ACX woes on MBP

Again, my steps are these:

1 - Record raw WAV audio tracks onto my Zoom F6 recorder (there’s no noise redux option on this recorder since it is a professional field audio recorder, and the only audio “shaping” available is the standard low cut filter - set, by the way to my standard 80Hz rolloff).

2 - Transfer my selected WAV takes into Reaper to assemble and edit. I like to lay tracks down as alternating tracks. This practice goes back to the days when I cut 16mm and 35mm audio tracks for film. The stage mixers liked having a gap between sounds so they could change levels between sounds if necessary.

3 - I do no audio processing of any kind as I lay the tracks down. The only adjustments I make are to tweak levels if I need to.

4 - When I’m happy with the timing I RENDER the track (again there’s no signal processing of any kind) as either WAV or mp3, depending on what the requirements of the destination are.

I generally don’t use mp3 as an output other than for our church’s sermons which are loaded onto the Sermon Audio website.

The only time I ever record mp3 on location is when the producer asks for an mp3 file for transcription purposes.

The pictures I sent of the REAPER tracks and AUDACITY are the product of the above steps.

You mention the H6. That’s not what I have. The H6 is a very different machine, and not something I’d use as a primary or back-up field recorder since I don’t like the layout of the machine. Also since it doesn’t take timecode it’s useless for me. It’s mainly for FX gathering, although there are folks out there who carry it as a backup in their audio kits.

I have the F6. It’s a professional six channel, 10 track recorder, with timecode. That’s my back-up recorder. That’s why it’s parked in my VO booth.

There is no noise reduction on the F6 other than high pass filters. It has an automix option but since I’m only recording one track I don’t need that.

My files are WAV files all the way through the process to the RENDER step where I choose to render as WAV or mp3 depending on my need and application.

I am a production sound mixer by trade, btw.
My main field recorders when I’m in production are either my Sound Devices 633, my SD 664s or my SD 788.

Zoom!! I got the wrong one. Extreme Caffeine Deprivation.

A local grocer has “Hu” chocolate at checkout. It’s recognized as an Addictive Substance by the State of California. A different grocer, sensing an opportunity, offered Jelina Chocolates.

There’s nowhere to run, nowhere to hide.


If you open Mastering up in a text editor (and expand your screen), you’ll see it’s a comment line and just three tools. 100Hz High Pass Filter, Loudness Normalization, and soft, gentle Limiter. Nothing that can affect noise other than what happens when you make overall gain go up and down.

Based on the note by DVDdoug, I think I would select a copy of one show, add noise, and render to a single track. With nothing else on the screen, drag-select a large portion out of the middle of the new track and run ACX-Check again.


I doubt it. You’re always going to have Mic Preamp noise and what noise that little FET up at the capsule makes.


Okay. I’m thinking there’s something wrong with Audacity somewhere.

I just took our Sunday sermon (WAV), cut heads and tails off and RENDERED it out as WAV. No processing at all, not even gain changes.
Again, this was from our church service yesterday. Mic on the pulpit. There’s definitely some street noise in the track.

Ran the ACX check in AUDACITY with the track input as a WAV track and got the following (No mastering macro was run, this is what the track simply opened in Audacity shows).

If I remember correctly, “too quiet” is actually an issue for ACX. I ran into that, I think due to Izotope Dialogue Isolate, or maybe a noise gate, which didn’t test well, but was accepted by ACX without issue.

By the way, for anybody working with ACX and Audacity, I found this to be an AMAZING resource. I’ve done DOZENS of projects via ACX over the years, but not until going through this course was I able to put all the pieces together and do a full project without outside help.

This is me writing that down. We are warned that selling stuff on the forum is forbidden.

Screen Shot 2024-03-04 at 1.32.51 PM

Anyway, back to the original problem.

That’s not what I said.

Keep the selection away from the ends of the track.



I went back to my original finished track, did the select as you suggested and got a pass.
Wierd that excluding the heads and tails makes it fail…but oh well.
At least I know everything is good.

Tony Smyles

It’s not seeing that as “silence” or “noise” and there is probably silence somewhere else in the file.

And the test noise track I added ran the entire length of the narration…so??

OK, well. Being obsessive, it’s not the worst idea to find out what’s going on at the ends of the tracks that’s causing problems. This is the first I heard of anything like this. If you have magic files or conditions, instability can be counted on to pop up again, and at the worst possible time.


More info:
Selecting the beginning of the track through almost to the end is fine.
Selecting to include the end, which has a bit of tails silence causes it to fail.
So guess I just have to cut tight to head and tail…

ACX-Check doesn’t care where the silence is. Any silence in the presented data counts. Can you tell why the tail is there? That’s not normal.

I did notice another oddity. Better than 85dB noise is most unusual in a raw clip. Most material comes in mid 70s or so.

Leaning on the electronics: The Mic Preamp has to boost the voice 60dB or so and still maintain a 60dB noise floor. That’s a pile of electronics with 120dB internal noise. That’s an atomic-level effect.

That’s not to say it’s impossible. Traditionally, it was done with [gasp] [cringe] expensive, heavy, physical transformers. Voltage boost with zero added noise.

The consumer/wallet-friendly version of that _ is _ Noise _ Gating. See: Zoom Meeting with nine people.


Again, I’m not using any gating or processing other than the Audiobook macro in Audacity.

As a sound editor I’m used to always leave a bit of tails, and usually a medium fade out after the room dies out.
At least in the music world if you cut tight to the end of a track you get whatever playback software you’re using to hop to the next track without a pause. Also, if its a musical performance it’s better to err on the side of a little long than to chop off ambience that might still be sounding in the location.

So, I guess the ACX world is different and now I know. There’s presently all of a second tail at the end of the file…okay 1.085 seconds of tail.

FYI attached are the input specs for the Zoom F6 and an article on it. It’s quiet. Oh that my old analogue tape recorders were that quiet.

My gain trim on the mic is +45dB…on the high side for me but just over 1/2 of the available gain. (My other recorders don’t need that much gain with this mic. And they are transformer inputs).

The F6 does have a 32 bit float mode (as well as simultaneous 32/24) but I hardly ever use that, only 24 bit. Certainly NOT in this application. My booth (closet) is small and this is what I’ve wound up with.

My demo producer also commented on how quiet the test track I sent him was and that I’d have to loosen it up some for audio books.

And nothing wrong with high quality audio transformers.

Article on the F6 and screenshot of the specs.

“A Weighted” is the phrase that pays. That’s easier to hit if your sample spectrum isn’t flat.

I don’t think any of the Audacity tools uses weighing. ACX defaults to RMS for loudness, too, not LUFS.

I don’t think you have to do anything special to the beginnings and ends except not show signs of having been messed with. Background Room Tone that rapidly dies out or pumps and throbs is the kiss of death.

Their motto is No Distractions.

Do you know if they will allow submission of a voice sample for Human Evaluation? They used to. Then came Covid and the only thing left was automatic ACX AudioLabs which doesn’t evaluate theatrical quality or noise.


This topic was automatically closed after 30 days. New replies are no longer allowed.