Audio Recording Myth Debunked!!! (preamp noise misconception) 

Absolutely, setting your gain is always a balance act. To much gain and you risk clipping. To little gain and you will increase the noise floor. So you always record with the highest gain possible but still leave enough headroom to not clip.

You're welcome :)

Danke für die netten Worte! Freut mich, wenn dir meine Videos gefallen :) VG Julian

@@JulianKrause Ehre, wem Ehre gebührt! Ich stehe trotzdem etwas auf dem Schlauch. Als Uni-Dozent muss ich seit letztem Semester meine Lehre als Video aufnehmen. Um den Ton zu optimieren habe ich mir ein Rode Procaster angeschafft. Zum Vergleich habe ich zwei Interfaces auf dem Tisch: Focusrite Scarlett Solo und Behringer UMC204HD. Stellt sich die Frage, welches ich behalten soll... So aus dem Bauch heraus - wozu würdest du mir raten? Bekomme leider keine standardisierte und demnach ordentliche Messung hin. Vielen vielen Dank!

Hey, zugegeben für Sprachaufnahmen sind beide Interfaces ungefähr gleich gut und ich bezweifle, dass du da einen Unterschied hören wirst. Daher würde ich dir empfehlen das Interface zu behalten, das die Features hat, die du brauchst. Ich würde es ale abhängig davon machen, ob du einen oder zwei Mikrofoneingänge brauchst und auch die Baugröße kann eine Entscheidungskriterium sein.

@@JulianKrause Vielen lieben Dank für deine Einschätzung. Ich habe in der Zwischenzeit herumprobiert und muss sagen, dass ich beim besten Willen in meinen Lehrvideos keinen Unterschied feststellen kann. Somit werde ich beim Behringer bleiben, weil es sowohl einen Poti für den Kopfhörer als auch einen für die Monitore gibt. Für meine Anforderungen ist das einfach praktikabler. Dein Tipp bestärkt mich in meiner Entscheidung. Dankeschön :)

Thank you!

whut. its not linear? holy shid

@@lleonardi the ADC noise is constant up to a certain point where the preamp noise will take over

The pad attenuates the signal before it is passed on to the preamp. So, if you have the gain on the preamp turned down all the way and your signal is still to strong, you could then use the PAD to lower the isgnal further. Cheers!

But can be, that the interface has sweet spot, I had Focusrite which had the best signal to noise ratio somewhere between 75 and 90% of gain.

@@Mr_G So in this case, if you record to 100% maybe you get a noisier signal that if you do it at 90%, right? Because the ratio becomes lower

Nope, proportional is the correct word. What I meant is that if you lower the gain by 1dB the noise also drops by 1dB. And when you lower the gain by 2dB the noise also drops by 2dB. So, when two values are related to another with a constant ratio it is called proportional. Linear was not the right word there but logarithmic is also not correct. Glad you liked the video and sorry for the confusion!

You are correct when you refer to scaling, this is proportional in part of the context you were talking about and in certain respects can also be called linear when comparing two things affected by the same scale, or proportion. There were certain parts where you were talking about the scaling of the gain, that's what I was referring to as log. Cheers, Dave.

most small audio interfaces/preamps shouldn't be cranked 100% anyways. There's always a sweet spot/range for optimal results.

but by increasing it in post, you're still gonna get the same amount of noise. did you listen to the video? lol

true.. When I listen to my favourite artists early work (Elliott Smith in the nineties), lots of his songs have a very obvious background hiss/white noise. He didn't seem to care at all, WHen there is enough sound it totally drowns out, I imagine because it's such a wide spectrum, it can "glue" to any noise, high or low. As a listener, these less than ideal Elliott recordings are among some of my favorite songs. People are overly critical on their own mixes, for things that aren't important enough to waste time on.. It just stifles creativity, which is a fleeting thing as it is. The other night I got way too deep into a mix, and it was sounding so bad, the noise, the tracks wouldn't blend.. Next day I \play it back and ti was sounding fine. Made a few panning tweaks and suddenly it was airy and balanced. Objectivity is so hard to maintain in mixing/writing. Frequent breaks help, not trying to compresss the same verse perfectly for hours on end. You dont know what you're even hearing at that point.

Hey, I use a hypercardioid small diaphram condensor mic boomed above me. The SM7B would be way to heavy and would not really sound that great at the distance required to get it out of the shot.

I assue you mean the "match loudness" window in Adobe Audition. There is no difference if you apply digital gain manually or amplify the audio with the "match loudness" window. In both instances digital gain is applied to your audio. When you raise your gain digitally it will amplify the noise floor of your recording by the same amount. There is no way you get around that.

@@JulianKrause Thanks Julian

The background noise will still be the same in relation to your voice, regardless of your gain setting. If you turn the gain down by 10dB, the background noise will be 10dB lower, but so will your voice. The only solution to increase the signal-to-noise ratio regarding the background noise, is to either make your voice louder (use it louder and/or get closer to the mic), or reduce the background noise acoustically.

Julian, just found this comment from you below. Was a bit surprised to hear this, but if this is the case ... my comment is moot: "Hey, that's exactly why it is a myth, because preamps do not become disproportionately more noisy at higher gain settings. The signal-to-noise ratio of a preamp is counterintuitively the best at the maximum gain settting. In all the years that I have done preamp testing in I haven't found a single device where this is not the case."

OK. I've just found this, which covers the lot! ... thanks again! ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-beXVfl1TSD0.html

lowering the gain and getting closer to the microphone works yeah

*needless to say, that with the hi-z engaged, my potentiometer is at 0 and still able to clip. According to the interface's indicator. In the DAW there's no clipping.

The clipping indicators of preamps are often set several decibels lower than the point where it really starts to clip, so it might not be alarming to see it flash every now and then. If you set the gain as low as possible and play as hard as you ever will, does the signal clip at the daw? If not, and if the signal sounds as it should, I'd say you're good to go.

@@JiihaaS I've sent it back months ago :)

I'm a beginner in this but I think the crucial parameter is "Maximum input level". I had UR22C which has max input lvl in hi-Z mode +9.0 dBV, same as UR24C. It kept clipping during palm mutes, I don't have too hot pickups (SD TB-4) but they're very dynamic. With EMG 81 it was fine. Then I tried UA Volt 2, which has max input lvl +12.5 dBV and it was significantly better but still not safe enough. I finally got Motu M2 which has +16.0 dBV and it's the perfect spot for me. DI box could be a solution with other interfaces but I wanted to keep my signal chain as simple as possible.

yeah it would on quiet mics. The DM1 dynamite is cheaper than both, and equally effective.

You could be having a grounding wire issue. Or maybe just face a different direction than your amp, that can reduce feedback. Or if you have noisy pickups its a matter or learning to tame them. Noise gates can help with that.

It can but it is not guaranteed. Depending on the implementation of 32bit, there might actually be no analog gain anymore you can set. In that case it does not matter where the digital trim is set to, as you can always increase or decrease the audio in post without without affecting the signal-to-noise ratio.

Hallo Björn, genau diese Gradwanderung macht man beim Einstellen des Gains. Wenn zu hoch einpegelst, kann es zu Clipping und Verzerrungen kommen. Wenn du zu niedrig einpegelst, fängt es an zu rauschen. Der Sweetspot liegt heutzutage bei vielen Geräten bei um die -12 dBFS. Wenn du dein Gain so einstellst, dass dein Level im Durchschnitt um die -12 dBFS peakt, erreichst du einen optimalen Signal-Rauschabstands. Eine leichte! Verstärkung in der Nachbearbeitung ist völlig in Ordnung und in den meißten Fällen standard.

@@JulianKrause OK, danke .. d.h. normalisieren auf 97-100% am Ende ist durchaus ok, ja?

@@B-S-K Absolut!

Every piece of equipment produces an electrical hum. It's usually quiet... But if you turn it up enough you will hear it. You want your signal to be as far away from that sound, that noise floor, as possible. That way you can edit the signal after, add volume, reduce volume, without worrying about the noise floor being raised too high. If you record too quiet, when you go to increase levels in post, you wil increase the noise floor, and soemtimes its audible by the time you get your signal close to 0 db, which is the clipping limit.

Hey in the video I talk about gain setting on recording device, not amplfiers. I'm not sure what you mean by "more powerfull speaker". A speaker is a passive device. Do you mean the speakers sensitivity or the max power rating?

Better mic pre :p

i have a weird ass hum in my interface, so i think i could relate to what you're talking about. i just lowered the output volume of the codec, and brought up the gain. my output volume is 50%, and my gain is around 1:00. if this didn't help, then i didn't understand the question

@@voicehead It was actually mic's fault, I have a Rode NT1 and it's awesome, making wonders ^^

@@voicehead likley due to low quality xlr cables and interfearence...

@@peachesasmr9366 the hum is present with and without cable. but i determined it's just a problem with the output, because it was only present when the output volume is over 60

Yes, a compressor lowers the peaks in your audio and brings them closer to your noise floor. If you then use the makeup gain in the compressor you also increase the noise. So, compression decreases the signal-to-noise ratio and should thus be used carefully.

A linear relationship would still look linear on a loglog graph, but the issue here isn't the logarithm, it's the division. We can calculate the signal to noise ratio (SNR) in decibels using the following formula: SNR = 10log₁₀(P_{signal}/P_{noise}) If we assume that the SNR remains constant (in reality it doesn't, it increases slightly as we increase the gain), we need the division inside the logarithm to remain constant, meaning that the power of the signal and the power of the noise must be scaled by the same multiplier. If we simply add the same value to each power level, the relationship is thrown off which means that the SNR has changed, which isn't allowed according to our assumption. Even if we disregard our assumption and accept that SNR increases with more gain, the power levels still have to be scaled by a multiplier, they can't be increased linearly. For example, if we set P_{signal} = m and P_{noise} = n, where m > n (if your noise is louder than your signal, you have bigger issues). If we assume that the amp increases power linearly by adding a power level of x to each power level, the expression for the SNR becomes SNR = 10log₁₀((m +x)/(n + x)). Since m > n, this expression is continuous for x >= 0, meaning that we can examine what happens as X approaches infinity. Firstly, we must reassure ourselves that the SNR is positive when x = 0. We can do this by using m > n again, which guarantees that m/n > 1 and that the SNR is positive. Now, let's see what happens when x approaches infinity. lim_{x -> ∞} SNR = lim_{x -> ∞} 10log₁₀((m+x)/(n+x)) Using L'Hopital's rule, we can see that lim_{x -> ∞} SNR = 10log₁₀(1) = 0, which means that the SNR decreases as x grows, which is not what happens in reality. That's why we don't add gain linearly. Hilariously, if we let x decrease, the SNR will approach infinity as x approaches -n from the positive side. This also doesn't work in real life. Analysing the limit of the SNR when x is applied as a multiplier instead is trivial since the x's will just instantly cancel out, leaving us with the same expression for the SNR as we had at the beginning, proving that the SNR is constant.

It is the exact same scenario. You should still set your gain to a proper level and NOT back it off any further. Turning down the gain lower than the proper setting, regardless on which audio recording device, does not lead to a better signal-to-noise-ratio. Of course, sometimes it is a bit difficult to judge what a proper gain setting is. For dialogue I recommend you set your gain so that you are peaking between -18 and -12dBFS.

@@JulianKrause Appreciate the response and the continued clear and concise content you put out. Kudos!

@@JulianKrause So this means the argument of some preamp getting noisier at 100% is false? I mean, in reality, the signal to noise ratio would not change from 90 to 100% of preamp gain? Thanks!

@@infectropodo Yes, that is correct.

Hey, that's exactly why it is a myth, because preamps do not become disproportionately more noisy at higher gain settings. The signal-to-noise ratio of a preamp is counterintuitively the best at the maximum gain settting. In all the years that I have done preamp testing in I haven't found a single device where this is not the case.

You are not the only one

Hey, which mic are you using?

Julian Krause C-1 behringer but I do not plug a mic to the interface yet it has that hiss sound when I press the record button on my DAW.

@@apollotwin Hey, that's very normal. An interface will have much more noise when no mic is plugged in. Please plug in the mic and record your audio and you will notice, that the noise is much quiter.

Julian Krause wow thanks a lot much appreciated!!! Big ups man.

The noise test was done with a 150Ohm dummy mic. This plug has a 150Ohm metal-film resistor across pin 2 and 3. When connected to a preamp this will result in about as much noise as a dynamic mic will produce with this combination. In my review of the UMC202HD I concluded, that you do not need a "pre-preamp" for this device because the UMC202HD has an Equivalent Input Noise (EIN) of -129.1dBu. When you use a Fethead with pretty much any interface you will have an EIN of about -130dBu. So, the preamp in the UMC202HD is already so low noise, that adding a Fethead will only result in a 1dB noise reduction. In the real world this difference is mostly inaudible and that's why I said that a "pre-preamp" for the UMC202HD is not necessary. This does not mean that a "pre-preamp" is always useless. In my review of the Fethead for example I used it on a Zoom H5. The H5 has an EIN of -120.5dBu which means that with a Fethead you can lower your preamp noise by about 10dB. That's huge! If a "pre-preamp" is worth it really depends on the device it is used with. Now onto your question why there are tests that show that a "pre-preamp" does work. Again, if they used a device which has a high amount of preamp noise then a Fethead or Cloudlifter will of course lower the noise floor. But some tests show some improvement even on devices, where it shouldn't make a big difference. I think this comes from not matching the gain properly between the tests. You have to match the amplification of both recordings in order to make a meaningful comparison. I do controlled tests to come up with reliable and consistent data. If I had to guess, I would say that the testing methodology of some people might introduce some variance which leads to the misinterpretation of the data. Shoot me a link to such a test and I'll have a look. Maybe I can tell you exactly what is going on. Cheers!

@@JulianKrause So this means the UMC202HD (-129EIN) and the Focusrite 2i2 (-128EIN) both with around +50db of total gain, would get a cleaner signal with a Shure SM7B than for example an Audient ID44 (-126EIN and +60db) which is supposedly a higher end interface? I'm kind of confused. Thank you!

@@infectropodo The EIN eliminates the gain from the equation, so it is directly comparable. So, yes -129 dBu is less noise than -126 dBu. BUT, my measurements are a-weighted, that's why it is dBu(A). As far as I know the EIN in the specs of the iD44 is unweighted, so it is naturally going to be higher. Because of this you cannot directly compare the EIN of the specs sheet from the Audient iD44 to my measurements. But you're in luck, I measured the Audient iD4 and it has an EIN of -130dBu(A). I expect the iD44 to have a similar amount of noise. This means that in the real world the iD44 is probably going to have about 1dB less noise than the UMC202HD.

@@JulianKrause That is very informative, didn't thought about the A weighting. Now it makes more sense, looking forward to your Audient review, also a SM7b noise test would be great too since there is a lot of people like me wanting to upgrade to that mic and worrying if the preamps are good enough for it! Every one is talking about the CloudLifter!! Thank you very much!

@@infectropodo I can say that the iD4 has such a good low EIN that you really don't need a Cloudlifter. And I guess it will be the same for the iD44. Cheers!

I set my gain so that the peaks hit on average between -18 and -12 dBFS. This will give you optimal performance in most scenarios.

@@JulianKrause What audio level is necessary in Audition before uploading to RU-vid?