Non-Clipping Amplifier

I have been working on an amp design, using an opamp to compare the output signal to the input signal and create a feedback signal to correct it. the crazy thing is regardless of what load I put on the amp It never clips, just increases the current, has anyone ever seen anything in a design like this. I simulated it at 8 ohms and got 140v pp and 185 watts output, when I replaced the load with a 1ohm resistor i got 140v pp and almost 1500 watts with no clipping. Has anyone ever seen somthing like this used before?
 

wg_ski

Member
2007-10-10 5:21 pm
There are a lot of things you can simulate in Spice, but never build in the real world. Simulators use ideal power supplies (voltage souces) and smoke never comes out of a transistor model when you overload it. And the amp *would* clip eventually, if you tried to put out more voltage than those ideal sources are capable of. Even at 8 ohms.

If you had a big enough transformer and enough output transistors in parallel, 1500 watts at 1 ohm is very doable.
 
Simulation results always require some common-sense reality checking. The described behaviour simply doesn’t sound realistic. Anyway, the overall idea is not that far-fetched: It is a rather common technique to compare the input and output signals and use the comparator’s output in driving a VCA or some other input signal-attenuating element. You can conveniently take these signals from the inputs of the differential stage. This design actually works (as clipping prevention) but you must note that besides clipping it will also trigger to things like phase shift or slewing of the amplifier - or any other anomaly that creates unequal differential signals. Commonly circuits like these tend to “overreact” at higher frequencies due to aforementioned reasons.
 
I understand that there are alot of thing that you can simulate but dont actuly work in the real world.

but when I simulate the amp using typical global feedback, it starts to clip with a two ohm load, when I use the feedback with the opamp circuit it never clips, because it reduces the gain to keep the signal exactly 5 volts under the rail voltage when the signal is at its peak poit. I was more intrested in the never clipping, not so much the output power.
 
heres how it works, I removed all of the global feedback, and got a square wave on the output, took the output and ran it through a voltage devider to get the same peak to peak voltage square wave as the input, used the opamp to find the diffrence of the two and then applied that as feedback into the voltage amplifier, to recreate the input, just 180 degrees out of phase. so it takes the amplified square wave and subtracts the inverted input to make an amplified sine wave. as long as the input voltage never goes above the designed level it never clips.
 
poynton said:
If it is limiting the gain to prevent clipping, then surely it must be compressing the input signal i.e. it is non-linear and therefore useless for audio.

I truly disagree with this statement. Clipping is the ultimate form of compression and all audio amplifiers clip (at least in some form) if overdriven. Are you saying that all amplifiers are useless for audio?

It's just a matter of choosing in which form the evident non-linearity takes place. Gradual compression ("soft clipping") is preferred by many over the sudden and harsh "flat topping" clipping (which may also create other annoyances like "rail sticking", oscillation etc.) Consider, for example, the popularity of tube amplifiers and the use of limiter/compressor solutions in PA and instrument amplifiers. Are you saying that these are useless for audio?
 

poynton

Member
2005-03-10 11:57 pm
UK
teemuk said:


I truly disagree with this statement. Clipping is the ultimate form of compression and all audio amplifiers clip (at least in some form) if overdriven. Are you saying that all amplifiers are useless for audio?

It's just a matter of choosing in which form the evident non-linearity takes place. Gradual compression ("soft clipping") is preferred by many over the sudden and harsh "flat topping" clipping (which may also create other annoyances like "rail sticking", oscillation etc.) Consider, for example, the popularity of tube amplifiers and the use of limiter/compressor solutions in PA and instrument amplifiers. Are you saying that these are useless for audio?

Do not take remarks out of context.


The use of limiters/compressor solutions in PA amplifiers and also non-linear instrumentation amplifiers is not what is being discussed here and represents a "special case" scenario.



ph_christensen said:
.... when I use the feedback with the opamp circuit it never clips, because it reduces the gain to keep the signal exactly 5 volts under the rail voltage when the signal is at its peak poit. ........


Your simulation does not clip because it assumes a limitless current supply and as you state " keeps the output voltage exactly 5 volts under the rail voltage " by reducing the gain.

Therefore, overall it is non-linear. Works fine on a simulated sinewave input but would be horrible with music.
 
ph_christensen said:
as long as the input voltage never goes above the designed level it never clips.

And that is exactly how you prevent a normal audio amp from clipping as well.

wg_ski said:
There are a lot of things you can simulate in Spice, but never build in the real world. Simulators use ideal power supplies (voltage souces) and smoke never comes out of a transistor model when you overload it. And the amp *would* clip eventually, if you tried to put out more voltage than those ideal sources are capable of. Even at 8 ohms.

If you had a big enough transformer and enough output transistors in parallel, 1500 watts at 1 ohm is very doable.

Totally agree.
 
Your simulation does not clip because it assumes a limitless current supply and as you state " keeps the output voltage exactly 5 volts under the rail voltage " by reducing the gain.


Say you have 75 volt rails, with a 1 vpp 1khz input signal you get a 140 vpp 1khz sine wave output regardless of the current flowing. that sounds pretty linear to me. I understad that in a simulation you have limitless current.

if you simulate an amp as you change the load resistance, you will see clipping, all I am saying is no matter what load I put on it it dose not clip. design an amp capable of 1500w at 1 ohm, and then put a 1/4 ohm load on it and in the simulation I bet you it will clip.
 
If it is limiting the gain to prevent clipping, then surely it must be compressing the input signal i.e. it is non-linear and therefore useless for audio.

Seems to that by definition, anything that limits or reduces clipping introduces distortion. It does seem to me that if you could design/choose each element of of the reproduction chain from source to loudspeaker (and room volume), there is the possablity that you could have an over system in which clipping would never take place or at least not at any level below the point that hearing loss begins.
 
Has anyone used a sensible clipping prevention method?

Anything that prevents clipping is preferable to letting the amp clip.

I'm looking into a good way to prevent the clipping, and I think perhaps something based on an opamp which has its gain controlled by some level sensing of the amp's output or some other level sensing, would be a reliable way to do this and it wouldn't cause too much distortion.

There are very good opamps that can be used for this and it shouldn't be overly complex.

I'm not referring to compression, although that could also be done. I think the gain could be controlled in a way that keeps the amp's input signal below a limit, set right below clipping. So the amp would never be driven into clipping at all.

Any ideas?
 
Anything which prevents clipping must by definition start limiting the signal before it reaches the point of clipping. This means distortion sets in at a lower output level, even if it is 'smoother' than clipping. Clipping itself is not especially noticeable, unless severe (which no clipping preventer can help) or accompanied by temporary blocking. This suggests two solutions:
1. if you have severe clipping than you need a bigger amp or more efficient speakers or a smaller room or a hearing aid.
2. if you have mild, occasional, clipping then you just need to ensure that the amp recovers quickly after each clip event.
 
Could the .ase file be posted so we can see what you are looking at?

The trick is "recovery" Much easier said than done. Keeping the outputs off the rail is the problem. Baker clamp helps one rail, and flying bias clamps help low loads. I was starting to look at a symmetrical VAS so as to have two baker clamps. Been a bit sidetracked with bias stability.

There are many ideas on input limiting. All cause considerable distortion as they by definition are either clipping or compressing. Self claims to have some bright idea but won't talk about it without getting paid. Can't blame him as patents are about useless.
 
Some horrible effects can be mitigated using Baker clamps on the VAS and even on the CCS. You don't need to make it symmetrical to implement a BC on the CCS.

ANother aid is to make sure that the VAS stage does not invert when saturated. The BC helps to stop saturation, but this also means making sure that the driver/output stage clips first. To make sure the driver/Output clip use higher supply rails for the VAS/CCS than the output.
 
Anything which prevents clipping must by definition start limiting the signal before it reaches the point of clipping.

Agreed. But it doesn't have to start acting way before the onset of clipping. The threshold can be set so it is right below where the actual clipping would occur.

Since clipping does cause distortion, the clipping prevention causing some doesn't make much of a difference, it only happens at a slightly lower level.

And more importantly, we want to prevent clipping at all costs, at least to protect tweeters, if not other things, including the ears.

A good clipping prevention that acts by limiting the gain at the input, would not cause so much distortion anyway, and that is preferable to what's caused by clipping.

This means distortion sets in at a lower output level, even if it is 'smoother' than clipping.

Just a slighty lower output level. It needs to be properly calibrated to act right before the real clipping occurs. And if the action is to limit drive level, then the extra distortion can't be that much.

Clipping itself is not especially noticeable,

I would think it is. When doing PA, sometimes the drive level can be pushed too far and noone really notices, when so many amps and speakers are used. And then tweeters can be harmed, or worse...

temporary blocking. This suggests two solutions:
1. if you have severe clipping than you need a bigger amp or more efficient speakers or a smaller room or a hearing aid.
2. if you have mild, occasional, clipping then you just need to ensure that the amp recovers quickly after each clip event.

If an amp is used for PA sometimes, and not only for private listening, clipping is likely to occur.

I will look into limiting the input drive level. This is the best way to go I think.
 
Could the .ase file be posted so we can see what you are looking at?

There isn't one, because what I'm searching for isn't aimed at a specific amp.

I am searching for a way to act on the input drive level, so it acts as a limiter and the amp never gets driven into clipping.

The trick is "recovery" Much easier said than done. Keeping the outputs off the rail is the problem. Baker clamp helps one rail, and flying bias clamps help low loads. I was starting to look at a symmetrical VAS so as to have two baker clamps. Been a bit sidetracked with bias stability.

This is conventional thinking, with ways to make the amp itself behave better under clipping conditions.

What I'm researching is something else. Thinking outside of the box (no box required, just thinking).

There are many ideas on input limiting. All cause considerable distortion as they by definition are either clipping or compressing. Self claims to have some bright idea but won't talk about it without getting paid. Can't blame him as patents are about useless.

That is what I'm looking for. Doing some behavioral taming on the amp or compressing isn't what I'm looking into, although I was considering compression a while back. Compression is a bit too complex, to do it right, and it does cause some distortion.

I think it can be done. We'll find a way.