economical home storage music production parts list proposal

[Nick2253]

It may be possible to do this all in RAM, but it would probably need hundreds of Gigabytes

Using your numbers: 192,000 samples per second * 24bits/sample * 40 channels = 184 Mbps = 23MB/s. If you have 16GB of memory in your system, and you can dedicate 8GB simply to audio stream buffering, you'd have about 6 minutes of buffer for this application. Considering that a typical hard drives can receive data on the order of 100 MB/s, and a typical SSD can do 3-5x that, you could easily handle latency on the order of minutes! Which makes patently absurd the idea that a few milliseconds one way or another matter for your storage in this day and age.

 

[rogerh]

Using your numbers: 192,000 samples per second * 24bits/sample * 40 channels = 184 Mbps = 23MB/s. If you have 16GB of memory in your system, and you can dedicate 8GB simply to audio stream buffering, you'd have about 6 minutes of buffer for this application. Considering that a typical hard drives can receive data on the order of 100 MB/s, and a typical SSD can do 3-5x that, you could easily handle latency on the order of minutes! Which makes patently absurd the idea that a few milliseconds one way or another matter for your storage in this day and age.

I bow to your superior knowledge of audio applications. When I was (slightly) involved in this they didn't cope well with this kind of buffering, and one would have needed the whole composition (perhaps up to 30 or 60 minutes) in RAM. If things work better now, that's good. And reduces very much the demands on real time storage. Still wouldn't like to do it over ethernet, though.

 

[ramar]

Using your numbers: 192,000 samples per second * 24bits/sample * 40 channels = 184 Mbps = 23MB/s. If you have 16GB of memory in your system, and you can dedicate 8GB simply to audio stream buffering, you'd have about 6 minutes of buffer for this application. Considering that a typical hard drives can receive data on the order of 100 MB/s, and a typical SSD can do 3-5x that, you could easily handle latency on the order of minutes! Which makes patently absurd the idea that a few milliseconds one way or another matter for your storage in this day and age.

thanks for taking the time to calculate and post this

i was just watching video of the NAMM show 2018 California (the National Association of Music Merchants) http://www.guitarcenter.com/NAMM.gc?source=4TP8ARBB&mscr=2098765_256573 and looking at the Boss GT-1000 Multi-Effects Guitar Processor and it uses a sample rate of 32 bit depth/96 K; i assume he's saying 96 KHz sample rate. i assume the bit depth is logarithmic in that 32 bit is way more than 24 bit. at the moment i am using 24 bit depth and 44.1 KHz, because of a conflict between Logic Pro X (DAW) and Focusrite Scarlette 18i20 (audio interface) and a concern about how much memory per project i am using. the incompatibility is that i am not able to change the sample rate. when i select a different sample rate, the computer (or interface) just sets it back to 44.1 KHz, as if i moved it back to that setting. as i am Mastering (finish mix) for CDs that also use a 44.1 KHz (but only 16 bit depth), i am ok with 44.1 KHz, for now, but now that i think about it, a lot of mastering is done at 96 and 192 KHz and the music industry may be moving to a 32 bit depth at some point. i am thinking that recalculating using 32 bit depth / 192 KHz might lower 8 GB dedicated audio RAM/ from 6 minutes to single digits of seconds. ?? perhaps seconds are fine because i am concerned with milliseconds of latency.

i don't know if you saw my last edits to my last post about the expensive CPU with a long list of functions that i haven't the faintest idea about. would you agree that 'i want Zamba (i don't know what Zamba is) to be "zippy"' for music production, if i were to use FreeNAS to file share and edit in from various OS platforms? my primary reason for using NAS is file sharing between OS platforms, my secondary reason is to create, protect and store pristine data, although i may reverse these priorities.

i was wondering about ECC memory, if it only notifies of errors, how are the errors corrected?

 

[danb35]

I am thinking that recalculating using 32 bit depth / 192 KHz might lower 8 GB dedicated audio RAM/ from 6 minutes to single digits of seconds.

You don't seem to understand how bits work--changing the sample size from 24 bits to 32 bits would increase your storage requirements (or decrease your time capacity) by 33%. Let's do the math. You have 192k samples/sec, each consisting of 32 bits, or 4 bytes. That means your datastream would be 768 kB/sec per channel. For 40 channels, that would be 30 MB/sec. If you use 8 GB of RAM for buffer, that's still 4.5 minutes.

I was wondering about ECC memory, if it only notifies of errors, how are the errors corrected?

ECC doesn't "only notify of errors", it also corrects them. If it can't correct them, it halts your system.

 

[Nick2253]

I don't know if you saw my last edits to my last post about the expensive CPU with a long list of functions that I haven't the faintest idea about.

You seem to be of a mindset that more expensive is better. The reality for 90% of applications is that more expensive == overkill (aka wasted money). Furthermore, unless you understand your application, it will be impossible to optimize. Are you CPU limited? RAM limited? Are your applications single threaded, and therefore limited by clockspeed, or are they multithreaded, and can take advantage of multiple cores?

You keep bringing up "milliseconds of latency", but where!? There are a good half-dozen handoffs within a computer during the process of recording audio from microphone to disk, and not all of them have such strict latency requirements.

Also, if you want to learn about these functions, just Google! For example, the first Google result for "Enhanced Speedstep" is literally an FAQ from Intel about Enhanced Speedstep. If you have specific questions, then we'd be happy to help, but again, most of this learning has to come from you.

 

[ramar]

your datastream would be 768 kB/sec per channel. For 40 channels, that would be 30 MB/sec. If you use 8 GB of RAM for buffer, that's still 4.5 minutes.

Really cool, thanks for the specific info and for showing what these type of calculations are.
i read that comparing 16 bit to 24 bit, in terms of the amount of info was equivalent to comparing the hight of a phone book to a 2 story building. not seeing myself as a scientific person i accepted that explanation as all that i needed to know the difference, which is silly. what you guys are showing is not only accurate, it is relevant in terms of the hardware i need and the amount of latency that is involved.
i realized i didn't know what bit depth was so i looked up an article that looked good. https://www.head-fi.org/threads/24bit-vs-16bit-the-myth-exploded.415361/
i skimmed through it and i do not have the interest to really think about this and read it and follow the links. someone one suggested i get a tech person if i don't understand or don't want to learn, i am still going to read a bit about this but i am going to need someone, but right away. i am in the middle of remodeling my house and learning Mixbus (a DAW to use instead of ProTools and Logic X, not instead of, but to migrate away from. i will still use them when it is more convenient. i might use VST plugins from a Windows OS, that is what prompted me to take a look at NAS to begin with)

ECC doesn't "only notify of errors", it also corrects them. If it can't correct them, it halts your system.

thanks for this correction, i was really hoping this was the case. i was reading somewhere in this forum something that i took to mean this wasn't the case, they we only discussing the shutting down aspect of ECC memory and saying how not worth ECC memory was for the price, or something like that. i have read enough of your posts to hold what you say in high regard. obviously i need to read about it, or hire someone to be in charge of it, for now i will take your word about that. this is a forum, after all, not a library or a search engine. thanks for responding :)

 

[danb35]

I read that comparing 16 bit to 24 bit, in terms of the amount of info was equivalent to comparing the hight of a phone book to a 2 story building.

Ignore binary for a second and think about decimal numbers. A three digit number can be as large as 999; a six-digit number can be as large as 999,999. The latter number is twice as long (i.e., twice as many digits) as the former, but it represents a value 1000 times greater. It works similarly in binary. A 24-bit number is eight bits (i.e., 50%) longer than a 16-bit number, but it represents a value 256x greater. In the case of audio, that doesn't mean that the values are larger (i.e., the waveform has a higher amplitude), but that the resolution is greater--the range is divided into more discrete values, which should result in more natural sound (all other things being equal).

 

[ramar]

You seem to be of a mindset that more expensive is better. The reality for 90% of applications is that more expensive == overkill (aka wasted money)

for most things that i have encountered, more expensive equalled better. however one of the purposes of this forum seems to be to try to save members money while steering the less well informed towards a superior system than they would otherwise go towards (without this forum) and i think the forum is functioning very well in this regards and i wanted to thank you for your efficiency in dealing with me along these lines.

i really cannot be wasting money and time with this.

i thought it interesting that the CPU in question costs over a $1,000 (maybe a few thousand, i do not recall at the moment) vs CPUs that i had been looking at in the $20 to $80 range and had all these features listed. keep in mind that i only first heard about NAS about a month ago and learned of ECC memory here, so i was curious about the possibility of other features that i might find as fascinating as ECC memory. perhaps it will search out talent and produce recordings? LOL i guess i will be satisfied with files sharing between OS platforms, ECC memory and data storage.

If you have specific questions, then we'd be happy to help

thanks. i will keep this in mind when i get to that point :)

 

[danb35]

for most things that I have encountered, more expensive equalled better.

That's close enough to accurate as a general rule, but there are a lot of exceptions. The biggest issue, though, is that "better," in most cases, implies a purpose. A 1000-watt PSU is "better" than a 500-watt PSU, but if you only anticipate needing 300 watts, the extra cost for the 1000-watt unit would be wasted. The same goes for other components of your system--"better", beyond what you have any real use for, is wasted.

We have some people whose systems are complete overkill (I'm one of them). There isn't necessarily anything wrong with that, but you should have a good awareness of your general "need" level so you're aware of that.

 

[Nick2253]

I thought it interesting that the CPU in question costs over a $1,000 (maybe a few thousand, I do not recall at the moment) vs CPUs that I had been looking at in the $20 to $80 range and had all these features listed.

The driver of performance for a CPU is clockspeed and core-count. All these features you are talking about are pretty low level capabilities that are pretty uniform across modern processors. The most prominent features that vary across product lines in the same generation are ECC support, virtualization support/optimization, and encryption. To be honest, the only reason you usually need to care about features of a CPU is because you have a particular need for a particular feature based on your application. For example, with FreeNAS, we typically recommend ECC memory, so you'd need ECC support. If you were doing encryption, you'd want AES support.

I really cannot be wasting money and time with this.

Obviously, you are the best judge of your time use, but if you are concerned about wasting time, building a DIY FreeNAS server seems like the wrong approach. Seeking a consultant with expertise in this area will probably save you significant time, and will likely save you money given the hardware you've posted already.