There is a limit to how much speed you can coax out of a Pi3. I have experienced the same limitations. Image downloading is definitely a bottleneck, so is astrometry. Solutions on the Pi can take as long as 90s.

For this reason, I have moved to a mini-PC. An important consideration is that you choose one that has USB3 capability, otherwise you will still have to deal with slower image downloads. You also should have at least 4 GB RAM and run the system form an SSD or eMMC. I use a Zotac 332pico and I am very, very happy with it. It runs a bit hot, so I stuck two heat sinks on it, that seems to have done the trick. Astrometry runs on that mini-PC (although I keep the astrometry.net files on an attached USB3 pocket hard drive).

The only drawback is that it draws twice as much power than the RPi3, so you need a larger battery to have it last through the night.

But not nearly as large a battery as you need for running the cooled camera, dew strips and the mount.

Well I have the rpi on my pier and run Kstars and Ekos on an Ubuntu laptop in the house and just run indiserver on the rpi, so all the work, including Astrometry is done on the laptop, I also solve offline as I have all the index files I need downloaded to the laptop, it all runs great...

Thanks El Corazon and Astronerd, I think I have worked out that I have the INDI parts running on the PI and Astrometry is, in my head, part of Ekos which is on the laptop. This is what I wanted. I just need to to some timings and see if I can find what is using up valuable photon counting time.

I have the PI and voltage regulators and PWM driver board (still have to do the arduino software for heaters), usb hub etc all in a neat box so would rather stick with that but I will consider a net top PC, strapped on the side as well.

I do have a couple of Tinker Boards which are a bit faster than the PI but I have not even looked to see if Indi will run on there.

Astronerd, I see you using similar gear to myself, SH SXVR H18, Lode Star, EQ5 and EQ6.

So I am curious if you have noticed any dead time between images, presumably either download time or network transfer time.

I am just setting up ready to try and measure those times taking some exposures of the dust cap in the day time.

No not really, I normally set a pause between exposures anyway just to let things settle, with the H18 the images take about 5 seconds to pop up in fits viewer and I have around a ten second pause...

I think I was worrying unduly, I was standing in the cold waiting for 10 second test images and it seemed to take forever. (I am an armchair astronomer after all)

So I have done some timings to put science ahead of guess work.

Raspberry PI-Raspbian-Indi and Laptop Ubuntu Ekos etc. H18 via USB2. Delay=0.

10 x 10 seconds took 212 seconds (53% photons lost)
10 x 300 seconds took 3145 seconds (5% photons lost)

Laptop Ubuntu Indi+Ekos etc. H18 via USB2. Delay=0.

10 x 10 seconds took 150 seconds (33% photons lost)

10 x 300 seconds took 3049 seconds (2% photons lost)


So for realistic exposure lengths, there is a small but not too big penalty to using the PI.
There is of course the advantage of less cables etc., that is I want to achieve.

And yes I also use a 10 second delay for the camera to dump download heat also heavy dithering.

All in all nothing to worry about.

Thanks to all for suggestions.

It's when you're downloading the image that the bottleneck occurs. When you take your picture, it's a raw file (or a JPEG, bad idea) or a *.fits which is sent back to your laptop. Depending on the size of your sensor, the size of your image can vary from 2 MB up to 40 MB and even more.
So the way you link your laptop to your Raspi is essential. If you use a WIFI cnx, depending on how you proceed, (802.11 b, g, n, ac) there's a huge difference in terms of time of transfer. One thing you should do is use this small and useful program called "iperf" to test the actual quality of your cnx. In my case, I reach 96Mb/s with a Raspi 2. That means my 40MB image will be sent in (40*/96=3.5 secs more or less. But if you have a lesser cnx, fetching your image will be much longer. So trying to optimize your link is the best call as a fist step in your quest for performance .
As for the plate solving, I think, you should stick to solve on the laptop with local database, and if you have ssd disks, it's even better.

So when using the rpi, was that remotely with the camera connected to the rpi, and using Ekos on the laptop..??, also how were you then connecting the rpi from laptop, WiFi or Ethernet cable..?

Yes, download speed of the connection is the biggest bottleneck. Second (depending on camera) is the USB connection. I have an ASI1600 with USB 3. RPi has USB 2. Both WiFi strength and USB3 port availability were for me reason to move from RPi to headless PC on the OTA. I opted for the Fitlet2, which runs Linux, great little computer, designed to be used outdoors in rough circumstances.
Overall I went from about 10-12s download time per image to around 5-6s download time, which for me was a worthwhile upgrade. For polar alignment, plate solving, focusing, you can further opt to take 2x2bins, which are 25% in size and download even faster.


Sent from my iPad using Tapatalk

I 100% agree with what wjdrijfhout said. USB3 + Gigabit ethernet + a decent CPU for solving make things MUCH more convenient. In my case readout of a frame plus transfer to the client combined takes 4s and plate solving typically takes 2s. The minor additional expense for quicker hardware is a one-off. The penalty for slow hardware however must be paid time and time again, on every single frame that is being downloaded and transferred.

Obviously you want the fastest computer with the fastest connection possible. That will differ which each user’s personal circumstances.
If you have access to a power outlet, you have more options than if you are in the field and running on batteries. In that latter case the Pi3 has an edge. Download speed throughout the night becomes a secondary factor if your Pi lasts all night on an external LiIon battery, while your PC has depleted the same battery in <2 hours.
There is no universal advice on this topic. Every setup has its pros and cons.

You’re correct El Corazon, to point out the power usage. And this is where small telescope mounted PC’s can vary quite a bit. The Fitlet2 uses 7W (Linux, not sure about windows, might be more intensive) which is more than a RPi, which is probably closer to 2 or 3 W. But in the total scheme of things, 7W is very little. By comparison, putting the ASI1600 cooling on 75% takes 14W, or turn on one dew heater for a 130mm scope takes 10W, and the mount takes 13W. For my setup, I use about 40W, of which I could save 3 or 4 by going with a RPi. I am using a 400Wh battery, which lasts me 10h of observation time. Unfortunately there are few nights where I need the full battery.