Gain Structure

[sir sanders zingmore]

So if I can summarise what I think I’ve learnt:

For (most) digital sources the gain structure is very important. You want as little digital attenuation as possible (because the noise floor remains the same even as you attenuate the signal)

 

For analog sources the gain structure is much less important as long as the devices are well designed. 
 

 

[tripitaka]

18 minutes ago, sir sanders zingmore said:

So if I can summarise what I think I’ve learnt:

For (most) digital sources the gain structure is very important. You want as little digital attenuation as possible (because the noise floor remains the same even as you attenuate the signal)

 

For analog sources the gain structure is much less important as long as the devices are well designed. 
 

 

I'm glad you started this topic!

 

Not sure what I've learned except maybe that as the signal grows in strength along the chain then you can afford to attenuate more savagely without sacrificing the signal/noise ratio.  ...but that's provided your output stages earlier on in the chain can deliver the more robust voltages associated with lower attenuation.

 

[davewantsmoore]

https://www.diyaudio.com/forums/diyaudio-com-articles/186018-gain-structure.html

[Steffen]

5 hours ago, davewantsmoore said:

Your mobile phone is not showing you the "peak volume"..... as the meter responds too slowly.   It's showing you more like an average volume.

On the “slow” setting, yes, but not on the “fast” setting, where it is only limited by the 48kHz sampling rate. But fast transients are beside the point, really. I find any average listening level above 75dB to be really loud, which puts 3-10W in the very loud bracket for any speaker, at least for my liking and in my room.

[davewantsmoore]

8 hours ago, Steffen said:

On the “slow” setting, yes, but not on the “fast” setting

Yes, on the fast setting too 

 

8 hours ago, Steffen said:

But fast transients are beside the point, really.

I couldn't agree less.

 

They can be the difference between "sounds amazing" and "sounds boring and strangled".   As I said, this depends greatly on the content.   4db and 24db are very different numbers for how much peak to average there is.

 

8 hours ago, Steffen said:

I find any average listening level above 75dB to be really loud

Can be, sure.

 

[Steffen]

7 minutes ago, davewantsmoore said:

I couldn't agree less.

 

They can be the difference between "sounds amazing" and "sounds boring and strangled".   As I said, this depends greatly on the content.   4db and 24db are very different numbers for how much peak to average there is.

 

Dave, you’re going off on the wrong tangent here. Nobody disputes that transients are important. I was responding to:

Quote

Yes the best performing region is from 3-10w so you would want to match speaker sensitivity that reaches the majority of your targeted listening spl level within that range. 

 

by pointing out that 3-10W is really quite loud, even with rather insensitive speakers.

[sir sanders zingmore]

36 minutes ago, Steffen said:

Dave, you’re going off on the wrong tangent here. Nobody disputes that transients are important. I was responding to:

 

by pointing out that 3-10W is really quite loud, even with rather insensitive speakers.

I'm with Dave on this one. At the risk of taking my own topic off topic, I think that with insensitive speakers 3-10 W means you are clipping the amp on transients (which need much more power). So whilst it may sound loud, I reckon it would sound better with more power.

Clipping on transients may not sound like obvious clipping. It will be more like the amp sounding a bit congested or even the tonal balance changing as the volume changes - hard to tell because amps clip differently. IMHO you are better off having enough headroom so that you don't have to listen to your amp's distortion

 

 

Edited July 22, 2020 by sir sanders zingmore

[Ittaku]

7 minutes ago, sir sanders zingmore said:

I'm with Dave on this one. At the risk of taking my own topic off topic, I think that with insensitive speakers 3-10 W means you are clipping the amp on transients (which need much more power). So whilst it may sound loud, I reckon it would sound better with more power.

Clipping on transients may not sound like obvious clipping. It will be more like the amp sounding a bit congested or even the tonal balance changing as the volume changes - hard to tell because amps clip differently. IMHO you are better off having enough headroom so that you don't have to listen to your amp's distortion

Count me in too. In many of the blind testing where differences were found between two apparently volume matched sources, when the signal was investigated it was found that many amplifiers would clip at surprisingly low levels despite their power rating. When volume matched and ensuring there is no clipping, amplifier differences are much harder to tell apart.

[sir sanders zingmore]

19 minutes ago, Ittaku said:

Count me in too. In many of the blind testing where differences were found between two apparently volume matched sources, when the signal was investigated it was found that many amplifiers would clip at surprisingly low levels despite their power rating. When volume matched and ensuring there is no clipping, amplifier differences are much harder to tell apart.

 

I did an experiment many years ago which I think I've mentioned before.

My old Zingali speakers were rated at something like 95dB - pretty efficient you'd think. 

I was driving them with a 50w Sonos Faber Musica amp. Just for fun I connected a CRO to the speaker terminals and watched the signal that was going to the speakers. At a reasonable volume level (only a little louder than my normal listening level but certainly not louder than my "crank it up, no-ones' home" listening levels) I saw the amp clipping. 

 

The waveform just seemed to hit a wall. It only lasted very briefly but it happened often. I sold the Musica and bought a Magtech

Edited July 22, 2020 by sir sanders zingmore

[Ittaku]

26 minutes ago, sir sanders zingmore said:

 

I did an experiment many years ago which I think I've mentioned before.

My old Zingali speakers were rated at something like 95dB - pretty efficient you'd think. 

I was driving them with a 50w Sonos Faber Musica amp. Just for fun I connected a CRO to the speaker terminals and watched the signal that was going to the speakers. At a reasonable volume level (only a little louder than my normal listening level but certainly not louder than my "crank it up, no-ones' home" listening levels) I saw the amp clipping. 

 

The waveform just seemed to hit a wall. It only lasted very briefly but it happened often. I sold the Musica and bought a Magtech

We're getting way off topic but it is your thread and it's up to you if you think it's inappropriate to continue, but I had a similar experience. This is why it's stupid to compare amplifiers with distortion levels of .01% versus .001% when we have trouble hearing 1%. Most of what we pay for in expensive amplifiers is for them to have a big and beefy - and therefore expensive - power supply, for them to deliver the power they're supposed to deliver. All these years of auditioning amplifiers people haven't been comparing the sound of its amplification but the sound of its clipping. This is why valve amps are still popular.

[Steffen]

55 minutes ago, sir sanders zingmore said:

I'm with Dave on this one. At the risk of taking my own topic off topic, I think that with insensitive speakers 3-10 W means you are clipping the amp on transients (which need much more power). So whilst it may sound loud, I reckon it would sound better with more power.

What? I have no idea where you all are going with this. Of course you need an amp that can deliver more than 3-10 watts, why is this even mentioned? I use a 200W amp for example, playing at max. 1-3W on average. All I was saying is that you don’t need very sensitive speakers to make the supposed amplifier sweet spot of 3-10W sound loud enough.

 

I think I will remove myself now from this discussion that I should never have gotten into.

[sir sanders zingmore]

8 minutes ago, Ittaku said:

All these years of auditioning amplifiers people haven't been comparing the sound of its amplification but the sound of its clipping. This is why valve amps are still popular.

That's it - I think I'll steal that quote for future reference

 

[sir sanders zingmore]

Back on topic

 

Thanks for this link @davewantsmoore

12 hours ago, davewantsmoore said:

https://www.diyaudio.com/forums/diyaudio-com-articles/186018-gain-structure.html

 

Just started reading it. It's very good

[Telecine]

55 minutes ago, sir sanders zingmore said:

 

I did an experiment many years ago which I think I've mentioned before.

My old Zingali speakers were rated at something like 95dB - pretty efficient you'd think. 

I was driving them with a 50w Sonos Faber Musica amp. Just for fun I connected a CRO to the speaker terminals and watched the signal that was going to the speakers. At a reasonable volume level (only a little louder than my normal listening level but certainly not louder than my "crank it up, no-ones' home" listening levels) I saw the amp clipping. 

 

The waveform just seemed to hit a wall. It only lasted very briefly but it happened often. I sold the Musica and bought a Magtech

People underestimate the power required to deal with dynamic swings. That is why headroom is important in an amplifier.

[AlurkA]

Hi All,

 

Was very interested to see this thread appear.  Kinda new to this topic and somewhat electrically impaired so forgive me if the update sounds kinda dumb.  I have just read the article https://www.diyaudio.com/forums/diyaudio-com-articles/186018-gain-structure.html which I have to admit I don't fully understand and was hoping to clarify it in the context of my own setup....hoping this may clear up my understanding of things and wondering if I can achieve a nicer sound.

 

DAC - Output  2.5V 50 Ohm Gain set to 0db

Pre - Max Input Level 6.0V, Max Output 7.0V Gain selector options of 12,18,24

Power - Input sensitivity 0.76V continuous avg output, 0.11V for 1 watt (8Ohm load gain at max), Gain selector options of -12db,-6db,-3db and Max, Amp is Class A rated at 90 per channel into 4Ohms.

Speakers - 4Ohm, 88db

 

I am trying to make sense of the numbers and whether there is a "better" way to configure the chain from a gain perspective.

 

Currently DAC - gain is 0db

Pre - gain is 12db

Amp - gain is Max (whatever that means 0db?)

 

Cheers

[davewantsmoore]

2 hours ago, Ittaku said:

This is why it's stupid to compare amplifiers with distortion levels of .01% versus .001% when we have trouble hearing 1%.

As you say, what we hear is clipping.... and the rise in noise floor, and time recovery.    All amps sound different when abused.

[davewantsmoore]

2 hours ago, sir sanders zingmore said:

50w amp

Depending on the impedane curve of the speaker, you need to understand how many volts and amps the amplifier can deliver.

 

Most people only focus on the amps side..... because current clipping can be much nastier.... and amplifiers are often specificed with simply an impedance range that can "handle". (rather than amps and/or volts it can make)  .... so the layperson understands the issue.   (Impedance down, current up = uhoh).

 

The voltage can be just as problematic.   At the high impedance peaks (not present on all speakers) ... the amplifier need to make a lot of voltage.

 

I would suspect/assume this is what you were looking at on the scope.

[Kanef]

Most straightforward way of putting the clipping concern to bed.  
check the rail voltage on the power amp feed. 
if you have 100 VDc across the two legs you’re not going to go into clipping any time soon on 95dB /w sensitivity bass drivers. The driver will be good for 105 + dB sound pressures in a normal listening room. Put them in a paddock and you could scare the cows. 

[Kanef]

If you have a well designed room you’ll find yourself immersed in 110db Peak sound pressures especially on well recorded hi dynamic range material. Atmos

[Ittaku]

48 minutes ago, Kanef said:

Most straightforward way of putting the clipping concern to bed.  
check the rail voltage on the power amp feed. 
if you have 100 VDc across the two legs you’re not going to go into clipping any time soon on 95dB /w sensitivity bass drivers. The driver will be good for 105 + dB sound pressures in a normal listening room. Put them in a paddock and you could scare the cows. 

Not enough. Check it with the speakers attached.

[tripitaka]

This has been an interesting thread.

In regard to current and voltage limitations on a device's output stage (let's say a preamp, but could be a dac), does anyone know whether the demands alter when feeding 2 poweramps (bass & treble) when biamping?

 

One might innocently assume that 2 amps shouldn't affect the demands on the preamp, since the speaker crossovers will divvy up the current flow. But maybe there is more to it than that and the poweramps do actually gobble something up before handing off to the speakers? 

 

Thanks for any knowledgeable responses!!

Edited July 25, 2020 by tripitaka

[andyr]

1 hour ago, tripitaka said:

This has been an interesting thread.

In regard to current and voltage limitations on a device's output stage (let's say a preamp, but could be a dac), does anyone know whether the demands alter when feeding 2 poweramps (bass & treble) when biamping?

 

The only issue, AIUI, with having a preamp feed 2 power amps ... is that, as a result, the preamp 'sees' a much lower load.  In 'normal' circumstances this won't produce a problem but if you are using a preamp which has a high Zout (a tube preamp, for instance) then parallelling the power amp Zins may mean the preamp no longer sees a load which is >10 times its Zout.

 

For instance, if your tube pre has a Zout of 5K ohms then this is no problem if it is feeding a (tube) power amp whose Zin is the standard 100K ohms.

 

But if it is feeding 2 of these amps then the combined Zin is 50K ohms ... which is only the minimum recommended 10x the preamp's Zout (not 20 times as it was before).  This 'load mismatch' may cause the preamp not to sound as good.

 

1 hour ago, tripitaka said:

when feeding 2 poweramps (bass & treble) when biamping?

 

One might innocently assume that 2 amps shouldn't affect the demands on the preamp, since the speaker crossovers will divvy up the current flow. But maybe there is more to it than that and the poweramps do actually gobble something up before handing off to the speakers?

 

Your words (the speaker crossovers will divvy up the current flow) suggest you  seem to be thinking about passive biamping - ie:

• preamp --> bass power amp  --> bass terminals on the spkr
• preamp --> treble power amp --> treble terminals on the speaker
• spkr's internal passive XO sends bass notes to the woofer and treble notes to the tweeter.

But this ('passive biamping') is of limited benefit - and you achieve little for the cost of the 2nd amp.

 

A much better way of biamping ... ie. active biamping.  This involves:

* removing or bypassing the spkrs' internal passive XO.

* this leaves the skrs with:

• bass spkr terminals directly connect to the woofer
• treble spkr terminals directly connect to the tweeter.

In this case, the signal chain becomes:

• preamp --> active XO
• active XO LP filter --> bass amp
• active XO HP filter --> treble amp
• bass amp --> bass spkr terminals
• treble amp --> treble spkr terminals.

 

In this case, no Zins are parallelled - so no load conventions are broken.  And, through the magic of active bi-amping, you get a significant increase in power.  (Read Rod Elliott's articles on active biamping, if you can't quite get your mind about this.

 

Regards,

Andy

 

[tripitaka]

Thanks @andyr for that excellent information.

 

In regard to ur second point, yes I guess I was talking about passive pre

amping.  As a follow-up question, I've often wondered when the speaker terminal 'jumper' connections are removed, whether the bass amp is then 'directly connected' to the woofer(s) or does the signal get filtered by some sort of low-pass? Or maybe this varies from speaker to speaker...

 

Thanks again!

 

[andyr]

1 hour ago, tripitaka said:

Thanks @andyr for that excellent information.

 

NP, T.    I've been actively tri-amping my Maggies for over 20 years so what is second nature to me ... is often not so obvious to others. 

 

1 hour ago, tripitaka said:

As a follow-up question, I've often wondered when the speaker terminal 'jumper' connections are removed, whether the bass amp is then 'directly connected' to the woofer(s) or does the signal get filtered by some sort of low-pass? Or maybe this varies from speaker to speaker...

 

Thanks again!

 

In general terms:

• inside a 2-way spkr there is a passive XO consisting of a LP filter (which is connected to the woofer) plus a HP filter (which is connected to the tweeter).
• if there is only 1 pair of spkr binding posts ... then both the HP and LP filters are joined at the spkr BPs.
• if the spkr has 2 pairs of binding posts ... then the LP filter is connected to the woofer BPs, and the HP filter is connected to the tweeter BPs.

In either case, there is a passive filter network between the spkr BP(s) and the drivers; the drivers are not 'directly connected ' to the drivers.  (That's the essential definition of an 'active ' speaker - the amps are directly connected to the drivers ... which means the XO must lie before the amps, in the signal chain.)

 

A spkr with 2 pairs of binding posts can be driven by:

• a single pair (+ve & -ve) of spkr cables ... plus the jumpers
• or a separate pair of spkr cables for each driver.

Note: the latter is the reason why spkr mfrs provide 2 sets of binding posts - it is not to enable them to be passively bi-amped!  The theory behind it is that, with separate BPs, the high currents going down the woofer cable can be separated from the light currents going down the tweeter cable, avoiding IMD.

 

 

Regards,

Andy

 

[tripitaka]

Very clear, Andy!

Thanks for taking the time to write all of that out in straightforward language