Cosmic Homebrew

Spindown Disks under Proxmox

Mon, 19 May 2025 14:09:26 +0200

Reduce your power consumption in your Homelab: Spindown Idle Disks

My Proxmox server is an older Ivy Bridge based Xeon E3-1225-v2 with 32 GB ECC RAM and four 3 TB harddisks as main storage and three smaller SSDs for various services. In the true sense of hyper-convergence this small server provides computing and storage resources. It uses about 65W when idle, which amounts to roughly 15€ in monthly energy costs.

Roundabout 25-30W come from the four harddisks being constantly on. But truth be told, the NAS services are actually only used one or two times a day.ä for watching movies or shows. For 20 hrs the disks are idling away. Spinning them down saves 1/3rd of the energy costs.

Many people are concerned about stressing their disks with spinning down and up a few times a day. My experience is that this is not a problem unless you spin them up a few times per hour. I have been spinning down disks since 2014 and never lost a disk. My latest array has been running for three years now. Even if I would loose a disk eventually, the saved money is more than enough to buy a new disk. My RAID array can handle loosing a disk. In my opinion common sense is much more important - for disks this means to mix vendors and use disks from different manufacturing dates. Buying all identical disks at once from the same shop puts you at much higher risks for multiple disk failures which cannot be handled by RAID levels.

In Proxmox/Debian it is very easy to spin down idle disks. I use the hd-idle package:

sudo apt install hd-idle
systemctl enable --now hd-idle

The configuration file needs updating to configure the correct disks for spindown:

nano /etc/default/hd-idle

At the bottom of the file you can find the options string. Below example shows the options to spindown two disks after 1200s of idle time. This works well for me, you might be happier with longer or shorter idle periods.

HD_IDLE_OPTS="-a disk/by-label/wwn-0x50014ee25ca35250 -i 1200 \\
              -a disk/by-label/wwn-0x50014ee2b1f6a196 -i 1200 \\
              -l /var/log/hd-idle.log"

You start the hd-idle service with systemd:

sudo systemctl enable —now hd-idle.service

The current status of disks can be checked with hdparm:

hdparm -C /dev/disk/by-id/wwn-0x5000039fe6f3c34e

Its best to use disk identifiers /dev/disks/by-id instead of /dev/sdX. Disk identifiers never change even when add or remove disks from your server.

Header image designed by Freepik

Galaxies

Sun, 27 Oct 2024 20:17:25 +0100

Raw Stacks

Bode’s Galaxy M 81

Bode’s Galaxy M81 is a great target for galaxy season in the northern hemisphere. It is high in the sky rather close to the NCP which makes tracking easy and keeps it visible for a long time during the night. It is one of the few objects where a lot of fine-structure can be seen inside the galaxy. The whole area has a lot of faint dust that my image below does not capture.

Raw stack from PixInsight can be downloaded here. It was calibrated with bias and flat frames. Standard settings from WBPP (2024) were used, incl. CFA drizzle at scale 1. It is slightly oversampled. This is 275x 180s at gain 101. Although I am generally not a fan of AI tools, I used BlurXTerminator for this image.

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

Fireworks Galaxy NGC 6946

The Fireworks Galaxy NGC 6946: This is the result from three nights. It’s amazing to see how fast you can swamp the readnoise with the ASI533MC. The sensitivity seems to be roughly 4x that of my old Nikon D3400. And this is with the uncooled model! Sadly, my seeing is mostly meh. It’s always 3“ at my place. Rarely it’s 2.5“ and similar rarely it’s 3.5“. Kind of boring.

This image was done with a Celestron C8, the original 0.63x reducer/corrector and an ASI533MC (uncooled). Everything was mounted on a Skywatcher HEQ5 (with Rowan Belt Mod) together with 15 kg of counterweights on my backyard pier. During processing I had to toss 40% of my frames from twining the nights due to wind and clouds. The HEQ5 is small for a C8 after all and it wasn’t the best weather. Integration time is a bit over 14 hrs from a three nights.

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

Spiral Galaxy NGC 891

The side-on-visible spiral galaxy NGC 891, also called the Outer Limits Galaxy. I consider it to be a very special object because it’s arguably one of my best images: I like the side-on view with lots of details in the galaxy and the nice colorful stars. This image was done with a Celestron C8, the original 0.63x reducer/corrector and a Nikon D3400 (unmodified). Everything was mounted on a Skywatcher HEQ5 (with Rowan Belt Mod) together with 15 kg of counterweights on my backyard pier. Integration time is a almost 9 hrs from three nights.

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

NGC 2403

The galaxy NGC 2403 s a great target for galaxy season in the northern hemisphere. It is high in the sky rather close to the NCP which makes tracking easy and keeps it visible for a long time during the night. It is one of the less popular but very beautiful objects.

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

Pinwheel Galaxy M 101 (widefield)

Another object that is high in the sky in the northern hemisphere: the Pinwheel Galaxy M101.

This image shows a widefield view of the area. It was acquired with a 20 EUR Sigma zoom lens at 195mm focal length (stopped down to f/7 for better quality) and a unmodified Nikon D3400. Everything was mounted on a Skywatcher HEQ5 (with Rowan Belt Mod). I collected almost 11 hrs over five nights.

Raw stack from PixInsight can be downloaded here. It was calibrated with bias and flat frames. Standard settings from WBPP (2024) were used, incl. CFA drizzle at scale 1. This is 128x 300s subexposures.

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

Nebulas

Tue, 22 Oct 2024 19:13:56 +0200

Raw Stacks

North American Nebula NGC 7000

This image shows a widefield view of the North American and Pelican Nebula are in Cygnus. It was acquired with a 20 EUR Sigma zoom lens at 195mm focal length (stopped down to f/7 for better quality) and a unmodified Nikon D3400. A ZWO Duoband filter was placed in front of the lens. Everything was mounted on a AZ-GTI in EQ mode. I collected data on three nights, some of the Ha data was captured in astronomical twilight.

Processing was done in a special way. Since the duoband filter passes light at Ha and OIII wavelength (12nm and 30nm width) I seperated the RAW data into its individual Bayer matrix components using SplitCFA in PixInsight: Red was assigned to Ha, Green to OIII and Blue also to OIII. The oxygen signal ends up in the blue and green channel because for a typical Bayer matrix, both color channels have a significant overlap. This allows to do a freestyle HOO color combination when processing.

Raw stack from PixInsight can be downloaded here. The zip files contains a stack of the red channel as Ha (132x 300s) and blue/green as OIII (195x300s).

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

Little Dumbell Nebula M 76

The Little Dumbell Nebula M76 is one of the smaller yet interesting objects

My processed imaged with details can be found on Astrobin. You can read about the this galaxy here on Wikipedia.

Star Clusters (Open and Globulars)

Tue, 22 Oct 2024 11:28:11 +0200

Raw Stacks

Double-Cluster_2x30x180s.zip

Double Cluster NGC 864 and NGC 869

The Double Cluster in Perseus consists of two open star clusters NGC 869 and NGC 869. They span around 1° on the sky, which makes them a joy to observe and image in smaller scopes. This image was done with a Celestron C8, the 0.63x reducer/corrector and an ASI533MC (uncooled). The small FOV necessitated making this a mosaic.

Individual stacks can be downloaded here. Each panel is 30x 180s at gain 101.

My processed imaged with details can be found on Astrobin. You can read about the Double Cluster here.

GoTo Modification for the Skywatcher SkyTee 2 Alt-Az Mount

Sun, 20 Oct 2024 17:44:38 +0200

A few people asked me to share how I motorized my manual SkyTee 2 alt-az mount into a full GoTo system using OnStep. The SkyTee 2 is sold under different brand names, but the mount is always the same. Mine is branded as the Omegon TwinMaster AZ.

Cost Overview

Component	Cost (approx.)
Mount	350 EUR
Tripod	150 EUR
Electronics	100 EUR
Total	~600 EUR

The beauty of this modification is that it requires no 3D printer and no custom parts. For around 600 EUR you get a mount that handles a C8 effortlessly. With upgraded clamps and proper balancing, I would put a C9.25 on it without hesitation.

Performance

Visually, the mount is a joy — focus wobble is negligible even at 500x. For astrophotography, it achieves 10 seconds unguided at 1300mm focal length, and with guiding I get 90-second exposures at 1300mm with a 95% keep rate (field rotation becomes the bigger limiting factor at that point). Some example images can be found in my alt-az imaging thread on CloudyNights.

Parts List

2× M4×14 screws (plus 2× M4×16 as backup)
2× NEMA17 mounting brackets
2× NEMA17 motors (200 steps/rotation, 1.5A)
2× 16T 6mm GT2 pulleys
2× 48T 6mm GT2 pulleys
1× 166T 6mm GT2 belt (ALT axis)
1× 164T 6mm GT2 belt (AZ axis)
1× OnStep controller (MKS Gen L 2.0 with TMC5160 drivers)

I used mounting brackets with the measurements below:

Mechanical Assembly

The SkyTee 2 has conveniently located mounting screws at the wormgear housings of both the AZ and ALT axes. These can be used to mount NEMA17 motor brackets directly. The screws are M4 — be careful not to strip the threading, as the metal is fairly soft.

Azimuth Axis

The AZ bracket fits directly and is straightforward to install. I only use one screw to attach it — I was skeptical at first, but 18 months of use has shown it is solid enough. There is very little torque applied during normal operation.

Altitude Axis

The ALT bracket is slightly more involved. The wormgear housing sits closer to the base, so you need to remove the back part of the NEMA17 bracket to make it fit. This is easy to do with a small metal saw.

Pulleys and Belts

Remove one slow-motion knob per axis — each is held by a small grub screw and comes off easily once loosened. The 48T pulleys replace the slow-mo knobs, and the 16T pulleys go on the motor shafts, giving a 3:1 reduction. The wormwheel has 144 teeth (identical to the EQ5), resulting in an overall 432:1 mechanical reduction.

Since both axes are slightly different lengths, the required belt lengths differ — 166T for ALT and 164T for AZ. That said, it is worth buying a small variety of lengths since they are inexpensive. Fit the belt onto both pulleys before gently sliding the assembly onto both axis shafts, and align carefully. Belts do not need much tension — being able to press the belt in by 5mm is perfectly fine.

At this point the mechanical part is essentially complete. Opening the clutches or disconnecting the motors from the controller allows you to continue using the mount fully manually.

Electronics and OnStep Configuration

I used an MKS Gen L 2.0 3D printer board with TMC5160 drivers. The TMC5160s are on the pricier side (~10 EUR each on AliExpress) but run cool and quiet. I bought 1.5A motors, which could not be driven reliably by the TMC2130s typically recommended for OnStep. Most OnStep users recommend 400-step, 0.9A motors — these work well with TMC2130s.

Key configuration parameters in OnStep:

Microsteps for tracking: 32 → 7680 steps per degree
Motor current (via SPI): 1200mA tracking / 1400mA GoTo
Microsteps during slews: 8 (improves GoTo speed on the slower MKS Gen L board)
Base slew rate: 2°/second

The controller runs on 12V. I control it via SkySafari and the built-in OnStep web interface. If you build your own board instead of using the MKS Gen L, you can add an ST4 port for a handcontroller. I also demonstrated a wireless handcontroller solution that has since been merged into the standard OnStep repository — meaning handcontroller support is available even on builds like this one.

Full OnStep documentation and a friendly community can be found at onstep.groups.io.

Final Result

The finished build has the controller neatly attached to the legs of the mount — a compact and clean setup that is easy to transport.