I am designing/DIYing a headphone amp. It has a LDO regulator which has lower PSRR than regular regulators, can it effectively filter out the broadband switching noise generated by switching P/S?
Can one hear the noise of a switching power supply?
- 737 Posts. Joined 3/2012
- Location: Scotland
- Select All Posts By This User
Not as well is the short answer. Dont like digital switching supplies It limits the frequency response of the amp because a lot have switching frequencies below 100KHZ so your FR must be lower not to "pick up" this noise.
- 1,098 Posts. Joined 5/2011
- Location: Devon, UK
- Select All Posts By This User
The answer is ... it depends.
The question is too broad. Which LDO regulator? Which SMPS? With what switching frequency? What do you mean by 'effectively'?
You can calculate or estimate the effect. You can build it and measure or listen to it. You can adopt a PSU known to work effectively with the kind of headamp you are proposing.
An LDO regulator will provide some attenuation of switching noise when used after an SMPS.
Thank you for all your answers. I know it's a broad question, that's why I posted here to get the answers. I have asked many friends and classmates, they all said that even the big brands are using switching P/S, it must be OK. I think the answer depends on how "deep" you are in audio. But for engineering, it's actually a compromise, as the weight of an EI core transformer scales up heftily with it's capacity, especially for Class A amps, the air shipping cost will be to high.
The thing is, those companies' engineers have done the engineering. (One hopes, anyway.) They didn't just ask a broad question, then someone said "yes," so they whacked a switcher in there. They knew the switching frequency, they knew their budget for filtering, they knew the amplifier's FR and PSRR...
Most importantly, I think these other engineers knew their quality goal: they knew what success would look like when they achieved it.
So, you can blindly follow these other engineers if:
- you're using the same SMPS
- the same filtering
- the same amplifier stage
- and have the same quality goal/definition of success
The fact is, it's simply a lot easier to succeed in DIY with a linear power supply.
On the other hand, it's a lot easier to succeed in the high-volume commercial world with a switcher, because size, cost and weight matter more in volume.
Let's say going with a switcher saves you $15 in parts per unit. Plus, it might save you another $1/unit in packaging, warehousing, and shipping due to the lower volume and weight. If you did it right, it might have cost you $5,000 of engineering time to solve all the problems you bought by choosing a switcher. If you sell 100,000 of these things, that NRE only amounts to a nickel per unit, so it's a clear win.
Now consider the same problem in DIY terms.
If you choose not to solve the problems you bought along with that switcher, you might ruin a perfectly good audio circuit. The easy fix is to buy/build a new linear supply, so you've wasted ~$10-25 on the switcher. Clear loss.
If you do choose to solve these problems, the $5000 in NRE time will look different. Let us say it amounts to three months of weekends instead, in DIY terms. And in the end, you've only solved the problem for one unit, unless you're making a project for others to build, in which case you'll still be lucky if the volume goes into the single-digit thousands. Is solving SMPS problems your idea of fun? If so, great, you have a new fun DIY project that'll keep you busy for months! If not, are you willing to trade three months of weekends for ~$15, a few ounces, and a few watts?
To answer your title question, yes, I have indeed heard the noise of a switching power supply. It was a $1 DIP-8 based switching controller that I DIY'd up on a solderless breadboard. It had a 10 kHz switching frequency, and I didn't filter it beyond what was recommended in the switching controller's data sheet. In other words, I'd done about everything I could to fail, and I succeeded in failing. I heard a high frequency buzz in the headphones when I turned the amp on. I chose not to spend the three months of weekends on the SMPS, to make it achieve the audible performance of a linear.
Is that another way of saying, "Don't use switchers?" No, it means, "Do the engineering."
I have direct knowledge of at least three highly-regarded headphone amplifier designs that used switching power supplies. And in all cases, they all got negative feedback (the other kind) when people realized it, because there is a knee-jerk reaction to SMPSes in the audio world. So my second piece of advice is, if you use a switcher, and you've done it right and know it, don't brag about it.
Edited by tangent - 3/1/13 at 9:37am
I forgot to address this in the previous post.
The PSRR of the regulator almost doesn't matter. Or rather, it matters at low frequency, which means it matters in purely linear-regulated power supplies.
What maters when trying to filter switching power supply noise with a linear post-regulator is the latter's frequency response. Take an LM317: beyond about 1 kHz performance drops off steadily, to the point where it might as well be a wire for all the good it will do to filter noise. Here's the relevant curve from TI's LM317M datasheet:
Most switchers operate in that 30+ kHz region you see where the curve starts flattening out. And they don't show the higher frequency performance, so you can't assume that the curve continues flat along at ~20-30 dB down, as TI's markerters hope. At some point, the regulator is simply going to run out of bandwidth, so its regulation can't be > 0 dB.
Edited by tangent - 3/1/13 at 10:44am
Thank you a lot. Your answer(s) clarifies my doubts. They are well thought out and opinions well supported. With it as the reference let me draw my own conclusion on this subject, correct me if I am wrong:
A well informed (audio) design engineer knows what he is doing. It's just the scale of production (or something else) decides which one, switching or linear, to use. A well designed and manufactured device with SMPS should be OK to use, only that if you replace the SMPS with a linear PS and have a thorough AB test, then you will hear the difference. May be more transparent, more details, more solid bass etc. The only way to achieve PERFECTION is DIY, when you can ignore all other restrictions and only go for the best!
No. A properly implemented power supply, regardless of method/topology, will not impair the function of the device.
A bad SMPS implementation is bad.
A good SMPS implementation is good.
A bad linear supply implementation is bad.
A good linear supply implementation is good.
What the smuggler said.
Look at it this way: if one type of power supply were unconditionally better than the other, we'd only have the one. We have both kinds because they both continue to have raisons d'être.
Here's a thought to consider, just off the top of my head: an SMPS is inherently a wideband device, whereas most off-the-shelf linear regulators focus on AC line frequencies and their first half dozen or so harmonics. Thus, you can imagine situations where an SMPS might actually perform better than a linear.
I don't mean for you to take that too seriously. As I wrote above, for DIY audio purposes, linears are pretty much where it's at. But, never get caught up in an absolutist mode, where you exclude options needlessly. SMPSes have their place.
I think I understand what you mean.
Corrections: PSRR is for op amps, I actually meant Ripple Rejection for LDO's, the LDO I used is LM2941T. I meant hear it in the headphone.
I don't own expensive equipments like Spectrum Analyser etc. Therefore I have to rely on college electronics theories, datasheets, googling, my ears and of course this forum.
I don't think it's wrong to use the term PSRR for linear regulators. Linear regulators contain op-amps. It was perfectly clear to me what you meant.
The reason there's a separate term has more to do with how and where you use naked op-amps vs regulators than with what is being measured, or how. The term ripple rejection suggests that they're concerned with reducing the ripple from a full-wave rectifier, whereas the term PSRR suggests rejection of any noise that comes from the power supply, after any regulator. Both terms refer to the same sort of measurement: dB of output noise as a ratio with noise from the supply rail.
As for hearing the noise in the headphones, as I said above, yes, you can hear the noise in the headphones in some circumstances.
If your SMPS has a 100 kHz switching frequency and there are no sub-harmonic effects going on, then you're not going to hear it even with a wideband headphone amplifier without feedback, hence no PSRR. That's not to say the noise isn't present, however.
Edited by tangent - 3/7/13 at 7:17am
- 2,885 Posts. Joined 8/2003
- Location: Hartford
- Select All Posts By This User
- 28 Posts. Joined 12/2011
- Location: Bremerton WA USA
- Select All Posts By This User
As an amateur radio operator, I have more than a few receivers and most of them can be overloaded by switching power supplies and wireless routers, which generally scream all the way from 10 KHz to 30+ MHz. Computers are a rich source of harmonics, odd and wandering digital noises. In cracking open more than a few computer cases, one can see the places where the various filters would be if the manufacturer had included them.
All of this is perfectly legal under Part 15 of FCC regulations, where by purchasing a device you agree that you accept whatever noise it may make or receive.
Nature regards every single wire as a potential antenna. I live 400 yards from an AM tower. Until it was removed, my chain link fence acted as a receiver and re-radiated the station signal quite effectively, as well as harmonics up to ~14 MHz. To receive AM stations I have to use tuned shielded loop antennas and keep the runs very short even after cutting down the fence.
Isolating the LDIII from the AM station required use of an audio isolation transformer and connecting the grounds on both sides together via a high value ceramic resistor. Even a 6" cable serves as a good antenna for 24/7 ESPN.
For those of you who live in an area subject to power fluctuations, outages or thunderstorms I strongly recommend putting lightning breakers on the panel and the use of a medical grade power conditioner on your dedicated audio power circuit.
Switching from a Monster power strip to a medical grade power conditioner has stopped the otherwise seasonal blowing of computer components in winter during storms.. If you are willing to drop a few hundred on your audio chain, consider the power conditioner cheap insurance. (I am using the 500 watt Tripplite.). Another option is the ICE line filter, used by amateur radio operators around the world to isolate their radio power supply from powerline noise and provide circuit protection. It doesn't provide voltage regulation, but it much better than a power strip at protecting a circuit, and has a noise ground. Every time your air conditioner, dryer, fridge or other high demand appliance cycles there will be a momentary spike and voltage drop across most if not all of the circuits on that panel that can blow LED lightbulbs.