less torque. Close enough though that we really don’t have to worry about the pitch torque being dramatically different than yaw.

Now, instead of having the hand on the camera to take exposures, Here I just manipulate the settings on the lens and camera as I would during normal shooting. I am changing the shutter speed, aperture, ISO, focus point, menu buttons, etc:

The peak torque is much higher. The average torque is roughly the same. Some settings on the camera required much more force to change than others.

I took data in the style of the graphs above for a number of camera and lens combinations. I am going to spare us both and not post any more of the torque graphs. There really isn’t anything to be gained. You are just going to have to trust me that they are all mostly the same. Instead, I recorded the average peak torque from each graph in the tables below:

So for each camera and lens combo, I have taken three sets of torque measurements, each of which consisted of roughly four exposures. I took the rough average max torque in each set. Those are then recorded, and themselves averaged. We aren’t looking for absolute precision here, just to test consistency. And reasonably consistency was in fact observed.

The last column “Necessary Stiffness” shows the amount of yaw stiffness a tripod would have to have in order to hold the camera and lens pixel-level still while my hand was taking the picture. These are pretty large numbers, and you may question these results based on past experience of obtaining sharp images on weaker tripods. Given the slow nature of the torque, the camera is also moving about somewhat slowly. Thus only shutter speeds in a certain range would show loss of image sharpness. This likely would be in the 1/15 – 2 sec range, but I don’t know for certain, and would depend on focal length. Additional testing needed.

Here are the results from observing the torque while adjusting the camera settings:

Again, these are the result of 20 second trials where I took the rough maximum from each. We observe significantly more torque, but we are still easily within the same order of magnitude. We also see more variance here. This is reasonable. While testing it was obvious that the controls on the smaller X-H1 camera were lighter and moved easier. This resulted in smaller torques being exerted.

I didn’t calculate a stiffness column here because we generally don’t care about keeping the camera still while adjusting the settings. We just want the vibrations from doing so to settle in a reasonable amount of time. The situation is the same for pressing the shutter button when using the camera’s built in timer. In this case though we want to make sure that any vibrations from hitting the shutter will die out within the 2 second timer that is most commonly used.

We will be using this data as a basis for actually calculating these damping times in a future post. We also want to take a look at the necessary stiffness of tripod for it to be used without a cable release or timer. Just use a cable release though.

Taking the question list from before and adding to it:

- How much torque will we see from lesser wind speeds, such as 15mph 10mph, 5mph?
- How much loss of sharpness do we actually see from wind on tripods of differing stiffness?
- At what shutter speeds do we see a loss in sharpness from hand pressing the shutter button?
~~How much torque is placed on the system from pressing the shutter button?~~- How much damping is necessary for that vibration to damp out in a reasonable time?
- Can we minimize wind torque with camera placement on the tripod?

First, here is the measured torque and calculated amount of tripod stiffness necessary to obtain sharp images for each camera and lens combination at 15 MPH wind speed:

This data is consistent with our previous set. The amount of torque is in each case less, and thus is the amount of tripod stiffness necessary. Again though, we see that holding the telephoto lenses stable can only be done with the largest, stiffest tripods available. For the smaller normal focal length lenses though, small lightweight tripods should be sufficient.

Here is the data for 10 MPH:

Again, the torques are lower as expected. The amount of stiffness required for the telephotos is approaching a more reasonable range.

And the data for 6 MPH wind:

The air pressure at 6 MPH is incredibly light. Only for the 400mm lens on the X-H1 do we need a particularly stiff tripod. Note that 0.1 N*cm is the smallest amount of torque that the torque meter can measure. It hardly matters at that level though. Basically any tripod will provide the necessary amount of stiffness.

Here is the required tripod yaw stiffness assembled for each wind speed:

The results are intuitive. The more wind speed and the larger, longer focal length lens used, the more tripod stiffness is necessary. The weight and sensor format of the camera are not particularly relevant. A stiffer tripod allows the shooter more flexibility in shooting longer focal lengths in windier conditions.

There isn’t a lot of fanfare around this post in particular, but it is going to be central towards forming this site’s recommendation for how much tripod stability you need based on your equipment. As we have shown here, this approximation will lack subtlety and precision but will be critical to the site’s message and usefulness. Most users aren’t carrying around an anemometer to measure wind speed, or know the moment of inertia for their camera. We thus want to form a general guide people can start from when selecting a tripod and have them end up with reasonable results.

Moving forward, we need to test the actual loss in measured sharpness while taking long exposures on some specific tripods with the wind speeds measured above. Also, wind is far from the only force acting upon the camera. We need to measure the effects from shutter shock, handling the camera controls, my dog’s tail hitting the tripod leg, or whatever.

Taking the question list from before and adding to it:

~~How much torque will we see from lesser wind speeds, such as 15mph?~~~~10 mph?~~~~5 mph?~~- How much loss of sharpness do we actually see from wind on tripods of differing stiffness?
- How much torque is placed on the system from pressing the shutter button?
- How much damping is necessary for that vibration to damp out in a reasonable time?
- Can we minimize torque with camera placement on the tripod?