Hi everyone,
Listen, this is a very "meta" question and I wasn't sure to post it here or under amplifiers. I've been an audiophile for a very long time, listened to a lot of systems, and also built several speakers, and was involved in motion picture sound systems and measurements. I want to ask a question to see if anyone can add to this.
As most of us understand it, we model the amplifier/cable/speaker chain as a hopefully ideal voltage amplifier, with an output resistance, which then adds to the cable impedance, and finally the speaker's own impedance curves.
As the math and many have commented, the impedance of even a modest 18 gauge speaker cable with most speakers of even down to 3 Ohms should easily be dealt with by most modern day solid state amplifiers. Even a 1 ohm output impedance on the amp shouldn't make a big difference, or so the impedance models tell us.
Based on my own listening though, a speaker with a 3 Ohm impedance can be noticeably hard to drive, and speaker cables seem to matter (but not enough for me to pay $50 for them!).
So here is my question: What if we are missing something in the amp/speaker impedance calculations we normally do? Is it possible that feedback, or some transient effect makes low impedance speakers (say bass between 3 and 4 Ohms) behave a lot worse than we think they should? Just pretend for a moment, IF we missed something what would it be?
Listen, this is a very "meta" question and I wasn't sure to post it here or under amplifiers. I've been an audiophile for a very long time, listened to a lot of systems, and also built several speakers, and was involved in motion picture sound systems and measurements. I want to ask a question to see if anyone can add to this.
As most of us understand it, we model the amplifier/cable/speaker chain as a hopefully ideal voltage amplifier, with an output resistance, which then adds to the cable impedance, and finally the speaker's own impedance curves.
As the math and many have commented, the impedance of even a modest 18 gauge speaker cable with most speakers of even down to 3 Ohms should easily be dealt with by most modern day solid state amplifiers. Even a 1 ohm output impedance on the amp shouldn't make a big difference, or so the impedance models tell us.
Based on my own listening though, a speaker with a 3 Ohm impedance can be noticeably hard to drive, and speaker cables seem to matter (but not enough for me to pay $50 for them!).
So here is my question: What if we are missing something in the amp/speaker impedance calculations we normally do? Is it possible that feedback, or some transient effect makes low impedance speakers (say bass between 3 and 4 Ohms) behave a lot worse than we think they should? Just pretend for a moment, IF we missed something what would it be?
Transistors are generally more linear when responding to voltage than to current. With low impedance loads you increase the ratio of current to voltage. So as a very spacious generalization, most analog amplifiers will have less distortion, or at least a more low-order harmonic spectrum, when driving a higher load impedance.
Whether this is what makes the difference in any individual case is difficult to determine and probably shouldn't be the first theory.
Another question is how well optimized are the motors in low impedance drivers? There always seems to be a difference in the Bl(x) curve when I've seen comparisons of the same driver with different impedances. It could just be that they spend more time getting the higher impedances just right and the less common impedances are made as good as they can be without changing the parts.
Whether this is what makes the difference in any individual case is difficult to determine and probably shouldn't be the first theory.
Another question is how well optimized are the motors in low impedance drivers? There always seems to be a difference in the Bl(x) curve when I've seen comparisons of the same driver with different impedances. It could just be that they spend more time getting the higher impedances just right and the less common impedances are made as good as they can be without changing the parts.
There are people who can hear 0.2 dB level differences and midband frequency response errors of the same order under double-blind conditions. Compared to that, the errors caused by a 1 ohm series resistance are big.
Non-linear effects are not covered by linear impedance models. Reduction of loudspeaker distortion when the driving impedance increases, for example, or an amplifier going into SOAR protection in the peaks of the music if you play very, very loud.
A non-resistive speaker input impedance can indeed draw current peaks from the amplifier that are
much higher (some have measured up to six times higher) than a resistive load of similar magnitude.
https://www.aes.org/e-lib/browse.cfm?elib=5198
much higher (some have measured up to six times higher) than a resistive load of similar magnitude.
https://www.aes.org/e-lib/browse.cfm?elib=5198
Last edited:
How the amp responds to a reactive load is often not very well tested or at least not known to most users.
The PowerCube test from AudioGraph is one way to try to show this in a concise way by mapping output across a range of impedances with resistive, capacitive, and inductive characteristics.
https://bostonaudiosociety.org/meetings/2014/14-11_meeting.htm
Years ago Rockford Fosgate tested many different amps using this method. Since then the idea hasn't gotten a lot of traction as far as I can tell. Or maybe manufacturers just don't share the data anymore.
Some amps from other manufacturers had shockingly low output capabilities into capacitive loads, especially when combined with low impedance.
The PowerCube test from AudioGraph is one way to try to show this in a concise way by mapping output across a range of impedances with resistive, capacitive, and inductive characteristics.
https://bostonaudiosociety.org/meetings/2014/14-11_meeting.htm
Years ago Rockford Fosgate tested many different amps using this method. Since then the idea hasn't gotten a lot of traction as far as I can tell. Or maybe manufacturers just don't share the data anymore.
Some amps from other manufacturers had shockingly low output capabilities into capacitive loads, especially when combined with low impedance.
Last edited:
I assume the OP is aware that the loudspeaker load is only resistive at resonance, and that it is also capacitive or inductive reactive to varying amounts over the rest of it’s bandwidth.
In a multi-way speaker design that doesn’t use power factor correction, the load will swing capacitive and inductive to varying degrees of severity at different frequencies. These asymmetric loads are rarely a problem, but in speakers where the designers had little regard for the amplifier driving them, the loads dip quite low and the phase may approach the amplifier’s phase margin. I’m looking at you, Wilson Audio, Sonus Faber and Infinity.
In a multi-way speaker design that doesn’t use power factor correction, the load will swing capacitive and inductive to varying degrees of severity at different frequencies. These asymmetric loads are rarely a problem, but in speakers where the designers had little regard for the amplifier driving them, the loads dip quite low and the phase may approach the amplifier’s phase margin. I’m looking at you, Wilson Audio, Sonus Faber and Infinity.
EPDR (equivalent peak dissipation resistance) can result in some uncomfortably low impedance figures at significant frequencies.
Large phase angles mean that the EPDR can fall below the DC resistance at more than one frequency.
I looked into EPDR in a previous thread:
More here: https://www.stereophile.com/content/heavy-load-how-loudspeakers-torture-amplifiers-page-2
Large phase angles mean that the EPDR can fall below the DC resistance at more than one frequency.
I looked into EPDR in a previous thread:
More here: https://www.stereophile.com/content/heavy-load-how-loudspeakers-torture-amplifiers-page-2
From the concluding paragraph of the Stereophile article >
"Many speaker manufacturers ... are apparently happy to throw the output-device dissipation problem over the fence for amplifier designers to deal with. Jim Lesurf ... jokingly postulated the existence of SCAMP - the Society for Cruelty to Amplifiers. If it existed, its membership would be thriving."
Yep very much aware of the reactive nature of impedance curves of actual speakers.
And I have seen users who have posted details on a speaker like the KEF Reference 1 Meta, which due to the impedance + phase angle are really close to two ohm speakers... but I think of them as an exception, rather than the rule. It just seems to me that overall, "4 Ohm" speakers tend to be more demanding than the simple math would imply... which makes me wonder if either we are missing something, because it's not just these exceptionally difficult speakers.
PS - I use Dayton DATS FTW and ease of use.
PS - I always thought the issue with Infinity was that they just didn't have the tools. If memory serves, computer spreadsheets (Lotus 1 2 3) had just become popularized, and being able to simultaneously tune for fr and impedance was very hard to do.
Now in the modern times I've seen Focal introduce impedance LOWERING circuits which apparently only exist to create a < 4 Ohm impedance in the bass.
Now in the modern times I've seen Focal introduce impedance LOWERING circuits which apparently only exist to create a < 4 Ohm impedance in the bass.
Many who use active systems don't address these issues but there's no reason they couldn't.
In any case, a small source impedance can have a significant effect. Not a mistake as much as an anticipated variation. I simulated this using a common reciever with a damping factor of 25 (320mΩ), and a random speaker (Revel).
Compared to flat, I'd think this would be audible.
I'm unclear how the amp would deal with (compensate?) the cable resistance.
In any case, a small source impedance can have a significant effect. Not a mistake as much as an anticipated variation. I simulated this using a common reciever with a damping factor of 25 (320mΩ), and a random speaker (Revel).
Compared to flat, I'd think this would be audible.
With regard to the discussion here, I don't think the short wires and band limiting are as big a factor as getting the speaker level crossover out of the design, especially if the active one is fully active (one amp per driver with low-level crossovers). The crossover/multiple drivers together on one circuit can significantly worsen the reactive characteristics if the speaker designer isn't concerned with that. I imagine most designers know when they are making a speaker that's difficult to drive, but they find some other characteristic more important.
The band limiting does help from a distortion standpoint at higher levels, but that is a separate issue to me. Linkwitz talked about that and another benefit of direct drive:
https://www.linkwitzlab.com/crossovers.htm
"I have a strong preference for line level active dividing networks ahead of the power amplifiers (Ref. 2, 12, 17). In this approach the power amplifier output is connected directly - except for a very low resistance speaker cable - to the voice coil of the driver. The amplifier takes maximum control over the motion of the speaker cone which gives a greater sense of clarity and dynamism compared to a passive dividing network between amplifier and driver. Active crossovers make much more effective use of amplifier power. A clipping woofer amplifier is not seen by the tweeter, which has its own amplifier. The clipping of the woofer amplifier may not even be noticed in this case. It would surely be heard with a passive crossover, where it might even overheat and damage the tweeter, because of the large amount of high frequency energy in the clipped signal."
Power = current x voltage x Cos(phase angle between the 2)
The last factor is 1 if the load is purely resistive, otherwise the cos() is less than 1.
dave
At the time the test mentioned was unavailable. But we had Dayton-Wright XG8s. We tried lots of amplifiers on them. 20-350 watt rated. A 70 w amplifer performed best and played loudest.
But even into a very resistive load (Magnepan MGA), we had a 20w rated amplifier playing louder before augiable clipping than a particualr 200w rated amplifier.
dave