Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 7, 2021 15:39:48 GMT -6
www.analog.com/en/analog-dialogue/articles/understanding-switching-reg-output-artifacts.htmlPSA time guys. Switcher artifacts will kill your audio. In addition to all the switcher noise and high frequency hash from cheap switching mode power supplies, when they're dying, they can leave these little spurs everything that can kill your recording. You'll have to manually edit out the spurs like it's tape. Izotope or spectral editing will not get rid of them. I had to split and delete the spurs out at the zero crossings and just soft clipped the other distorted artifacts on transients with the Molot GE and VCL-373 limiters. This track I'm mixing was engineered by a fairly well known producer a few years ago with dying equipment. Never use dying equipment. Dying or defective switching mode power supplies will destroy your recordings and make your life hell. Some manufacturers are notorious for their poor power supplies and only a few of them sell spare parts for diy repairs. Repair your gear or get better gear. These switcher spurs were all from the same kick track. These were all done in one or two takes and nothing was comped. I have two more kick tracks like this bash my brains out on after this. Another big one: A smaller spur: 
A very tiny one: 
|
|
|
|
Post by EmRR on Feb 7, 2021 15:46:35 GMT -6
ugly!
|
|
|
|
Post by matt@IAA on Feb 7, 2021 16:19:29 GMT -6
Hm. What frequency was that showing up at?
Trying to figure out how a switch spur shows up in audio. Cant be a class A circuit, cuz then the ripple would be constant. But you’d expect an op amp or something to have a ton of filtering + PSRR so that high frequency crap like that wouldn’t make it into audio.
|
|
|
|
Post by svart on Feb 7, 2021 17:06:41 GMT -6
Looks like a normal kick track to me..
Anyway, you can't really discern switching harmonics modulated into an audio track in time domain. You have to use frequency domain for that.
|
|
|
|
Post by svart on Feb 7, 2021 17:14:46 GMT -6
Hm. What frequency was that showing up at? Trying to figure out how a switch spur shows up in audio. Cant be a class A circuit, cuz then the ripple would be constant. But you’d expect an op amp or something to have a ton of filtering + PSRR so that high frequency crap like that wouldn’t make it into audio. You can't, at least not like this. You won't see a "spur" on a signal in time domain, you'd simply see a modulated waveform with higher frequency noise. I think this is a case of a little knowledge being dangerous. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 7, 2021 17:17:46 GMT -6
Hm. What frequency was that showing up at? Trying to figure out how a switch spur shows up in audio. Cant be a class A circuit, cuz then the ripple would be constant. But you’d expect an op amp or something to have a ton of filtering + PSRR so that high frequency crap like that wouldn’t make it into audio. I presume it was some interface with a dying power supply and dying parts. It could just be Logic weirdness? Some of the earlier recorded songs for the record in the session do not have BIP in the raw multitrack file names so they might not have Logic weirdness. They still have weirdness but nothing as bad as the later ones from the session with the "bounce in place" indicator in the filenames.
The smallest spur is 14.7 khz peak  Others: |
|
|
|
Post by svart on Feb 7, 2021 17:20:01 GMT -6
Honestly the track looks rather normal to me, but the initial transient had quite an amount of overshoot.
It looks more like when someone "upgraded" to a rf-capable opamp and didn't properly dampen the feedback bandwidth and it's on the edge of oscillation.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 7, 2021 17:34:53 GMT -6
Honestly the track looks rather normal to me, but the initial transient had quite an amount of overshoot. It looks more like when someone "upgraded" to a rf-capable opamp and didn't properly dampen the feedback bandwidth and it's on the edge of oscillation. Yeah it's normal except for the overshoots on heavier transients and these little spikes. The earlier recorded kick tracks just have transient weirdness. The later ones are covered in the little spurs and almost full scale overshoots. It seems to have been recorded on a bunch of older M-Audio stuff. I could see it being a poorly done mod or dying equipment. |
|
|
|
Post by svart on Feb 7, 2021 17:39:07 GMT -6
There's also no way you're getting smps hash down in the hundreds of hz.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 7, 2021 17:45:14 GMT -6
There's also no way you're getting smps hash down in the hundreds of hz. Not here on the kick but all kinds of weird stuff can happen with bad switching mode power supplies. I’ve only had similarly weird stuff, but not this weird, from an RME interface that needed a new power supply |
|
|
|
Post by svart on Feb 7, 2021 17:57:49 GMT -6
There's also no way you're getting smps hash down in the hundreds of hz. Not here on the kick but all kinds of weird stuff can happen with bad switching mode power supplies. I’ve only had similarly weird stuff, but not this weird, from an RME interface that needed a new power supply I spent 4 years designing switch mode supplies and motor controllers. I now design rf with a particular focus on signal integrity. One of the daily struggles is keeping smps harmonics out of my signals. I've seen all kinds of strange stuff, but I've never seen what you're seeing. Just saying. |
|
|
|
Post by svart on Feb 7, 2021 18:15:58 GMT -6
Those overshoots remind me of reflections on cables and traces with poor impedance matching, which is something I see a lot of.
It's also dangerous of me to even suggest that since it's a common problem with engineers "seeing" the common things they deal with in every situation.
It ruins the objectiveness.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 7, 2021 18:17:19 GMT -6
Not here on the kick but all kinds of weird stuff can happen with bad switching mode power supplies. I’ve only had similarly weird stuff, but not this weird, from an RME interface that needed a new power supply I spent 4 years designing switch mode supplies and motor controllers. I now design rf with a particular focus on signal integrity. One of the daily struggles is keeping smps harmonics out of my signals. I've seen all kinds of strange stuff, but I've never seen what you're seeing. Just saying. I don't know what's going on here. I thought it might be switcher issues. I've only experienced weird stuff like this (but not really like this) from dying equipment. Nothing this weird but weird behavior. Carelessness + dying cheap gear = more weirdness. The guy who recorded the band never did anything else like this and told them that it could all be fixed in the mix, ie replace everything with samples because he flat-lined clipped some tracks |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 7, 2021 18:25:57 GMT -6
Thanks svart for your replies. I guess the only way to mitigate it is to edit out the worst of the worst like I've been doing |
|
|
|
Post by the other mark williams on Feb 7, 2021 18:47:01 GMT -6
Hm. What frequency was that showing up at? Trying to figure out how a switch spur shows up in audio. Cant be a class A circuit, cuz then the ripple would be constant. But you’d expect an op amp or something to have a ton of filtering + PSRR so that high frequency crap like that wouldn’t make it into audio. I presume it was some interface with a dying power supply and dying parts. It could just be Logic weirdness? Some of the earlier recorded songs for the record in the session do not have BIP in the raw multitrack file names so they might not have Logic weirdness. They still have weirdness but nothing as bad as the later ones from the session with the "bounce in place" indicator in the filenames. I don't see why this would be Logic related, unless it's the original recordist using Logic poorly, which wouldn't be Logic's fault. The "BIP" (bounce-in-place) indicator from the filename suggests to me that he (the recordist) wanted to render plugins or comp multiple takes into a final comp take and render that. I wouldn't be surprised if those are artifacts from where he made a poor cut from one take to another (i.e., not at a zero crossing). Either that, or some random plugin he used (and then rendered) causing a spike. Of course, regardless of how he did it, the fact is you're having to clean up the mess now, which sucks. |
|
|
|
Post by tkaitkai on Feb 7, 2021 21:46:28 GMT -6
@tomegatherion (and svart , regarding the smps harmonics) Any chance you guys could share some clips of this stuff? This is really fascinating and I'd love to know what to look out for in the future. |
|
|
|
Post by EmRR on Feb 7, 2021 22:39:01 GMT -6
In DP if you instigate a new plug on a track while recording you get mutes and spikes like this. It’s not a path problem because the plug is playback, but the memory demand punches a hole in incoming audio. Seems like bad code to me, there’s no observable hit on CPU or RAM.
|
|
|
|
Post by askomiko on Feb 8, 2021 3:58:30 GMT -6
Ok, so what causes that thing you mention? I saw similar spikes on initial transients of some lead vocal consonants on a track I got yesterday, I thought it was vocal compressor clipping badly or something weird. I remember editing stuff like this in Wavelab around 1999, it had a function where you could just draw the waveform with mouse. Does it still have that? Does some other, cheaper program have that nowadays? Edit: OK now looking at your examples they are different, never mind. The first image of yours looks the same as incorrect sync between units, the rest are different.  |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 8, 2021 10:09:25 GMT -6
Ok, so what causes that thing you mention? I saw similar spikes on initial transients of some lead vocal consonants on a track I got yesterday, I thought it was vocal compressor clipping badly or something weird. I remember editing stuff like this in Wavelab around 1999, it had a function where you could just draw the waveform with mouse. Does it still have that? Does some other, cheaper program have that nowadays? Edit: OK now looking at your examples they are different, never mind. The first image of yours looks the same as incorrect sync between units, the rest are different. Spectrally edit it out by hand or use Izotope. If that fails, limit or clip it. If that or automation fails, cut out the transient at the nearest zero crossings. |
|
|
|
Post by christopher on Feb 8, 2021 10:14:37 GMT -6
Is this kind of thing from USB 1/2 interfaces? I can’t remember since I didn’t use those very long, went back to FireWire.
|
|
|
|
Post by jeremygillespie on Feb 8, 2021 11:46:19 GMT -6
So, what does this sound like?
|
|
|
|
Post by svart on Feb 8, 2021 12:18:01 GMT -6
Ok, so what causes that thing you mention? I saw similar spikes on initial transients of some lead vocal consonants on a track I got yesterday, I thought it was vocal compressor clipping badly or something weird. I remember editing stuff like this in Wavelab around 1999, it had a function where you could just draw the waveform with mouse. Does it still have that? Does some other, cheaper program have that nowadays? Edit: OK now looking at your examples they are different, never mind. The first image of yours looks the same as incorrect sync between units, the rest are different. If you notice the squished initial transient in the picture, that's most likely a clipped mic output. |
|
|
|
Post by svart on Feb 8, 2021 12:33:13 GMT -6
Hm. What frequency was that showing up at? Trying to figure out how a switch spur shows up in audio. Cant be a class A circuit, cuz then the ripple would be constant. But you’d expect an op amp or something to have a ton of filtering + PSRR so that high frequency crap like that wouldn’t make it into audio. I presume it was some interface with a dying power supply and dying parts. It could just be Logic weirdness? Some of the earlier recorded songs for the record in the session do not have BIP in the raw multitrack file names so they might not have Logic weirdness. They still have weirdness but nothing as bad as the later ones from the session with the "bounce in place" indicator in the filenames.
The smallest spur is 14.7 khz peak Others: These look like they might lay near harmonics of rectification, i.e., 360hz, 720hz, 1080hz, 1560hz, and so on. I'd bet that this unit has a linear supply with a bad/leaky diode in the rectification section and caps going bad due to the ripple heating them up. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 11, 2021 13:52:11 GMT -6
I presume it was some interface with a dying power supply and dying parts. It could just be Logic weirdness? Some of the earlier recorded songs for the record in the session do not have BIP in the raw multitrack file names so they might not have Logic weirdness. They still have weirdness but nothing as bad as the later ones from the session with the "bounce in place" indicator in the filenames.
The smallest spur is 14.7 khz peak Others: These look like they might lay near harmonics of rectification, i.e., 360hz, 720hz, 1080hz, 1560hz, and so on. I'd bet that this unit has a linear supply with a bad/leaky diode in the rectification section and caps going bad due to the ripple heating them up. Perhaps the kick was frontloaded was something bad on the way in. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 11, 2021 14:19:44 GMT -6
tkaitkai jeremygillespie small spike    Basic Instinct ice pick    Sound files of these but keep in mind these have been edited out of 24-bit fixed tracks in 64-bit float and then dithered down to 24-bit fixed again through PSP X-dither with no noise shaping on high dither and with clip on. we.tl/t-tHPMuQPjyZEditing out the spikes at the zero crossings is what I usually had to do to neutralize them. The Basic instinct ice pick required me to remove the entire hit.
|
|
