Mobile stream box BOYS!!! A must-read ^^

Joined
Sep 24, 2021
Messages
52
Let me know mods if there is a better section for this.

So, I am looking into possibly building a mobile stream box. My goal is to build a box that can handle this:

Stream Box Requirements
*Support portable monitor via DP/USB/HDMI (DP preferred I think. It has latches so cable less likely to come out)
WIFI would be best but I don't know if a product like that exists yet. If you know of a portable wifi monitor I am all ears.
*Adlar Lake lake or 16 core ryzen
*USB PD or DC power
*enough power to support 4x1080P streams to youtube/odessy. The goal is to make the 2-4x1080p GoPro Hero10 streams into a single 4K stream, which I believe is possible.
*will probably use 4G/5G tethering for data
*up to 100w/h draw for power limit

I plan on building a 400-600 W/h battery pack probably using 21700 cells so giving myself around 4-8 hours of stream time. I will most likely use LiitoKala cells because they are fantastic and quite affordable. I have their 18650s and 26650s and they are always slightly better than their rating and also have solid reviews from trusted sources.
21700 3.7v 5000mah same energy density as 18650 and ~$4.00 per cell including shipping. So 30 cells will be +/-555 w/h capacity for ~$120. An amazing deal. Hit me up if you know better~
https://www.aliexpress.com/item/32998232055.html?spm=a2g0o.productlist.0.0.78904abaym5WYY&algo_pvid=de3a480f-9db7-4967-b6ee-2e38d6ae92ff&algo_exp_id=de3a480f-9db7-4967-b6ee-2e38d6ae92ff-9&pdp_ext_f={"sku_id":"66958238438"}&pdp_npi=1@dis|USD||4.32|||||@0b0a556b16509873015136659e1e2c|66958238438|sea

Currently, the only 2 options I have found are an ASROCK tiny box (what are these things called), which looks promising unfortunately no 1280P the best is 1260P or simply building a mini ITX while using something from this site to convert DC to DC for regular ITX MB connectors.

What do you gents and ladies think of these 2 options? Do you have any other ideas?

I plan on doing extensive testing and tweaking of any CPU to optimize it like underclocking (forcing lower frequencies). I personally have done extensive testing in the past with IB with finding optimal frequencies for CPUs so I know I can squeeze a ton of performance per watt out of any CPU. If you take the time to test each frequency to find the optimal GHz/Mhz for the CPU you can cap the max frequency of the entire CPU. Since streaming is really well threaded you can find the perfect frequency and make that the max frequency of the CPU and then do extra tuning like lowering the voltage as low as possible where its still stable. This will allow you to get very good performance per watt vs just limiting the total power limit like what techpowerup did in this test. You can still see though even with just limiting the power limit efficiency still went up a tangible amount but you can still do substantially better with further tweaking. With a 12900K:
50w=51.9% of total performance
75w=65.4%
100w=76.4%
So you can see that my 100w max can actually get me near 100% performance of a 1260P 4p/8e and 76.4% of a 12900K with proper tweaking you can get substantially more efficiency. So ideally I would like a Raptor Lake with 4-6p/16e cores or a 5950x properly tweaked but I am not sure what's my best option/what's available and could use advice.
1650988950550.png

Here is a chart from my unpublished work with IB. My goal was to test and compare IB vs HW with FIVR to show how FIVR worked but health issues got in the way at the time, but I plan on finally doing that study over the next year and including testing ADL against future RL with the "ghetto FIVR" (I forgot the name of it maybe LIVR). So I'll redo this when all the CPUs are out and I can test them all at once. The ideal sweet spot depends on the application, some will run more efficiently +/- .xGHz. This was just a preliminary test I ran many years ago and I was planning on running 1c,2c,3c,4 to show how with and without FIVR worked but in this situation, I plan on running all cores at once with streaming and this chart shows what I am talking about albeit on a different CPU but the concept applies to all CPUs. You can see with Prime95 the most efficient was ~2000 GHz. IB clocked poorly and CPU process and architecture technology are a lot better now so I am curious to see if maybe ADL or Ryzen sweet spot might be more like 2.5GHz or maybe 3GHz vs ~2GHz on IB. The test below was also just using Intel stock power/voltage settings. Efficiency can be considerably improved by tweaking those settings. At least in the past Intels power management has been over-aggressive and not very good in regards to efficiency. Setting a max CPU frequency is much more efficient than just setting a wattage power limit. On top of that, you can still tweak voltages to get even more efficiency. If you set ~21/22w power limit like TechPowerUP did. You will actually get lower performance vs setting max CPU frequency because Intel power management will apply much higher voltages and do weird stuff with frequency and end up with lower performance vs just setting max frequency from my experience. Plus my testing method shows much more data than TPUs method but mine is much more difficult to test. Granted, CPU single thread is more limited in my method but that doesn't really matter for what I am doing and looking for. You might be able to set and let a single core clock high but that could tank efficiency significantly since these don't have FIVR. Every use case is different and that might be a plausible option for some situations.
1650984353949.png

The other option like I said above is going full custom ITX and using something like below so I can use a custom battery pack to power it. Granted, it might be wisest to use one of their products in any build since it has a built-in voltage regulating and conditioning to a degree. As the battery loses capacity and voltage drops the "regulator/adaptor" will adjust the voltage/amps from the battery pack to make sure it's giving a consistent voltage/amps to the device, which is pretty awesome. (assuming I understand this properly). Let me know what you think!
https://www.mini-itx.com/2017/09/05/guide-choosing-the-right-dc-dc-psu

What are everyone's thoughts? Does anyone have recommendations? Has anyone else done this in the past and has a previous build I can see or am I the first person building a battery-powered mobile streaming box to live stream multiple go pros to youtube/Odessey?

The other question is there any program that would allow me to use multiple cellphones in a load-balanced manner to give myself not just more bandwidth but also redundancy.
 
Last edited:
To use multiple cell networks, I would look into a cradle point router as that will just have a nice neat ethernet directly to your rig and will handle all the multi-wan functions. I don't know if peplink makes one of theirs for cellular yet, but they are the other king of multi-wan routing.
 
To use multiple cell networks, I would look into a cradle point router as that will just have a nice neat ethernet directly to your rig and will handle all the multi-wan functions. I don't know if peplink makes one of theirs for cellular yet, but they are the other king of multi-wan routing.
do you have a link or name of a specific product?
555whrs would only get you half an hour powering a 1000w system. Obviously won't be using that much juice, but even a 500w system would only last an hour or so under heavy load.

Also, that power is going to need to be converted from 3.x volts to whatever your computer PSU takes as input. 500w at 12v is 41A -- almost a third of your Ah capacity, before losses.
what are you talking about? I clearly said I would be looking at running an underclocked/undervolt CPU at 100w/h draw max AKA under 100w TDP and provided data (mine showcasing how CPU tweaking works (IB chart) and TPU showing 12900K efficiency and the link ASROCK stream box) showing that the CPU will be limited to 100w/h power draw MAX.The ASROCK box is technically 90w/h or possibly going a 5950x ITX build. Where are your numbers from?

Also, I clearly gave links to a website that provides DC-DC volt converters here is the link again:
https://www.mini-itx.com/2017/09/05/guide-choosing-the-right-dc-dc-psu


I also provided a bunch of embedded links for your ease to click on. Did you not see the half dozen hyperlinks? Did you read this?

Keeping the total system under 150w max is a cake. I'll probably be under 100w/h....just read what I wrote and look at the links and graphs.

I am honestly, so confused about what you typed. Like where is 500w from? I can't even use 500w even if I tried.

I appreciate any and all help but I recommend you re-read this. You might be pretty surprised by what I have found so far especially if you aren't familiar with this stuff. The mini-itx.com is super interesting and something everyone should bookmark.
 
Last edited:
so that mini-itx site has very little useful information on power supplies. What you are really going to want is to have your power source as close to your power needs in voltage. Any time you convert voltage you waste energy in the process. With a buck or boost converter, energy is used to charge the inductor during switching, with a LDO you burn that energy off as heat to get to the voltage you want.

That said you want to change voltage as least amount of times as possible. And in a simple 100% efficiency math you have power in = power out. so if you want 19V (90W) 4.73A. Your battery pack must put out 90W 3.7V 24.3A.

But real world is different, you get voltage drops in connectors, wires, battery temperature increase.

I like Nobu 's suggestion of creating a series/parallel arrangement to get to what you want.

Another idea is use power tool battery packs. Could create the setup you want with off the shelf parts and easily swap out and be up and running again instead of building multiple packs.
 
So, going by the specs of that asrock box for example. It has a 90w/19v power supply. Assuming it was fully utilized and 100% efficient, that'd be 4.74A of current being drawn from the batteries (at 19v, but those batteries aren't 19v batteries). Because you need 5 batteries in series to make that voltage, you'd need 30x5 = 150 batteries to maintain the same capacity, or use a transformer/boost converter to increase the voltage (with some efficiency loss).
Again, as I clearly stated, they are referencing the AC-DC power supply that comes with it. I doubt that's the max power the box can use. Granted, the question is how good is the cooling system? I really don't get how you can't understand that I would be powering this off of a DC battery directly so the power supply that comes with it is irrelevant.

Again, if you actually read my post and my reply with the link to a potential DC-DC adapter (Not sure what the term is for the product they sell). I am still researching it. All you are doing is going on about irrelevant stuff. If you know or understand the product I have referenced now 4 times in this thread or want to help me figure out if its the perfect solution I welcome the help but all you are doing is posting nonsense.

https://www.mini-itx.com/2017/09/05/guide-choosing-the-right-dc-dc-psu
 
so that mini-itx site has very little useful information on power supplies. What you are really going to want is to have your power source as close to your power needs in voltage. Any time you convert voltage you waste energy in the process. With a buck or boost converter, energy is used to charge the inductor during switching, with a LDO you burn that energy off as heat to get to the voltage you want.

That said you want to change voltage as least amount of times as possible. And in a simple 100% efficiency math you have power in = power out. so if you want 19V (90W) 4.73A. Your battery pack must put out 90W 3.7V 24.3A.

But real world is different, you get voltage drops in connectors, wires, battery temperature increase.

I like Nobu 's suggestion of creating a series/parallel arrangement to get to what you want.

Another idea is use power tool battery packs. Could create the setup you want with off the shelf parts and easily swap out and be up and running again instead of building multiple packs.
All you said is obvious and I don't get why you wrote it. Using power tool packs is not ideal and not helpful. That would be huge and cumbersome and have lower-quality batteries and have a terrible weight-to-energy ratio. It would be better to just buy a premade battery pack off aliexpress or Amazon at that point and where is this "multiple packs" coming from? I never said anything about that. I would build a single pack and get my own PCB/BMS or whatever. I could use what that site sold if it works for my needs. I am not sure yet if I need my own PCB/BMS and their product or if their product includes that PCB already. Still working on figuring that out. If you got more helpful solutions I am all ears but I could have gotten that info from a google search or Facebook.....

If you clicked through you would see this page that explains in more detail. The page I shared was a rough overview just in case someone didn't understand the topic at all but there was obviously a link to the PSUs
https://www.mini-itx.com/store/psu
 
Last edited:
This particularly looks promising. I assume I still need a PCB/BMS on the battery. AFAIK, it seems to check off everything I need for either the ASROCK or any other product. If I built a custom ITX build there are some simpler options like ones with actual 20/24 pins. This is pretty neat, especially with the wide input range and the ability to turn off the computer when the battery is low.

https://www.mini-itx.com/store/psu?c=10#DCDC-USB-200

this too is an interesting product and with higher limits.

https://www.mini-itx.com/store/psu?c=10#M4-ATX

These supposedly have 94-95% efficiency but I would budget for 90% to play it safe.
 
Last edited:
I know about those, I've used their picopsu in one of my builds, posted on this forum. That USB-200 won't work with your batteries without putting at least two in series (same with the m4-atx), meaning you would need to double the number of batteries to maintain your stated target runtime. I won't explain again why, since apparently that's not helpful. Have fun, please don't let me spoil it for you -- I'm not trying to discourage you, nor am I trolling. But I don't appreciate your attitude, so I'm done...not helping.

Edit: Well, as I said before, you could use a boost converter and keep the same number of batteries, you'd just lose another 5-10% or so (in addition to the losses from the buck-boost converters on those dc-dc solutions). Okay, that's all the help I give, last bone I'm throwing.

Edit2: okay, I feel bad for deleting all my other posts. There are a couple products on mini-box.com that would work. iirc, a couple of the UPS solutions go down below 3v input. Dunno how well they'd work, though.
 
Last edited:
I know about those, I've used their picopsu in one of my builds, posted on this forum. That USB-200 won't work with your batteries without putting at least two in series, meaning you would need to double the number of batteries to maintain your stated target runtime. I won't explain again why, since apparently that's not helpful. Have fun, please don't let me spoil it for you -- I'm not trying to discourage you, nor am I trolling. But I don't appreciate your attitude, so I'm done...not helping.
Nothing you said explained how that would not work. You rambled about stuff that was elementary and was not relevant to the discussion. I had to post the link to that site 3 times for you to read it and now magically you claim you have used them before? Why not mention that in your first post?

We might have had a cogent discussion but instead, I had to post the link 3 times, which is why I am so irritated/annoyed because you sound like you are trolling and not being genuine. How can I take you seriously when you can't even read/see a link? Please explain that to me. I don't have an attitude. I just have a limited amount of patience and don't like people wasting my time when I spend so much of it trying to be patient. In a previous thread, you couldn't even follow along so there is a pattern here.

Now you delete your previous posts? childish.

So, going by the specs of that asrock box for example. It has a 90w/19v power supply. Assuming it was fully utilized and 100% efficient, that'd be 4.74A of current being drawn from the batteries (at 19v, but those batteries aren't 19v batteries). Because you need 5 batteries in series to make that voltage, you'd need 30x5 = 150 batteries to maintain the same capacity, or use a transformer/boost converter to increase the voltage (with some efficiency loss).
Are you talking about this post? This was partly gibberish, lacking context, and sort not relevant. If I am understanding you right were you talking about using batteries directly? Where the hell are these 150 batteries coming from? That's not what I am doing, which actually does work with laptops....that's how all backup battery packs work but I assume in laptops there are some regulators or whatever that clean/boost the power in some way to make it stable that doesn't exist in a standard computer but unsure. But again that's not what I would be doing like wtf

Since you claimed to have used them before please tell me how a DC-DC PSU with a 6-34v input range will not work with a battery pack of lets say this:

I assume since their peak charge is 4.2v I can't have 9S batteries since that would cause 37.8v, which is above the 34v max.

Battery size 21700 3.7v 5000mah in this config:

Pack 1
8S
3P
24 cells total
3.7v*8=29.6
4.2v*8=33.6
444 Wh

Pack 2
8S
4P
32 Cells
3.7v*8=29.6
4.2v*8=33.6
592 Wh

Please tell me how I would need 150 batteries and how this won't work? I could be wrong but 150 batteries makes no fucking sense. You are the expert though so please tell me. I want to know how I am so far off.

How will this explode :sneaky:
 
Last edited:
The minimum input voltage on those devices is 6v. they will not work at all at 4.6v or even 5.2v.
 
Also, batteries in series don't add to capacity, so the first config is 55.5whr, and the second is 74whr. That's where 150 batteries came from.
 
To be honest I was reading this thread in the wee early hours during an insomniac episode.

I think next time it would help you and the readers if you expounded more on the WHYs of the project versus the HOWs right at the start. There is a lot of information to read here asking for details and opinions, but not a lot of info to arouse interest so you are going to have less responses because TLDR. Sorry, just trying to help.

I'm no streamer, but wouldnt it made sense to transmit the 2-4 feeds to a home PC with more power to then merge/stream? You could go for a much lower power setup that way, but with delays obviously. Ffplay can combine streams with a low power requirement too.

Or do you truly want this machine to be self-sufficient? I would probably just rip up a laptop and add the battery pack to it, but you do you. Laptops are designed for mobility, PC parts not so much.

I'm not going to touch on the battery topic, but you can get portable monitors with their own batteries in them like the AOC 16T2. Touch screen, vesa mount, won't draw from the system unless you choose it to. Other options around too.

Good luck with the project.
 
The minimum input voltage on those devices is 6v. they will not work at all at 4.6v or even 5.2v.
WTF are you talking about? Can you use coherent thoughts? I never even mentioned 4.6v or 5.2v you are making shit up. You clearly didn't read what I wrote and have no idea what you are talking about. You should be ashamed for even posting that nonsense and wasting my time while I have been patient with you.
Also, batteries in series don't add to capacity, so the first config is 55.5whr, and the second is 74whr. That's where 150 batteries came from.
You obviously don't understand how batteries and power work so I am just going to ignore you. Holy crap, that's an insane amount of ignorance or blatant stupidity while pretending to know what you are talking about. Thanks for wasting my time on your idiocy. I am printing this and saving this for later use.

But here is a website explaining the basics just so people can learn and understand that you clearly have no idea what you are talking about.

https://en.wikipedia.org/wiki/Kilowatt-hour
https://www.wikihow.com/Calculate-Kilowatt-Hours
 
Last edited:
To be honest I was reading this thread in the wee early hours during an insomniac episode.

I think next time it would help you and the readers if you expounded more on the WHYs of the project versus the HOWs right at the start. There is a lot of information to read here asking for details and opinions, but not a lot of info to arouse interest so you are going to have less responses because TLDR. Sorry, just trying to help.

I'm no streamer, but wouldnt it made sense to transmit the 2-4 feeds to a home PC with more power to then merge/stream? You could go for a much lower power setup that way, but with delays obviously. Ffplay can combine streams with a low power requirement too.

Or do you truly want this machine to be self-sufficient? I would probably just rip up a laptop and add the battery pack to it, but you do you. Laptops are designed for mobility, PC parts not so much.

I'm not going to touch on the battery topic, but you can get portable monitors with their own batteries in them like the AOC 16T2. Touch screen, vesa mount, won't draw from the system unless you choose it to. Other options around too.

Good luck with the project.
Obviously, bandwidth and data limitations.....Do you not understand how much bandwidth and data it would be to stream 4 unprocessed streams via any connection let alone on 4G/5G in a remote area. The point of a local stream box is to reduce that data into a practical amount that can be sent over 4G/5G since there will be no hard lines. 1Gbps fiber connections are not just magically everywhere. Do they grow on trees? If so awesome. If I had unlimited data and a 1Gbps connection sure I could send all of this data to a stream server unprocessed but that's not practical outside.

It is just clear that this forum does not have competent people that can help me. I will have to find another place for help. This forum has clearly gone downhill over the years.
 
So first off, I want to apologize again. I understand the math, and I know well what all the terms mean, but somewhere my understanding of the application went awry. I did all the math and then my dog ate it. Suffice to say, I kept doing it over and over again, and the numbers seemed to bear out my logic every time...well it was flawed logic based on bad reasoning. Sorry for causing all that frustration -- I was just as frustrated, though I doubt that makes it any better.

Sorry, I really was trying to be helpful.

For some reason I was ignoring that input power equals output power, and I never got to calculating that because I was so sure I was right. Something about arrogance or whatever. Anyway, Watts don't change (ignoring efficiency), current depends on voltage, is the gist. Current should be the reason for wiring batteries in series, not voltage, and I was missing that. Addendum: Capacity does increase with each battery, duh. Otherwise current draw wouldn't drop per cell (in series). I'm so retarded.
 
Last edited:
Watts equals volts times amps lol

Like I said, I'm not a streamer but thanks for the sweet comments, they touched me deeply.

I've ran plenty of 1080p and 4k security camera systems and can stream dozens of feeds simultaneously on a 4G connection to monitor the situation but what do I know. Yes upload is more important in your scenario but the feeds are compressed when viewing a dozen+ at once, my tired old phone can handle them fine over 4G but what the hell do I know. And it can handle the 4k feed just fine. Bit rate and frame rate are important, the latter you didn't specify and the former will affect transfer and quality. 4k 120hz-30hz. You also said 2-4x 1080p feeds too. 3840x1080 at 30hz takes a lot less bandwidth than 3840x2160 120hz irregardless of compression.

Have fun, I hope no lithium cells blow up on you. I have lots of experience with BMS and lithium but you're not very receptive to input so have fun, good luck.
 
Back
Top