Thought experiment

[glenn]

So I was thinking, if anyone would use a wireless microphone with no input gain adjustment, and the audio was very low, lets say for example that only 8 bits of the A-D converter were being utilized. In this case the audio would be represented by no more than 256 discrete levels. This works out to a dynamic range of 48 dB. Since we can not change the mic gain on the transmitter we can not boost the signal to the A-D in the transmitter to get the proper transmitter A-D resolution of let say 96 dB. So we have lost about 48 dB of dynamic range.

 

If we were to boost the audio signal at the receiver, recorder or at a mixer, we can add 48 dB of gain to come up with a signal that we can mix into other signals at a proper level. When we do this we are now utilizing about 64000 discrete levels. The fact is the audio will never actually have more than 256 discrete levels no matter how much gain is added down stream. All that happens is that the 256 levels are in this case multiplied by 256 to make the signal louder not better.  In this example signal and noise are boosted at the same time and the distortion of only 8 bits of data representing the original audio remains at exactly 8 bits or 256 discrete levels. It makes no difference if the recording and or transmission is 24 bit fixed or 32 bit float. This is exactly why adjusting the mic preamp level to match a microphone to the proper level for a A-D converter is essential to achieve the best audio quality. Nothing will restore the audio quality that was lost by not having a transmitter with adjustable gain.  Suggesting that 32 bit digital connections to devices that can add gain later in the chain to make up for the lost A-D dynamic range of the transmitter input gain mismatch is promotional sales nonsense and is no different than selling snake oil at a side show.

[Dalton Patterson]

You mean this???

 

 

KRON) — The first Bay Area sideshows took place in the parking lot of a McDonald's, across the street from Oakland's Eastmont Mall. They involved young men, some drug dealers, flush with cash showing off their flashy cars while locals from the neighborhood gathered to watch.

 

https://m.youtube.com/watch?v=XjBaNryXCWc

[Tim Glynn]

That's only true for a single analog to digital convertor, no?  The implementation from other manufacturers describe using multiple convertors, each targeting different gain ranges, combined back into a single 32 bit float representation that contains all of the headroom without adding noise.

 

 

Isn't that also exactly how NeverClip on the Nomad was advertised?  Dual ADCs and an internal 32 bit float representation to ensure no clipping with no added noise and no compression artifacts?  

[chrismedr]

I‘m not sure where this thought experiment is going except to tout once again how zaxcom is superior to all other wireless 😕

 

as Tim says, pretty much all systems which use 32bit float recording will have a dual or multi ADC which means the signal will never be only 256 discreet levels. How exactly implement this in a wireless system for transmission is certainly a challenge, but instead of bringing up strange thought experiments and accusing other people of selling snake oil, the better way would be to just test the system and see if we can adjust the signal level without loosing quality.

 

obviously such a test is better done by an independent party then from a competitor. 

[glenn]

The purpose is to ask people to think about the tools they use. It does not matter how many A-Ds are used in the example as 2 A-Ds with no gain control is effectively like having an input gain control with only 2  settings low and high.  If the level coming into an A-D is very low (around 48dB above the noise floor), there could easily be around 256 levels or less.  I am sure it happens all the time. What happens to low level audio when it is adjusted later in the chain rather than at the pre-amp is the point to consider.  There is a lot of 32 bit floating point information and mis-information on this group.  Some companies have adjustable gain and others don't. What gear to use is a personal decision and people will choose what is best for them.

[Patrick Farrell]

47 minutes ago, chrismedr said:

the better way would be to just test the system and see if we can adjust the signal level without loosing quality.

 

This. Rather than a thought experiment, how about a listening experiment? Borrow a couple systems that are as you describe and see how they sound. I personally find they sound better than some systems which utilize adjustable preamps in the transmitter.

[BAB414]

I'm still wondering what's wrong with using my digital hybrid wireless transmitters? They sound fantastic. I mix movies and TV shows and it's rare that I run into issues that would necessitate the ridiculously wide dynamic range that 32 bit supposedly has. In those rare cases, I throw up an extra mic. Maybe this tech is best for pro-sumers or documentaries where you only get one take.

[Derek H]

7 minutes ago, BAB414 said:

I'm still wondering what's wrong with using my digital hybrid wireless transmitters? They sound fantastic. I mix movies and TV shows and it's rare that I run into issues that would necessitate the ridiculously wide dynamic range that 32 bit supposedly has. In those rare cases, I throw up an extra mic. Maybe this tech is best for pro-sumers or documentaries where you only get one take.

This thread is obviously aimed directly at the Sound Devices Nexus and A10/A20 system. 
 

I suspect they’ve thought about this scenario and there is something happening behind the scenes with the A20 input stage that allows their gain forward concept to work well even dealing with low levels. If it didn’t work well at low levels then rule out using it on almost any TV show or movie because that’s all we seem to get out of many actors is low levels. Maybe someone who owns one of these systems and has used it on a real scene can comment on how it performs. 

[rich]

1 hour ago, Derek H said:

This thread is obviously aimed directly at the Sound Devices Nexus and A10/A20 system. 
 

I suspect they’ve thought about this scenario and there is something happening behind the scenes with the A20 input stage that allows their gain forward concept to work well even dealing with low levels. If it didn’t work well at low levels then rule out using it on almost any TV show or movie because that’s all we seem to get out of many actors is low levels. Maybe someone who owns one of these systems and has used it on a real scene can comment on how it performs. 

they perform great.

i have not noticed any noticeable difference since swapping from my A10 to A20 mini for cast radios.
the A20 mini are handling the quiet and loud sounds from cast equally well. but without the necessity of changing gain on the tx to get good levels at the mixer, which makes my workflow, and that of my team sooooo much less stressful.

when testing, the noise floor sounded fairly similar between an A20 mini with 45dB of gain on an A10rx, and an A10 with 22dB of gain and 23db of gain at the mixer. my normal settings for A20 mini are 24-28dB and A10rx at 0, and A10 tx at 22dB with between 0 and +4 at the mixer.

[Tim Glynn]

This thought experiment nerd-sniped me, so I thought I would see if I could turn it into a real experiment.  Some caveats first, then a TLDR, then the details for reproducability.

 

 

Caveats:
I don't have a particularly fancy or accurate way to measure SPL.  SPL readings came from the NIOSH SLM app on my iPhone and should be taken with grain of salt.
I used a pretty crummy studio monitor (Yamaha HS5) for my source; the frequency fresponse for that monitor is definitely not flat, and it's a little guy, so low frequency reproduction is questionable.
I don't have the quietest room in the world; my floor is somewhere around 29 or 30 db(A) according to the NIOSH SLM app.  So I'm not sure that this necessarily gets to the heart of the thought experiment for truly quiet sources, but I figured that it's real world enough until someone with an anechoic chamber volunteers.

 

TLDR:
For a DPA 4060 in a room with an ambient SPL of ~30 db(A) and a target signal in the neighborhood of 60 db(A), adding up to 51 db of digital gain from GainForward on a Sound Devices 888 doesn't appear to increase the noise floor any differently than applying that gain to a regular old preamp.  

 

The test:

I took two DPA 4060s and attached them in front of a speaker.  My test audio track was a tone (for alignment), 30 seconds of silence, 60 seconds of pink noise, 30 seconds of silence, and final tone for alignment.

I started with a control - both lavs hooked up with the DPA XLR adapter to inputs 1 and 2 on my 888.  I set levels based on the pink noise, so that I was peaking around -18.  That happened to be 50 db, based on the loudness of my speakers and exact position of the mics.  Putting my phone between the two mics, I measured the pink noise at around 60 db(A) (plus or minus whatever, since this isn't a super accurate way to measure sound pressure).  I did two takes with MicA in input 1 and MicB in input 2, then I swapped them and did two more.  These control samples gave me my margin of error - I measured the peaks and RMS for the clips as a whole, the silence sections and the pink noise sections.

Then I swapped in an A20 Mini, switching the microdot on one of the 4060's to the lemo connector.  I didn't change the levels, since when I swapped from the XLR connection to the A20 mini, I noticed that it was peaking in the right place to match the mic coming in VIA XLR.  It seems like 36.5 db of digital gain (which was the same position of the trim knob) gave me the same level.  I ran through two takes, then again swapped Mic A and Mic B (so Mic A went from the transmitter to the 888 XLR input, and vice versa) and ran it twice more.

 

Then, just to see what happened, I turned the gain up as high as I could on both until I was just shy of clipping (the final clips peaked at -0.5 db when I got them into my DAW).  That was preamp gain of 65 db for the XLR input and 51.5 db of digital gain for the A20 mini.  I ran through two takes as well.

 

Pulling all the takes in and matching their peaks, running through an A20 mini and adding digital gain does not appear to add any to the noise floor.  Again, see all the caveats above - this was not a perfect test, I don't have great SPL measuring equipment, or even a particularly quiet room, but I thought it was interesting enough to share.

 

[glenn]

Thanks Tim for your post. My original post example focused on distortion mostly not specifically noise. I do think an exact gain setting for a microphone is the best way to get the best results for noise and distortion rather than a fixed transmitter gain. This thread is directed toward the idea that 32bit floating point transmission somehow improves the results of audio quality when not adjusting the gain of the transmitter at the transmitter.  The thread is not directed at any one company as multiple companies seem to be going this way. It is a commentary on the method as a whole and what I see as a poorer choice than transmitter gain adjustment as used by multiple companies as well.

[Derek H]

Tim, thanks for posting your test. We need more of this kind of thing here. 

[Philip Perkins]

Thanks for doing a real test.   There is a long tradition of location soundies doing their own gear tests, no matter how "cowboy" they are so that they then really KNOW what their gear does.  Well done.

 

[Patrick Tresch]

Bad preamp with a lot of gain on it, don't get you more DR. It's just loosing the DR of that preamp. It won't get you more DR out of your preamp (noise to clipping range). Ideally you should have the best preamp in the world without gain to keep the whole 24bit DR (135db?) more you push you gain less headroom (DR) you get before you clip. But who mixes at -60db ?... So you put some gain to get a workable sound level.

 

Is my thinking right?

[DSatz]

Glenn, your hypothetical situation seems to assume an entirely undithered a/d converter. In a properly dithered system none of your conclusions apply. Very low-level signals may be close to the noise floor, but they won't suffer any distortion due to the small number of bits that are actively flipping in any given time span.

 

--best regards

[henrimic]

I just finished a movie where the actors were whispering and then suddenly screaming in the same scene.

I used the A20 mini's with the A20RX plugged in AES3 to my Sonosax R4+. The mikes were DPA6061.

The Gain forward principle worked perfectly for these tricky situations.

With my old Wisycom systems, I should have set the gain very low at the transmitter to avoid clipping in the loud parts, and thus would amplify the noise in the quiet parts.

I don't know how SD achieve this, but it was a life saver in this situation.

I have to mention that I used the older A10 for the booms. In these scenes, I set 20 dB of gain at the transmitter, and amplify the quite parts at the recorder, and it worked very well too.

 

 

[sonyslave]

Interesting...

A question for your question: Would an analog recording that was as equally weak and as equally clean as your digital signal, then "cleaned up" in post using the latest  computer processing yield a cleaner final product than a "cleaned up" version of the low bit utilization audio?    Perhaps the "infinite" levels of analog would offer more raw material  for the super computer (and by that I mean  your phone)  to weave into usable audio.

 

So, should we return to analog wireless with analog recording ... then dump it to the super computer sort it out?  ...of course this would be for "under gained" situations only.😉

 

[BAB414]

3 hours ago, sonyslave said:

Interesting...

A question for you question: Would an analog recording that was as equally weak and as equally clean as your digital signal, then "cleaned up" in post using the latest  computer processing yield a cleaner final product than a "cleaned up" version of the low bit utilization audio?    Perhaps the "infinite" levels of analog would offer more raw material  for the super computer (and by that I mean  your phone)  to weave into usable audio.

 

So, should we return to analog wireless with analog recording ... then dump it to the super computer sort it out?  ...of course this would be for "under gained" situations only.😉

 

I love this. I love analog whenever I can.