advertisement


A look at the Harbeth P3 crossover

I genuinely wonder what he means by ‘quality parts/good parts’.

I’m hugely skeptical of this as I’ve been down the road of thinking expensive film caps are “better” only to take them out again and realise the original designer knew exactly what they were doing and actually voiced the kit perfectly using the components they selected.

I select and decide to own kit because I like it. If I want something different I’ll just go and buy it. It is a trajectory to hell to assume I (or any random self-professed expert off the internet or YouTube) can do better than Peter Walker, Guy Fountain, the BBC or whoever. If you like the sound of Harbeth or whatever just enjoy them. Alan Shaw voices speakers that sell well and hold value. Filling them up with boutique components will have entirely unpredictable results, will cost a lot of money, and will certainly destroy any resale value. I now try to the best of my ability to restore/refurbish kit that is suffering the signs of age, I’m absolutely not trying to alter or tweak it.
 
I sometimes think a lot of what gets bantered about in the end really matters far less than some want it to be.
 
Some interesting reading here about the B110 and cone materials in general - http://www.nutshellhifi.com/library/speaker-design2.html

Quote "The BBC-derived designs always employed notch-filter equalization to flatten the Bextrene driver in the midband; the most famous (or infamous, depending on whether you were the listener or the designer) driver was the KEF B110 used in the BBC LS 3/5a minimonitor. Not everyone knows that this speaker, which is legendary for its sweet midrange, employs a deep notch filter with 6dB of attenuation at 1.5kHz to correct the B110."

Frequency response of a B110 here. I think this was the Linn Kan version - http://ww.dibirama.altervista.org/h...f-b110-sp1057-mid-woofer-5-8-ohm-50-wmax.html
Doesn't look like there's a problem at 1.5khz in other versions of the B110 in post #14 either.
Not sure how useful this is given I measured them in free space without a baffle from a distance of 1 metre, but here's the raw response of a pair of 8ohm KEF B110 SP1003 compared to a similar vintage 16ohm Audax HD13 and modern 16ohm Monacor SP-135TC:

https://pinkfishmedia.net/forum/threads/jr150-restoration-thread.190943/page-5#post-3690303
 
Not sure how useful this is given I measured them in free space without a baffle from a distance of 1 metre, but here's the raw response of a pair of 8ohm KEF B110 SP1003 compared to a similar vintage 16ohm Audax HD13 and modern 16ohm Monacor SP-135TC:

https://pinkfishmedia.net/forum/threads/jr150-restoration-thread.190943/page-5#post-3690303

It's showing the differences above about 2khz, but below it's hard to tell what's a driver issue or diffraction. Obviously, as the baffle gets smaller the baffle diffraction peak raises in frequency (shorter wavelengths), and a circle is the worst possible shape because the diffraction happens at the same frequency / wavelength, so I'm not surprised at the sharp 1 to 2khz peaking.

I've heard of people keeping the drivers in the cabinets, then burying the cabinet so it's flush with the ground, and then measure directly above as you did. This is a better way to isolate the drivers from the cabinet to see what they are really doing. Or you can flush mount them in a large piece of mdf or plywood etc.
 
I genuinely wonder what he means by ‘quality parts/good parts’.

Capacitors especially, can make a difference, but if your speakers have been very carefully voiced using certain parts, and then you change them, there's a chance you could make them sound worse, and who's to say his go-to parts are guaranteed to sound better than the original anyway. I've never tried the tube connectors that he's always banging on about, but I'd be quite surprised if anyone could notice any difference.

I have about half a dozen of pretty much every value sandcast resistor you can get. Mills do sound slightly sweeter, but I'm not convinced they are more accurate (I like a bit of 'bite').
 
Last edited:
Here’s the BBC Research departments paper on the design of the original LS35/A. Details of the crossover are on page 2. Notice that the crossover is adjustable “to allow different relative sensitivities of individual l.f. and h.f. Units to be matched”. The paper also mentions that it was necessary to redesign the speakers between batches of drive units.

http://downloads.bbc.co.uk/rd/pubs/reports/1976-29.pdf
 
It's showing the differences above about 2khz, but below it's hard to tell what's a driver issue or diffraction. Obviously, as the baffle gets smaller the baffle diffraction peak raises in frequency (shorter wavelengths), and a circle is the worst possible shape because the diffraction happens at the same frequency / wavelength, so I'm not surprised at the sharp 1 to 2khz peaking.

I've heard of people keeping the drivers in the cabinets, then burying the cabinet so it's flush with the ground, and then measure directly above as you did. This is a better way to isolate the drivers from the cabinet to see what they are really doing. Or you can flush mount them in a large piece of mdf or plywood etc.

TAPiKOz.jpg
 
Here’s the BBC Research departments paper on the design of the original LS35/A. Details of the crossover are on page 2. Notice that the crossover is adjustable “to allow different relative sensitivities of individual l.f. and h.f. Units to be matched”. The paper also mentions that it was necessary to redesign the speakers between batches of drive units.
http://downloads.bbc.co.uk/rd/pubs/reports/1976-29.pdf
http://downloads.bbc.co.uk/rd/pubs/reports/1976-29.pdf

I posted a screenshot of the crossover page earlier: https://pinkfishmedia.net/forum/threads/a-look-at-the-harbeth-p3-crossover.261066/#post-4513984
 
I feel I should make a couple of comments:

Firstly, a passive crossover should be as simple as possible. As someone earlier pointed out, the more component it has, the more likely the deleterious effects (insertion losses, being one of them) are. The selection of driver complement, in itself, is a bit of an art. Get that right, and the crossover will be much easier to design for simplicity. I'm a firm believer that using notch-filters to tame resonant peaks, or conjugate filters to flatten impedance are unnecessary with sympathetically chosen drivers. My very own E-IX has a fourth-order Linkwitz-Rliey acoustic crossover, and yet uses only two inductors, two resistors and three capacitors. Yet, it has near perfect relative phase tracking, and sounds better than anything that tiny has any right to. I have to say that, of course, but so do a number of PFMers who have made or acquired a pair for themselves.

Secondly, component quality, as do most things, make a difference. The question is how audible are the differences. I have used cheap Solens and less cheap Mundorf, Jantzen, Sonicap etc. in many of my designs. They make a slight difference to the perceived quality of HF when used on the high-pass section of the crossover. Basic Solen do sound grainier than Jantzen Z-Superior, but if basic components is what the crossover is designed with, then it may or may not sound better with higher quality parts. I have specified Z-Superior on my E-IX's high-pass, and I expect it sounds exactly as I intended when the crossover is built accordingly. Someone wishing to use mega-buck Duelund caps may do so to please themselves, but I can't guarantee it'll be a better result - mainly because there will be other electrical attribute (ESR, for instance) that could change my design.

James
 
I feel I should make a couple of comments:

Firstly, a passive crossover should be as simple as possible. As someone earlier pointed out, the more component it has, the more likely the deleterious effects (insertion losses, being one of them) are. The selection of driver complement, in itself, is a bit of an art. Get that right, and the crossover will be much easier to design for simplicity. I'm a firm believer that using notch-filters to tame resonant peaks, or conjugate filters to flatten impedance are unnecessary with sympathetically chosen drivers. My very own E-IX has a fourth-order Linkwitz-Rliey acoustic crossover, and yet uses only two inductors, two resistors and three capacitors. Yet, it has near perfect relative phase tracking, and sounds better than anything that tiny has any right to. I have to say that, of course, but so do a number of PFMers who have made or acquired a pair for themselves.

Secondly, component quality, as do most things, make a difference. The question is how audible are the differences. I have used cheap Solens and less cheap Mundorf, Jantzen, Sonicap etc. in many of my designs. They make a slight difference to the perceived quality of HF when used on the high-pass section of the crossover. Basic Solen do sound grainier than Jantzen Z-Superior, but if basic components is what the crossover is designed with, then it may or may not sound better with higher quality parts. I have specified Z-Superior on my E-IX's high-pass, and I expect it sounds exactly as I intended when the crossover is built accordingly. Someone wishing to use mega-buck Duelund caps may do so to please themselves, but I can't guarantee it'll be a better result - mainly because there will be other electrical attribute (ESR, for instance) that could change my design.

James
Why not just go active?
 
Why not just go active?
Passive crossovers are cheaper than multiple amplifiers, for one.

Active is undoubtedly better if designed for the driver complement in-situ, but I find it easier to tinker with unpowered circuits.
 
Surely you mean low pass.

There's a multitude of reasons to not run mid-bass drivers unfiltered. Break up can be tamed with enough doping, but beaming cannot.
 
The Tannoy 607s that lived in my mancave system for a while had no crossover on the bass driver and I believe some Epos 2 ways possibly did the same thing.
 
The Tannoy 607s that lived in my mancave system for a while had no crossover on the bass driver and I believe some Epos 2 ways possibly did the same thing.

Given modern materials and computer modelling I can’t see why anything beyond a simple safety HPF on the tweeter is required for any company with the ability to actually design a whole loudspeaker from scratch (i.e. not an off-the-shelf driver box-stuffer). Roll-off and bandwidth should be definable mechanically within the driver itself to a very large degree. If it was possible way back in the Epos ES14 era it should be simple with current technology. The 12” Tannoy Silver ran the 12” paper bass cone wide open back in the 1950s and that’s considered one of the best sounding speakers ever made by many. I don’t understand why so many modern crossovers are so complex. I get it with the box-stuffing brands, but the proper manufacturers such as Kef, Spendor, IAG etc etc should be able to do the design at the driver stage, surely?
 
Given modern materials and computer modelling I can’t see why anything beyond a simple safety HPF on the tweeter is required for any company with the ability to actually design a whole loudspeaker from scratch (i.e. not an off-the-shelf driver box-stuffer). Roll-off and bandwidth should be definable mechanically within the driver itself to a very large degree. If it was possible way back in the Epos ES14 era it should be simple with current technology. The 12” Tannoy Silver ran the 12” paper bass cone wide open back in the 1950s and that’s considered one of the best sounding speakers ever made by many. I don’t understand why so many modern crossovers are so complex. I get it with the box-stuffing brands, but the proper manufacturers such as Kef, Spendor, IAG etc etc should be able to do the design at the driver stage, surely?

The only off-the-shelf driver that I can think of that can be run without a crossover is the Seas A26 - http://www.seas.no/index.php?option=com_content&view=article&id=475:seas-a26-kit&catid=66:seas-d

The Eminence B102 has a reasonably smooth roll-off, but I'm pretty sure it would need an inductor for baffle-step-correction. I have a pair here but have never used them (does anyone here want to buy them?)

You require a soft cone material for a smooth roll-off, and some would argue softer cones are less detailed than harder ones. With good quality parts, and a reasonably simple crossover, I'm not convinced that they degrade the sound of a woofer, It's mostly the tweeter that shows the difference between parts. So for me, I'm not bothered about finding drivers with a smooth roll-off, just as long as It won't need lots of parts to correct it.
 


advertisement


Back
Top