I can't recall ever building a power amp with a gain much higher than 10X. Going over 10X seems like the first step towards the troubles described. OTOH, trying to go too low can lead to stability problems; notice how many high bandwidth opamps have a minimum gain of 5X, or you have to add phase compensation networks to avoid building an oscillator.
Signal to noise performance is set at the first stage of any amplifier. It can get no better than that downstream, though it can certainly get worse, as the article shows. Complexity is bad; I've no experience with true pro audio, but it seems obvious that complicated (many stages) systems get a huge benefit from standardizing on larger signals. It's a wonder modern PC stuff works as well as it does, given 5V power rails.
[begin dumb idea] We have an environmental chamber at work with a refrigerator and a heater. It controls the temperature by running the fridge full blast all the time, and simply changing the heater power to get the desired temperature. Maybe the ultimate answer is to run the audio system wide open, just below clipping, and put a power attenuator right at the speaker! The numbers work out well unless efficiency is one of them. [end dumb idea]
Signal to noise performance is set at the first stage of any amplifier. It can get no better than that downstream, though it can certainly get worse, as the article shows. Complexity is bad; I've no experience with true pro audio, but it seems obvious that complicated (many stages) systems get a huge benefit from standardizing on larger signals. It's a wonder modern PC stuff works as well as it does, given 5V power rails.
[begin dumb idea] We have an environmental chamber at work with a refrigerator and a heater. It controls the temperature by running the fridge full blast all the time, and simply changing the heater power to get the desired temperature. Maybe the ultimate answer is to run the audio system wide open, just below clipping, and put a power attenuator right at the speaker! The numbers work out well unless efficiency is one of them. [end dumb idea]
Ha! Actually I was going to suggest just such a thing, if only as an example. 
Run it all flat out (barring clipping) and turn it down at the speaker with an L-Pad or transformer attenuator. Inefficient, and probably won't sound good, but it should keep the S/N ratio pretty high.
But then you run into problems such as trying to mix consumer and pro levels, such as the DCX2496 example in the article. If you are using the analog inputs, your CD player signal may still be a little weak. Use the digital input and you'll get more than enough level out to run any amp.
Run it all flat out (barring clipping) and turn it down at the speaker with an L-Pad or transformer attenuator. Inefficient, and probably won't sound good, but it should keep the S/N ratio pretty high.
But then you run into problems such as trying to mix consumer and pro levels, such as the DCX2496 example in the article. If you are using the analog inputs, your CD player signal may still be a little weak. Use the digital input and you'll get more than enough level out to run any amp.
What a great article and thread. One of the most useful I have read here. Wish I'd been more aware of this a year or so ago. After moving to a tube amp (an OTL Atma-sphere S-30), I was delighted that I was able to get jumpers to replace 2 of the 6SN7s and reduce the gain of my amp. Speakers are much quieter as a result and I get a greater range of travel in my preamp.
To get more fine-tuned control from my preamps attenuator, I have also lowered the output voltage from my DAC to 1.35V. But after reading this thread, I feel that the sound quality might benefit from bumping it back up (options are 5.48, 2.74, 1.78 or 1.35V output). With the lowered amp gain, maybe I will have an acceptable sensitivity in my volume control and improved sound quality from having a hotter source?
To get more fine-tuned control from my preamps attenuator, I have also lowered the output voltage from my DAC to 1.35V. But after reading this thread, I feel that the sound quality might benefit from bumping it back up (options are 5.48, 2.74, 1.78 or 1.35V output). With the lowered amp gain, maybe I will have an acceptable sensitivity in my volume control and improved sound quality from having a hotter source?
GAIN
I guess I am confused here about at all the post regarding Amp gain....did I miss something all my professional career???
Did your favorite XYZ bands, sound company's/Pro audio designers, recording studios, etc. miss the collective boat?
WHY is amp gain such a big issue - did the rest of the world skip a beat???? Did the DIY counsel find something the rest of the audio engineering world missed??? Help me out here...please! Dang these industry pro's sheesch
- Dave
I guess I am confused here about at all the post regarding Amp gain....did I miss something all my professional career???
Did your favorite XYZ bands, sound company's/Pro audio designers, recording studios, etc. miss the collective boat?
WHY is amp gain such a big issue - did the rest of the world skip a beat???? Did the DIY counsel find something the rest of the audio engineering world missed??? Help me out here...please! Dang these industry pro's sheesch
- Dave
Why? Because of the costs of gain. Why have 40X voltage gain if you only need 4x?
The DIY crowd tend to be frugal. Frugal with their money and also their designs. Many like to uncomplicate things.
The costs of more gain than you need include things like noise, distortion, higher parts cost, more complicated circuits, etc. While I can't swear that low gain systems are great just because they are low gain, many of the best systems I've heard were low gain. Certainly noise is low in such systems.
And there is a certain elegance in "enough, but not too much." Nothing wrong with that as a goal.
The DIY crowd tend to be frugal. Frugal with their money and also their designs. Many like to uncomplicate things.
The costs of more gain than you need include things like noise, distortion, higher parts cost, more complicated circuits, etc. While I can't swear that low gain systems are great just because they are low gain, many of the best systems I've heard were low gain. Certainly noise is low in such systems.
And there is a certain elegance in "enough, but not too much." Nothing wrong with that as a goal.
That's cool that you were able to do that.
So basically, you should have a power amp with short signal path, that you could adjust the psu voltage on...?
So I could use for example 2x3-30VDC power supplies for the power amp (in series, middle to ground), and replace the pot for voltage adjust with a stereo lightspeed attenuator, or just a regular stereo pot (one part for each psu).
But what kind of power amp would handle such a variable psu?
Edit:
I noticed that this has been discussed before:
http://www.diyaudio.com/forums/soli...able-power-supply-replace-volume-control.html
So I could use for example 2x3-30VDC power supplies for the power amp (in series, middle to ground), and replace the pot for voltage adjust with a stereo lightspeed attenuator, or just a regular stereo pot (one part for each psu).
But what kind of power amp would handle such a variable psu?
Edit:
I noticed that this has been discussed before:
http://www.diyaudio.com/forums/soli...able-power-supply-replace-volume-control.html
Last edited:
Let's not get bogged down in amplifier design, it's not the subject of this article. There are many threads that cover the subject.
The idea is good management of signal levels all along the path. Choosing devices with modest gain is often a good way to do this for a home system. Finding the right amount of gain and attenuation is what you really want.
The idea is good management of signal levels all along the path. Choosing devices with modest gain is often a good way to do this for a home system. Finding the right amount of gain and attenuation is what you really want.
Yes, the right amount of gain in the right place.
I was just thinking how you could apply that in a practical way. You have to be able to adjust the gain somewhere, I have no intention of listening to music on 10 different audio systems, pending on the volume i desire.
The consensus seems to be that high gain sources are nice, so that would translate to voltage buffers - stepped attenuator (because a lot of potmeters cannot really cope well with more than a few volts) - current gain stage.
But then the gain would be higher in one stage than the other, and regulated by variable resistance on top of that. Isn't some of the idea that the gain should be somewhat balanced throughout the entire chain?
Unless you have some way of adjusting the gain directly, by for example adjusting the supply voltage of the circuits, provided that they can handle that without going into "crappy snr mode".
I may be lost in some chain of thought here, but the article got me thinking, which is completely foreign to me.
I was just thinking how you could apply that in a practical way. You have to be able to adjust the gain somewhere, I have no intention of listening to music on 10 different audio systems, pending on the volume i desire.
The consensus seems to be that high gain sources are nice, so that would translate to voltage buffers - stepped attenuator (because a lot of potmeters cannot really cope well with more than a few volts) - current gain stage.
But then the gain would be higher in one stage than the other, and regulated by variable resistance on top of that. Isn't some of the idea that the gain should be somewhat balanced throughout the entire chain?
Unless you have some way of adjusting the gain directly, by for example adjusting the supply voltage of the circuits, provided that they can handle that without going into "crappy snr mode".
I may be lost in some chain of thought here, but the article got me thinking, which is completely foreign to me.
LOL. Thinking can hurt sometimes.
All you need to do is find or design a system that goes as loud as you want, and then a bit more (headroom), then optimize the gain throughout. The volume control is your friend. You just don't want to be stuck at the bottom of its range all the time, or overdrive some of the components while under-driving some others. It's about balance.
That would be BAD gain structure. There is no need to do that at all, it's wildly impractical.
All you need to do is find or design a system that goes as loud as you want, and then a bit more (headroom), then optimize the gain throughout. The volume control is your friend. You just don't want to be stuck at the bottom of its range all the time, or overdrive some of the components while under-driving some others. It's about balance.
good article.
I've often wondered why some "source selector with volume attenator" units have a fixed gain before the volume knob and others fixed gain after the volume knob... I figured it was more for impendance matching either with the previous stage or the next stage.
driving a car keeping your foot hard on the gas and hitting breaks from time to time or driving a car with the foot on the breaks and hitting the gas to move ... very different passenger experience
my signal path:
cd/tuner -> passive transformer attenuator -> SET(8W) -> speakers
I've often wondered why some "source selector with volume attenator" units have a fixed gain before the volume knob and others fixed gain after the volume knob... I figured it was more for impendance matching either with the previous stage or the next stage.
driving a car keeping your foot hard on the gas and hitting breaks from time to time or driving a car with the foot on the breaks and hitting the gas to move ... very different passenger experience
my signal path:
cd/tuner -> passive transformer attenuator -> SET(8W) -> speakers
Nicely done article, Michael... a good read for all diyers.
In a domestic system setting, I think we can do away with the gain contributed by preamps. With the current crop of digital Sources and DACs at 2Vrms or more voltage output and not too high an output impedance the active preamp is dispensable. This combination of voltage and impedance specs will have enough current to drive amps to healthy levels of sound even with a lowish voltage gain of 4x (12dB) and loudspeakers in the 88dB/2.83V sensitivity range. Unless of course your domestic setting is a stadium-like great room! And that of course your amps are current capable with your fancied speakers' impedance specs.
Also on headphones, I find running some numbers, that headphone amps too are a waste of gain. Again digital Sources with the standard voltage output with a lowish output impedance coupled to a capable volume control with enough range of levels will be more than enough to go very loud. With Noise Induced Hearing Loss pegged at 85dB and headphones with sensitivities at 98dB/1mW or more, I say we don't need additional gain. I will be posting a spreadsheet soon to prove my point.
All the best,
blues
In a domestic system setting, I think we can do away with the gain contributed by preamps. With the current crop of digital Sources and DACs at 2Vrms or more voltage output and not too high an output impedance the active preamp is dispensable. This combination of voltage and impedance specs will have enough current to drive amps to healthy levels of sound even with a lowish voltage gain of 4x (12dB) and loudspeakers in the 88dB/2.83V sensitivity range. Unless of course your domestic setting is a stadium-like great room! And that of course your amps are current capable with your fancied speakers' impedance specs.
Also on headphones, I find running some numbers, that headphone amps too are a waste of gain. Again digital Sources with the standard voltage output with a lowish output impedance coupled to a capable volume control with enough range of levels will be more than enough to go very loud. With Noise Induced Hearing Loss pegged at 85dB and headphones with sensitivities at 98dB/1mW or more, I say we don't need additional gain. I will be posting a spreadsheet soon to prove my point.
All the best,
blues
Thanks Blues.
Yes, agreed. In most home systems an active preamp isn't needed. Thus the popularity of the integrated amp. Also agree on the headphone amps. The 32 ohm types may need a little voltage gain.
I can say, tho, that I've heard some preamps or line stage with minimal voltage gain but good current gain improve the sound. Using a somewhat low value pot also seems to work well, such as 10K - if your source will allow it.
The real trouble comes when you start getting more devices in line. Things line phono preamps, active crossovers, EQs and such.
Yes, agreed. In most home systems an active preamp isn't needed. Thus the popularity of the integrated amp. Also agree on the headphone amps. The 32 ohm types may need a little voltage gain.
I can say, tho, that I've heard some preamps or line stage with minimal voltage gain but good current gain improve the sound. Using a somewhat low value pot also seems to work well, such as 10K - if your source will allow it.
The real trouble comes when you start getting more devices in line. Things line phono preamps, active crossovers, EQs and such.
noob here with a question
I've moved to wifi preamps in music playback the most part. I'm aware of "passive attenuators" vs "preamps" (sometimes in the same box.)
I'm feeding a sub-through/direct to amp connection as well as direct to amplifier (etc.)
I suppose it's more accurate to set the digital player to 100% and use passive attenuation before the amp, but how much am I really suffering, quality wise, using the "digital attenuation" option?
Paul
I've moved to wifi preamps in music playback the most part. I'm aware of "passive attenuators" vs "preamps" (sometimes in the same box.)
I'm feeding a sub-through/direct to amp connection as well as direct to amplifier (etc.)
I suppose it's more accurate to set the digital player to 100% and use passive attenuation before the amp, but how much am I really suffering, quality wise, using the "digital attenuation" option?
Paul
Thanks Michael, that makes a lot of sense! Also makes me pleased I'm going down the route of using buffers in my active crossover/preamp (ie there will be no gain).
It does make me want to lower the gain on my chipamp (currently 32X) I can drive it into clipping with only 50% output on my pc sound card. The only problem there is that it is p2p and I didn't make it easy to subsequently modify! In fact I just checked the build pics and I would say pretty close to impossible Looks like I'm going to have to start on that Baksa!
Tony.
It does make me want to lower the gain on my chipamp (currently 32X) I can drive it into clipping with only 50% output on my pc sound card. The only problem there is that it is p2p and I didn't make it easy to subsequently modify! In fact I just checked the build pics and I would say pretty close to impossible
Looks like I'm going to have to start on that Baksa! Tony.
Pano this is a great article, well and clearly explained.
Pski using digital attenuation lead to different plague...
You are dealing with quantification problems.
If your source file is 16 bit (CD red book standard) you have approximately 65000 steps to express dynamic content of music (approx 90db usable dynamic range). If you attenuate in digital domain you'll loose many of this steps during the process, and digital being referenced to maximum out level (0dbfs=0db full scale) you are loosing definition with quiet sounds and increase quantification noise (digital background noise). With highly dynamic music content with low subjective level (classical music) you are loosing many information along the process. Mainly 'delicate' contents are lost (infos about the recording space, harmonics, ...).
If you are using 24bit source files the problem is less effective: you have approximately 2 000 000 steps allowed for dynamic reproduction ( approx 120db dynamic usable) range. And reference level, 0dbfs, being the same max output level the increase in resolution is focused on the other 'end' of the digital scale, where the 'delicate' informations are. Attenuation in digital domain will reduce again the number of steps allowed for reproduction of dynamic but this will leave many steps for low level information and with increased usable dynamic range, quantification noise level is lower too. So definition is better but still not optimal.
Depending of music style it can be really a problem (low subjective level -rms- as classical music) or something you won't notice too much (high subjective level -highly compressed/high rms- like metal/ hip hop/pop music).
So for me, buffered analog attenuators are better choice quality wise in the gain structure.
Pski using digital attenuation lead to different plague...
You are dealing with quantification problems.
If your source file is 16 bit (CD red book standard) you have approximately 65000 steps to express dynamic content of music (approx 90db usable dynamic range). If you attenuate in digital domain you'll loose many of this steps during the process, and digital being referenced to maximum out level (0dbfs=0db full scale) you are loosing definition with quiet sounds and increase quantification noise (digital background noise). With highly dynamic music content with low subjective level (classical music) you are loosing many information along the process. Mainly 'delicate' contents are lost (infos about the recording space, harmonics, ...).
If you are using 24bit source files the problem is less effective: you have approximately 2 000 000 steps allowed for dynamic reproduction ( approx 120db dynamic usable) range. And reference level, 0dbfs, being the same max output level the increase in resolution is focused on the other 'end' of the digital scale, where the 'delicate' informations are. Attenuation in digital domain will reduce again the number of steps allowed for reproduction of dynamic but this will leave many steps for low level information and with increased usable dynamic range, quantification noise level is lower too. So definition is better but still not optimal.
Depending of music style it can be really a problem (low subjective level -rms- as classical music) or something you won't notice too much (high subjective level -highly compressed/high rms- like metal/ hip hop/pop music).
So for me, buffered analog attenuators are better choice quality wise in the gain structure.
Thanks Guys, glad you liked the article.
Tony- yes we often have more gain than we need. My horns are an extreme example as running loud they are 4-6dB below unity gain. All I really need is current gain. Most systems just don't need 30X gain from line level.
krivium: I do think that analog attenuation has its advantages, but I use digital. My player does all manipulation in 64 bit floating, then recalculates that to 24 bits for my sound card. Not a big loss - and I usually don't run more than about 10dB of attenuation anyway. I can not hear a difference between that and good analog attenuation. YMMV.
Tony- yes we often have more gain than we need. My horns are an extreme example as running loud they are 4-6dB below unity gain. All I really need is current gain. Most systems just don't need 30X gain from line level.
krivium: I do think that analog attenuation has its advantages, but I use digital. My player does all manipulation in 64 bit floating, then recalculates that to 24 bits for my sound card. Not a big loss - and I usually don't run more than about 10dB of attenuation anyway. I can not hear a difference between that and good analog attenuation. YMMV.
I start with the average spl required for the program material, calculate the input voltage needed for the amp/s to work in the 10-60% output range. Then.. get to that voltage as early in the signal path as possible, then set the rest of the gains to passthrough. Then, I spend time listening to the background noise with headphones while onsite equipment is turned on and off and adjust my power and cable runs to minimise the interference.