• Thanks for visiting the Kaleidescape Owners' Forum

    This forum is for the community of Kaleidescape owners, and others interested in learning about the system, equipment, services, and the company itself.

    It is run by a group of enthusiastic Kaleidescape owners and dealers purely as a service to this community.

    This board is not affiliated in any way with Kaleidescape, Inc.
    For official technical support, product information, or customer service, please visit www.kaleidescape.com

    You can dismiss this box forever by clicking the "X" in the upper right corner of this message.
  • You are currently in "Guest" mode and not logged in with a registered account.

    The forum is free to use and most of the forum can be used by guests who are not registered....

    ... but we strongly encourage you to register for a full account. There is no cost to register for a full account.

    Benefits of registering for a full account:

    • Participate in the discussions! You must have a registered account to make posts on the forums. You will be able to start your own thread on a topic or question, or you can reply to other threads/discussions.
    • Use the "Conversation" feature (known as "private messaging" on other forums) to communicate directly with any of the other users here.
    • Access the Files area. The "resources" area of the forum contains many "Favorite Scene" and Script files that can dramatically increase the enjoyment of your Kaleidescape system. Go directly to great scenes in your favorite movies, created by other owners, and add automation to playback of your system with Scripts.
    • You won't see this annoying notice at the top of every screen!😊

    It's easy and free to register for the forum. Just click the "Register" button in the upper right corner of this page, and follow the instructions there.

    This box goes away for registered users.

190" MicroLED Wall Build

We'll all need to trek to your place once you get a Strato-K to see the difference it makes ;-)
 
@madolfsson

Wonderfully documented! Thank you for sharing!

I was wondering, if you thought about buying spare whole panels as well. At my workplace, we have some sort of similar video wall screen systems that are used as digital set pieces. I remember, when I was on a tour of the studio, that they kept a number of hot spares running so that they could swap in a panel that would have matching characteristics as the other panels that had been running for X number of hours.

I don’t know enough to know if this was simply because they need to have a fast replacement that is easy, but was wondering why you didn’t seem to have any spare units prepurchased in case of any sort of failure.

Having spare units running along side the active ones means the you can swap a failed panel without worrying about calibrating a new one to match the ones that have been running. During your research, did you find out how this is handled in commercial systems and decide not to go for full spares?

Cheers and take care,

-Jonathon
 
Thanks for the great feedback.

The system did come with quite a few spare parts, although not complete cabinet panels.

- 12 spare "panels"
- 2 spare cabinet "mother boards"
- 3 spare power adapters
- 3 spare receiving cards

We will see if that is enough... If enough fails at the some point in the distance future I could just scale down the screen to 7 wide 6 tall cinemascope style and have 7 spare cabinets of parts.
 
I believe the spares will look different than the ones you had in operation. Just using the factory calibration data will not be sufficient.
 
Are you planning on getting a Strato K and documenting the experience on this wall?
 
I believe the spares will look different than the ones you had in operation. Just using the factory calibration data will not be sufficient.
I damaged five tiles during installation and used spares in those locations. After syncing the calibration data they are indistinguishable from the surrounding panels. I don't quite understand the calibration process - I think the data is stored in flash on the receiver card but also has to be read into the MX2000 Pro.
 
I damaged five tiles during installation and used spares in those locations. After syncing the calibration data they are indistinguishable from the surrounding panels. I don't quite understand the calibration process - I think the data is stored in flash on the receiver card but also has to be read into the MX2000 Pro.
Yes, brand new. But over time, the spare tiles will look different than the ones that have been in operation for years.
Nothing that can´t be solved with re-calibration, but it won´t be as easy than just replacing them.
 
HDR UPDATE: ADDING THE madVR ENVY CORE

IMG_5001.jpeg

In my original post, I spent a lot of time explaining how to make HDR look the way I wanted on this wall. The NovaStar accepts HDR10 and HLG, handles the PQ signal correctly, and gives you several controls for shaping how that signal is mapped to the wall. But by default, the HDR picture is too dark for daytime viewing.

My original workaround was to use ST2086 linear mapping and manually lower the MaxCLL value. On a wall calibrated to roughly 1070 nits, telling the NovaStar that the source peak was something like 300 nits effectively scaled the image upward and gave me the punch I was looking for. It worked surprisingly well, but the tradeoff was exactly what I described in the original post: the lower I pushed MaxCLL, the brighter the overall image became, but the more highlight information I sacrificed above that point. In other words, brightness and clipping were tied together.

That was always the compromise.

The NovaStar also has additional tools that I didn’t explore deeply enough at the time, including Clip Level and Shadow Compensation. Clip Level in particular can soften the transition near the top of the PQ range rather than simply hard-clipping it, so I probably could have gotten somewhat more sophisticated results than my original MaxCLL-only approach.

But all of those controls still share the same fundamental limitation: they’re static.

You choose a curve and a set of values, and those settings are applied to everything. The controller isn’t looking at the actual content and changing its mapping from scene to scene. That’s where the Envy changes things.

The chain is now:

Sources -> HDFury VRROOM -> madVR Envy Core -> NovaStar MX2000 Pro -> LED wall

The Envy now receives the HDR signal, performs the tone mapping, and outputs SDR to the NovaStar. Instead of me choosing one MaxCLL compromise that has to work for an entire movie, the Envy analyzes the actual video as it plays and continuously adjusts the mapping to fit the display target. With my old NovaStar-only approach, I was effectively saying:

“Make everything brighter by this amount, and I’m willing to lose whatever falls above this point.”

With the Envy, I’m instead telling the processor what kind of display target I want, and it dynamically works out how to fit each scene into it. A scene with very bright highlights can be compressed differently from a dark interior scene. I no longer have to choose one static point that inevitably works better for some scenes than others.

It also solves one of the biggest annoyances I mentioned in the original post: HDR10 no longer needs its own per-title MaxCLL adjustment. Previously I had presets like 100, 200, 300, 400, 500 and Off because different HDR10 titles could respond very differently to the same override. One movie would look great at 300, another would clip badly, and another might still look too dim. That’s gone.

The Envy analyzes the picture itself rather than depending on me to pick a static override based on the title’s metadata.

The same applies to my LLDV setup. Previously the HDFury could have the source perform dynamic Dolby Vision tone mapping first, but I was then applying my own static brightness boost at the NovaStar afterward. That got me the brightness I wanted, but it also partially defeated the point of the upstream dynamic mapping because I could still clip the result with my MaxCLL override.

The Envy replaces that two-stage compromise with one dynamic mapping process. Instead of choosing between midtone brightness and highlight preservation with one static setting, I can now push the overall presentation toward the brighter look I prefer while letting the processor continuously decide how to handle the actual highlight information in each scene.

That distinction becomes especially important because of what I did next with automatic brightness.

PSEUDO Dolby Vision IQ or HDR10+ Adaptive AUTOMATIC BRIGHTNESS

Once the Envy was in place, I realized I could finally automate brightness in a way that made sense for both HDR and SDR. There are two separate brightness controls in the system, and they do very different jobs.

For HDR, the main control is hdrNits on the Envy. This is the tone-mapping target, and in my setup lower values produce a brighter, punchier image. Around 1000 nits is my reference setting; around 100 nits is much brighter for daytime viewing.

For SDR, hdrNits does nothing, so the only meaningful control is the actual LED drive level on the NovaStar, which I can change dynamically over Art-Net.

That gave me enough control to build an automatic mode based on room brightness.

An Aqara FP2 in the room reports ambient light in lux to Home Assistant. I run that through a 10-minute rolling average, then map the result to an Envy hdrNits target using a simple piecewise-linear curve:

0 lux = 1000 hdrNits
10 lux = 1000 hdrNits
50 lux = 700 hdrNits
150 lux = 400 hdrNits
300 lux = 180 hdrNits
500+ lux = 100 hdrNits

So in a dark room, the Envy stays close to my reference setting. As the room gets brighter, the hdrNits target drops and the HDR image becomes progressively brighter to compensate for the ambient light. Everything between those points is interpolated and then rounded to 25-nit steps. Dampening logic ensures changes are smooth.

There is a separate rule for SDR. Since the Envy’s hdrNits setting has no effect on SDR, SDR content in a dim room drops the NovaStar LED drive to 50%. HDR content, or SDR viewed in brighter ambient light, runs the wall at 100%.

The result is that I basically stopped touching the brightness control. At night, the wall gradually settles back toward my reference HDR setting and lower SDR LED output. During the day, opening the blinds causes the system to progressively increase the perceived brightness so the image still holds up against the room.

Conceptually it is similar to what Dolby Vision IQ or HDR10+ Adaptive does on a television, except the ambient-light logic is running in Home Assistant, the tone mapper is external, and I control the curve myself.

THE NEW REMOTE

IMG_5002.jpeg

The other big change is the remote. I’m now using a Sanytron Astrion HA100, a small Android-based touchscreen remote that integrates with Home Assistant.

I liked the hardware, but the stock software was too limiting for what I wanted to do. The interface was restricted to a fixed set of card types, and the physical buttons were particularly frustrating: they could either behave as global navigation controls or perform page-specific actions, but not both. So I replaced the stock remote software with a Kotlin/Compose app. It’s open source here:

github.com/marcusadolfsson/astrion-custom

The home screen gives me direct access to the things I use most: activity selection, the Dim / Normal / Bright / Auto brightness modes described above, shades, climate, and the rest of the room controls. The VOICE button is probably my favorite part. Holding it streams audio from the remote’s microphone to Home Assistant, which decides where to send it based on the current activity.

For Apple TV, the audio is routed to Siri, so the remote effectively regains the Siri functionality I lost when I stopped using the Apple TV remote.

For Kaleidescape, the same button becomes a voice movie search. I can say the name of a movie, ask for movies by a particular director, or even describe a movie I can’t remember the name of. Home Assistant searches my actual Kaleidescape library and can start the matching title.

Being able to say something like “the one where the president fights terrorists on a plane” and have Air Force One start playing is probably my favorite trick in the house.


I also added a volume/mute OSD directly to the remote. Pressing a volume button gives immediate visual feedback in your hand rather than requiring an overlay on a 190-inch screen.

The end result is that the Astrion hardware is basically just the front end now. Home Assistant provides the state and configuration, while the custom app handles the UI, physical buttons, microphone, and feedback. That gives me one remote that can control the entire room without requiring anyone using it to know what’s happening behind the scenes.
 
HDR UPDATE: ADDING THE madVR ENVY CORE

View attachment 11294

In my original post, I spent a lot of time explaining how to make HDR look the way I wanted on this wall. The NovaStar accepts HDR10 and HLG, handles the PQ signal correctly, and gives you several controls for shaping how that signal is mapped to the wall. But by default, the HDR picture is too dark for daytime viewing.

My original workaround was to use ST2086 linear mapping and manually lower the MaxCLL value. On a wall calibrated to roughly 1070 nits, telling the NovaStar that the source peak was something like 300 nits effectively scaled the image upward and gave me the punch I was looking for. It worked surprisingly well, but the tradeoff was exactly what I described in the original post: the lower I pushed MaxCLL, the brighter the overall image became, but the more highlight information I sacrificed above that point. In other words, brightness and clipping were tied together.

That was always the compromise.

The NovaStar also has additional tools that I didn’t explore deeply enough at the time, including Clip Level and Shadow Compensation. Clip Level in particular can soften the transition near the top of the PQ range rather than simply hard-clipping it, so I probably could have gotten somewhat more sophisticated results than my original MaxCLL-only approach.

But all of those controls still share the same fundamental limitation: they’re static.

You choose a curve and a set of values, and those settings are applied to everything. The controller isn’t looking at the actual content and changing its mapping from scene to scene. That’s where the Envy changes things.

The chain is now:

Sources -> HDFury VRROOM -> madVR Envy Core -> NovaStar MX2000 Pro -> LED wall

The Envy now receives the HDR signal, performs the tone mapping, and outputs SDR to the NovaStar. Instead of me choosing one MaxCLL compromise that has to work for an entire movie, the Envy analyzes the actual video as it plays and continuously adjusts the mapping to fit the display target. With my old NovaStar-only approach, I was effectively saying:

“Make everything brighter by this amount, and I’m willing to lose whatever falls above this point.”

With the Envy, I’m instead telling the processor what kind of display target I want, and it dynamically works out how to fit each scene into it. A scene with very bright highlights can be compressed differently from a dark interior scene. I no longer have to choose one static point that inevitably works better for some scenes than others.

It also solves one of the biggest annoyances I mentioned in the original post: HDR10 no longer needs its own per-title MaxCLL adjustment. Previously I had presets like 100, 200, 300, 400, 500 and Off because different HDR10 titles could respond very differently to the same override. One movie would look great at 300, another would clip badly, and another might still look too dim. That’s gone.

The Envy analyzes the picture itself rather than depending on me to pick a static override based on the title’s metadata.

The same applies to my LLDV setup. Previously the HDFury could have the source perform dynamic Dolby Vision tone mapping first, but I was then applying my own static brightness boost at the NovaStar afterward. That got me the brightness I wanted, but it also partially defeated the point of the upstream dynamic mapping because I could still clip the result with my MaxCLL override.

The Envy replaces that two-stage compromise with one dynamic mapping process. Instead of choosing between midtone brightness and highlight preservation with one static setting, I can now push the overall presentation toward the brighter look I prefer while letting the processor continuously decide how to handle the actual highlight information in each scene.

That distinction becomes especially important because of what I did next with automatic brightness.

PSEUDO Dolby Vision IQ or HDR10+ Adaptive AUTOMATIC BRIGHTNESS

Once the Envy was in place, I realized I could finally automate brightness in a way that made sense for both HDR and SDR. There are two separate brightness controls in the system, and they do very different jobs.

For HDR, the main control is hdrNits on the Envy. This is the tone-mapping target, and in my setup lower values produce a brighter, punchier image. Around 1000 nits is my reference setting; around 100 nits is much brighter for daytime viewing.

For SDR, hdrNits does nothing, so the only meaningful control is the actual LED drive level on the NovaStar, which I can change dynamically over Art-Net.

That gave me enough control to build an automatic mode based on room brightness.

An Aqara FP2 in the room reports ambient light in lux to Home Assistant. I run that through a 10-minute rolling average, then map the result to an Envy hdrNits target using a simple piecewise-linear curve:

0 lux = 1000 hdrNits
10 lux = 1000 hdrNits
50 lux = 700 hdrNits
150 lux = 400 hdrNits
300 lux = 180 hdrNits
500+ lux = 100 hdrNits

So in a dark room, the Envy stays close to my reference setting. As the room gets brighter, the hdrNits target drops and the HDR image becomes progressively brighter to compensate for the ambient light. Everything between those points is interpolated and then rounded to 25-nit steps. Dampening logic ensures changes are smooth.

There is a separate rule for SDR. Since the Envy’s hdrNits setting has no effect on SDR, SDR content in a dim room drops the NovaStar LED drive to 50%. HDR content, or SDR viewed in brighter ambient light, runs the wall at 100%.

The result is that I basically stopped touching the brightness control. At night, the wall gradually settles back toward my reference HDR setting and lower SDR LED output. During the day, opening the blinds causes the system to progressively increase the perceived brightness so the image still holds up against the room.

Conceptually it is similar to what Dolby Vision IQ or HDR10+ Adaptive does on a television, except the ambient-light logic is running in Home Assistant, the tone mapper is external, and I control the curve myself.

THE NEW REMOTE

View attachment 11295

The other big change is the remote. I’m now using a Sanytron Astrion HA100, a small Android-based touchscreen remote that integrates with Home Assistant.

I liked the hardware, but the stock software was too limiting for what I wanted to do. The interface was restricted to a fixed set of card types, and the physical buttons were particularly frustrating: they could either behave as global navigation controls or perform page-specific actions, but not both. So I replaced the stock remote software with a Kotlin/Compose app. It’s open source here:

github.com/marcusadolfsson/astrion-custom

The home screen gives me direct access to the things I use most: activity selection, the Dim / Normal / Bright / Auto brightness modes described above, shades, climate, and the rest of the room controls. The VOICE button is probably my favorite part. Holding it streams audio from the remote’s microphone to Home Assistant, which decides where to send it based on the current activity.

For Apple TV, the audio is routed to Siri, so the remote effectively regains the Siri functionality I lost when I stopped using the Apple TV remote.

For Kaleidescape, the same button becomes a voice movie search. I can say the name of a movie, ask for movies by a particular director, or even describe a movie I can’t remember the name of. Home Assistant searches my actual Kaleidescape library and can start the matching title.

Being able to say something like “the one where the president fights terrorists on a plane” and have Air Force One start playing is probably my favorite trick in the house.


I also added a volume/mute OSD directly to the remote. Pressing a volume button gives immediate visual feedback in your hand rather than requiring an overlay on a 190-inch screen.

The end result is that the Astrion hardware is basically just the front end now. Home Assistant provides the state and configuration, while the custom app handles the UI, physical buttons, microphone, and feedback. That gives me one remote that can control the entire room without requiring anyone using it to know what’s happening behind the scenes.

That is an amazing setup. I think I am going to order that same remote. Is there a good place to learn more about the custom setup you did on it?
 
Well done and great write up. Someone told me that videowalls were the next big thing and while I'm sure they will leak into the consumer market, I think they are a long way off from wide adoption by a few years. Your experience sounds fantastic but not many of us can drop that kind of coin on a display device.
 
Huge respect to @madolfsson for documenting this build so meticulously.

My own journey into this madness actually started right here on this thread. My friend and AV integrator Michael Boeker (@HighImpactAV) forwarded me Marcus’s post a while back. I read it front to back, had my mind blown, and reached out to Marcus directly. We had an awesome conversation, he even jumped on a video call to walk me through his space, show off the wall, and talk through the realities of the build. That conversation gave me the itch, and I decided to dive in headfirst.

Naturally, I started down the exact same path: I called Canbest first. But as I started digging into fine-pitch COB specs, it quickly turned into an obsession. I wanted to understand the absolute ceiling of the technology, what the best of the best looked like, what silicon made it tick, and what it would actually cost to build.

That led to about a month of living on Shenzhen time. I lost count after talking to probably 50+ sales reps on WhatsApp at 3:00 AM, wading through endless claims of "we are the direct factory."

That rabbit hole taught me some fast, hard lessons about the overseas supply chain:

  1. "Factories" vs. Trading Companies: A huge percentage well over 80–90% of the vendors pitching direct on WhatsApp are actually just trading companies. They don't own cleanrooms, SMT lines, or optical calibration equipment. They are middlemen sitting in office buildings who just take an order, place it with a real ODM, slap a markup on it, and ship it out. When something fails or calibration doesn't sync, they have zero technical ability to support you.
  2. Going Straight to the Source: It took weeks of filtering out the noise until I managed to get onto Zoom calls with actual NovaStar engineers and true Tier-1 ODMs. That’s where I learned what actually separates a passable commercial board from a true residential reference canvas things like custom Bill of Materials (BOM) locking, pairing specialized driver ICs running at relaxed 1/30 scan rates to eliminate dark-scene banding and unlock true grayscale depth, and identical-batch spare vaults.
The biggest shock came when I started cross-referencing what true pinnacle hardware actually costs against what standard domestic luxury distributors are charging.

When I looked at equivalent spec systems stateside from domestic brands, quotes were coming in north of $130,000 to $160,000+ for the hardware alone. When you run the actual factory BOM math, it becomes very clear how much sheer distributor padding is baked into the market.

That didn't sit right with me.

Like Marcus pointed out, buying direct as an individual is totally viable if you have the engineering stomach for it, but you are the integrator. You have to design the sub-wall, route dozens of network home runs, figure out commercial EDID/audio extraction, calculate power, and pray your factory calibration coefficients don't get corrupted in transit. For most enthusiasts, that level of friction is a massive wall.

Going through that entire ringer—unraveling the supply chain, building relationships directly with genuine ODMs and NovaStar engineering, and seeing the massive price gap in the domestic market—convinced me to jump in full-time to bridge that gap. We shouldn't have to choose between a 300% markup or spending six months acting as an unpaid electrical contractor and logistics broker.

Marcus, your madVR Envy update and the ambient-lux Home Assistant curve are brilliant. You built a world-class system and paved the road for a lot of us. Honored to have compared notes with you!
 
Huge respect to @madolfsson for documenting this build so meticulously.

My own journey into this madness actually started right here on this thread. My friend and AV integrator Michael Boeker (@HighImpactAV) forwarded me Marcus’s post a while back. I read it front to back, had my mind blown, and reached out to Marcus directly. We had an awesome conversation, he even jumped on a video call to walk me through his space, show off the wall, and talk through the realities of the build. That conversation gave me the itch, and I decided to dive in headfirst.

Naturally, I started down the exact same path: I called Canbest first. But as I started digging into fine-pitch COB specs, it quickly turned into an obsession. I wanted to understand the absolute ceiling of the technology, what the best of the best looked like, what silicon made it tick, and what it would actually cost to build.

That led to about a month of living on Shenzhen time. I lost count after talking to probably 50+ sales reps on WhatsApp at 3:00 AM, wading through endless claims of "we are the direct factory."

That rabbit hole taught me some fast, hard lessons about the overseas supply chain:

  1. "Factories" vs. Trading Companies: A huge percentage well over 80–90% of the vendors pitching direct on WhatsApp are actually just trading companies. They don't own cleanrooms, SMT lines, or optical calibration equipment. They are middlemen sitting in office buildings who just take an order, place it with a real ODM, slap a markup on it, and ship it out. When something fails or calibration doesn't sync, they have zero technical ability to support you.
  2. Going Straight to the Source: It took weeks of filtering out the noise until I managed to get onto Zoom calls with actual NovaStar engineers and true Tier-1 ODMs. That’s where I learned what actually separates a passable commercial board from a true residential reference canvas things like custom Bill of Materials (BOM) locking, pairing specialized driver ICs running at relaxed 1/30 scan rates to eliminate dark-scene banding and unlock true grayscale depth, and identical-batch spare vaults.
The biggest shock came when I started cross-referencing what true pinnacle hardware actually costs against what standard domestic luxury distributors are charging.

When I looked at equivalent spec systems stateside from domestic brands, quotes were coming in north of $130,000 to $160,000+ for the hardware alone. When you run the actual factory BOM math, it becomes very clear how much sheer distributor padding is baked into the market.

That didn't sit right with me.

Like Marcus pointed out, buying direct as an individual is totally viable if you have the engineering stomach for it, but you are the integrator. You have to design the sub-wall, route dozens of network home runs, figure out commercial EDID/audio extraction, calculate power, and pray your factory calibration coefficients don't get corrupted in transit. For most enthusiasts, that level of friction is a massive wall.

Going through that entire ringer—unraveling the supply chain, building relationships directly with genuine ODMs and NovaStar engineering, and seeing the massive price gap in the domestic market—convinced me to jump in full-time to bridge that gap. We shouldn't have to choose between a 300% markup or spending six months acting as an unpaid electrical contractor and logistics broker.

Marcus, your madVR Envy update and the ambient-lux Home Assistant curve are brilliant. You built a world-class system and paved the road for a lot of us. Honored to have compared notes with you!
Wow! Very cool! So, the $64,000 question. Are you done? Did you build it? How much was it?
 
Haha, that really is the $64k question (or in this market, the $150k+ question)!

The timing of your question is uncanny—I literally just received these photos and videos from the factory floor showing the completed wall running through its mandatory 72-hour live burn-in and calibration aging test! (attached below)



To answer your questions directly:

Are you done? Did you build it?

We are right at the finish line of the factory build. After spending time comparing notes with Marcus and dissecting the BOM with the factory engineers, I locked in a custom 205" setup. In the photos and clips, you can see the fully assembled canvas undergoing point-by-point calibration and soak testing on the factory rig before crating. We also verified the silicon architecture directly on the receiving cards (spec’ing custom TB Silicon drivers running at a relaxed 1/30 scan rate to lock in deep shadow detail without banding). Once the aging test clears, the crates ship out to Texas, and we’re staging the space for on-site commissioning with factory engineers.

Speaking of the source chain: my older Strato S recently went out, so a huge shoutout to Michael Boeker (@HighImpactAV) for jumping on it and getting me set up with the new Kaleidescape Strato K paired with a 30TB Terra server so quickly! The difference with uncompressed 4K cinematic delivery on the Strato K is unbelievable—to my eyes, the jump in clarity and detail felt just like going from 1080p to regular 4K all over again. Feeding that pristine, high-bitrate signal chain straight into an uncompressed MicroLED processor is going to be something else; I can’t wait to fire it up on the 205" screen once it's hung.

How much was it?

For the raw hardware, custom mounting frame, NovaStar MX2000 Pro processing, extra same-batch spare modules, and international freight direct from the ODM, my landed build cost is sitting right around ~$56,000 all-in.

That number is the exact catalyst for why this evolved from a personal passion project into launching Eclipse MicroLED.

When I shopped the exact same tier of fine-pitch display through traditional domestic residential brands and distributors here in the US, quotes were coming in between $130,000 and $160,000+ for the hardware alone. Once I peeled back the supply chain layers, I realized domestic clients were frequently paying upwards of $70,000 to $100,000 purely in distributor markups and middleman padding.

Marcus proved that buying direct as an individual is totally doable if you have the engineering grit for it, but you are the integrator. Navigating overseas time zones, language barriers, power distribution, audio workarounds, and calibration sync is a massive barrier for most people.

Our model with Eclipse down in Texas is simple: eliminate the 200–300% middleman markup, lock down true tier-1 silicon and custom BOMs directly with the manufacturers, and handle the local logistics and on-site factory commissioning so clients can get a reference MicroLED canvas at fair value without having to spend months acting as their own general contractor.

I’ll share more footage and photos once the crates arrive in Texas and the wall goes up on the home structure!
NOTE: i have videos to share but they wont upload, not sure why.
 

Attachments

  • 9e4570e6-58dd-4d4a-aff0-364f93264e4f.jpg
    9e4570e6-58dd-4d4a-aff0-364f93264e4f.jpg
    28.6 KB · Views: 17
  • 811bba23-6684-4f7b-8fbb-4aaf8e6134a9.jpg
    811bba23-6684-4f7b-8fbb-4aaf8e6134a9.jpg
    64.8 KB · Views: 17
  • 545e35b7-a58b-4c4d-9338-389c04a54b53.jpg
    545e35b7-a58b-4c4d-9338-389c04a54b53.jpg
    87.7 KB · Views: 17
Huge respect to @madolfsson for documenting this build so meticulously.

My own journey into this madness actually started right here on this thread. My friend and AV integrator Michael Boeker (@HighImpactAV) forwarded me Marcus’s post a while back. I read it front to back, had my mind blown, and reached out to Marcus directly. We had an awesome conversation, he even jumped on a video call to walk me through his space, show off the wall, and talk through the realities of the build. That conversation gave me the itch, and I decided to dive in headfirst.

Naturally, I started down the exact same path: I called Canbest first. But as I started digging into fine-pitch COB specs, it quickly turned into an obsession. I wanted to understand the absolute ceiling of the technology, what the best of the best looked like, what silicon made it tick, and what it would actually cost to build.

That led to about a month of living on Shenzhen time. I lost count after talking to probably 50+ sales reps on WhatsApp at 3:00 AM, wading through endless claims of "we are the direct factory."

That rabbit hole taught me some fast, hard lessons about the overseas supply chain:

  1. "Factories" vs. Trading Companies: A huge percentage well over 80–90% of the vendors pitching direct on WhatsApp are actually just trading companies. They don't own cleanrooms, SMT lines, or optical calibration equipment. They are middlemen sitting in office buildings who just take an order, place it with a real ODM, slap a markup on it, and ship it out. When something fails or calibration doesn't sync, they have zero technical ability to support you.
  2. Going Straight to the Source: It took weeks of filtering out the noise until I managed to get onto Zoom calls with actual NovaStar engineers and true Tier-1 ODMs. That’s where I learned what actually separates a passable commercial board from a true residential reference canvas things like custom Bill of Materials (BOM) locking, pairing specialized driver ICs running at relaxed 1/30 scan rates to eliminate dark-scene banding and unlock true grayscale depth, and identical-batch spare vaults.
The biggest shock came when I started cross-referencing what true pinnacle hardware actually costs against what standard domestic luxury distributors are charging.

When I looked at equivalent spec systems stateside from domestic brands, quotes were coming in north of $130,000 to $160,000+ for the hardware alone. When you run the actual factory BOM math, it becomes very clear how much sheer distributor padding is baked into the market.

That didn't sit right with me.

Like Marcus pointed out, buying direct as an individual is totally viable if you have the engineering stomach for it, but you are the integrator. You have to design the sub-wall, route dozens of network home runs, figure out commercial EDID/audio extraction, calculate power, and pray your factory calibration coefficients don't get corrupted in transit. For most enthusiasts, that level of friction is a massive wall.

Going through that entire ringer—unraveling the supply chain, building relationships directly with genuine ODMs and NovaStar engineering, and seeing the massive price gap in the domestic market—convinced me to jump in full-time to bridge that gap. We shouldn't have to choose between a 300% markup or spending six months acting as an unpaid electrical contractor and logistics broker.

Marcus, your madVR Envy update and the ambient-lux Home Assistant curve are brilliant. You built a world-class system and paved the road for a lot of us. Honored to have compared notes with you!
How is that different from what JVC and Sony do? You're still paying an enormous premium for technical expertise and support from your dealer/integrator and then next level support from the distributor and vendor. If you don't want that exposure then consider a Barco wall.
 
That’s a completely fair question, but there’s a fundamental difference between traditional consumer hardware distribution and the current directview MicroLED market.

With companies like Sony, JVC, or Christie, you’re buying an engineered, single-chassis finished product designed specifically for residential consumer use. There's real R&D in their proprietary optical blocks, consumer friendly menus, and established dealer margins that support local integrators. Nobody faults a dealer for making a fair margin on technical expertise and local support.

Direct-view fine-pitch LED in residential, however, isn't a plugand play projector.

Because this technology originated in commercial broadcast and virtual production, what frequently happens in the US residential channel is multi-tiered relabeling. A domestic brand will source fine-pitch COB cabinets from an overseas ODM, slap their badge on the frame, package it with off the shelf NovaStar or Megmeet processing, and mark it up 200% to 300% at the distributor level before the local integrator even adds their labor and margin.

That’s how a canvas with a landed BOM of $50k–$60k ends up with a $150k–$180k retail invoice for the exact same underlying silicon and receiving cards.

You mentioned Barco. Barco’s TruePIX and high-end processing are extraordinary, but a Barco wall at this 190"–205" scale will easily run $300,000 to $400,000+. For clients where budget is irrelevant, that’s a turnkey option.

But for cinema enthusiasts who want reference-grade MicroLED performance, there is a massive chasm between a $350k Barco setup and a grueling, high-friction DIY direct import where you're left to troubleshoot commercial EDID tables, calibration coefficients, and power distribution on your own.

We aren't arguing against the value of integrator expertise or support, because that's critical. The goal is simply engineering direct with tier-1 ODMs, locking in the custom BOMs and silicon upfront, and providing on-site commissioning and warranty coverage without forcing the client to pay an extra $100k purely in distributor markups.
The bottom line is simple: we are delivering the exact same tier-1 silicon, optical resin COB glass, and NovaStar processing architecture that domestic luxury brands sell—except their business model relies on a 200% to 300% distributor markup, whereas ours caps margins well below the raw hardware cost. You get the exact same reference display, but without paying a $100,000 middleman tax to get it. We aren't cutting corners on technical expertise, either: we fly factory engineers on-site for laser alignment, maintain a local Texas spare parts vault, and back the commissioning so the client gets the full reference experience at true fair value.
 
Back
Top