Hello everyone,
I haven’t found much information regarding Touptek-cameras and software, maybe there is somebody in this forum that can help answering some questions.
We are using USB3 CMOS cameras from Touptek for live-streaming microscopy-images on a 4K monitor.
The cameras currently in use feature the Panasonic 16MP 4/3 sensor that can be found in many 4/3 mirrorless cameras. This models do not have internal processing (transformation to video formats),
instead they send a stream of full resolution uncompressed images to the PC-system.
Theoretically this sensor can produce a frame rate of 22,5 fps at full resolution in 12bit.
The cameras have electronic shutters only, there is no vibration or noise.
Touptek cameras are sold under different names and are also used in astrophotography. They offer two different software packages, one for micro and one for astro. Both are free and work (as far as I know) with all cameras. I didn’t know that and bought the first camera labelled under a different brand name and sold as a camera for „astrophotography“. It came with the astro-software on CD only and so I used this for a while (it has the same live-image function). Later I found that there is a download-link for the micro-software on the Touptek homepage and gave it a try – it turned out that the software works flawless also with the rebranded camera. That was a nice surprise, especially because the software has impressive features (stacking, stitching, HDR, calibration, measuring functions, ROI,binning, color composition from mono images,.…). The system is very comfortable to use and delivers high quality images.
But there is one thing unclear at the moment: using an AMD Ryzen 3400 processor (quadcore with built in graphics, 16MB RAM, fast M2 SSD) we can reach a frame rate of max. 11fps . That’s the frame rate shown by the software, but it seems there is also some additional lag somewhere in the system so the fps finally shown on the monitor are maybe less than half of this value (coarse estamination).
It would be nice if we could increase the framerate, but I’m not sure where the bottleneck is and how far we could go with faster components.
My impression is that most of the work is done by the CPU and little by the GPU. The built in GPU of the ryzens are also relative fast, so buying a dedicated expansive graphics card has maybe not an great effect (?).
It is also unclear if the software makes use of processors with higher core numbers.
There is a new generation of Ryzen processors (series 4000/5000) also on the market now, but I don’t want to buy and change components by trial and error.
Maybe someone here has experience with this software on a very fast system and can help to get an idea what’s possible.
Thanks and best wishes,
Horst
Touptek cameras / software and framerate
Moderators: rjlittlefield, ChrisR, Chris S., Pau
Re: Touptek cameras / software and framerate
Hi,
I have no experience with that camera system but a bit with machine vision cameras in general.
first thing I would do is find out more about the bottle neck.
do you know if the camera transmits the RAW images or does it a internal debayer and sends out full RGB images?
if it's the latter, that might be your bottle neck since 16MP x 3channels x 11fps = 528MB/sec which pretty much fits the max of USB3.
with 12 bit RAW you could get 22fps, but then your CPU will have to do the debayer.
so next question is does it use multiple cores of your CPU or only one?
you should be able to see that in the task manager under resource monitor.
if the program doesn't multi thread well you'll only see a low load there, and then getting a CPU with best single core performance would be the way to go (last time I checked the high end intels were better in that respect but it might have changed in the mean time)
just some ideas.
ps: also check if the frame rate changes if you don't save the frames to disk.
500MB/sec is possible with a fast SSD hocked up over a fast M2 slot, but if there's a problem with drivers or trim etc you might be limited by that.
I have no experience with that camera system but a bit with machine vision cameras in general.
first thing I would do is find out more about the bottle neck.
do you know if the camera transmits the RAW images or does it a internal debayer and sends out full RGB images?
if it's the latter, that might be your bottle neck since 16MP x 3channels x 11fps = 528MB/sec which pretty much fits the max of USB3.
with 12 bit RAW you could get 22fps, but then your CPU will have to do the debayer.
so next question is does it use multiple cores of your CPU or only one?
you should be able to see that in the task manager under resource monitor.
if the program doesn't multi thread well you'll only see a low load there, and then getting a CPU with best single core performance would be the way to go (last time I checked the high end intels were better in that respect but it might have changed in the mean time)
just some ideas.
ps: also check if the frame rate changes if you don't save the frames to disk.
500MB/sec is possible with a fast SSD hocked up over a fast M2 slot, but if there's a problem with drivers or trim etc you might be limited by that.
chris
Re: Touptek cameras / software and framerate
Hi Chris,
thanks for your tips.
There is an option to set image parameters to RAW, RGB24 or RGB48 - this defines in which format an image taken by the snapshot-function is saved, but I can not see any difference in the live image or framerate. I assume the data coming from the camera is RAW (with this model, there are also cameras with internal processing available).
I was able to tweak the system a bit. The lag between the shown fps and the framerate actually on the screen is now very small (there is an option to make the software use also the GPU which is switched off by default). Focusing, or shifting the image on the screen with the mouse is now very smooth.
The software is using more than one core, but it is unclear to me in which way - looking at the core activity of two slightly different processors (both 4C/8T) they showed very different activities. The first processor was working near 100% on one core with just live image and when an additional function was activated (displaying live histogram for example) another core, in this case core 6, did the work. On the second system the workload was more evenly distributed. Maybe the drivers and operating system are deciding factors.
However, it' easy the bring the processor to it's limit - with the live histogram activated the framerate drops from 11 to 6-7 fps. Sharpening the live image brings the fps down to 1.
Some functions are not critical, changing contrast, gamma or color settings didn't show any significant effect on the framerate.
The nice thing with this new generation AMD processors is, that they have very high performance and at the same time low power consumption and heat dissipation. This allows for a small form factor and nearly noiseless operation, the PC almost disappears from the desktop, even one for such demanding operations. High resolution displays, extremely capable camera sensors and powerful processors come together at the moment - for me this is clearly the next step in imaging after the change from film photography to DLSR, then to mirrorless. Even photography and video is melting together.
Just an example - there is a Live-EDF function implemented which aligns and stacks the stream of images automatically, producing an image with extended deph of field. This may not produce the same quality as a set of carefully captured and manually stacked images (which is also possible) but it is super simple to use and done in seconds - all you have to do is activating the function an slowly turning the focus knob. In the screen capture below it is shown how it works, the smaller image is the actually live image (already out of focus) and the larger image shows the stacked result. I didn't even know that the software can do this - an intention for the new year is to occasionally read manuals 
.
thanks for your tips.
There is an option to set image parameters to RAW, RGB24 or RGB48 - this defines in which format an image taken by the snapshot-function is saved, but I can not see any difference in the live image or framerate. I assume the data coming from the camera is RAW (with this model, there are also cameras with internal processing available).
I was able to tweak the system a bit. The lag between the shown fps and the framerate actually on the screen is now very small (there is an option to make the software use also the GPU which is switched off by default). Focusing, or shifting the image on the screen with the mouse is now very smooth.
The software is using more than one core, but it is unclear to me in which way - looking at the core activity of two slightly different processors (both 4C/8T) they showed very different activities. The first processor was working near 100% on one core with just live image and when an additional function was activated (displaying live histogram for example) another core, in this case core 6, did the work. On the second system the workload was more evenly distributed. Maybe the drivers and operating system are deciding factors.
However, it' easy the bring the processor to it's limit - with the live histogram activated the framerate drops from 11 to 6-7 fps. Sharpening the live image brings the fps down to 1.
Some functions are not critical, changing contrast, gamma or color settings didn't show any significant effect on the framerate.
The nice thing with this new generation AMD processors is, that they have very high performance and at the same time low power consumption and heat dissipation. This allows for a small form factor and nearly noiseless operation, the PC almost disappears from the desktop, even one for such demanding operations. High resolution displays, extremely capable camera sensors and powerful processors come together at the moment - for me this is clearly the next step in imaging after the change from film photography to DLSR, then to mirrorless. Even photography and video is melting together.
.