For those on a budget

E

ecc83

Well-known member
I post here because newbs often ARE on a budget. The cheap part of the recording is a Mackie EM91-C L*DC microphone. They are available now under $50 although we payed closer to £70 for ours IIRC.
The interface is not so budgety! MOTU M4 but I would expect similar results from all but the most dire interfaces? Son is playing the Bach and that is actually his "definitely second best" classical guitar which someone gave him and he repaired and he is definitely NOT a luthier! His £2000 #1 guitar is still here as he flew back to France without it.

I only got the recordings this morning and am yet to speak to him but I expect him to apologize for the squeaks! The long period of silence is for some tests we are doing.

Hope somebody finds this interesting?

* on the L because it is not a full 25mm cap capsule.

Dave.
 

Attachments

  • Bach EM 91C MOTU01.mp3
    1.3 MB
rob aylestone said:
That's a nice recording Dave!
Click to expand...
Thank you Rob, praise indeed coming from you! I have another recording I can post, this time done with the same microphone but through a very early ZED10 USB mixer. Son was a bit devastated when I told him it was a 16bit only recorder (A&H very quickly went 24 bits) hence the two recording to compare. The problem is the dolt left the HPF in on the mixer so it sounds rather "thin". Hopefully I can press him to do again at full bandwidth!

Oil be bek.
 

Attachments

  • Bach EM 91C USB01 m.mp3
    3.7 MB
I'm very responsive to tone - but 16-24-32 bit5 recordings I often find difficult to compare unless you have both and can swap to one or the other quickly. Shame this one was cut off! You seem to have a good combination of mic, room and instrument - and he seems to be able to play with very quiet fingers. Some people make so much noise with their left hand swiping the strings.
 
rob aylestone said:
I'm very responsive to tone - but 16-24-32 bit5 recordings I often find difficult to compare unless you have both and can swap to one or the other quickly. Shame this one was cut off! You seem to have a good combination of mic, room and instrument - and he seems to be able to play with very quiet fingers. Some people make so much noise with their left hand swiping the strings.
Click to expand...

Ah! Son is not at all technical and so was bit worried when I told him that a 16 bit recording will not be AS low noise as a 24 bit one. I tried to console him by saying that IN PRACTICE he is never likely to tell the difference and I think the two recordings support that assertion? Shame they are 'tonally' dissimilar. I can only attach top Q 320k MP3 but both the .wav recordings hover around the -74dBFS mark and I can hear no noise on headphones (but then I am mutton!) That is why the recordings are incomplete. They are just essentially a noise test.

I am gratified that the recordings support something I often tell newbs? If they have a decent mic and interface and set it all up properly, the noise floor is more than likely set by the 'room' not the electronics. We are rarely aware of room noise when it is low but it does get picked up on mics. Only those with sub 25dB studios really need the very best kit.

He will also be pleased with your comments on his technique Rob. He has put in a tremendous amount of work learning 'classical' left hand technique. He has found he needs jeeeust the right fretting pressure or else intonation goes to ****! Also the bass strings can rattle against the nails unless one is very careful. Nail care of the right hand is also a continuous effort.

Dave.
 
but if I can't hear it .................?

what I mean is that I have plenty of old CDs I digitised into the computers over years, and I cannot tell which are the original and which are the remastered versions from later? I get the physics, but 16 bit still sounds great. I've made worse sounding recording in 32 bit fp!
 
There have been plenty of excellent recordings made at 16 bits - in fact just about every digital recording before the late 1990s. You just need to be a bit more careful when setting levels than with 24 bits. Of course, you need more bits for intermediate calculations when processing but for many years 16 bits was all we had for storage.
 
jamesperrett said:
There have been plenty of excellent recordings made at 16 bits - in fact just about every digital recording before the late 1990s. You just need to be a bit more careful when setting levels than with 24 bits. Of course, you need more bits for intermediate calculations when processing but for many years 16 bits was all we had for storage.
Click to expand...
I confess to not knowing the intricacies of digital processing but my understanding of the main benefit of 24bit processing is that the ~90dB dynamic range of 16 bits is expanded (DOWN! It is not 'headroom') to a theoretical 144dB and that puts the DR over 20dB better than any analogue system. Only a very few interfaces manage a DR better than 120dB and even then only by a dB or two. Thus we can record at -20dBFS or even lower confident that the signal will never clip.

I can just about see where increased SAMPLE rates help with processing? Reduced aliasing and lower latency. Not sure how a greater word length helps? There is BTW a comprehensive feature on "upsampling" in the current issue of SoS. Most of it is beyond me!

Dave.
 
Slouching Raymond said:
Nice confident playing. The sound stopped suddenly around 1 min.
Wonder if that was my end.
Click to expand...
No, not you Ray. Son was sending me two files, one 24 bits done on the M4 and one 16 bits done through a zed10 USB mixer. We wanted to compare the noise floors. As it turned out and as I suspected there was nothing in it.

I then thought the recordings would be of interest here especially considering the low cost of the microphone?

Dave.
 
Actually increasing the number of bits does exactly allow for higher headroom. Since you start at 0 (no sound at all), the theoretical limit is 98dB for 16 bits. 146dB for 24 bits and 194 dB for 32 bit if you use fixed point. Using 32bit floating point can do 1,528 dB of range. Of course you'll never record 1,528 dB unless you are sampling something like nuclear explosions or exploding stars!

I think the reason has much more to do with avoiding the case of someone setting a level, say boosting a signal by 50dB, and then a sound that is 100dB hits the mic, and you end up with 150dB or 54dB more than a 16 bit system can handle. This is from a cell phone recording that I did at a concert. It completely overloaded a DAC, which confuses it and it gives bogus numbers.

Clipped audio.jpg

Using a 32 FP system, this is avoided. Any distortion you get will be from the analog section, not the digital section. Plus I've read that there are advantages when applying effects from a mathematical standpoint.
 
ecc83 said:
I can just about see where increased SAMPLE rates help with processing? Reduced aliasing and lower latency. Not sure how a greater word length helps? There is BTW a comprehensive feature on "upsampling" in the current issue of SoS. Most of it is beyond me!
Click to expand...
If you want to multiply two 16 bit numbers together you could end up with a 32 bit number if both 16 bit numbers are close to full scale. There are plenty of multiplications in signal processing so you need to allow plenty of headroom. 32 bit floating point numbers get around this by extending the range but reducing the precision. Sometimes this reduction is precision is audible which is why most DAWs use 64 bit floating point internally now.
 
Thank you James. My mention of "headroom" was the simpler concept that any AI must have a maximum output at 0dBFS. For most smaller interfaces that will be between +10 and +16dBu. Thus, it matters not a jot whether you are running at 16 or 24bits, max out is max out. The extra dynamic range of 24 bits then just lowers the noise floor...'leg' room if you will.
This of course allows you to record at a lower level, -20dBFS or even much lower.

Dave.
 
Slouching Raymond said:
Guess what? Whatever resolution your software thinks it is using, the actual floating point processor in a PC does everything in 64-bit.
Click to expand...

From what I can see from a quick search, the Microsoft C++ compiler still treats a float data type as 32 bits and a double as 64 bits so the programmer has to specify that they want to use 64 bits for the compiler to use it. I understand that the processor may work at a higher precision but the results are stored at the precision that the programmer specifies so there is still the danger of rounding errors with a string of 32 bit calculations.
 
In the olden days when floating point processors were separate chips, some had more resolution than 64 bits, and could be more accurate.
It is the loading and saving of data from the FPU registers that may be done at smaller bit lengths, according to the instruction parameters used.
If transfers were all done at the FPU's 64 bit length, slow processors wouldn't have the bandwidth for complex jobs.
I know about this, because I designed and coded thorough floating point test software back in 93. It was a surprise when some FPU brands were more
accurate than others.
 
Slouching Raymond said:
In the olden days when floating point processors were separate chips, some had more resolution than 64 bits, and could be more accurate.
It is the loading and saving of data from the FPU registers that may be done at smaller bit lengths, according to the instruction parameters used.
If transfers were all done at the FPU's 64 bit length, slow processors wouldn't have the bandwidth for complex jobs.
I know about this, because I designed and coded thorough floating point test software back in 93. It was a surprise when some FPU brands were more
accurate than others.
Click to expand...
Now, I ask this in all ignorance Ray, how much does such accuracy matter for audio work? Since the human ear has a "resolution" no better than 2dB, how many iterations would even a complex audio file have to go through until a difference could be detected? Thousands? Billions?

Dave.
 
How much calculation needs to happen to convert a midi file to an orchestra, or to simulate a spacial reverb from an impulse response? How much error can you tolerate? Like you, I'm asking in all ignorance, but I have to assume that the people who are creating these program and plugins know what's needed, not just some marketing guy telling the engineering dept "we need to be twice as good as the competition. If they have 32 bits, I want 64 bits".

I think the goal is to just eliminate any possible way that an audio signal might be corrupted in the digital domain. Then it's all up to the analog system to keep up.




Raymond, I remember the days when we got an 80286 CPU and an 80287 FPU for some of our instrumentation at work. Computers were pricey and the FPU just added to the cost. The bean counters wanted to know why we needed it. $3000 for a computer was a TON of money in the mid 80s!
 
Last edited:
You must log in or register to reply here.
Back
Top