DC Offset Cancellation in Windows 10

Hello:

I am using audacity 2.1.3 and am finding the issue around DC offset to be a bit confusing. I have read the manual thoroughly and can’t tell whether or not I should be concerned or not, with my system. I am running Windows 10. When I record LPs or Tapes into Audacity, the result sounds fantastic. Yet, different online postings have suggested ALWAYS use DC offset correction, and others suggest only using DC offset correction if you have a problem with it appearing. The sample shown in the audacity tutorial at:

http://manual.audacityteam.org/man/dc_offset.html

…does not appear to be happening with my recordings, nor do I have significant sections of the recording where this is happening. Only occasionally, here and there, do I have very small micro-sections (I have to zoom in quite deep to see anything) that appear to have some offset.

Per the document address above, it suggests checking to be sure the DC offset correction of my recording device is enabled, however, when I follow the directions:

Newer Windows PCs may have a DC offset cancellation feature when recording from the built-in sound inputs. To check for or enable this:

Step A: Right-click over the speaker icon in the System Tray > Recording Devices, or click Start > Control Panel > Hardware and Sound > Sound, then the “Recording” tab
Step B: Right-click over empty space, show disabled and disconnected devices, then right-click over each device and enable it
Step C: Right-click or select each device, choose “Properties” then look in the “Enhancements” tab
Step D: If there is no “Enhancements” tab, look in the sound device’s own control panel in “Hardware and Sound”.

…I can’t find any DC offset switch in Steps A through C or D under Control Panel. And actually…Steps A-C and D (by itself) appear to take you to the same window where there is no DC offset switch to check or uncheck.

Does anyone know how to find this in Windows 10?

Also…should I even be concerned? I like using the ‘Amplify’ effect and my recordings sound fantastic, but should I always be using DC offset first (and NOT normalizing), and then using ‘Amplify’ to boost my tracks to -2db or -1db as I see fit? Is there an advantage to using 'Normalize’s amplify effect vs. the ‘Amplify’ effect by itself after using DC offset? The document below seems to suggest that using ‘Normalizes’ amplitude settings will set each channel to the same level (which is not what I want). Alternatively, the ‘Amplify’ effect description seems to suggest that it will amplify each channel by the same amount, thus preserving any variance in the channels that exists (this is what I want). Am I reading this correctly? Am I getting what I want by using the latter over the former? And if so…in the event that I do need to use the DC offset correction at some point, is there anything wrong with using DC offset correction first, and then amplifying it with the ‘Amplify’ effect rather than the ‘Normalize’ amplification effect?

http://manual.audacityteam.org/man/amplify_and_normalize.html

Input and/or advice on this matter would be greatly appreciated.

Thank you all.

DC Offset happens when the electronics housekeeping gets into the sound by accident. “DC” for example, is what turns the lights on and amplifies your microphone if you have a USB Microphone interface. DC doesn’t sound like anything but it makes everything else possible. The batteries in your flashlight are DC.

Putting DC in the sound by accident—even a little bit—can cause confusion later on if you need editing, effects or corrections.

DC looks like a mistaken vertical shift in the blue waves, most noticeable when the show goes silent. This picture is of a bad edit. The sound on the left has DC Offset and the sound on the right is normal.

You can inspect your show for this error by recording a bit of silence (lead-in groove) and then boost the volume way too loud just for a test. Say, Effect > Amplification, 20dB, Allow Clipping. That should boost your sound so loud it’s terrible, but if the middle of the blue waves stays more or less around 0 (measured on the left) you don’t need offset correction.

Edit > UNDO.

Koz

…does not appear to be happening with my recordings, nor do I have significant sections of the recording where this is happening.

If you don’t have a problem I wouldn’t worry about it. Some DC offset correction algorithms can actually cause a DC offset while attempting to “balance” the top-half & bottom-half of the waveform, while throwing-off silence and quiet parts in the process.

True DC offset will be present throughout the waveform and noticeable during silence. Some real-world sound waves are not symmetrical and it’s not unusual to have a positive peak greater than the negative peak or vice-versa.

And, some “cheap” soundcards will [u]clip[/u] unevenly. (Analog-to-digital converters should clip at exactly 0dB on the positive & negative side.) If you get that problem, the best solution is to record at a lower level to prevent clipping… Don’t try to balance the waveform with offset correction.

Sounds like you have nothing to worry about

Don’t worry about that. It can be a (no cost) convenience if you have a bad sound card, but the better solution if you have a bad sound card is to replace / upgrade the sound card.

Properly implemented, non-reak-time DC Offset correction (like in Audacity’s Normalize effect), when applied to audio that has no DC Offset, does nothing.

“Real time” DC Offset correction (like the setting that some sound cards have), apply a low frequency “rumble filter” that reduces very low frequencies (“DC” is in effect a “frequency” of zero Hz). This may or may not be a good thing.

Need to be careful about terminology.
Audacity uses the term “channel” to refer to the left / right streams of a stereo track (and the one data stream of a mono track). A “track” may be mono or stereo, so each “track” may have one or two “channels”.

By default, the Normalize effect, when applied to a stereo track, will amplify both channels by the same amount, thus retaining the relative level differences in the channels that exists. The effect therefore uses the highest peak in either of the two channels to calculate the amount of amplification to apply, and then applies that (one) amount of amplification to both channels.

The Normalize effect has an option (off by default) to process each channel individually. If this option is selected, then the amount of amplification is calculated for each channel in turn, and may apply a different amount of amplification to each channel so that both channels end up with the same peak level.

When applying Normalize to multiple “tracks”, the effect will set each “track” to the same level.

The Amplify effect applies a specified amount of amplification (“gain”), so everything that it is applied to goes up or down by the same amount, retaining their relative differences in level.

Thank you all very much for your prompt response. Much appreciated!

Koz,

I tried what I believe you were suggesting: Record silence from an album/tape and then boost the volume above 0db to see if the sound waves stay around the ‘0 Line’ mark. I tried to do it as you suggested, i.e. record at +20db, but when I did, the wave diagram was pure red, so I scaled it down to +10db so that I could see something. The top image below is a more macro view of that recording, while the image below it is a more micro view of it. More or less, I would say that it seems to be remaining around the center line. Am I correct? From this, does it appear to you as though I have a DC Offset problem?


Below, I have also attached a WAV file photo sample of a portion of a rock album. The macro view shows a ‘silence’ intermission between two songs (notice the flat line between the blue waves) so that you can see if the waves are off center from the 0 Line. They don’t appear to be.

The micro view right below it, describes what I was talking about in my original post. Sometimes the waves seem to go ‘astray’ at the mega micro-view (see the red circled section), and I don’t know whether this means I have a DC Offset problem or not. As you can see, it is but a tiny fraction of the macro view shown above it.


In your opinion, guys, does this mean I’m okay? Are these simply examples of what you (DVDdoug) were talking about with regard to ‘real world’ sound waves not being perfectly symmetrical, or something worse? From what you are talking about DVDdoug, it doesn’t appear that there is DC offset throughout the signal; only in the circled section, if at all. Is that what it looks like to you guys? Is the circled section in the photo above even DC Offset, or just a bit of a harmless, ‘funky’ wave form?

Steve, if I am reading your response correctly, it sounds as though you are saying that if I do DC Offset through the ‘Normalize’ effect when I don’t have any DC Offset to correct, no harm can be done. Is that correct? And if so…why wouldn’t everybody do DC offset all of the time, since in best case scenario you might be correcting some DC offset you didn’t even know you had, and in worse case scenario, nothing at all (i.e. no harm)?

Additionally Steve, just to be sure I understand correctly about the differences between ‘Normalize’ and ‘Amplify’, it sounds as though you are saying, if I leave the Normalize Channels Independently box UNCHECKED, it is the same thing as Amplify with one exception: Amplify will increase the volume in the left and right channel (assuming two channel stereo) by the same amount in all tracks (i.e. songs), as measured from the loudest track’s (i.e. song) channel up to the amount desired (for example: -1db). By contrast, ‘Normalize’ will increase EACH track (i.e. song) to the volume desired (desired db level. i.e. -1db) as measured from the loudest channel in EACH track. Is that correct?

Finally, if I need to use DC Offset, per what I have shown you here, do I HAVE to use Normalize to amplify my recordings, or can I use Amplify instead? If I am correct in the paragraph above, after diligently reading the user manual, I would prefer to use Amplify after DC Offset. Is there any reason I shouldn’t be doing this?

Thank you all, in advance, for your input. I just want to be sure I am getting this right, before I spend thousands of hours recording hundreds of albums.

Kind Regards,

BlethiK82

In this screenshot:

the region that you have marked shows a “thump” or “hum” of low frequency noise. Looking at the time scale, one cycle (up, down and back up again) takes about 0.01 seconds (1/100th of a second), which is equivalent to 100 Hz (frequency = 1/wavelength). It’s not unusual to get thumps like that from vinyl records, so perhaps that tape was made by recording a vinyl record? I can only see the start of it, so I don’t know how long it goes on for. If it is not much longer than is shown in the image, it will probably sound like a low thump. If it continues for a long time then it will probably sound like a low hum, in which case it is most likely caused by “mains hum”.

Because it is very low frequency, and I presume at a fairly low level, it will not be very audible at normal listening levels. My guess is that that noise is present on the tape recording, and is not any kind of fault with your digital recording setup.

If it is a constant hum, it can be removed with a notch filter.
If it is just a slight “thump”, I’d probably ignore it.

Hi Steve,

Thank you for the reply. Now that I see my post, I can tell that some things are out of order. My apologies. I don’t know how that happened. Must be first-time user error. The order I wanted to post these images was as follows:

To respond to Koz, the images should have been the following recording of tape hiss/silence between songs:

The photo below is a tape hiss micro view set to +10 db, allow clipping:

The next photo is a micro view of the same tape hiss set to +10db, allow clipping:

The images you were commenting on, Steve, are below:

The first image (below) is a macro image of the end of one song and the beginning of another on a rock album. It is a direct recording from a record. It is not a tape, as was used in the demo for Koz above.

The next image (the one you were commenting on- FYI- I don’t know why it got cut off, but the circled section is about 2X the size shown in the cutoff image) is a micro of an actual song, it is not a view of the ‘silent’ space between songs, as is shown in the macro view above. My apologies for not making this clear. There is no ‘hum’ coming through my system.

Seeing all this, Steve, do I have anything to worry about with DC Offset? In addition: I would really value your input on my analysis of the differences between Normalize and Amplify at the end of my previous post. I will post it again in a follow-up reply, as I cannot view it on the post screen to copy and paste.

Kind Regards,

Blethik82

Steve…here is that re-posting of the follow-up section I was talking about. I would be interested in your commentary on my questions below when you have time.

Steve, if I am reading your response correctly, it sounds as though you are saying that if I do DC Offset through the ‘Normalize’ effect when I don’t have any DC Offset to correct, no harm can be done. Is that correct? And if so…why wouldn’t everybody do DC offset all of the time, since in best case scenario you might be correcting some DC offset you didn’t even know you had, and in worse case scenario, nothing at all (i.e. no harm)?

Additionally Steve, just to be sure I understand correctly about the differences between ‘Normalize’ and ‘Amplify’, it sounds as though you are saying, if I leave the Normalize Channels Independently box UNCHECKED, it is the same thing as Amplify with one exception: Amplify will increase the volume in the left and right channel (assuming two channel stereo) by the same amount in all tracks (i.e. songs), as measured from the loudest track’s (i.e. song) channel up to the amount desired (for example: -1db). By contrast, ‘Normalize’ will increase EACH track (i.e. song) to the volume desired (desired db level. i.e. -1db) as measured from the loudest channel in EACH track. Is that correct?

Finally, if I need to use DC Offset, per what I have shown you here, do I HAVE to use Normalize to amplify my recordings, or can I use Amplify instead? If I am correct in the paragraph above, after diligently reading the user manual, I would prefer to use Amplify after DC Offset. Is there any reason I shouldn’t be doing this?

Thank you all, in advance, for your input. I just want to be sure I am getting this right, before I spend thousands of hours recording hundreds of albums.

Kind Regards,

BlethiK82

Provided that you correct DC Offset correction to all of the audio in the project, and the correction is applied before any other processing, then that is correct.

If there is a small amount of DC Offset throughout a recording, then although it should not really be there, it does not do any harm. However, if you are editing and some of the audio has a bit of DC offset and some hasn’t, then at every edit where the DC level changes, there will be a click, as shown in the illustration that Koz posted:

Some people do, though it can take an appreciable amount of time on very long tracks, and it is “one more thing to do”, and it is pointless (does nothing) if DC offset is not present.

There is however a risk:
If the audio has already been processed, for example with a Noise Gate, it is possible that the could be DC offset during the actual audio, but no DC offset during silences. In this case, applying DC offset correction will cause the silent parts to be offset from zero. I’ve attached a short audio sample to demonstrate:

You have no DC problems.

You can’t see DC in waves with sound. That’s why I said use a lead-in groove or just not put the needle on the record. All the quiet waves you present settle nicely around the zero point.

We beat this up a while ago. Remove DC isn’t perfect and there are (rare) conditions where it can cause problems. The way Audacity Remove CD works is the least worst of all the available techniques.

There is a tendency to pile on effects, filters and corrections. Resist that tendency. The gold standard, admittedly difficult to do, is announce into a microphone, export the file, post to an audiobook and go to lunch. No processing. ACX has a quality control failure they call “Overprocessing.” New readers trigger that failure all the time.

My voice is naturally non-symmetrical. It just is. I found one broadcast announcer whose waves I can pick out just from looking. No sound.

Koz

Oh. One more. Phonograph turntables go through RIAA processing because bass notes do not fit in a groove. The cutting process reduces bass notes so they fit and depends on the player to boost them back to normal.

So every proper turntable has bass boost.

Koz

Koz and Steve,

I deeply appreciate both of your responses to my concern, but I really want to be done with this and have confidence that I have done what you have requested in enabling your ‘blessing’ that I should have no concerns.

I hate to keep harping on this, but I am simply trying to cover my bases to be sure I am doing this right.

Koz, My apologies, I didn’t understand that what you meant by ‘lead in groove’ was to let the record run with the needle lifted. In the following images I have done that…as you suggested. The first photo is a macro image where the recording volume on Audacity was set to maximum (100) and then the sound was ‘Amplified’ to to +10db (going higher makes the image waves far less easy to discern due to the ocean of red that occurs) using Audacity’s ‘Amplify’ effect.

The second image is the same recording in a micro view.

Is this showing what you are talking about? From these images, do I have a problem with DC Offset?

I realize that this does mean that there is an extremely low level of hum coming through the analog audio system (which I can live with, since it is only very slightly audible at the most extreme amplification levels- which I will never use)

But does this show a signal clean from DC offset? I just want to be sure. After everything I have read, and all that we have discussed on this forum, it sounds as though having a system good enough to not need DC Offset Correction is the best option. Does my sound card make the grade from what you see here?

Thanks in advance to each of you for your reply.

Kind Regards,

BlethiK82

The lead-in groove is the silent groove on the outer edge of the record before the first song. When you play it, it should be almost entirely silent. It’s a good sample of what the whole system is going to do.

At least for now, turn off View > Show Clipping. That will get rid of the red bars and bands. You should turn it back on for actual record transfers.

You are zooming in too far. Make it so three or four seconds of sound are on the screen at the same time. The goal, even if you boost the volume a bit, is for most of the blue wave to settle around zero on the left scale. Not too high and not too low.

Koz

Hi Koz,

Once, again, thank you for your reply.

Okay. I see what you are saying about the ‘lead-in’ groove. I chose to do the ‘Needle Lift’ option to record the silence. Below, I redid the recording, as you suggested, by turning off the ‘Show Clipping’ button and then ‘zooming out’ to allow for a four second view. The recording was done at 97% Audacity recording volume (Audacity’s meters don’t even register the hum (or whatever it is) coming through my audio equipment when record volume is lower than 70% of max).

The first image below is the recording amplified with Audacity’s ‘Amplify Effect’ to -1db. The second image, shows the the same recording amplified to +10db, allowing clipping.


Hopefully, Koz, these give you a clear enough picture of what the prospects are for my system having DC offset issues. Based on what I see here, I would say I’m okay. The waves seem to center nicely around the 0 mid-line. What’s your opinion?

If I understand the point of this test correctly, it is performed because DC offset, if present, is most prevalent in the ‘silence’ areas between songs. Is that correct?

Therefore, as I venture on through my recording project, If I want to avoid the presence of DC offset in my recordings, I should look for it in the ‘silence’ areas between songs. Is that correct, Koz?

Finally, a question using the correct recording volume in Audacity: When recording, if the recording volume bars are turning yellow, or sometimes red, but DO NOT hit the wall at 0db, I am not clipping the recording am I? Usually my the bars are maxing out at around -3db. The bars turn red at this point, only to indicate that you are getting CLOSE to 0db (and thus, potential clipping), correct? It does not actually indicate clipping, right?

Thank you, in advance, once again, for your reply. I value your input.

Kind Regards,

BlethiK82

The waves seem to center nicely around the 0 mid-line. What’s your opinion

I agree. All evidence is you have no D.C. Problems at all.

D.C. damage, if you had any, is there all the time, but it’s only easily visible when the show is quiet.

Koz

The digital system assigns numbers to your sound. If the meter goes all the way up or the blue waves smack 100%, the system runs out of numbers and just stops following the show. Worse, it starts making up its own sound. That’s the cracking, popping, and buzzing of overload.

Keep the high peaks no higher than about -6dB or so. Play the loudest part of the record briefly to set the volume, then don’t touch anything during the transfer.

Koz

Koz,

Thank you for your prompt response. Much appreciated.

I appreciate the hard lesson learned on DC Offset by running all of those samples. Thank you. I simply don’t have your experience on this type of software and needed some guidance there.

Now, on the volume issue, just so I am clear. Yes…that is what I am doing…setting it to peak around -6db, but sometimes it slips to -5, or -4, because…as I record, louder samples from other songs come in (songs that I did not ‘peak’ test) than was determined from the one song I did test for max record volume.

My question is: If the volume does get as high as -5db or -4db, for example, the meter bar tips start turning red, but…does that mean that it is actually distorting? I’ve always thought that distortion ONLY occurs if the bars actually HIT the wall (0db), in which case, the tiny red indicators go on at the far right of the record volume meter, showing that you have ‘clipped’ in your recording. If the tiny blue vertical ‘max’ indicator bars that measure peaks of the volume bars are actually getting to -5db or -3db, for example, though, it is not actually clipping, right? Am I correct in that assumption?

In the end, after the recording is complete and I have added the songs, I like to amplify it to -1db anyway (leaving the ‘Allow Clipping’ box unchecked), so that the volume of the tracks played through my stereo or MP3 player more closely matches those of my burned CD tracks. So, I am assuming that if it goes as high as -3db during the recording process, I can’t do any harm to the sound of the recording, since, in the end, the tracks that come from it will be boosted with the ‘Amplify’ effect to -1db. Am I correct in that assumption? Does that make sense to you?

Thank you, in advance, for your reply, my friend.

Kind Regards,

BlethiK82

The meters are set to start turning colors as a warning and they turn red before actual sound damage.

After you capture your work, you can do almost anything you want. Almost. It makes everyone uncomfortable, but you could Amplify or Normalize to 0 rather than a slightly lesser value. That’s fine until you decide to post an MP3 or other compressed sound format to a friend, on-line or personal music player. Because compression re-arranges sound within a song, it’s completely possible the new song is going to overload—try to go into positive dB numbers. So you should stay away from those last one or two dB.

Nobody can hear those anyway.

Overload is an interesting programming challenge. It’s impossible. Since there are no positive dB values, programmers can’t measure them to know when something is overloading. And 0 is a valid sound value.

So now what?

The best we can do is anticipate overload.

Koz

Hmmmm. Okay, Koz.

Your point is well taken, my friend. I never would have thought of that. So, you are anticipating me correctly…I am turning them in to WAV tracks where I save them at three levels:

1. 32Bit 44,100 Hz files (a higher level file that I could work with if I needed to)
2. 16Bit 44,100 Hz files (in the event I want to make a CD someday), and
3. MP3s that I can stick on my I-Pod.

Does that rule ONLY apply to the MP3 format? In other words, can I set my Audacity ‘Amplify’ effect to -1db for the WAV files and say…-3db or -2db for the MP3 files without damaging either of them?

As I mentioned, I have been using -1db, and was happy with it, but now having conversed with you, it sounds as though I may be damaging my MP3 music files in doing so.

What would you say is the best setting for MP3 Amplification? If I set it to -2db am I going to be okay, or should I bump it all the way back to -3db?

Additionally, are the WAV files okay if ‘Amplified’ to -1db?

Thank you, again, Koz. Very much appreciate your patience.

Kind Regards,

BlethiK82

Running everything at -1 is probably fine. If you think you made a damaged MP3, open it in a fresh Audacity and see if you get the red lines in the blue waves. Set View > [X] Show Clipping.

MP3 is the only one of the three that’s subject to “magic” volume shifting. The other two are perfect formats and stay where you put them.

Koz