Electrostatic speakers?

[gnarly]

Agreed! A tall electrostatic panel approximates a line source far more closely than a same-height array of small drivers.

Also, for a hybrid electrostat (tall panel + dynamic woofer), the line-source-approximating panel will roll off more slowly with distance than the point-source-approximating woofer does. This results in the tonal balance changing with listening distance, and can handicap a hybrid electrostat if it's compared to a conventional speaker at the wrong listening distance (I suspect this happened to the Martin Logan in the Harman speaker-shuffler room). One solution would be to make the relative levels of panel and woofer user-adjustable.

Yes, I think the continuous nature of panels and ribbons makes the difference between any array of small drivers..

Yes again, I think speaker systems should have user-adjustable relative levels between main and sub. I'll go so far as to say, it's plain silly not to.

 

[gnarly]

I thought they summed a lot of spherical sources ( the measured points) to recreate the sound field. This would also recreate the cylindrical sound field of the line array.

I dunno. Appears the math is forced to a spherical solution no matter the actual wavefront propagation. Which is the issue in and of itself if any line source wavefront exists.
But again, I dunno.
I did see in the NTK thread, that there is an additional NFS module for multiple drivers in an array. Maybe it's needed for line arrays/planars??

 

[Hayabusa]

I dunno. Appears the math is forced to a spherical solution no matter the actual wavefront propagation

Amir has measured the Magnepan and the system also recreated the narrow vertical dispersion. So clearly there is more going on.

 

[MattHooper]

Absolutely agree. And while their early groundwork was very useful and it is always a pleasure to discuss matters with Dr. Toole, something tells me their testing parameters might have stepped into the same circle of confusion I had been in, by not taking the home listening conditions into account and producing results which stand diametral against what people prefer when doing listening tests as a base for their personal purchasing decision.

I mean, if these results were accurately and reliably predicting buyer´s preference based on measurements, developing high end speakers based on these findings would have inevitably been a money-printing machine for Harman with JBL, Revel and Lexicon completely dominating the market with no-one ever buying a pair of stats.

I’m not saying, I agree to the letter with the above - the researchers involved were very clever, and certainly took plenty of variables into consideration in their tests.

But for one thing, it is interesting to consider how successful or not the project ultimately was for Harman Kardon. To what degree did all that time and research translate into speaker sales, and where they commensurate with expectations that they had produced what blind testing told them were among the very best speakers available? I’m not familiar with their speaker sales.

Another issue is that perhaps many of the speakers that did not perform as well as as Revels didn’t translate into real world use as particularly poor sounding. For instance, some loudspeakers I seem to recall weren’t that bad on axis, but they’re off access frequency may not have been as smooth. The results being that they would be more difficult to integrate into an average room than a Revel, but that wouldn’t bar them from actually sounding quite good if the audiophile was careful in their set up. Probably lots of audiophiles got speakers that didn’t perform textbook to actually sound pretty good with trial and error in their room and seating positions.

As to the Quad 63s, those things just sounded out of this world to me when I owned them, especially when paired with the Gradient subwoofers (and a CJ tube amp). Many nights I had the sensation of listening right in to a concert hall to a real person performing or singing.
Wasn’t just me. Guys I work with in post sound, mixers, editors, used to drop over for lunch and we’d spin tunes on the system and they were always completely blown away.

 

[tengiz]

Considering our ears are spaced horizontally, wouldn’t a vertically extended source make more perceptual sense? Point sources might be easier to study, but are they really the best fit for how we actually hear?

Do point-source speakers come out ahead in research because they’re genuinely better for stereo, or simply because they’re easier to design, measure, and iterate on - and so have received more attention and refinement over time? The whole point-source concept also seems to overlook the fact that most home rooms are reliably symmetrical in one axis: vertically, between the floor and ceiling.

Line dipoles, pumping out tall, mostly flat-in-the-vertical-dimension waves with an almost equally tall sweet spot, minimizing floor and ceiling reflections, and delivering a slower 1/R falloff over distance instead of the 1/R² drop you get from a point source - combined with their side nulls, they start to look like a pretty compelling package.

 

[Arindal]

Considering our ears are spaced horizontally, wouldn’t a vertically extended source make more perceptual sense?

In theory yes, they can particularly achieve more accurate and narrower phantom source localization, and there are concept of narrow line sources on the market. Another advantage is you can in theory combine higher directivity index (in the vertical dimension) with broad listening window in the horizontal one.

My personal opinion is that a linesource should not be too tall in order to keep vertical localization stable. Which leads to a concept like a truncated linesource of hybrid thereof, which many desigers seem not to like.

It is still a mystery to me why such concepts are dominating the sound reinforcement market but not hi-fi (or studio monitoring).

Do point-source speakers come out ahead in research because they’re genuinely better for stereo, or simply because they’re easier to design, measure, and iterate on - and so have received more attention and refinement over time?

Define ´point source´ pls! There are too many different concepts on the market, and the word itself is sometimes used in acoustics for an indefinitely small soundsource, which ist not what a loudspeaker is.

If you mean conventional coaxials: There are anything but easy to design and manufacture. And in my understanding they are popular because everyone can understand the advantages of the concept and they usually keep what they promise in terms of imaging stability. Unfortunately, some tradeoffs are not widely discussed, like uneven dispersion.

One more thought: your comment sounds as if linesource and pointsource are incompatible concepts. They are in fact not, in theory it should be possible to merge both, with would be something like the perfect speaker according to my personal philosophy.

 

[Duke]

... in theory it should be possible to merge both [line source and point source], with would be something like the perfect speaker according to my personal philosophy.

I'm intrigued. Could you elaborate?

 

[DavidEdwinAston]

I'm intrigued. Could you elaborate?

Can I ask you Duke, if there are Soundlab speakers as small as, or possibly smaller than the current largest Quad esl?

 

[Newman]

As usual, you vastly overstate the actual results of the Toole-Olive listening tests.

Far from it: I do my best to fairly represent what Toole and Olive have themselves said about their listening tests.

You are the one who seems to on a campaign of misrepresenting their tests to draw your personal convenient conclusions that just happen to be pretty much the opposite of what Toole and Olive have themselves concluded on the overall topic, ie integrating numerous test results and various experiments, by themselves and others, to a general conclusion on a topic. You don't like that conclusion, so you are on a campaign to white-ant their views, for example by cherry-picking one graph below and completely twisting a desired conclusion from it, as I have discussed below the graph.

It has crossed my mind to wonder why you whould choose the "anti-science / weaponise science against itself" approach, given that you are after all a PhD in Physics at Stanford with 40+ years of experience as a professor at an R1 university and you have authored many research papers. My normal experience of highly regarded scientists is that they are personally modest and would never dream of stomping all over the findings of scientists in a field which is not their field. But I guess they are not all so self-effacing.

Here is a key figure from Toole's book (p.385, 2008 edition):

View attachment 465273

As can be plainly seen, the results of blind vs sighted listening are very similar in terms of relative ranking, with only the ranking of the "small sub/sat system" changing significantly (from a tie with "US competitor" when heard blind to not as good when heard sighted, but both ranked below the two European speakers when heard both blind and sighted).

So: while the data shows that there is a visual bias effect (that varies with both the situation and the person), your claim that it "dominates" judgment of the actual sound is not supported by data.

...and here is Toole's own text around the above graph that you say refutes claims that sighted listening is not about the sound waves:
"A widespread belief among audio professionals is that they are immune to the influences of brand, price, appearance, and so on. They persist in conducting listening evaluations with the contending products in full view.
... Holt commented as follows:
As far as the real world is concerned, high-end audio lost its credibility during the 1980s, when it flatly refused to submit to the kind of basic honesty controls (double-blind testing, for example) that had legitimized every other serious scientific endeavor since Pascal. [This refusal] is a source of endless derisive amusement among rational people and of perpetual embarrassment for me, because I am associated by so many people with the mess my disciples made of spreading my gospel.
...In general, though, what listeners saw changed what (they thought) they heard."

And Toole's own summary of that section of the book (my bold):
"Summarizing, it is clear that knowing the identities of the loudspeakers under test can change subjective ratings.

• They can change the ratings to correspond to presumed capabilities of the product, based on price, size, or reputation.
• So strong is that attachment of “perceived” sound quality to the identity of the product that in sighted tests, listeners substantially ignored easily audible problems associated with loudspeaker location in the room and interactions with different programs."

and,
"These findings mean that if one wishes to obtain candid opinions about how a loudspeaker sounds, the tests must be done blind."

How much clearer can he be? And yet your lesson from it was your post above????

As any third party reading this can see (I am not confident that you can), Toole's extensively contextual commentary around that graph leads to the diametrically opposite conclusion to the one you attempted. Whereas my summary, which you say is vasty overstated, is actually very accurate.

I don't know why you are doing this. After all, you posted exactly the same misinterpretation of exactly the same graph in 2022, and I corrected it with Toole's own commentary of his own graph. To which you replied, "I ignored the words, looked only at the graph". I must say I admire the confidence of a non-participant observer who contradicts the researcher's own interpretation of his own data, based on one graph. Not only that, but the graph itself actually supports Toole's argument (what a surprise). After all, two of the three large speakers were visually identical, and their ranking relative to one another was exactly the same sighted vs unsighted. Whereas the score of each speaker, sighted, changed a lot from unsighted, going way up for the large expensive home brand speaker, up a bit for the large expensive home nation but competition brand speaker, and down for the small cheap speaker.

Or maybe I do know why you are doing this. Back in 2021 you expressed the view, "Of course it's possible to tell good sound from bad sound by listening alone." And you meant sighted listening, because you backed it up with sighted listening anecdotes and the conclusion, "I believe this experience was real." If you still stand by those conclusions, and you still stand by the use of sighted listening anecdotes as evidence that sighted listening is trustworthy, then you will do things like this, looking for little slivers of data from which you can extract a personal interpretation to support your intuitive/anecdotal assumption.... even if means insisting that a top researcher cannot interpret his own graphs, and that when I correctly summarise that researcher's conclusions, I am "vastly overstating" him.

 

[NTK]

...
First thing is technical...having studied NTK's NFS thread and Erin's interview with Klippel's Bellmann, it seems Klippel is using only spherical wave math imputed out to the acoustic far field's -6dB falloff, to then determine frequency response. (@NTK, if you read this, any extra light appreciated.)
It seems to me, owing to my thoughts about stats behaving as line arrays, that at some point, the upper spectrum needs to be analyzed as cylindrical waves, which changes far-field convergence distance(s) and associated response calcs.

Can a planar or line array even be measured/portrayed/condensed into a spherical point source? (which I see the NFS/spinorama type measurements, to be essentially).
Take a flat full range diaphragm,.... on-axis direct is pretty much the same over a fairly wide area perpendicular to the diaphragm. The off-axis set that goes with each chosen reference axis are pretty much the same too, until moving towards the panels edges. How do we really describe off-axis? It's kinda unchanging until we get to the end's of the panel/lines, where in the case of VHF the change is dramatic. These things are what I've measured with mine .What is the correct way here to evaluate off-axis.....I dunno.

Another thing is the well discussed dipole issue....how does NFS measure any speaker that has been purposely built for specific boundary conditions?
I know from having built straight-line floor to ceiling line arrays, and also CBT lines, that anechoic measurements are non-sense when reflections are essential to the basic acoustic design.

The spherical wave expansion used by the Klippel NFS is much less efficient when dealing with non-compact sound sources, i.e. the span (edge to edge distance) of the active sound radiating area is larger than the wavelength. That means it will require much higher order cutoff (N). Since the number of coefficients (C_mn) goes up with (N+1)², the number of measurement points required to compute those coefficients goes up with it.

In Amir's Magnepan LRS review, he measured >2000 points, more than twice those for the typical boxed speakers he tested, and the fitting errors are still unacceptably high for ~3 kHz and above.

The method used by the Klippel line array option is to measure each element of the array one at a time, and then take the (vector/complex) sum the individual responses to obtain the combined response of the array. This won't work for panels as you can't divide the panel into multiple little panels to measure one small (compact) portion at a time.

The NFS gives the free field response. If you want to predict how a (any) speaker perform in a reflective room, you can feed the NFS results into an auralization program.

 

[Duke]

Can I ask you Duke, if there are Soundlab speakers as small as, or possibly smaller than the current largest Quad esl?

The smallest SoundLabs are still a bit bigger.

My understanding is that the Quad 2912 is about 54" tall by 27" wide. The smallest SoundLabs are about 60" tall by 30" wide.

 

[tengiz]

Define ´point source´ pls! There are too many different concepts on the market, and the word itself is sometimes used in acoustics for an indefinitely small soundsource, which ist not what a loudspeaker is.

One more thought: your comment sounds as if linesource and pointsource are incompatible concepts. They are in fact not, in theory it should be possible to merge both, with would be something like the perfect speaker according to my personal philosophy.

What I meant is a “standard” piston-type transducer — one that’s small enough relative to the shortest wavelength it radiates, or shaped in a way that allows it to be approximated as a point source. Not omnidirectional in full 3D, but more like having rotational symmetry around its main forward axis. The goal is smooth directivity across a wide enough angle - no lobing, no strange frequency-dependent behavior. In a multiway setup, the transducers don’t have to be collinear, just close enough and individually point-source-like so that the overall response stays coherent.

An almost "true" point source, of course, would be something like a piezoceramic sphere - common in hydroacoustics. Works beautifully in water thanks to the mass loading etc., but unfortunately, that approach doesn’t easily translate well to air, of course.

That’s a whole different design philosophy from a contiguous line transducer, which isn’t aiming for that kind of rotational symmetry at all — at least not from the listener’s perspective. And even grid or array-based designs don’t really count unless the elements are spaced closely enough to behave collectively like a single coherent source.

 

[gnarly]

The spherical wave expansion used by the Klippel NFS is much less efficient when dealing with non-compact sound sources, i.e. the span (edge to edge distance) of the active sound radiating area is larger than the wavelength. That means it will require much higher order cutoff (N). Since the number of coefficients (C_mn) goes up with (N+1)², the number of measurement points required to compute those coefficients goes up with it.

In Amir's Magnepan LRS review, he measured >2000 points, more than twice those for the typical boxed speakers he tested, and the fitting errors are still unacceptably high for ~3 kHz and above.

The method used by the Klippel line array option is to measure each element of the array one at a time, and then take the (vector/complex) sum the individual responses to obtain the combined response of the array. This won't work for panels as you can't divide the panel into multiple little panels to measure one small (compact) portion at a time.
View attachment 465563
The NFS gives the free field response. If you want to predict how a (any) speaker perform in a reflective room, you can feed the NFS results into an auralization program.

Thanks for all that...makes sense per what I've gathered about NFS so far.
And I realize the NFS goal is to derive free field.

My big picture take is that NFS is best suited for conventional speakers that are not overly large, and are fit for typical small-space home audio rooms.

 

[gnarly]

What I meant is a “standard” piston-type transducer - one that’s small enough relative to the shortest wavelength it radiates, or shaped in a way that lets it be approximated as a point source.

Makes sense to me. And I think that simplicity can be extended past one transducer, to a collection of traducers that fit the size vs wavelength criteria to try to act like a collective point source. Well known examples being conventional co-axials.

My best DIY effort towards a full range 'collective point source,' has been this MEH/synergy horn with a 75H x 60V pattern.
The rectangular area encompassing all contributing ports measures only 24" x 16".
Radiating out of that area are: a coaxial compression driver, four 4" drivers, two 12"s, and two 18"s. So a full range 5-way.

 

[tengiz]

The spherical wave expansion used by the Klippel NFS is much less efficient when dealing with non-compact sound sources, i.e. the span (edge to edge distance) of the active sound radiating area is larger than the wavelength. That means it will require much higher order cutoff (N). Since the number of coefficients (C_mn) goes up with (N+1)², the number of measurement points required to compute those coefficients goes up with it.

In Amir's Magnepan LRS review, he measured >2000 points, more than twice those for the typical boxed speakers he tested, and the fitting errors are still unacceptably high for ~3 kHz and above.

The method used by the Klippel line array option is to measure each element of the array one at a time, and then take the (vector/complex) sum the individual responses to obtain the combined response of the array. This won't work for panels as you can't divide the panel into multiple little panels to measure one small (compact) portion at a time.

The NFS gives the free field response. If you want to predict how a (any) speaker perform in a reflective room, you can feed the NFS results into an auralization program.

Right, the theory holds - given accurate pressure or velocity on an enclosing surface, the exterior field is fully determined, in line with Huygens, Kirchhoff–Helmholtz, and all the rest. But I keep wondering about the limits when the source doesn’t behave like a mechanical radiator.

It seems the Klippel's measurement philosophy implicitly assumes a spatially band-limited source - something mechanically realizable. But if you’re using digital beamforming to generate something unholy, with sharp spatial features, deep nulls, or subwavelength structure, the reconstruction starts to feel fragile - not because of noise or inadequate sampling per se, but because the spatial content violates assumptions baked into the model.

Even with a fine enough grid, if the sampling and reconstruction pipeline assumes “speaker-like” behavior, can it really capture fields that don’t follow those rules? It starts to feel like a spatial adversarial case - not a breakdown in theory, but something that slips past what the system was built to see.

I’m not saying electrostatic panels can actually do this - it's more of a general consideration. I do remember Yamaha had a digitally beamforming consumer soundbar back in the 2000s. Makes me wonder: how well would something like a Klippel system handle a device like that, taken to the extreme? Would it still provide meaningful data, or would it be quietly misled by a source that doesn't radiate like any conventional transducer?

 

[Arindal]

I'm intrigued. Could you elaborate?

There are compact P.A. systems on the market consisting of two line arrays, one for midrange and one for the tweeters (for the latter, you oftentimes find ribbon tweeters which are line sources by themselves, or on array of smalls horns/domes). Some place the drivers like this (example by Christie):

It is a linesource, and by the way the midrange drivers and tweeters are arranged, it has from localization and lobing point the properties of a pointsource as well (and I don't mean pointsource in the sense of hypothetic infinitely small soundsource creating a perfect spherical omni wavefront). For hi-fi purpose this geometry would not be a good idea as the tweeter array is too broad and would show restricted treble dispersion. But it might be doable in miniaturized form or with different drivers like horns, partly overlapping each other. As mentioned, this is just a thought experiment. I have no idea if this concept would meet an unsolvable dilemma or is doable.

The goal is smooth directivity across a wide enough angle - no lobing, no strange frequency-dependent behavior

Smooth or constant directivity?

Have some personal sympathy both for coaxial concepts and for planars but they are both pretty fuzzy when it comes to delivering constant directivity (which I find to be as important), as both have the inherent tendency of getting narrower and narrower in dispersion pattern with higher frequencies without the usual countermeasures like horns helping here.

Multi-entry horns could be a solution indeed, but they usually get broad, I have no idea how that translates to localization.

A line source has the advantage that you can make it very compact in one dimension (width), and very tall in the other.

But if you’re using digital beamforming to generate something unholy, with sharp spatial features, deep nulls, or subwavelength structure, the reconstruction starts to feel fragile - not because of noise or inadequate sampling per se, but because the spatial content violates assumptions baked into the model.

Does not even have to be that complex in terms of beamsteering. A simple active cardioid with just 2 or 3 sources delayed and inverted makes it seemingly difficult to calculate the anechoic behavior.

 

[tengiz]

Smooth or constant directivity?

Looks like I keep misusing the terminology/jargon - being from a different filed: physical acoustics completely unrelated to audio. When I said “smooth,” I meant something that varies gradually within a limited range before falling off - not necessarily constant directivity

Does not even have to be that complex in terms of beamsteering. A simple active cardioid with just 2 or 3 sources delayed and inverted makes it seemingly difficult to calculate the anechoic behavior.

Indeed! I hadn’t really thought of it that way as a prime example. I started working in my field after digital beamforming - outside of military applications - had already become the norm in civilian research, so that’s the framework that shaped my intuition.

 

[MattHooper]

It has crossed my mind to wonder why you whould choose the "anti-science / weaponise science against itself" approach,

If you could put your torch down for just a moment… somebody disputing the interpretation of a graph does not make them “anti-science.”

You’ve been at times incorrect in interpreting what Floyd Toole has written. Does that make you “ anti-science?”

Nobody’s disputing that there’s a problem of sighted listening bias, or that blind testing doesn’t help overcome this variable and can therefore be useful.

What is being disputed is your particular tendency to overstate these issues, by continually saying that sighted listening will be GUARANTEED to be DOMINATED by the sighted listening effect.

It’s been pointed out to you why that is an unsupportable claim, and that at the very least you need to dial back or explain to what degree you think this “guaranteed” bias will distort perception.

The problem you face with this level of exaggeration has been put to you over and over. … without your having come up with a satisfactory way to resolve the level of scepticism you ascribe to sighted listening.
And the way that you avoid examples that are inconvenient to try and explain based on your conclusions is telling.

You quote Toole :

"Summarizing, it is clear that knowing the identities of the loudspeakers under test can change subjective ratings.

I’ve helped you out by emphasizing the relevant word. “can”

“can” is very different from GUARANTEED to be DOMINATED by the sighted listening effect.

Does Toole give examples showing where sighted listening effects played a part in influencing speaker preference ratings to make his point? Of course.

But he’s also in these pages said that sighted listening is not useless. Because of course he understands bias effects can’t be GUARANTEED to DOMINATE sighted listening impressions because if that were the case blind listening tests and measurements would have no relevance to how we actually listen to loudspeakers.

Therefore, the coherent, reasonable, non-dogmatic, non-torch bearing interpretation of this:

and,
"These findings mean that if one wishes to obtain candid opinions about how a loudspeaker sounds, the tests must be done blind."

Would be but if one wants to obtain (the most) reliable opinions about how loudspeaker sounds the test should be done blind.

Which it seems pretty much everybody agrees with, but which is different from your overreaching statements about sighted listening guaranteeing inaccurate perception.

Not only that, but the graph itself actually supports Toole's argument (what a surprise).

It supports Toole’s more careful positions. What it doesn’t support is YOUR statements.

After all, two of the three large speakers were visually identical, and their ranking relative to one another was exactly the same sighted vs unsighted. Whereas the score of each speaker, sighted, changed a lot from unsighted, going way up for the large expensive home brand speaker, up a bit for the large expensive home nation but competition brand speaker, and down for the small cheap speaker.

Hold on. In that particular test: the majority of the scoring relationships remained consistent from sighted to blind.

Maintained between sighted and blind, listening : The two European speakers were rated best, higher than the other two speakers, with speaker one and two maintaining the same preference relationship in both tests (design 2 rated slightly higher than design 1). Both rated significantly higher than the American speaker - and the American speaker ratings changed very little between sighted and blind. It was the sub/sat system that displayed the most significant change between sighted and blind listening. As Toole has pointed out that speaker system was “ small and plastic” and clearly produced a stronger sighted bias effect than the other speakers, which is certainly not hard to believe.

But Mark was correct in pointing out that test clearly contradicted your claim that sighted bias effects are guaranteed to dominate the actual sound - like he wrote, there was an effect, but not nearly as dominant as your claim suggests.

Otherwise, you’d have to explain the relative consistency in the ratings. And you’d have to explain the relevance of blind test testing to the consumer selecting speakers. As always.

 

[MarkS]

The spherical wave expansion used by the Klippel NFS is much less efficient when dealing with non-compact sound sources, i.e. the span (edge to edge distance) of the active sound radiating area is larger than the wavelength. That means it will require much higher order cutoff (N). Since the number of coefficients (C_mn) goes up with (N+1)², the number of measurement points required to compute those coefficients goes up with it.

It should be possible to do better by using a more apt coordinate system. For a large tall planar, elliptic coordinates in the horizonatal plane plus a height coordinate z should be close to optimal. Of course Klippel would have to write the approprite software to implement this.

 

[tengiz]

Yesterday I went to listen to a pair of QUAD ESL-63s that had just popped up at one of the local hi-fi consignment stores. They were said to have been recently restored.

I’d never heard QUADs before, and given that they’re a genuine piece of audio history, I was excited to finally experience them. The older salesman set everything up and started playing some music.

Unfortunately, it just wasn’t a good experience. I’m pretty sure the amplifier - a tube unit - wasn’t the issue, but I didn’t feel comfortable asking him to swap it out. I didn’t want to bother him more than I already had. The amp was huge, probably quite heavy, and the speaker cables were just barely long enough to reach the terminals, so swapping it would’ve been a hassle for the old man.

I tried a few tracks I know well, but it just didn’t work. The sound was plain bad - distorted and thin. There was some imaging, but nowhere near what I’m used to from other electrostatic speakers I’ve listened to for decades.

Given the reputation these speakers have, I’m fairly sure the restoration didn’t fully resolve whatever issues are still present in the speakers themselves.

For those here on the forum who’ve enjoyed ESL-63s in the past, I’m guessing this particular pair still needs proper servicing. I could theoretically take that on myself - one of my degrees is in electrical engineering - assuming it’s even feasible for someone who’s never opened a pair before. The big question is whether the necessary parts can still be sourced, or if there are workable substitutes for anything that’s no longer available.

That said, the price is very reasonable. I can easily afford to take the risk. Even if I fail to restore them properly, it wouldn’t break me.