Cheap DIY Remote Controlled Flat Panel - Complete Guide

[IanL]

I've finally managed to get round to building a remote control flat panel (suitable for use in SGPro and any other software that supports Alnitak Panels). Total cost was less than £25 - I used one of those LED tracing panels that you can find all over Amazon and eBay. Surprisingly the first one I bought turned out to be a real winner - A4 sized, USB power, 60 LEDs and dimmable for £15. The key difference seems to be that many of these tracing panels have a grid of LEDs close behind a diffuser panel which creates dark and light spots, or the are edge-illuminated from one side only.

The one I got has two strips of 30 white LEDs along the long edges - internally it has a reflecting layer, a clear acrylic sheet which the LEDs shine in to and then a diffusing layer sandwiched on top.  The light seems very uniform and I didn't need to add any futher diffusing or similar elements to make it usable. I hacked out the original controller and hooked everything up to a 5V Arduino Nano clone (£5) and a MOSFET (£2) and used some Alnitak emulation code written by one of the SGPro developers.

It is now mounted on my observatory wall and I just have to park the scope and SGPro can take care of the flats for me for a fraction of the cost of the cheapest automated flat panel. At full brightness I can do flats for LRGB filters in a couple of hundredths of a second, and about half a second for narrowband filters. This is ideal for my ASI 1600MM-C (not Pro), since longer exposures use a readout mode that creates gradients in your flats. Used it this week and results are as good as using my DIY manual diffuser and cloudy sky method, with less risk of floating away in a garden under 2 inches of water!

Anyway, full write up with shopping list, photos and diagrams is available here:

https://www.blackwaterskies.co.uk/2020/03/cheap-diy-remote-controlled-flat-panel/

[8324689]

This guide is fantastic and am about to dive in. 
I cannot find any lighboxs on amazon that say they have any amount of leds let alone 60. Has anyone bought one lately they could recommend please ? 

[james_screech]

On 12/03/2020 at 15:39, IanL said:

I've finally managed to get round to building a remote control flat panel (suitable for use in SGPro and any other software that supports Alnitak Panels). Total cost was less than £25 - I used one of those LED tracing panels that you can find all over Amazon and eBay. Surprisingly the first one I bought turned out to be a real winner - A4 sized, USB power, 60 LEDs and dimmable for £15. The key difference seems to be that many of these tracing panels have a grid of LEDs close behind a diffuser panel which creates dark and light spots, or the are edge-illuminated from one side only.

The one I got has two strips of 30 white LEDs along the long edges - internally it has a reflecting layer, a clear acrylic sheet which the LEDs shine in to and then a diffusing layer sandwiched on top.  The light seems very uniform and I didn't need to add any futher diffusing or similar elements to make it usable. I hacked out the original controller and hooked everything up to a 5V Arduino Nano clone (£5) and a MOSFET (£2) and used some Alnitak emulation code written by one of the SGPro developers.

It is now mounted on my observatory wall and I just have to park the scope and SGPro can take care of the flats for me for a fraction of the cost of the cheapest automated flat panel. At full brightness I can do flats for LRGB filters in a couple of hundredths of a second, and about half a second for narrowband filters. This is ideal for my ASI 1600MM-C (not Pro), since longer exposures use a readout mode that creates gradients in your flats. Used it this week and results are as good as using my DIY manual diffuser and cloudy sky method, with less risk of floating away in a garden under 2 inches of water!

 

Anyway, full write up with shopping list, photos and diagrams is available here:

https://www.blackwaterskies.co.uk/2020/03/cheap-diy-remote-controlled-flat-panel/

Nice writeup, you mention that the ASI1600MM-C give gradients on longer exposure flats, that's exactly the problem I've trying to solve for weeks since mounting a EL flat panel on the wall of my new observatory. Do you know what exposure is needed to avoid the gradients?

[IanL]

5 hours ago, james_screech said:

Nice writeup, you mention that the ASI1600MM-C give gradients on longer exposure flats, that's exactly the problem I've trying to solve for weeks since mounting a EL flat panel on the wall of my new observatory. Do you know what exposure is needed to avoid the gradients?

I can only speak for the earlier versions of the 1600, i.e. the newer "Pro" version has onboard RAM to overcome these issues as it speeds up the readout greatly. With the older versions of the 1600 there is a change in readout modes for exposures longer than 2 seconds (if using USB3) or longer than 5 seconds (if using USB2).  Apparently this is to reduce the noise in longer exposure images but ZWO never really explained how or why.

Shorter exposures below these limits effectively stop the exposure at read-out, but longer exposures read out progressively which means that pixels at the top of the sensor are exposed longer than those at the bottom. For lights with relatively little signal, this does not matter much, but for flats where there is a bright light source you can end up with a gradient of maybe 5% from top to bottom of the flat if the exposure is longer than the cut-off.

The solution is to either set the brightness of your flat light source and/or higher gain to achieve a short exposure less than the cut-off (no gradient) or do the opposite and use a dim source / lower gain to achieve a much longer exposure. You probably wouldn't be able to dim the panel enough for LRGB filters, but possibly for narrowband., so you might need to use some material over the panel to reduce output, e.g. sheets of white paper or ND film. The idea is that once the exposure is long enough, so the ratio of flat exposure time to unwanted exposure time during progressive read-out is sufficient to reduce the gradient to a negligible percentage of the median brightness.

[8324689]

I got this.....

https://www.amazon.co.uk/gp/product/B07B93R956/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&psc=1

 

fingers x'd

[james_screech]

39 minutes ago, IanL said:

I can only speak for the earlier versions of the 1600, i.e. the newer "Pro" version has onboard RAM to overcome these issues as it speeds up the readout greatly. With the older versions of the 1600 there is a change in readout modes for exposures longer than 2 seconds (if using USB3) or longer than 5 seconds (if using USB2).  Apparently this is to reduce the noise in longer exposure images but ZWO never really explained how or why.

Shorter exposures below these limits effectively stop the exposure at read-out, but longer exposures read out progressively which means that pixels at the top of the sensor are exposed longer than those at the bottom. For lights with relatively little signal, this does not matter much, but for flats where there is a bright light source you can end up with a gradient of maybe 5% from top to bottom of the flat if the exposure is longer than the cut-off.

The solution is to either set the brightness of your flat light source and/or higher gain to achieve a short exposure less than the cut-off (no gradient) or do the opposite and use a dim source / lower gain to achieve a much longer exposure. You probably wouldn't be able to dim the panel enough for LRGB filters, but possibly for narrowband., so you might need to use some material over the panel to reduce output, e.g. sheets of white paper or ND film. The idea is that once the exposure is long enough, so the ratio of flat exposure time to unwanted exposure time during progressive read-out is sufficient to reduce the gradient to a negligible percentage of the median brightness.

Thanks for the information. However that's not my problem then as my exposure is less than 1 sec.

[IanL]

1 minute ago, james_screech said:

Thanks for the information. However that's not my problem then as my exposure is less than 1 sec.

You may have done this already, but the first thing I'd try taking a flat (call it "A"), taking a second flat (call it "B") then physically rotating the panel 90 degrees and taking another one (call it "C"). Then:

- Compare the median ADU values of "A" and "B". They should be pretty much the same give or take a few ADU or tens of ADU. If they're markedly different then you either have a problem with the light source (flickering at a high frequency that is sufficient to create different brightness levels from frame to frame and may appear as a gradient), or perhaps the camera/electronics but seems less likely.

- Verify this by subtracting "B" from "A" using PixelMath in PixInsight or something similar.  To do this properly you'd probably want to use an expression like : (A + 0.1) - B so that noise doesn't end up clipping loads of pixels to negative values that are then truncated to zero. All being well you should end up with a result frame with a median value of close to 0.1 (using the normalised real range in the statistics tool). If you end up with a value that is closer to zero or well above 0.1 then your flats have different illumination levels (e.g. banding caused by the panel or camera). 

- Stretch "A" and "C" and examine them visually. If you can see the gradient it should be the same in both, in which case there may not be a problem, just that there is genuinely a gradient in your imaging train that the flat will correct. If the gradient is rotated by 90 degrees between A and C then the source of the problem is the EL panel itself which isn't producing a flat field, or perhaps you haven't got the panel orthogonal and equally spaced between the two setups.

- Verify this by subtracting the unstretched versions of "C" from "A" using PixelMath in PixInsight or something similar.  Again to do this properly you'd probably want to use an expression like : (A + 0.1) - C so that noise doesn't end up clipping loads of pixels to negative values that are then truncated to zero. All being well you should end up with a result frame with a median value of close to 0.1 (using the normalised real range in the statistics tool). If you end up with a value that is closer to zero or well above 0.1 then again your light source is not flat or it is mechanically shifted with respect to the optical axis between A and C.

If it's any comfort, I've always struggled with getting decent flats using any type of device. The best ones have always been taken using strongly overcast daylight sky with a diffuser on the end of the OTA, but is isn't convenient at all. This LED panel is the best I've managed with a device of any sort, but even so it is not as reliable and 100% clean as using daylight.

[james_screech]

I've tried rotating the panel and this had no noticeable affect on the gradient, I've also subtracted one master flat from another and I get a gaussian distribution of noise cantered around 0. So I think the panel is fine.

I might just have to go back to using a tracing panel (similar to yours) with the scope pointing straight up instead of horizonal. I can only assume something in the image train is moving but I can't figure out what as I can't feel any movement when I put pressure on the camera. The gradient is dark in the bottom left of the image and brightest in the top right (3%-4% difference) as the panel in approximately to the east of the mount, this corresponds to an external vertical gradient due to the sensor orientation. I've checked and checked again that the scope is pointing directly at the panel.

I had been hoping to use the EL panel wall mounted and controlled by a USB relay so I could control everything remotely from indoors.

[IanL]

39 minutes ago, james_screech said:

I've tried rotating the panel and this had no noticeable affect on the gradient, I've also subtracted one master flat from another and I get a gaussian distribution of noise cantered around 0. So I think the panel is fine.

I might just have to go back to using a tracing panel (similar to yours) with the scope pointing straight up instead of horizonal. I can only assume something in the image train is moving but I can't figure out what as I can't feel any movement when I put pressure on the camera. The gradient is dark in the bottom left of the image and brightest in the top right (3%-4% difference) as the panel in approximately to the east of the mount, this corresponds to an external vertical gradient due to the sensor orientation. I've checked and checked again that the scope is pointing directly at the panel.

I had been hoping to use the EL panel wall mounted and controlled by a USB relay so I could control everything remotely from indoors.

Other light source adding to the flat maybe - daylight or a street light?

[steppenwolf]

On 12/03/2020 at 15:39, IanL said:

Anyway, full write up with shopping list, photos and diagrams is available here:

https://www.blackwaterskies.co.uk/2020/03/cheap-diy-remote-controlled-flat-panel/

Excellent project and resource - thanks!

[james_screech]

3 hours ago, IanL said:

Other light source adding to the flat maybe - daylight or a street light?

No, the same with the observatory closed at night and no lights on in the observatory.

[IanL]

15 hours ago, james_screech said:

No, the same with the observatory closed at night and no lights on in the observatory.

OK I was doing some flats yesterday which weren't satisfactory. There was a light band across width of the frame (about 5% of the total height of the frame), followed by a narrower dark band and then a lighter band shading in to a more 'normal' looking flat. This was across R, G, B and L filters. I also thought it might have been daylight leaking in so made a cardboard shield to block/reduce any light coming in from the narrow gap around parts of the observatory wall/roof meeting. Made no difference.

I had set the flat panel to 255 (maximum) brightness and had calibrated the exposure time to produce 30,000ADU using SGPro's flat calibration wizard, giving me exposure times of 0.01 - 0.03 seconds depending on filter. This had worked pretty well for previous images, so not sure what has produced this change.

I suspected banding due to either PWM flicker (in theory the circuit should have the LED on 100% of the duty cycle at that setting but who knows?), or possibly some kind of camera/AC mains frequency issue. The dark stripe on the red filter was in a slightly different position than the L, G and B - and the red has about 3x the exposure of the L, supporting my theory. I set the flat panel to 10 brightness (very dim) and re-ran the flats calibration wizard, producing exposure times of 0.13 to 0.51 seconds depending on filter.

This is still enough to avoid the 1600MM (non-pro) long exposure gradient issue, and I was hoping the longer exposure would average out any flickering or similar. Just captured some new flats and they seem to be free of the artefacts found with the much shorter exposures. I don't know if it will help you, but I'd definitely recommend trying similar by either dimming the EL panel or using some kind of diffuser/filter over it.

Edited September 23, 2020 by IanL

[james_screech]

I could try dimming the panel, I already have an opal acrylic sheet diffuser.

[8324689]

Is this ok even though it has the mesh type background? 
 

[IanL]

The mesh effect is a pretty standard part of the diffuser on these panels. Really the only test that matters is to take some flats and check them before you start tearing it down.

[AndyThilo]

I'm going to make a panel, is there a recommended panel from Amazon? I can't find any link to the one used by Sub Dwarf.

[8324689]

I asked the same question but Didn’t really get an answer but it seems that anyone is fine. I still need to put some sheets of white paper in front of mine to dim the light a bit. That diffuses the light nicely

[AndyThilo]

Got mine working at the weekend. Used this panel from Amazon - https://www.amazon.co.uk/gp/product/B07KLY84NB/ref=ppx_yo_dt_b_asin_title_o05_s00?ie=UTF8&psc=1

Did some test flats inside with my GT81 and ASI294. In order to get the flats to 3secs I had to drop the percentage in NINA to 2% lol. However, as suspected, it revealed virtually zero vignetting and only one dust mote (I keep my optics clean) and imaging from the GT81 so far has shown zero gradients in images. I'll test on my RC8 if the panel fits, I'm sure it'll be a different story!

I put the Arduino and MosFET in a small plastic box with the USB cable coming out via a cable clamp. The box is stuck on the back so I can just plug it in an go. 

Resulting Flat master from Pixinsight. Tried to stretch it out but there's just nothing there apart from small dark dust mote at the top right.