Changes in noisefloor during recording

Just started out on audiobooks with ACX so am careful in doing the ACX checks every so often. Within a 10 minute recording, the first 3.5 mins were fine (noise floor -64.3, could be better, I know) but the next 6.5 minutes showed a -53.2 reading. This was while I was recording - I had done nothing different during the recording. Does anyone have any suggestions for a reason/solution? I was using Audacity 2.3.1 at the time but have since downloaded 2.3.2, but I don’t think this was the problem. I’m using a Rode NT1 condenser, with the Rode interface, into a laptop running Windows 10. The laptop is not in the room I record in, so I’m using a tv monitor in the room. My auditions and first 15 minute uploads so far have been fine - this problem has developed in the past 24 hours. Thank you.

Where is the interface? The connection between the microphone and the interface is three-pin XLR, right?

That is designed to go relatively long distances with no ill effects. You may have gotten a 20 foot microphone cable with the interface. That’s a normal, generic microphone cable.

The USB connection needs to be about 6 feet or less. Long USB cables can create problems. That and they are not designed to be constantly connected and disconnected. The little gold contacts inside will eventually wear and some very nasty things happen when the connection quality goes down.

A little hiss in the performance is normal, and depending on your performance technique, environment, etc., it can be a problem or not.

Having it change is not normal and that can mean about a million different things. Did your refrigerator or air conditioner go on? Metrobus drive by?

Are you trying to read directly into ACX standard? I note you are missing describing the post production processing and mastering. Whose mastering process are you using? We publish one, but it’s not the only way to do it.

Record a sound test and post it here. We could guess at this for the next two weeks, but an actual sound test is very valuable.

Also post the interface model number or any other information.


Did you post and pass a quality test with ACX? They allow you to post a short test presentation which does two immediate things. It lets them inspect your technical and theatrical quality (those are different) and your ability to follow instructions.

That last is not obvious. You’re supposed to post using their standard book chapter configuration. It’s not that unusual for someone to get bounced because they put the wrong number of seconds of room silence at the end of their submission. Wouldn’t you feel silly if you went to all that effort and failed basic instructions.

Everybody with a pulse is trying to read for audiobooks, so their test evaluation process can take between one and two weeks.

This forum is a lot faster.


Hello Koz - I’m using the 20’ XLR cable that came with the audio kit and the USB connection that also came with it is less than 3’ long. The interface is approx. 6" from the laptop. The interface model number is the Rode A1-1.
The post production sequence follows the exact one prescribed by Audacity (EQ for voice roll-off, RMS Normalize and Limiter, in that order). The only tests passed with ACX were those when I submitted the first 15 minutes of the two audiobooks I am currently working on and I have heeded the instructions re. prescribed silence at the beginning and the end of each one. These did not have the problem I’m currently experiencing.
I’ll post the sample I’m currently working on, as requested, thank you. I’ve only applied the post production sequence and haven’t removed any clicks, mouth noises etc.
There is no traffic where I live and I turn off the heating system when I’m recording. The freezer is in the next room and the room I’m in has been effectively sound treated and is enclosed.
HOWEVER - on my second try this morning, I’ve noticed the problem occurring immediately after a break using the punch roll and record method - the break point is at 14.05.05 - this is where the hiss gets greater. Although the whole section passes ACX, the affected part’s noise floor doesn’t.
Thank you.

Sorry - just seen the instructions re. the clip - I’ll do that instead. Thanks.

The interface model number is the Rode A1-1.

I did that, too. The actual part number characters are a-i-1. Audio Interface One. Don’t feel bad. Some learned web pages got it wrong, too.

Terrific voice. I couldn’t help wondering what would happen if you did read the milk carton in full theater. You could cause a run on milk at the Tesco/7-Eleven.

You didn’t say that you were applying Noise Reduction. Are you? I can’t get you anywhere close to passing without it.

You are presenting at very low volume. How did you set the volume when you unwrapped the system?

How far from the microphone are you? If you’re too close you can get all sorts of proximity effects such as boosted sibilance and “announcer” boom. If you’re too far away, you get what you got. Perfect quality voice at a small fraction of the volume it’s supposed to be. There is always system noise (ffffff) down there waiting. Your job is to get your voice loud enough so it doesn’t matter.

A fuzzy rule of thumb spacing if you’re using the pop and blast filter is a power fist from the screen.

Where did you set the AI-1 volume control? It’s not unusual for “home interfaces” like this to run all or most of the way up.

There is a test you can do. There is a little light on the AI-1 just above the volume control. It turns green when it senses any useful volume at all, yellow or yellow-green if you’re just about right and red when you’re too loud. What does your light to?

Can you see the Audacity bouncing sound meter? You can make the meters larger by grabbing one side and pulling. The other little windows will scurry out of the way. Your raw reading should peak about -10dB or so (fuzzy numbers). Yours peaks at -30dB. Sound goes double and half every 6dB.

If it is a separation problem, a simple adjustment of your spacing may do it. It may be bumping the volume control up a bit.


I bet you thought I was never going to address the actual question. I don’t expect the Rode equipment to affect noise on its own, but you can get Windows to do all sorts of nasty tricks. This has the hallmarks of a Windows process trying to automatically set volume to “help you.”

I’m not a Windows elf, but here are a couple of places to look for tools that could affect the sound.

Do you use Skype or other conferencing or chat applications? They famously grab sound control away from you and the computer for their own designs. Make sure there is nothing else running when you record.

Oddly, one of the recommendations if you manage to go over about two weeks struggling with the computer is stop using it. There are some very good stand-alone sound recorders out there, and they have none of the computer baggage.

I know this is a pain to deal with, but try not to miss any questions in my posts. It’s irresistible to answer the last question and ignore everything else.


Hello Koz - first, thank you very much for taking the time to answer my query - I’m kind of trying to make VO/books a second career and I want to get it right.
In no particular order: I don’t use Skype or other such apps, so I don’t think that would be a problem To be honest, I think I knew volume might have something to do with it - the waveform on the file I sent you looks pathetic. The gain on the A1 is currently at ‘11 o’clock’ (it doesn’t have numbers) and the Audacity recording volume is at 60%.
I’m not using Noise reduction as I know ACX aren’t too keen, apart from a little 6-6-6 now and again.
I will have a look at your other suggestions and experiment again, along with the link you’ve sent me. Just out of interest, I stopped using the Punch roll record method this afternoon and the problem didn’t occur - my ACX check on a 35 minute recording showed peaking at -3.50, RMS at -20.6 and Noise floor at -64.60. But I’d rather get the volume issue right.
Thank you,

One of the reasons your knob has no numbers on it is you’re not supposed to be paying attention to the numbers. The knob is in the right place when the show volume is perfect.

The only time you get concerned is if the knob goes either direction too far. That can mean something is broken or connected wrong. Any setting in the middle 3/4 is fair game.

So forget Audacity for a minute. Can you make your little AI light happy? Occasional yellow?

There is a more vicious test. Can you force the light to turn red? A good recordist knows where both extremes of volume are, not just one. You know how to have low volume and noise problems. What do you have to do to have overload and distortion problems? Never blow into a microphone, but you can speak as loudly as you want. Your microphone is rated to work perfectly sitting in front of an operating jackhammer.

There is one difference between us I’m concerned about. When I master the clip you sent, I get this.

Screen Shot 2019-05-22 at 13.41.18.png
-53dB noise is not even remotely quiet enough even with Noise Reduction. How did you get -64dB on that same sound test?

And just to cover the bases, you are using ACX Check, the program we publish, right? It was designed to mimic the tests that ACX uses for audiobook acceptance.

There are ways to measure noise that make you seem a lot quieter than you are.


Why would you get a 4dB boost in background noise when you performed a Punch and Roll…???

Can you shoot a test? You may have turned up an Audacity program error.

Shoot a 20 second test just like you did the last one, but then go back about half-way and Punch and Roll a correction. I want to watch what happens to the background noise. Leave about a second of no voice at the end like you did this one.



Hello Koz - I’ve increased the gain on the interface so that I get green/yellow and even to the extent that I get red, just so I know what ‘too loud’ looks like - I take your point about no numbers. I’ve also increased the recording level on Audacity to maximum and I’ve changed my speaking position to a ‘fist width’ from the pop shield - the result is great and I don’t why I haven’t done this up to now.
Regarding the readings you’ve sent me - I was referring to my audiobook effort this afternoon - not the test I sent you earlier, so apologies for not being clear about that. Yes, I only use the Audacity ACX check plug-in, just to be clear.
Thanks for your assistance - it’s given me a great improvement.

I’ll do the ‘Punch and roll’ test tomorrow, if that’s ok - it’s bedtime in my part of the world and I’ve had too much red wine for one evening.


Yes. It’s easy to forget the rest of the world is getting ready for bed.

I’d love to say I’m headed to the beach to catch the last rays of the Warm California Sun®, but it’s actually raining like crazy and I walked to lunch.


Here is the punch and roll test as requested. The file hasn’t been processed at all.

The file hasn’t been processed at all.

We can’t take processing out of a clip, so when someone sends us a damaged clip after processing, it’s rough to know who damaged what.

As you noted, this sound and technique sails right through audiobook mastering and doesn’t need any further corrections. Your readjustment of volumes and process worked handsomely.

However, the test clip was not what I was expecting. I need to sit in a dark corner and beat it up a little more. Yes, the silence at the beginning and the end don’t match, but why they don’t is very odd. I think your system has DC Offset where the battery that powers everything inside the AI is leaking into the sound. I have a puzzled puppy look as to why you don’t always have this error. It’s a known defect, but it’s not supposed to come and go.

Thanks for the clip.

Can you briefly describe the behavior of the light? There’s different ways to make it easier to see. You can’t just jam a light on sound because that’s dim and flickering and nobody can see that. There’s a BBC technique where it gets bright very rapidly and then slowly fades out. There’s just make it sloppy and it throbs rather than rapidly flickering, etc.

I wonder if there’s a video of this anywhere…