TR Forums

I have a speaker question that I'm struggling with. I have a pair of Yamaha NS-A636s that I've been working on. You guys gave me some great advice on these units a few years ago and the speakers sound worlds better. The mods you suggested transformed them into entirely different speakers.

My question is this... The speaker is a three-way unit, rated at 8 ohms nominal. All three drivers are listed as 8 ohms. When I took the speakers apart, I could see how they were wired. They were not wired in series. Each speaker was wired, separately, to the speaker connector that attached to the outside back of the cabinet. So, that would mean a parallel wiring. According to my understanding of how ohms work, how could this wiring arrangement lead to 8 ohms total for the speaker? My calculation shows the speaker should be 2.66 ohms. If I were to wire the speaker in series, I should have a 24 ohm speaker. So, how can the speaker be rated at 8 ohms nominal with three drivers in parallel?

My guess is the caps used for the tweeter and midrange? How much do the caps change things up, if they do? The tweeter uses a 1.5uf cap and the mid uses a 5.2uf cap. Are those caps playing a role in raising the ohms from 2.66 to something closer to 8 ohms? Or am I missing something?

The standard for measuring speaker impedance is measured at the speaker input terminals, not the connections to the individual drivers. The crossover network presents the load to the amp; if it presents an 8 ohm load it really doesn't matter to the amp what happens behind the crossover network.

Simplest way of looking at what the drivers are doing: Each capacitor is acting, in series with the driver voice coil, as a simple high-pass filter.  And each speaker will stop reproducing and "roll off" at some upper limit of its own ability to reproduce, which acts as a rudimentary block, also.  So the audio spectrum is more or less being partitioned into three chunks and each chunk is being handled by a single driver.  5.2uF and 8R gives a midrange crossover point of about 3800 Hz, and 1.5uF and 8R gives a tweeter crossover point of 13000 Hz (which is a bit too high IMO).

This is also the cheapest possible way to build a speaker crossover. More complex designs will also add inductors, which block high frequencies from going places where they are not needed.

Note that if you put a multimeter at the speaker terminals and measure impedance, you may get some brief blips as those capacitors charge, but in the end, you will only be reading the impedance of the woofer, because you are essentially testing at 0Hz -- which only passes through the woofer.

Okay, so if the individual drivers are not connected to each other, but only combined, at the terminal itself, that would not present itself as a parallel connection? Would it present as a serial wiring arrangement? I'm trying to figure out how three 8-ohm drivers , together, come together and the combination is still 8 ohms. All of the diagrams I've seem show that combining drivers, depending on how they are connected to each other and to the amp, either add or divide the ohms . Three 8 ohm drivers connected in parallel should create a 2.66 ohm load, and in series, the ohms are added up.

I'm not accounting for the caps since I have no idea how those effect the final ohm load, but I am assuming they do, somehow.

ludi wrote:

Simplest way of looking at what the drivers are doing: Each capacitor is acting, in series with the driver voice coil, as a simple high-pass filter.  And each speaker will stop reproducing and "roll off" at some upper limit of its own ability to reproduce, which acts as a rudimentary block, also.  So the audio spectrum is more or less being partitioned into three chunks and each chunk is being handled by a single driver.  5.2uF and 8R gives a midrange crossover point of about 3800 Hz, and 1.5uF and 8R gives a tweeter crossover point of 13000 Hz (which is a bit too high IMO).

This is also the cheapest possible way to build a speaker crossover. More complex designs will also add inductors, which block high frequencies from going places where they are not needed.

Note that if you put a multimeter at the speaker terminals and measure impedance, you may get some brief blips as those capacitors charge, but in the end, you will only be reading the impedance of the woofer, because you are essentially testing at 0Hz -- which only passes through the woofer.

So, the caps, even though it's a parallel connection, sort of "add back" some of the ohms that are lost in the division that occurs when then three drivers are connected in parallel? That would be my guess, since the caps add resistance at certain frequencies. In other words, we're not talking about three woofers in full range, connected in parallel. I guess if that were the case, it would be a flat conversion since the caps would not be part of the equation. Am I correct?

Right, you were thinking in DC terms, which is how we typically express speaker impedances for quick reference. But audio is an AC (oscillating) signal and those are passed or blocked by reactive components, such as capacitors or inductors.  Here, the existence of the capacitors is obvious enough; less obvious is that the speaker voice coils have some inductance.  Capacitors and inductors do not have a "fixed" impedance at all frequencies, the way a resistor does.  Their impedance changes in response to changes in frequency.


But, to think of it in DC terms, each capacitor only "switches on" when the incoming frequency exceeds its crossover point, while each driver "switches itself off" when a frequency arrives that's too high to reproduce.  If you were to play a nice 100 Hz tone (fairly low rumble), both capacitors block it and the amplifier only "sees" the woofer for that frequency.  If you play a 5,000 Hz tone (moderately piercing and slightly higher than you can whistle) the woofer will actually self-block most of it and the tweeter capacitor also blocks pretty much all of it, but the midrange capacitor will allow it.  So the amplifier only "sees" the midrange for that frequency.  And so forth.


A real-world audio signal usually has a whole range of frequencies arriving at once, but these still divide through the speaker network in the same way.

Yup, basically what ludi said.

A speaker coil is an inductor, and as such has an impedance that increases with frequency. The impedance of a capacitor, OTOH, decreases with frequency. So with an ideal crossover, the impedances of the speaker coils and capacitors interact in a way that causes each driver to "see" only the frequencies it is designed to reproduce. This relationship works both ways, with the amplifier only "seeing" the corresponding driver at any given frequency.

That said, real-world crossovers aren't ideal, and resonances (both electrical and mechanical) will affect the impedance via feedback effects, so if you plot impedance vs. frequency there will be peaks and valleys in that curve. But for an 8 ohm (nominal) speaker, the impedance should be reasonably close to 8 ohms, throughout the audio frequency range. (Where the definition of "reasonably close" is a judgement call on the part of the designer...)

If you want a crash course, look into Electrical Impedance. Basically impedance Z = R + jX, where R is your standard resistance (resistors) and X is reactance determined by inductance, capacitance, and frequency. Since your frequency is actually a range, there's no good way of representing the impedance without a graph, which isn't as easy to include on a spec sheet.

If you tried to measure the impedance of the speaker with an ohmmeter, the results probably confused you. What you got was the resistance of the wire in the woofer's coil (or other drivers' coils if you measured them separately). It's several times lower than the speaker impedance and that's normal. It could as well be zero if we could make speakers with superconducting wires - the speaker would work, it would even have better efficiency and its impedance would not be zero.
@superjawes - R is not just resistors. A speaker is an electromechanical transducer. A part of R and a part of X are due to the fact that a part of electrical power is converted into sound.

Wirko wrote:

@superjawes - R is not just resistors. A speaker is an electromechanical transducer. A part of R and a part of X are due to the fact that a part of electrical power is converted into sound.

I said it was a crash course...

Although the transducer would still be modeled (electrically) as a combination of resistors, capacitors, and inductors...because that's just what EE's do.

Just getting past the mental hurdle of understanding non-linear, frequency-dependent impedances is a pretty big leap.  I had college coursework, but I still didn't get a firm grasp on it until I had been involved DIY audio for several years.

ludi wrote:

Just getting past the mental hurdle of understanding non-linear, frequency-dependent impedances is a pretty big leap.  I had college coursework, but I still didn't get a firm grasp on it until I had been involved DIY audio for several years.

If you think that's bad... how about characteristic impedance (and impedance discontinuities) of wires/traces/connectors carrying high-frequency signals? Although I majored in CS, I took some EE coursework as well, but really struggled with it. Once you get up into the multiple-MHz range things get REALLY strange, really fast. Just because you have a wire from point A to point B doesn't mean an electrical signal will get there!

It finally started to (sort of) make sense a couple of jobs back when I was tasked with figuring out why an LVDS video signal in a system I was working on was getting garbled. I was the software guy, but all the EEs were too busy to deal with it and I was told "just take some time and see if you can figure anything out". Armed with a schematic, Wikipedia, and Google I gave myself an "RF Transmission Lines 101" crash course... I eventually found multiple errors in how the wires were being routed, both internally in the system and in the external cabling. Some of the errors were made by our own engineers, and others had been made by subcontractors. I nearly gave the director of our facility a heart attack one day when he walked in on me taking an Exacto knife to a cable to get a closer look at its internal construction. Director: "WTF, that's a custom prototype cable, it cost us $2,000!" Me: "It doesn't work!" (That cable was indeed part of the problem; the internal LVDS pairs that were supposed to be twisted... weren't.)

just brew it! wrote:

If you think that's bad... how about characteristic impedance (and impedance discontinuities) of wires/traces/connectors carrying high-frequency signals?

Oh, you mean Voodoo Blackmagicraft?

Congrats on figuring it out, particularly solo. I've only ever gotten a dusting of it. Interestingly, it affects high voltage power lines -- a switching event or lightning strike can send an RF-type surge down the line where it behaves as a traveling wave, rather than a simple voltage spike.  And may then double in amplitude when it encounters a change in impedance.

Don't mind me, I'm just subscribing to this thread.
Lurk lurkety lurkey lurk.....

Chrispy_ wrote:

Don't mind me, I'm just subscribing to this thread.
Lurk lurkety lurkey lurk.....

It's good that you came, the OP has run away already!