Undoubtedly Parallel sounds better.
my case is lm7293. they sound as smooth as butter.
reasons I percieve:
less heat generated on the die
more conduction due to additional parallels .
more current headroom hence higher dynamics from speakers
Higher damping.
less vulnerable to short circuits or overdrives.
Looks beautiful !
my case is lm7293. they sound as smooth as butter.
reasons I percieve:
less heat generated on the die
more conduction due to additional parallels .
more current headroom hence higher dynamics from speakers
Higher damping.
less vulnerable to short circuits or overdrives.
Looks beautiful !
chipamps
I've made a single, a bridge and a parallel. Obviously the bridge runs very much hotter so you need good thermal sinking to have a reliable amp. I think they all sound pretty good, but I tend to use the parallel more probably because in the back of my mind I like the additional current drive. The bridge has more sparkle and the single is nice and crisp; bridge seems smoother.
I've made a single, a bridge and a parallel. Obviously the bridge runs very much hotter so you need good thermal sinking to have a reliable amp. I think they all sound pretty good, but I tend to use the parallel more probably because in the back of my mind I like the additional current drive. The bridge has more sparkle and the single is nice and crisp; bridge seems smoother.
I found that the parallel sounded better -- but my implementation was via the LM4780 which is 2 lm3886 dies in one package.
One of the best things you can do with these chipamps was demonstrated at the NJ Audio Society by Bob Cordell -- he used a buffer/inverter to drive the inverting input of a LM3886. It really sounded nice driving 2 SEAS Thors which I had built, and a K-Works up-modded version of the Thors. I don't know whether he used his "Klever-Klipper" in between (as provisioned in his book.) Didn't ask him if he used a servo.
One of the best things you can do with these chipamps was demonstrated at the NJ Audio Society by Bob Cordell -- he used a buffer/inverter to drive the inverting input of a LM3886. It really sounded nice driving 2 SEAS Thors which I had built, and a K-Works up-modded version of the Thors. I don't know whether he used his "Klever-Klipper" in between (as provisioned in his book.) Didn't ask him if he used a servo.
In a bridged circuit much of the common mode noise / distortion is cancelled. Distortion is lower on a bridged circuit.
2nd harmonic cancels pretty reliably, the higher harmonics no. Unfortunately, this doesn't often improve sound unless the 2nd was a screamer with very little else, which is rare. Some people even seek dominant 2nd harmonic, which would be their first reason to reject a bridge circuit.
To get the most extreme second harmonic cancellation, a perfect mirror is required, meaning two identical amplifier circuits, one fed with a perfect inverse signal. This could be acheived to a high degree with a balanced (differential) input signal as from a microphone, DAC, or maybe phono cartridge. If you start with a single ended source and use an inverting signal amplifier, or worse, set up one of the power amps as inverting, results aren't quite as good... that is, if youre looking for second harmonic cancellation.
Last edited:
The point being made was that the overall distortion would be lower in a bridged circuit than using a single chip. I agree that not all the harmonics would cancel as well as some of them.
Paralleling of two high feedback low output impedance amplifiers is VERY not recommended. The slightest difference in gain, DC offset, noise or distortion will cause a tug of war between the two outputs, which may well result in instability and/or blowout. The least it is likely to cause is the parts getting hotter than they need to.
The point being made is that there may be MORE noise and MORE distortion, than in the single chipamp when it is implemented properly.
Linkwitz uses the 3886 in both valid ways iin his Pluto biamp'd speaker system, and the schematics for it are on his website, linkwitzlab.com. One chip for the tweeter, and two bridged for the woofer. Linkwitz is very sharp. I recommend using his circuits as a reference anyway. If I remember correctly, his schematic didn't show the output zobel (?) filter (which feeds the speakers), which I do recommend. It reduces the probability of spurious oscillations, by somewhat isolating the reactance of the speaker load from the feedback function.
No matter if parallel or bridged, when you have a greater number of output devices driving the same load as before, linearity *could possibly* be improved. This infers somewhat lower voltage for the bridge amp, but that is workable. Also, bridge amplifiers are effectively in series, meaning that it will require more care in the implementation; but, that is doable. In trade, the high current return from the speaker will not strike the same ground reference as the small signal input and that *could* be an advantage. Even so, a bridge amp will require more elegance/care in order to do a similar job as a non-bridged amp.
8 ohms does not always mean 8 ohms, except in a resistor.
If the load has severe impedance swings a parallel application is generally best able to handle those. Obviously the exclusion of the paralleling resistor and the resulting summing currents, from the feedback loop leads to a higher level of thermal noise in the output, not sure if this is audible.
If the load has a smooth impedance curve and is properly impedance compensated, a single chip usually sounds best.
Bridged amplifiers have more headroom and can go louder at the cost of heat and dissipation, or be used to overcome low supply voltage (among other shortcomings). Also, there is an excellent point made on one of the old chipamp threads about the inability of the 'master' to properly see the 'slave' amp's ground reference due to the increasing impedance of the amp at high frequencies, so if the amp is not properly compensated the results will be much worse than a single chip. I have no idea where that post is now and who the author was.
You have to remember the 3886 was not designed as a high quality chip. If your expectations are modest, I would go ahead and build one of each and pick one to live with, don't bother too much about theory. However, I concur with the fact that a 'perfectly balanced' amplifier will usually lead to better cancellation of harmonics. Note that odd order harmonics sound much harsher than even order harmonics (which theoretically cancel in a bridged/balanced chain). It is not possible to achieve that without proper test equipment and a measurement setup.
Just remember that 'sound better' is a subjective opinion, is rarely based on scientific evidence, and is also heavily influenced by the amp topology and what an individual thinks of it. For example, I favour parallel-bridge configuration driven by end-to-end balanced inputs. There is still some imbalance due to the loose 1% tolerance that is used in component selection, but it provides lots of headroom and drive. Not as 'pristine' as a single chip, but has its uses.
When it comes to discrete topologies, multiple devices can lower distortion if properly used. But for modular amplifiers, there are already lots of limitations that the manufacturer is working with. A specific implementation may be used to overcome some of these limitations, but may not be sonically superior always. Since we listen to the system as a whole, it is possible that the solution we choose may have good results, but not be the optimal solution. For example, we may feel that an amp with increased current capability is better, but maybe the problem is speakers with unkind impedance swings. Impedance compensation of the speaker may help in extracting more performance out of the system, without adding more thermal noise sources.
I know you might be frustrated by now, but the point is about building amps. Coming to your posted schematic, that OPA627 will not be happy with that load. And the 220pf between the inputs is a little dangerous. That value is to be determined mathematically, not chosen at random or it causes oscillation in some cases. Read the LM3886 application notes in the main datasheet to see why, and to see some solutions.
And I would dump the servos and go back to a fully AC-coupled structure with caps in the feedback return, but that's just me. Servos require careful layout because the expected risetime is very short (I'm not good at the math, but a 1/t of 4x the amplifier's dominant pole is suggested - which means capable layout into the MHZ range, far beyond my level of talent).
If the load has severe impedance swings a parallel application is generally best able to handle those. Obviously the exclusion of the paralleling resistor and the resulting summing currents, from the feedback loop leads to a higher level of thermal noise in the output, not sure if this is audible.
If the load has a smooth impedance curve and is properly impedance compensated, a single chip usually sounds best.
Bridged amplifiers have more headroom and can go louder at the cost of heat and dissipation, or be used to overcome low supply voltage (among other shortcomings). Also, there is an excellent point made on one of the old chipamp threads about the inability of the 'master' to properly see the 'slave' amp's ground reference due to the increasing impedance of the amp at high frequencies, so if the amp is not properly compensated the results will be much worse than a single chip. I have no idea where that post is now and who the author was.
You have to remember the 3886 was not designed as a high quality chip. If your expectations are modest, I would go ahead and build one of each and pick one to live with, don't bother too much about theory. However, I concur with the fact that a 'perfectly balanced' amplifier will usually lead to better cancellation of harmonics. Note that odd order harmonics sound much harsher than even order harmonics (which theoretically cancel in a bridged/balanced chain). It is not possible to achieve that without proper test equipment and a measurement setup.
Just remember that 'sound better' is a subjective opinion, is rarely based on scientific evidence, and is also heavily influenced by the amp topology and what an individual thinks of it. For example, I favour parallel-bridge configuration driven by end-to-end balanced inputs. There is still some imbalance due to the loose 1% tolerance that is used in component selection, but it provides lots of headroom and drive. Not as 'pristine' as a single chip, but has its uses.
When it comes to discrete topologies, multiple devices can lower distortion if properly used. But for modular amplifiers, there are already lots of limitations that the manufacturer is working with. A specific implementation may be used to overcome some of these limitations, but may not be sonically superior always. Since we listen to the system as a whole, it is possible that the solution we choose may have good results, but not be the optimal solution. For example, we may feel that an amp with increased current capability is better, but maybe the problem is speakers with unkind impedance swings. Impedance compensation of the speaker may help in extracting more performance out of the system, without adding more thermal noise sources.
I know you might be frustrated by now, but the point is about building amps. Coming to your posted schematic, that OPA627 will not be happy with that load. And the 220pf between the inputs is a little dangerous. That value is to be determined mathematically, not chosen at random or it causes oscillation in some cases. Read the LM3886 application notes in the main datasheet to see why, and to see some solutions.
And I would dump the servos and go back to a fully AC-coupled structure with caps in the feedback return, but that's just me. Servos require careful layout because the expected risetime is very short (I'm not good at the math, but a 1/t of 4x the amplifier's dominant pole is suggested - which means capable layout into the MHZ range, far beyond my level of talent).
Parallel or bridge
A bridge circuit would most likely sound better because a bridged circuit has lower distortion. Also a bridged circuit would give you more power in an 8 ohm load due to higher voltage swings than a parallel circuit.
A bridge circuit would most likely sound better because a bridged circuit has lower distortion. Also a bridged circuit would give you more power in an 8 ohm load due to higher voltage swings than a parallel circuit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.