SUMMER MUSINGS: Paired speaker testing, more crossover thoughts, running tests ourselves and for ALL speaker reviews!

Paradigm Signature S8 v.3 speakers with grilles off. miniDSP UMIK-1 microphone for some simple testing with Room EQ Wizard (REW).

Well it's summertime which means (hopefully) a bit more opportunity to enjoy the nice weather around here in Vancouver with the family, more time to listen to music, maybe a little trip out of town on weekends.

As for my audiophile interests, it also gives me time to think about the broad topics and maybe run a little bit of testing here and there. Often these summer articles flesh out ideas and results covered over the last few months I may have left aside or on the "cutting floor" when writing.

This post will be along those lines in that these results were actually captured back in the January-February 2025 timeframe when I was doing the ferrofluid replacement for my tweeters, and then subsequent discussions about "audiophile" crossover parts and addressing beliefs of people like GR-Research/Danny Richie.

Since I grabbed these measurements and pictures when I was considering the importance (or not) of crossover part affecting sound quality, I thought I'd post them with some further summertime-meandering and related discussions.

To start... How symmetrical are our speaker pairs?

I've often wondered when reading reviews or watching videos especially about very expensive speakers such as bespoke creations and exotic brands that don't roll off assembly lines with tight quality controls. That was also the main question I had for myself after the ferrofluid change and while thinking about the topic of crossovers (especially hand-assembled kits being sold) and how this applies to my own listening with the front Paradigm Reference Signature S8 v.3 speakers which I bought back in 2013.

Here is one of the tweeter crossovers from my speakers, easily examined by pulling off the binding post cup:

According to the company it's a 3rd-order filter at 2.0kHz. We've got a small air core inductor, 6.8μF polypropylene capacitors, non-inductive metal oxide resistors (looks like Jantzen 10W, perhaps just 5% tolerance), not the most pretty soldering job but electrically it's fine, with basic push-on wiring clips to the drivers.

The internal wiring is reasonably thick 14 and 16AWG, appropriately the thinner 16AWG going to the tweeter. It looks like tin coated copper which makes soldering the connectors easier and reduces tarnishing. As usual, despite audiophiles getting excited about fancy garden-hose-sized speaker cables, interconnects, and power cords, who knows if internally one's speakers, amps, or whatever wire connecting the power connector to the power supply even use high-quality copper conductors.

The other crossover (2nd order, 230Hz) for the midrange and woofers lives inside the cabinet which can be accessed only if I remove the midrange/bass drivers like this picture from a Paradigm Signature S6 v.3 article a number of years back:

We see some electrolytic capacitors, ceramic resistors. Maybe it's worth replacing the electrolytic caps if there's a noticeable issue down the road (when speakers 20+ years old?). With the mild and cooler temperatures here in Vancouver, I think electronics devices keep pretty well, I have many devices with electrolytic caps around the house 30+ years old running without problem.

[Of interest, I see that Danny Richie had a look at the smaller Paradigm Signature S6 speakers in 2021, not sure which version he was sent. I agree with his findings around the pretty good off-axis response which is similar to these S8, and the effect of the grilles as well. His measurements of the impedance seem a bit odd when he found an average of very low 2.6Ω (I'll show you my results from the S8 below) - is this normal or a defective unit? Hard to tell if he only has one speaker to examine! 

As usual, be careful with these videos, they might not actually be representative of the actual device performance without corroborative evidence. Sounds like the owner that sent the S6 to him didn't want to proceed so Danny didn't bother with the crossover redesign. But then again, it didn't sound like there was much he could do with it since he didn't actually complain too much about the measurements otherwise.

Notice that Danny said nothing about how the single speaker sounded. Just some measurement interpretations and vague comments about the potential of crossover parts bringing sound quality to "another level". As usual, this is all very speculative and predictable as a seller of products and promoting faith in his perspective is part of his game.]

As for the binding posts, they are the typical non-magnetic brass variety with flat, solid, jumper plates (also brass) between the bi-wireable connectors - the banana connectors work well, sturdy:

DIY Canare 4S11G-based RED DRAGON cables. There was a time I though "bi-wiring" was cool. Like many audiophile ideas/myths, IMO, it just doesn't matter anymore after listening and testing over the years.

Examining crossovers and criticizing the parts used as if this explains sound quality to any great extent makes no sense to me as discussed in my previous article. Since a speaker is literally the sum of its parts, to put so much emphasis on the crossover parts assuming good frequency division, when it more than likely contributes less than the drivers and cabinetry, seems very simplistic. Besides, if we're very serious about the best sound quality, why spend hundreds on things like the GR-Research passive kits when one could use more advanced DSP active crossover for more control and clarity over the frequency and time domains? Or maybe use that money for room treatments to improve the reverb time of the room, reduced distortions from wall/floor/ceiling reflections, etc.

Turning to measurements, let's have a look at both of my Paradigm Signature S8 speakers in frequency reproduction:

Assuming we have a reasonably competent system that can handle intended output levels without gross distortions, frequency response is the most important factor when it comes to sound quality.

While each of us can enjoy whatever speaker tonality moves us, for the sake of high-fidelity - that is, to adhere to a standard that is considered more "transparent", not distorting the sound of the source - we typically would like to see something like a reasonably flat (quasi-)anechoic response like what we get at 1m on tweeter axis with the speakers and microphone unobstructed and not too close to boundaries (walls, floor, ceiling, furniture...). Note that this does not mean the speaker should measure flat at the listener sweet-spot, a -6 to -10dB "tilt" down to 20kHz is often desirable and reduces listener fatigue.

Basic frequency response of a speaker is quite easy to measure these days and any self-professed hardware audiophile who claims they "can't" do it is either technically or intellectually lacking or to be honest, just lazy; not wanting to do it is the other option. 😉

Notice that the above data is graphed from 200Hz up with both speakers. Even without the full spectrum down to 20Hz plotted, the graph already tells us a lot about speaker matching, how "flat" the speakers perform across the midrange and into the treble, integration between the drivers and crossovers on tweeter axis level. While the UMIK-1 microphone is only able to record up to 24kHz (48kHz internal sample rate), it still gives us an idea of whether there are some "near-ultrasound" anomalies typically related to tweeter cone breakup just over 20kHz. This is usually not a problem with beryllium tweeters as used in these speakers. 

Below 200Hz is of course also very important but the results are highly dependent on the room and for those of us integrating subwoofers, a big topic in itself with other considerations. Regardless, I would certainly encourage reviewers to show a frequency response at the listening sweet spot if possible to contextualize the subjective descriptions. For example, this guy has had all kinds of expensive speakers in his room, the fact that he has never shown any measurements of the frequency response is actually a shame. An educated man of science can't run a "lab" without measurements, right? 😅

Frequency response for my speakers looks well-balanced in the right and left channels at average ~87dB SPL @ 1m (around 90dB SPL would be good for these measurements). Good to see tight concordance between the two speakers at least to 10kHz to cover the spectrum where the human ears are most sensitive. We can see a good "flattish" response all the way to 20kHz with roll-off into 24kHz.

Notice that I used 1/12-octave smoothing in the graphs which provides more detail as compared to GR-Research/Danny's typically excessive 1/3-octave smoothing in his videos which I think could "hide many sins".

Let's look at the distortion amount:

While I would not consider the UMIK-1 microphone to be lab-grade for distortion measurements (noise level can be quite high for example), it's still able to show us troublesome frequencies. For my speakers, I see 1.3% maximum THD above 100Hz, peaking around 1.5kHz, speakers playing at ~87dB SPL at 1m. Almost all of the distortion for both speakers are 3rd harmonic in nature.

As far as I am aware, there is no clear recommended rule-of-thumb threshold for speaker THD to be considered "hi-fi". Something like 2% or less above 100Hz at 90dB SPL at 1m is probably reasonable for modern high-fidelity speakers to at least show technically "accurate" performance.

Years ago, I'm guessing early 2000's, Axiom wrote some research on audibility of distortions in speakers by using overlaid "distortion" tones when playing music. The conclusion was that the ear is very "crude" at hearing such distortions below 40Hz. <10% distortion can be audible above 250Hz. And we might be able to hear 1% distortion above 8kHz. Listening to real music masks a lot of distortion, even in the form of tonal 'noise' added to the signal as used in this experiment. The human ears-brain have amazing abilities, but there are some very real, measurable limits to even the best "golden ears". 

I notice that the graphs do not show up on the Axiom web page reliably. Here is the summary graph with distortion limits and frequency, some % levels added:

Only one study, important not to treat the numbers as gospel.
Be mindful of the magnitude of distortion we're looking at
in order to be detected when music listening!
BTW, I'm not sure if the linear correlation makes sense in light of
non-linear human auditory acuity.

Realize that even if we detect distortions, this does not imply that it's objectionable. Just ask vinyl lovers who can easily detect noise floor limitations and other crackles and pops but can still enjoy the music, even preferring playback with such anomalies. Needless to say, the amount of measurable distortion from speakers easily overwhelm the comparatively small anomalies from solid-state amplifiers, CD players and DACs. This is why speakers and your room are the main determinants of sound quality once we have good-sounding music and a decent music player.

In the last couple of decades we have seen an accentuated focus in the audiophile literature on "time domain" performance. As I suggested above, frequency is more important. But that doesn't mean I don't want low-jitter DACs, or good time-domain performance out of my speakers. What is important is to realize that we should not become overly obsessed to the point of getting too impressed about rather small technical matters like impulse response graphs for example! This kind of unnecessary obsession is what people like MQA/QRONO want to capitalize on, and allows snake oil "ethernet jitter" to be hyped.

Using the REW frequency response data, we can have a look at the "waterfall" plot to give us an idea about how the sound might "linger" over time coming from our speakers. Again, to escape from inevitable room reflections, let's avoid the bass region, plotting above 500Hz:

Nothing scary on a 3ms waterfall of 30dB drop from ~80dB SPL. Notice that the above was captured with grilles on which is how I usually listen to the speakers.

Again, it's nice to compare performance with both speakers. This will show us if we have anomalies or asymmetries between drivers, the speaker cabinets, and inevitably if we have crossover issues. In the case of these speakers, notice that the grilles accentuated the energy around 8-10kHz. Good to know, but not a big deal given the rapid >30dB attenuation by 3ms. Application of a little DSP to smooth the bumps would be easy.

Since we're examining the time domain performance, let's also look at the step response of the two speakers overlaid:

Notice beyond 4ms we're seeing effects from room reflections. These speakers have inverted tweeter and mid-range polarity easily seen even if I don't believe ears have the resolution needed to hear this!

Again, we're seeing good stereo matching, nothing scary in the step response captured with sub-millisecond resolution. DSP can optimize that step response at the listening sweet spot which no passive crossover (including those from GR-Research) would be able to achieve. Despite what might be considered "cheesy" crossover parts including electrolytic capacitors, budget "sand cast" resistors, and the like, I see no reason to be concerned even though the speakers are about 12 years old at this point!

BTW, I find it fascinating that Danny/GR-Research doesn't put up measurements of his crossover kits showing the step response difference pre-and-post upgrade. Even if he didn't fool around with the speaker frequency response, if the improvement in parts alone result in significantly better sound quality, maybe thanks to time-domain improvements, we would be able to see that effect here. As usual, it's best for audiophiles to not talk about things with dramatic huffing and puffing like Danny does as if he "hears" all kinds of "next level" differences - show us so we can believe you.

For completeness, it's good to know the speaker's impedance to understand how difficult it might be for the amplifiers to drive. Paradigm advertised these Signature S8 speakers as "Compatible with 8 ohms":

Again, good confirmation of symmetrical speaker impedance (determined by the crossovers, drivers and the effect of the ported box). The minimum impedance for these floorstanding Paradigm S8's is a little over 3.5Ω at around 90Hz, so it's good to suggest that the amp be able to easily handle 4Ω despite the company's 8Ω recommendation. (I discussed impedance measurements back in 2019.)

And let's compare the phase angles (relative lead/lag between voltage and current waveforms in an AC circuit):

Well-matched results for the two speakers from 10Hz to 24kHz, demonstrating good crossover (and driver) pair-matching. Again, all without particularly expensive components, "cheesy" parts in the eyes of GR-Research!

Putting the impedance and phase angle together, here's the EPDR for the left channel with a minimum down at 1.6Ω:

The right channel looks almost exactly the same (not shown) with minimum EPDR at 1.7Ω.

Based on all that, is there anything worrisome?

I don't think so... Despite being an obsessive audiophile (but not fetishistic of course!), let's just say that based on what I hear and measure with this pair of speakers, there's nothing to compel me to rip out the crossovers and update to expensive resistors, banish any ferromagnetic core inductors I might find, and insist on polypropylene/foil capacitors! If you get a chance to measure your speakers and find similar results, then I suggest you not worry either. 😉

So, with all this attention in the last few years, should speaker manufacturers care to upgrade crossover parts? Maybe for show and advertising to differentiate themselves in response to consumer sentiment, otherwise probably not. I would argue that manufacturers should not if based on their R&D, doing so makes no difference performance-wise and reliability is already good for many years with standard parts. For value-conscious products (most things), I don't think consumers would appreciate significant price increases for something they won't see hidden inside and with no evidence they would hear any improvement either.

Having said this, I think the perspective should be different for premium luxury products. "High end" audio / luxury-oriented companies obviously should not skimp on the crossover components and internal wiring especially when demanding what must be very healthy margins on the MSRP from the customer! The optics obviously will look very poor if the customer is expecting top-tier luxury and the components are too cheap.

"Cost-no-object" designs from the likes of Wilson Audio, Magico, Børresen, Sonus Faber, Göbel, etc. of this world are really the ones that audiophile buyers should insist on 1% tolerance parts, exclusively air core inductors, and appropriately fat Sonicaps. Maybe some "No Rez" and "tube connectors" could help too according to Danny. 🙄

So let's end this summertime discussion with three final thoughts that might be controversial for some but which I think are quite reasonable if we meditate about it:

1. Be it resolved: Trust no one in the audiophile world. Audiophile websites, audio magazines, manufacturer ads, YouTube videos, forum discussions about supposed sound quality have been "tainted" with too much fantasy, snake oil, and unsubstantiated subjective-only claims with inadequate fact-checking over the decades.

Over the years of writing this blog where I've routinely pointed out questionable magazine articles, ridiculous YouTube personalities, countless fanciful claims and beliefs from manufacturers, perhaps at the end of the day we need to just start from a "trust nobody" position when it comes to audiophile sound quality claims. This is especially true when the asking price can so blisteringly, unexpectedly humongous.

[Indeed, I don't even particularly care if anyone trusts that anonymous audiophile Archimago guy. 😂]

To start in a position of skepticism is not paranoia. If over time you know the person to be of sound mind and sound character, like any relationship, we'll develop trust based on time and experience with knowing that person. So too, to some extent here in the online audio hobby. That's assuming you want sound judgments, for entertainment value, anything goes!

2. To be respected as a higher tier audiophile, it's not based on how much money the person spends. IMO, respect can reasonably be bestowed on the audiophile who does his/her own testing, can discuss knowledgeably by demonstrating effect, and appears to have a well-considered reality-based perspective.

Given how accessible testing equipment has become these days, an audiophile who has the knowledge to run tests/measurements, correlate these results for themselves, and can speak with some authority based on something more than just ephemeral subjective impressions for me automatically elevates my appreciation for such a hobbyist.

However, as with the example of Danny Richie, just because a person can run measurements doesn't automatically mean that the person can be trusted if there are other motives (more often than not, financial in nature). As far as I can tell, many of Danny's claims are without merit and are not supported by evidence - more just "UFO Believer" fantasies with commercial interests.

There are also others such as Milind N. Kunchur we discussed a few years back where he tries to dazzle with "sciency" claims, performing studies, writing wordy research papers, but jumping to speculative conclusions. These research papers are really IMO not constructed or argued with a "sound mind" when we take a step back and consider his experimental methodology and apparent poor insight regarding such lapses.

3. IMO, all speaker reviewers should at least do a simple 1m tweeter-axis measurement with both speakers and measurement of full frequency response (from 20Hz) at the listening position or "sweet spot" for context.

An audiophile hobbyist not wanting to run measurements but just sharing their experience on a casual chat forum is to me more acceptable than a supposedly more serious loudspeaker reviewer ignoring such a basic task while acting as an "influencer" as if others should trust him.

I believe all reviewers at least should be able to show us this basic level of technical performance for the product when doing a write-up or video. Once one knows what to do, it literally takes minutes to get some worthwhile information on the speakers while also giving us a look at the frequency response with the amplifier in the room that the subjective impressions are based on. Other than simplistic AI-generated purely-subjective fluff, I'm sure it takes hours to write an article or edit a review video. So why not spend a little bit of time on a basic measurement as an objective anchor?! No, a simple 1m on-axis capture or sweet spot measurement would not be as good as a suite of off-axis, (quasi-)anechoic, or "Spinorama" data. But I think this is way better than nothing at all!

Above, I mentioned the Jay's Audio Lab YouTube channel. I'm fascinated by the amount of money the guy spends on this stuff - I hope it's worth it for him running this as a business! But practically speaking, beyond breathless praise for products (including insanely priced DACs), can we see what those $310k Marten Mingus Orchestra Statement loudspeakers do in your room, Jay? If those diamond tweeters are not too bright as some commenters wondered, then show it with a simple measurement rather than talking about it and recording audio for YouTube as if that sounds in any way accurate to how the audio is experienced in the room. The viewer really only is hearing the audio on our speakers and headphones transmitted over lossy, typically Opus-encoded YouTube compression (possibly only useful for A-B listening comparisons as discussed here)! This fact should be painfully obvious in 2025 for the viewership.

Anyhow, hope you're having a great time as we enter July, dear audiophiles...

Make sure to enjoy the music. Recently I've been getting into Teddy Swims' I've Tried Everything But Therapy (Complete Edition) (2025) for those into pop/rock. Here's "Bad Dreams":

For those interested in EDM, the new 51-track, 2.5hrs long, Armin van Buuren Breathe (2025, which includes the tracks on 2024's Breathe In) is fun, obviously no need for deep introspection. Here's "Let It Be For Love":

And those audiophiles more into an acoustic, blues-tinged British Americana sound - check out Mark Knopfler's 10th studio album One Deep River (2024); here's a cut:

For those of a more classical taste, here's "Schindler's List" with John Williams & the Saito Kinen Orchestra - off John Williams in Tokyo, Live at Suntory Hall 2023:

Now for those of you interested in guilty-pleasures, manufactured pop, here's Tate McRae's "Just Keep Watching" off the new F1 summer flick soundtrack. Technically, if you have an Atmos system, it's quite impressive what can be done in the studio these days with multichannel sound!

F1 The Album is an interesting example of how much better the multichannel/Atmos stream is even though it's lossy EAC3-compressed at 768kbps. Only 8 of the songs (the potential singles, I guess) are available in Atmos on Apple Music, but they're averaging DR12, not grossly dynamically compressed down to DR5/6 like those on the 2-channel 24/48 "hi-res" stream! As a result, pump up the volume when you listen to them and you'll appreciate noticeably more nuances than on the lossless 2-channel mix. Examples like this remind us that just because music is "lossy" does not imply poor quality if the bitrate is adequate (as discussed here, no need for a neurotic lossless-only orthodoxy as media becomes more complex).

To whomever did the 2-channel mix, come on guys, can't you spare a music-lovin' brother at least 3dB higher average dynamic range? DR8 is still plenty loud for a soundtrack, right? There has been research suggesting high dynamic range compression can damage hearing (animal data). So even from the perspective of public health, maybe let's give 2-channel music more dynamic headroom for the sake of reducing deafness, plus quality-loving audiophiles will respect you for it.

All the above titles are available in multichannel/Atmos which has become my preferred mix for most of the new music I listen to these days (for reasons discussed, plus improved downmix dynamic range). 

Happy Canada Day to fellow Canucks, and Independence Day to the American friends soon!

Here's a closing tune from the summer of 1997 - a dedication to the "UFO Believer" audiophiles and salesmen out there; be mindful of the MIB - often what you thought you heard, you did not hear: 😎