cageymaru

Fully [H]
Joined
Apr 10, 2003
Messages
22,062
Audio Science Review tackles whether it is better to use high gain or low gain settings on headphone amplifiers. Conventional wisdom says that the high gain setting on an amplifier adds noise as "there is no free lunch." A Schiit Magni 3 and JDS Labs Atom were analyzed during the testing by an Audio Precision APX555. The Schiit Magni 3 took a 3 dB hit to performance when comparing the low gain setting to the high gain setting. The JDS Labs Atom dropped from 115 dB SINAD to 101 dB SINAD when switching from low gain to high gain. It seems that low gain is superior!

The test matrix here is infinite in scope. What volume does one choose for each gain to test? After pondering for a second or two, I decided to go the defensible route of setting low gain to max and then matching the same in high gain. Both of these amplifiers have analog volume controls and in high gain, they can be touchy as far as getting accurate levels out of them but I managed to get close enough. For these tests, I chose to use 300 ohm test load as that is in my analyzer and hence, higher fidelity than my external dummy load.
 
There is a hell of a lot more to audio than the noise floor.
Yeah but that's one of the main things that this will impact, I mean obviously you should be setting the volume in each case so that you have the same output or you are testing entirely other aspects, but an ideal amp would have identical performance at both gain settings when the volume was matched (at least for the range where that is possible). But yeah, you also need to consider the peak voltage swing as well as the ability to drive enough current into your load such that the voltage doesn't collapse causing the amp to clip...
 
Ummmm..... Is there anyone out there who didnt already know that higher gain adds noise if used with efficient speakers?!??!?!?!

The only reason to use additional gain is if you cant get the volume high enough.... If you have efficient speakers, the lowest gain setting will be fine. If you dont have efficient speakers, adding gain to get the output volume you want will still not add noise as there was lower noise to begin with (due to lower volume levels). Now, adding gain to highly efficient speakers will increase noise/distortion as the affects are multiplied. Think of adding gain at the amp and both speaker level.
 
Yeah, as long as you can get loud enough for your tastes without the amp clipping, it's always better to stay in low gain.
 
Noise is directly proportional to the feedback resistor; it's a standard formula; higher resistance gives more gain, as Vo=Rf/Ri.

Or Vo=1+Rf/Ri for people that don't use current feedback amps properly, lol.

Presenting this as 'Research" does disservice to the term.

Read this as: "Idiots re-discover Nyquist's criteria."

https://en.wikipedia.org/wiki/Johnson–Nyquist_noise

Metal film resistors are great, but mosfet channels are better, using a fet in the feedback loop is harder, but it's even lower than a metal resistor.

I've done 100dB at 1.1GHz, Audio is easy, lol.
 
As one who dealt with SNRs in the range of 70-80 for cassette tape with Dolby enabled, either setting on this gizmo seems pretty damned good.

The only problem with cassette is the slow tape speeds; 1 7/8 ips tape speeds means all the hiss is in athe audio band; reel to reel with 15 ips is hissless. :)

Unfortunately, the amps were way worse, electronically. I want a nice reel to reel to play with, to see how good it can get, but that tech is dead. :(
 
As one who dealt with SNRs in the range of 70-80 for cassette tape with Dolby enabled, either setting on this gizmo seems pretty damned good.


Have you tried going back to that recently?

My old Volvo wagon has a tape deck. I find it completely impossible to listen to.
 
Noise is directly proportional to the feedback resistor; it's a standard formula; higher resistance gives more gain, as Vo=Rf/Ri.

Or Vo=1+Rf/Ri for people that don't use current feedback amps properly, lol.

Presenting this as 'Research" does disservice to the term.

Read this as: "Idiots re-discover Nyquist's criteria."

https://en.wikipedia.org/wiki/Johnson–Nyquist_noise

Metal film resistors are great, but mosfet channels are better, using a fet in the feedback loop is harder, but it's even lower than a metal resistor.

I've done 100dB at 1.1GHz, Audio is easy, lol.


Audio is easy, but you don't seem to get it. You may have spent too much time at RF. Either way, what you wrote is completely wrong.

First, "current feedback" has nothing to do with whether or not you are using an inverting or a non-inverting amplifier. Most headphone amplifiers don't use a CFA. CFA typically have higher noise and lower open loop gain than VFB. An audio application does not need the high gain-bandwidth product of a CFA op-amp.

Second, the increase in noise at a higher gain has virtually nothing to do with the change in resistance values used. I can make a gain of 2 amplifier using resistance values in the MOhm range which will lead to a high noise contribution of the resistors, or I can make a gain of 2 amplifier with 100 ohm resistors and have very little contribution. It is the ratio that sets the gain, not the absolute value... and the noise gain of the entire circuit increases at a higher gain setting obviously.

Third, for an op-amp, the noise is defined by much more than just the Johnson noise of the resistances involved. All standard op-amps will specify Input Referred Voltage Noise (En) and Input Referred Current Noise (In), in V or A / rtHz. Bipolar input devices will have lower voltage noise and higher current noise, so they work better with low source impedances. JFET and CMOS input devices will have higher voltage noise and typically a higher 1/f corner frequency as well, but lower current noise. Some op-amps are quiet, like the AD797 which has an equivalent input noise voltage near what a 50 ohm resistor yields. The noise of the entire circuit is NOT just simply proportional to the thermal noise of the resistances, that is a gross oversimplification.

https://www.analog.com/media/en/reference-design-documentation/design-notes/dn015f.pdf

Also, mosfet channels are poor resistors. I don't know where you learned that but it's completely wrong. Bulk metal foil resistors are the most linear and lowest excess noise resistances you can have. Any kind of polysilicon resistor or transistor channel is still subject to the same Johnson noise as a regular resistor except they also have other noise sources and non-linearities that a metal film or bulk foil resistor do not.

If you're gonna act like a know it all, you better know it all.
 
Yeah, as long as you can get loud enough for your tastes without the amp clipping, it's always better to stay in low gain.

Yep. Also, lower gain settings usually have lower distortion. Maximum feedback is applied at unity gain for a negative feedback amplifier.

It depends how the "gain" switch is implemented though. There are some products which run at a fixed gain internally and the gain switch is really an attenuator pad.
 
Last edited:
I just want to buy a Schitt headphone amp and some Shinola headphones. Would ask my friends over to listen and ask if they could tell Schitt from Shinola...OK, I will see myself out now.
 
You could amplify to a high level, apply noise filters, then attenuate to a lower level to get clearer audio with more powerful amplifiers, but you'd have to deal with the excess heat that generates and it wouldn't help much if you don't use quality components.
 
Audio is easy, but you don't seem to get it. You may have spent too much time at RF. Either way, what you wrote is completely wrong.

First, "current feedback" has nothing to do with whether or not you are using an inverting or a non-inverting amplifier. Most headphone amplifiers don't use a CFA. CFA typically have higher noise and lower open loop gain than VFB. An audio application does not need the high gain-bandwidth product of a CFA op-amp.

Second, the increase in noise at a higher gain has virtually nothing to do with the change in resistance values used. I can make a gain of 2 amplifier using resistance values in the MOhm range which will lead to a high noise contribution of the resistors, or I can make a gain of 2 amplifier with 100 ohm resistors and have very little contribution. It is the ratio that sets the gain, not the absolute value... and the noise gain of the entire circuit increases at a higher gain setting obviously.

Third, for an op-amp, the noise is defined by much more than just the Johnson noise of the resistances involved. All standard op-amps will specify Input Referred Voltage Noise (En) and Input Referred Current Noise (In), in V or A / rtHz. Bipolar input devices will have lower voltage noise and higher current noise, so they work better with low source impedances. JFET and CMOS input devices will have higher voltage noise and typically a higher 1/f corner frequency as well, but lower current noise. Some op-amps are quiet, like the AD797 which has an equivalent input noise voltage near what a 50 ohm resistor yields. The noise of the entire circuit is NOT just simply proportional to the thermal noise of the resistances, that is a gross oversimplification.

https://www.analog.com/media/en/reference-design-documentation/design-notes/dn015f.pdf

Also, mosfet channels are poor resistors. I don't know where you learned that but it's completely wrong. Bulk metal foil resistors are the most linear and lowest excess noise resistances you can have. Any kind of polysilicon resistor or transistor channel is still subject to the same Johnson noise as a regular resistor except they also have other noise sources and non-linearities that a metal film or bulk foil resistor do not.

If you're gonna act like a know it all, you better know it all.

I've had AD make opamps for me. Ive been doing this for 30 years.

You are still learning, but you mistake much of what I said; A CFA can be precisely controlled for bandwidth, just by changing the FB resistor. You can use valuse above the recommended level, if you're careful, and do it correctly.

Capacitors in the FB path are hard to stabilize, but most people will tell you you "Can't use caps in the FB of a CFA", that's just wrong and they don't know how.

Some CFA's have much less noise than VFB amps, but they weren't designed for that and are too expensive for production; I could care less for my stuff.
If it costs $20, it costs that.

There are a number of noise sources; you have to sum them all by RMS to know how it works out, that hasnt changed since tubes. :)

In VFB amps, the biggest contributor is usually the feedback resistor, unless you did it wrong.

I have a very nice Phono preamp I built for recording my albulms, it has lower noise and distortion than I can measure; I have some very nice equipment sitting here. :)

The input is a bunch of paralleled transistors, fwiw.

I'll never know it all; but I know my part.

I've designed amps from 100G ohm FB resistors, to CFB amps over 10GHz. YOu use the best you have for what you have to do.

The Burr-Brown amps with the 100G FB could count 200 electrons per second, lol. And outputted a 1mV signal. :)
I got a patent on those.
 
The Burr-Brown amps with the 100G FB could count 200 electrons per second, lol. And outputted a 1mV signal. :)
I got a patent on those.

Are these the ones Creative used on the X-Fi Titanium HD? I remember launch articles saying they had some fancy Burr-Brown parts in them.
 
The only problem with cassette is the slow tape speeds; 1 7/8 ips tape speeds means all the hiss is in athe audio band; reel to reel with 15 ips is hissless. :)

A high school buddy had his dad's old reel to reel deck. Fun to mess around with, but I don't remember it sounding "hissless" or "good". He had some good old stuff though, Zepplin and Pink Floyd etc.
 
A high school buddy had his dad's old reel to reel deck. Fun to mess around with, but I don't remember it sounding "hissless" or "good". He had some good old stuff though, Zepplin and Pink Floyd etc.

At 7.5 or 15 ips, the hiss is shifted out of the audio range.

I have a nice RtoR of LedZep2 in quadraphonic that rocks. :)

Zara, IDK what's on Creative boards, but Burr-Brown rocked, before they were bought by TI.

I talked one of their schedulers into moving up a fab date by begging, and then sending her a dozen rare roses; Barb Zhini rocked!! (Thanks, Barb!)
 
Are these the ones Creative used on the X-Fi Titanium HD? I remember launch articles saying they had some fancy Burr-Brown parts in them.

No, very different part for different purpose. They made thousands of parts. They were talking about PCM1794 DAC. That part was not even designed by what is left of the Burr Brown group within TI. That DAC was designed by a TI group in Japan.
 
I've had AD make opamps for me. Ive been doing this for 30 years.

You are still learning, but you mistake much of what I said; A CFA can be precisely controlled for bandwidth, just by changing the FB resistor. You can use valuse above the recommended level, if you're careful, and do it correctly.

Capacitors in the FB path are hard to stabilize, but most people will tell you you "Can't use caps in the FB of a CFA", that's just wrong and they don't know how.

Some CFA's have much less noise than VFB amps, but they weren't designed for that and are too expensive for production; I could care less for my stuff.
If it costs $20, it costs that.

There are a number of noise sources; you have to sum them all by RMS to know how it works out, that hasnt changed since tubes. :)

In VFB amps, the biggest contributor is usually the feedback resistor, unless you did it wrong.

I have a very nice Phono preamp I built for recording my albulms, it has lower noise and distortion than I can measure; I have some very nice equipment sitting here. :)

The input is a bunch of paralleled transistors, fwiw.

I'll never know it all; but I know my part.

I've designed amps from 100G ohm FB resistors, to CFB amps over 10GHz. YOu use the best you have for what you have to do.

The Burr-Brown amps with the 100G FB could count 200 electrons per second, lol. And outputted a 1mV signal. :)
I got a patent on those.


You are clearly a smart guy, I was a bit rude in my first post.

Not challenging your experience, but you are oversimplifying a bit and talking about things that do not apply to best practice line level at audio bandwidths.

About CFAs - Yes, every datasheet will show you what happens when you increase the FB resistor. The stock recommended value often allows a lot of peaking in the frequency response. The ability to stabilize them as an integrator is also well known- you are not the first person to have done this. I do not commonly see it though. I don't come across too many low-pass filters built with CFAs. You can also sort of get a similar effect by increasing the feedback resistor value.

CFAs are not the best choice for most circuit positions in audio work. Almost all CFAs have high input current noise, high input bias currents, mediocre DC specs like offset. They need to be driven from a very low source impedance. They are mostly low voltage noise but you should look at the voltage noise density graphs and you will see that they tend to have a 1/f rise higher than lower speed low noise VFB op-amps. There is no CFA that will be as quiet as AD797, ADA4898, OPA211, LT1028 etc. CFAs also tend to have poorer CMRR because the inverting input has low input impedance. The only place I would use a CFA for audio is inside the feedback loop of a VFB op-amp (composite amplifier) to drive heavy loads. http://www.ti.com/lit/an/sboa002/sboa002.pdf describes the tradeoffs well.

The best audio op-amp for a low source impedance is probably OPA1611. The best for high source impedance, or any other time you need a FET input is probably OPA827 or OPA1642. You are free to sprinkle CFAs where you want but your end product not measure as good as it could. I have used TI TPA6120 (rebranded THS6012) for audio as a headphone driver before, but that's only in the output stage and probably the only place they are suitable for audio. I measure better performance from a composite amplifier with BUF634 or LME49600 anyway.

What transistors are you using in your preamp? Toshiba JFETs (2SKxxx/2SJxxx), BF862, or is it bipolar?
 
Last edited:
They're bipolar; ssm2210's.

I actually bought a book about Audio Phono Preamps, "The Sound of Silence", by Burkhard Vogel; I believe it's available online, and it has some very nice descriptive math.

My main experience is more feeding high speed electronics.

I've designed low noise high bandwidth preamps for nuclear detectors for the last dozen years or so, mostly.

Getting 16bit accuracy at 1GHz is pretty hard; it then gets digitized and fed to an FPGA for processing.

My first prototype preamp for audio was disappointing, so I did what I always do, buy a couple of textbooks. :)

If you don't have Robert Pease's electronics books, you should buy them.

I ended up following Vogel's design fairly closely, but with more modern amps.

My AD rep gives me stuff that isn't out yet, so the numbers aren't in datasheets I can link you to, unfortunately.

If they hit the market, I'll post them.


The ADA4700 is the last opamp they made for us; then Linear tech stepped in with the LTC6090, and then AD bought LT, lol.

The 100W amp in the LT datasheet isn't too shabby, but the distortion is high as it's drawn, I got that down another 20dB on the top end with a few resistors/capacitors in the right spots. :)

We needed a 100V opamp for a project. IT will run +-50V all day, within its thermal limits. Not exactly low noise, tho. :)
 
Back
Top