ATEM Mini Pro deinterlacer quality

Just got an ATEM Mini Pro (updated to 9.5.1 firmware), and one use case I have involves working with native 1080/50i sources. As the ATEM Mini Pro has 1080/50i-capable inputs and outputs 1080/50p - I'd assumed, naively, that the 50Hz motion of a 1080i source would be preserved.

I have set my ATEM Mini Pro to 1080/50p output - and the recordings made to USB-C storage are 1080/50p in the 70Mbs Hyperdeck High setting.

However early tests seem to show a major quality issue with the deinterlacers being used for 1080i inputs. It appears that 1080/50i sources are being de-interlaced to 1080/25p and then frame doubled to get to 1080/50p, rather than being properly deinterlaced to 1080/50p with full source motion preserved and 50 unique frames each second?

If you are running a 1080/25psf input (1080/25p in a 1080/50i signal) you won't notice this - but if you have a native 1080/50i source, with motion between fields in each frame, the result is really nasty and juddery?

Would an UpDownCross HD Mini Converter upstream converting the 1080/50i to 1080/50p before feeding the ATEM Mini Pro sort this (do they properly deinterlace 50i to 50p) - or do the UpDownCross HD's use the same compromised half-field rate deinterlacing with frame doubling as the ATEM Mini Pro seems to?

Is this a know issue and a compromise to hit the price point in the ATEM Mini Pro?

** EDIT - have fed a 1080/50p source into the ATEM Mini Pro in 1080/50p (the same 1080/50i content I played in on a Hyperdeck - but this time software deinterlaced in ffmpeg to 1080/50p prior to playback) and confirm that doesn't get dropped down to 1080/25p and frame doubled and has full 50Hz motion - so it does look just to be interlaced sources that don't get full motion ?**

You may be right , that 50i is deinterlaced to 25p and doubled to 50 on output .. ON bigger atems , you can output 50i .. on multiview is picture tragical .. but output is OK ..

Built in deinterlacer is poor.
50i and 25PsF (or 60i/29.97PsF) is also poor.
UdX marginally better - but doesn't seem to do a proper pulldown on PsF so you still see dithering on fine details.
Teranex 2D does it properly.

Built in deinterlacer is poor.
50i and 25PsF (or 60i/29.97PsF) is also poor.
UdX marginally better - but doesn't seem to do a proper pulldown on PsF so you still see dithering on fine details.
Teranex 2D does it properly.

Ignoring pSF does the UDX do full 50i to 50p motion, or take you down to 25p and frame double like the Atem Mini Pro seems to?

One related deinterlacing quality issue that bugs me is the degradation of powerpoint slide sharpness when the laptop/pc output into the Atem Mini happens to be 1080i. In the attached image you can see the difference in the "o" of Plato. The rounded top of the "o" is sort of a straight line when 1080i.

what is the reason to send LAPTOP output at 1080i ? Lot of laptops cannot 1080i , only 1080p .. Last Nvidia drivers is progresive only too ..

One related deinterlacing quality issue that bugs me is the degradation of powerpoint slide sharpness when the laptop/pc output into the Atem Mini happens to be 1080i. In the attached image you can see the difference in the "o" of Plato. The rounded top of the "o" is sort of a straight line when 1080i.

Ouch - that vertical resolution drop in 1080i when converted to 1080p suggests the ATEM deinterlacer is not only not deinterlacing 50i to 50p, but it's deinterlacing by discarding/ignoring alternate fields so effectively treating a 1080/50i source as 540/25p (not even 540/50p)

It appears it’s taking 1080/50i, discarding alternate fields to generate a 540/25p signal that is then frame doubled to 540/50p and crudely upscaled to 1080/50p by line-doubling (or something not much better than that)?

If I get a chance I’ll bung some 1080/50i test signals from a Phabrix through it.

A 1080i zone plate should tell us quite a bit more.

OK - did a simple test.

Created a clip on a 1080/50p timeline in Resolve that had alternate frames of a big number 1 and a big number 2. Exported this in 1080/50p and 1080/50i ProRes HQ and played it on a Blackmagic Ultrastudio Mini 4K feeding the HDMI output of the Ultrastudio into the Atem Mini Pro.

The 1080/50p clip has alternate FRAMES of 1 and 2. The 1080/50i clip has alternate FIELDS of 1 and 2 (when displaying the interlaced frame weaved you see alternate lines of the 1 and the 2 with combing)

Different deinterlacers will handle the interlaced fields differently - some will merge the two fields to create a static image of 1 and 2 superimposed, others will display flashing 1 & 2 etc. However all decent deinterlacers should display contributions from both fields (i.e. show you elements of both number 1 and number 2)

When I played the 1080/50p clip into the ATEM Mini Pro I got what was expected - alternate frames of 1 and 2 (i.e. 1 and 2 flashing at 25Hz) on the HDMI 1080/50p output.

When I played the 1080/50i clip, I just got the 1 displayed and never saw the 2...

This looks to prove that the ATEM Mini Pro is simply discarding one of the two 540 line fields if fed an interlaced 1080i signal, scaling up the single retained 540 line field to 1080 lines and repeating it to get it to output the 50fps if needed. (I guess the ATEM is really designed for 30fps and below streaming?)

Presumably this also means that 1080/25psf outputs from cameras that output 1080/25psf (i.e. 1080/25p in a 1080/50i signal) rather than native 1080/25p will also be displayed in half-resolution - as just one field will be used, rather than both fields used to recreate the original 1080/25p resolution?

I discovered that you need to match each input ant the atem output to get the best possible quality.

Had one camera set to 1080/50 and another one to 1080/25. Result was that the image of one camera was chrystal clear and the other image got jagged edges.

After setting all inputs and ATEM output to the same values the quality of all inputs was perfect.

I discovered that you need to match each input ant the atem output to get the best possible quality.

Had one camera set to 1080/50 and another one to 1080/25. Result was that the image of one camera was chrystal clear and the other image got jagged edges.

After setting all inputs and ATEM output to the same values the quality of all inputs was perfect.

Were you running any native interlaced inputs in any situation (i.e. 1080/50i aka 1080/i25) or were all of your inputs 1080p?

As you can't run the output of the Atem in 1080/50i (aka 1080/i25) you can't switch the output to the input format in that situation.

I'm guessing you switched from 1080/50i or 1080/25psf (aka 1080/25p in a 1080/50i signal) to 1080/50p or 1080/25p native (not psf) for all inputs and outputs - which is fine if you can do that. But if you have a native interlaced 1080/50i source (such as a Decklink playing 1080i content) or a camera that only outputs psf you are slightly stuffed?

(Interlace is still very much with us - the UK standard delivery format for all UK transmission masters for broadcast is 1080/50i... Feels a bit disingenuous to describe the Atem format converters as handling 1080/50i inputs when in reality they take this content down to 540/25p and then scale it and frame repeat it back up to 1080/50p. But I get it - you are getting what you paid for. The Atem Minis hit a ridiculously low price point.)

Unfortunately we've found quite a few BMD products don't really 'do' interlaced, despite saying they support it. For example the MultiView 4 seems to internally deinterlace by crudely combining fields (not anything clever like double-rate bob/bwdif deinterlacing) and discarding vertical resolution. You can end up with a seriously degraded picture from an interlaced input source, and they also don't properly reinterlace on output.

All the BM multiviewers I've tested with seem to incorrectly handle interlaced incoming signals, regardless of how you configure them.

We stopped attempting to use BM multiviewers for important applications, and went back to the Decimator Designs products when we need an updowncross (MDHX) or multiviewer (DMON range). We know we can trust those to correctly handle and scale interlaced content. It's unfortunate given the price disparity, but Blackmagic don't seem interested in improving some of their product designs to correctly deal with interlaced signals.

Unfortunately we've found quite a few BMD products don't really 'do' interlaced, despite saying they support it. For example the MultiView 4 seems to internally deinterlace by crudely combining fields (not anything clever like double-rate bob/bwdif deinterlacing) and discarding vertical resolution. You can end up with a seriously degraded picture from an interlaced input source, and they also don't properly reinterlace on output.

All the BM multiviewers I've tested with seem to incorrectly handle interlaced incoming signals, regardless of how you configure them.

We stopped attempting to use BM multiviewers for important applications, and went back to the Decimator Designs products when we need an updowncross (MDHX) or multiviewer (DMON range). We know we can trust those to correctly handle and scale interlaced content. It's unfortunate given the price disparity, but Blackmagic don't seem interested in improving some of their product designs to correctly deal with interlaced signals.

Yes - I've heard bad things about BMD multiviewers and interlaced sources. Real shame.

We are working in 1080i50 and multiview is 2160p25 .. Its not usable ..

what is the reason to send LAPTOP output at 1080i ? Lot of laptops cannot 1080i , only 1080p .. Last Nvidia drivers is progresive only too ..

Good question. Occasionally I run across a situation at conferences where 1080i is the resolution used to feed the projectors in the room (because of limitations with SDI cable length is what I've been told) and the only feasible output available for me to grab for the Atem mini recording/livestream is a branch off of this feed.

Maybe this will become a less likely scenario as time goes on.

Sorry for the late reply. I forgot to subscribe to the thread.

OK - did a simple test.

Created a clip on a 1080/50p timeline in Resolve that had alternate frames of a big number 1 and a big number 2. Exported this in 1080/50p and 1080/50i ProRes HQ and played it on a Blackmagic Ultrastudio Mini 4K feeding the HDMI output of the Ultrastudio into the Atem Mini Pro.

The 1080/50p clip has alternate FRAMES of 1 and 2. The 1080/50i clip has alternate FIELDS of 1 and 2 (when displaying the interlaced frame weaved you see alternate lines of the 1 and the 2 with combing)

Different deinterlacers will handle the interlaced fields differently - some will merge the two fields to create a static image of 1 and 2 superimposed, others will display flashing 1 & 2 etc. However all decent deinterlacers should display contributions from both fields (i.e. show you elements of both number 1 and number 2)

When I played the 1080/50p clip into the ATEM Mini Pro I got what was expected - alternate frames of 1 and 2 (i.e. 1 and 2 flashing at 25Hz) on the HDMI 1080/50p output.

When I played the 1080/50i clip, I just got the 1 displayed and never saw the 2...

This looks to prove that the ATEM Mini Pro is simply discarding one of the two 540 line fields if fed an interlaced 1080i signal, scaling up the single retained 540 line field to 1080 lines and repeating it to get it to output the 50fps if needed. (I guess the ATEM is really designed for 30fps and below streaming?)

Presumably this also means that 1080/25psf outputs from cameras that output 1080/25psf (i.e. 1080/25p in a 1080/50i signal) rather than native 1080/25p will also be displayed in half-resolution - as just one field will be used, rather than both fields used to recreate the original 1080/25p resolution?

Very interesting! Thanks for taking the time to do this test!