Why Phase IV?

Just before going to college (1979) Graham, who worked at B&W, sold me a couple of 801 bass drivers. Phase I used these (they were really wonderful units) with Kef B110bs for the mid range and DM7 stand alone tweeters. The crossovers were stock 3-way units that crossed at 200 and 4k (at least as far as I can recall). I have no pictures of the Phase Is.

Phase II was a rebuild of the Phase I cabinet but used the same drivers and crossover. The result was fairly unsuccessful since I limited the bass response by making the air volume too small. Again, no pics.

Phase III was the intervening 30 years.

Phase IV was my attempt to build speakers using much cheaper drivers (my father, bless him, threw out the 801 drivers when he was cleaning out his attic in 1990). "Phase IV" (1974) was also a fairly memorable science fiction film for which "Silent Running" (1972), a slightly more memorable (also science fiction) film, was the supporting B-feature when I saw it (probably in 1976).

That there is a Mk1 and a Mk2 was an accident. I hadn't planned to build a Mk2, but as I noted earlier in the blog, it was a spur of the moment thing.

Aren't you sorry you asked?

Sparsness revealed

Listening to Pat Metheny "Bright Size Life" I realize for the first time how sparsely 'populated' it is. Having never looked closely at the sleeve (I listened to it mostly - but not exclusively - in the car for some reason) I see now that it's a trio! Of course having Jaco Pastorius on bass helps give the illusion of many more people, but even so...   

So why does this matter? Well, that I noticed I think suggests that the MkIIs are reveling much more detail than the DM6s. When you're listening for differences it's not hard to imagine ones that sometimes aren't really there. But this time I wasn't; I just put on the disk for something to listen to and the instrument separation leapt off the page. They may not be perfect but they're good enough to keep.

One month on..

A month of living with the Mk IIs and I'm not disappointed. They're not as clear as perhaps they could be, and spatial separation doesn't leap of the page, but they do reveal more detail than the DM6s. This may be a function of reflection in a small room. Of course, you might rightly suggest that a pair of 40 year old loudspeakers is hardly the best benchmark. Not only is physical deterioration potentially a problem, but technology and design have moved on since the 1970s.

That being said, it's all I've got (other than a pair of Kef bookshelf speakers I bought in 1990 - which I don't know I can find anyway). So I'm going to use the staircase approach. The Mk IIs are better than the B&W. The Phase V should, if I get things right, be better than the Mk IIs.

Of course, I'm not entirely sure what happens when the next design is worse than it's predecessor. However, using better drivers (I have already bought the AC 250F1s for the woofers) in the Phase V should reduce the likelihood of that happening.    

Phase IV complete

Sunday evening. The painting is finished, the cabinets polished. All that remains is mounting the drivers: Which was going fine until the last unit, which was taken from the original Mk 1 prototype (left rear, white cab). These drivers were 6 months or so older than the ones in the Mk 2s (red) and turns out weren't exactly the same size. They were significantly (about 1 mm) larger in diameter, which meant the openings in the baffle had to be re-routed. Without anything in the middle to locate the router this is a bit of a problem. Silver Flute has a long way to go in terms of manufacturing consistency.

However, 6am Monday morning, before it got too hot, I cut a plug to go into the baffle opening and re-routed the hole. By 8:30 they were finally finished.

Both were tested. The graph below shows the two Phase IV units (one black, one red) and one of my old DM6s (blue). I was (am) concerned about one of the Phase IVs having a 9dB drop around 162 Hz but then THe DM6 has a drop of 14dB @ 248 so the problem isn't quite as horrible as I first thought. Still it's not great; and worse yet I don't know what's causing it. I thought for a moment it was a polarity problem between the woofers and the mid-range but reversing the woofers polarity made things worse so that wasn't it. Most likely it's a feature of the cabinet which is a bit of a shame since at this stage I don't really know what I can do.

However they sound very different in the mid and upper mids from the B&Ws. I think they resolve a lot more detail but I'm so accustomed to the B&W sound - rather warm and soft like an old pillow - that the Phase IVs will take a little getting used to.

Finally, I'd like to thank Mark Zachmann for creating Speaker Workshop which has been an invaluable tool.

This will be the last post for a while. I have a three more projects lined up (replacing the B&W crossovers, building a smaller version of the Phase IV with only one 4ohm woofer, the then the Phase Vs for which the design work is done). But for the moment I'm going to spend some time getting used to the Phase IV Mk2s. 

Progress update

The (new) second cabinet is now built and waiting the final coat of paint. The first crossover is built and mounted in the first cabinet, which is now in use. The second should be up and running in two weeks.  One mildly interesting anecdote: while I had the speaker on it's back (the crossover is mounted inside the base), one of our cats decided to sit on the tweeter. Light for a cat she may be, but 8lbs crushes a soft dome tweeter. Somewhat mortified, I removed the driver, and then the dome and voice coil from the magnet. From the back I gently pushed the dome back into shape and smoothed out the wrinkles (!) with my thumbnail. (High tech stuff, speaker building). Without much in the way of expectation I remounted the driver and ran some tests. I then ran the same test with an undamaged tweeter of the same make and model. Result? There is really no noticeable difference in the frequency response. So it turns out that you can crush a tweeter's dome and as long as it's dome shaped it seems not to matter.

Setback

Today I realized the second cabinet has to be rebuilt. Painting it before it was finished structurally cause the wood to swell and bow, and go out of true. The baffle, which was not yet in place, was now too wide at the top and bottom and too narrow in the middle. Neither the baffle not the cabinet could be planed into shape any more, and no amount of coercion was going to get the baffle to bed into the cab properly. While I could have simply made a new baffle, the bow on the cabinet walls was such that I decided that it would be easier to start anew. Having dismantled the cabinet with a mallet (most satisfying), and looking forward to another three weeks worth of work, while I know that was the right thing to have done, have to say I'm a little despondent about the setback. The mistake was the result of a false economy in time and money.Live and learn I suppose.

Conception to realization

Conception

Realization

One done

One of the cabinets is finished. Looks better (I think) than the Mk1. Sounds OK through the first order Mk 1 crossover, but measurement and testing begins this afternoon.

Several lessons learned this time. First, don't use a light colored primer under the black spray on the front. The smallest scratch and it shows through. Second, make absolutely sure that joints are sealed. I discovered that the bass cab isn't air tight. The hole must be fairly small but it's enough. Whether this matters much I don't know yet.Another is don't use Duck masking tape to prevent the wood getting painted. It makes a real mess coming off leaving glue and bits of paper all over the wood. I've still got to figure out the painting. The red spray pain has left a fine stippled effect which is not what I wanted.  And I spent some much time sanding and priming and re-sanding and re-priming... 

This design uses two silver flute 8" drivers in parallel for the bass but one is a 4 ohm with a 3mH inductor in series with the driver. The ideas was to give a little boost to the very low (<100Hz) frequencies which these drivers don't handle overly well. We'll see this afternoon whether the idea worked.

Uneven progress

At the weekend, things seemed to be going slowly until I made several breakthroughs; the first was the decision add a base and then to increase the height of the base to allow the crossover to be mounted under the cabinet but not inside. That meant I could finish the wiring and not have to wire the binding post thought the inside of the cab. Aside from the last set of connectors the wiring is now finished. I also decided to glue the baffle in place which mean the mid cab had to be braced and filled from the top; which it was, and the top is now on and glued and the whole things trimmed and planed into shape. Filler was applied to the exposed pieces of particle board and the first unit is ready for sanding and painting. When I get the additional wood needed for the base I'll be able to complete that at which point the unit will be reading for measurement and testing. I'm waiting for the binding posts and terminals but, Fedx permitting, next weekend, I will be mounting the drivers. 

Slow progress

The Phase IV Mk2 is proving much more complicated to build than the Mk1. Partly it's the shape, but what seems to be holding me up is sequencing. I roughed out the top and bottom and these need to be trimmed. I probably need to paint the cabinet (which will be white) before gluing in the baffle (black). I need to make sure that the crossover will fit though the woofer's hole in the baffle which means getting the mounting board for the crossover built, located and then removed again... I am still figuring out details as I go such as how to mount the crossover, how to connect it to the external links in the base that connect to the drivers. I also learned something about quality: the more linked steps there are in a manufacturing process, the more quality matters. Screwing up (for me) is almost inevitable; but it matters more when a large number of antecedent steps are wasted as a result.

Phase IV Mk2 construction underway

After very little thought, I decided to see if I could improve on the Phase IVs. A new cab design was created in Sketchup (which I'm now getting better at using) in fairly short order. Judith was asked for her thoughts on the three variants I generated (each with slightly different baffle widths) and she chose the right one.

A weekend of fairly hard work, and not all of it plain sailing. I built part of a new pair of cabinets (this time I decided to build a pair). I'm not cabinet maker so I'm learning as I go, and almost all is trial and error. On Sunday there was a lot of error. The plate joiner joints between the side panels and the corners weren't at all clean and didn't have any structural strength, something which became very clear when I kicked one of them (to test the gluing, not out of frustration that the join wasn't mating cleanly - honest guv).

I'm going back and forth between the real artifact and Sketchup, updating the drawing as I make small adjustments in the process of building. For example, I added the cabinet bracing to the drawing and found that there may be a problem with the brace hitting the back of the mid range driver. We'll see soon enough. I'm quite pleased with structure in the back half of the cabinet. Now, I need to think about how best to build the fronts.

I ordered the drivers today (two 8" Silver Flutes for the bass @ $36, a 5 1/2" Silver Flute mid $20, and I've got some Vifa D27TG-45 tweeters to use up). I also ordered some ferrofluid in case the poor HF response (6db drop >15kHz) of the D27TG-45s has something to do with the old ferrofluid. I've no idea but it's worth a try. If that doesn't work I'll try replacing the voice coil.

One nice thing though; I was looking at either the Vifa XT25BG60-04 or the XT25TG30-04 as tweeter if I can't the the D27TG-45s to brighten up at the top and the mountings are identical. So if I can't get a pleasing sound from the D27TG-45s, I can swap them for either of the XTs without rebuilding the baffle (which is good because this time they're going to be glued not screwed in place).

The other change I'm anticipating is putting the crossovers inside the cab (a bit radical, I know). The drivers will be here by the weekend (I hope), an incentive to get the cabinets finished.

One week on

I've been listening to the Phase IV with the second order crossover for about a week now. The sound is still fairly pleasing, but isn't as crisp at the top as I'd like. Sounds a little like the tweeter is covered with a pillow. The bass is tight but does struggle at the very low end. So I'm about to put the first order cross over in and try that for the coming week.

A tale of two crossovers

Friday saw the arrival of two boxes of components from Erse. Definitely the place to go; their components are cheaper and have narrower tolerance bands than those I was buying from Madisound. The 4.0mH 14 gauge coil is, in the vernacular, "stonking".

The two crossover designs were both based on the same near-field measurements. The aim was to test the theory that 1st order networks did less harm to the phase than 2nd order network. Both should have fairly flat response curves if the simulation is to be believed.

Measurements in Speaker Workshop suggested that the frequency response wasn't too far from simulation for both. However, phase was another thing altogether; although the 1st order network looked better in theory between 500Hz and 5kHz, empirically there was little to tell the two apart. The 2nd order had a slightly flatter response so for the moment I'm staying with that one. 

The 1st order seemed to have slightly better performance below 50 Hz, and the 2nd order has a 9dB peak at about 250Hz - so on reflection I think I'll fry a 1st order on the Phase Vs (not yet built). The upper panel is the 1st order network the lower panel, the 2nd order.The black lines is the simulated response, red is measured. The dotted lines show the phase.

Another factor favoring first order is that there are fewer parts, particularly coils (copper isn't cheap, particularly when you're talking about 14 gauge). If I can damp the mid and bass driver oddities at the top end of their ranges by stuffing the cabs, then perhaps the slower roll off isn't such a bad thing...

So it looks like I can get fairly close with the measurement setup as it stands. Clearly measuring the driver performance in the cabs they'll be working in and building the crossovers accordingly is (quite obviously) the way to go. So for the phase V's I'll spend more time measuring the raw driver performance in situ and tweaking the physical design before turning to the simulation and the electrical design. 

Sitting here listening to the results, I'm surprised by what one can get out of the two Silver Flutes which were the cheapest cast frame woofers Madisound had.  Had I not already bought the AC 250s I'd be tempted to build the second cab and stop here. Interestingly I spent more on crossover parts (having now built four variants) than on the drivers, although the parts in the final 2nd order network I'm currently listening to came to about $55. Excluding experimentation and waste the total for half a pair was about $260. Since I think they're now at least as good as my DM6s (which I've seen advertised on the web for $1,000 the pair) which I bought second hand in 1979 for about $2,000 in today's money, I'm feeling fairly smug. 

Finally the philosophical question for the evening (post a smallish celebratory Blue Moon beer). Should I do the measurements in the room in which I will be using them (my home office) where there are all sorts of reflections that confuse things? Or, alternatively I could relocate all the measurement stuff (PC, amp, mic and pre-amp) to the shed and do the measurements in the open air - no walls to mess things up...?

Phase IV rebuild complete

Over the weekend I finished the rebuild of the Phase-IV baffle. Actually the baffle was done a while ago but since my wife was unhappy having something so ugly in the house, I decided to try to improve how it looked by giving it a coat or ten of paint. There's a I need to learn about finishing.

The mid range and tweeter are now properly located in close proximity so the interference fringing around the crossover frequency should be fixed. By way of comparison, here's what the speaker looked like before. Not great, but a lot less like a building site.

The bass cab is now filled with a fair amount fiberglass insulation - I know there are better materials but this was cheap ($15 and I've used less than a quarter of the roll). Since this changes the bass response, I'll need to rebuild the crossover. So next weekend the measurement equipment comes out again.

Volume estimation

I am still going back and forth over the size of the cabinet; is 5000 cubic inches (about 80 liters) enough for the sealed bass enclosure? In building the Phase IV, I had used the volume information from the DM6 which has in internal volume of "between 51 and 60 liters" according the manual. I decided to look at the original B&W 801; based on the published overall dimensions and some pictures, I estimated the internal volume to be somewhere between 108 and 122 liters depending on the thickness of the cabinet walls (according to B&W's literature, the Nautilus 801 walls are 1.5 inches thick - I used 1 inch and 1.5 inches). Big difference!

The actual air volume is likely considerably smaller since there is a fair amount of internal bracing (the 'matrix' used in the Nautilus could take 10%-15% off this figure). However, since the dimensioning calculation I was doing ignores the internal bracing, I decided to compare only the raw internal volume.

The 801 bass cab volume (~115 liters ) was over double that of my Phase IV (54 liters) and suggested a complete redesign of the cabinet was needed. It probably also contributes to the poor performance below 100 Hz, though the small linear excursion of the Silver Flute drivers is clearly another factor. The Phase V needs a wider baffle to accommodate the 10" AC-250F1 driver, but this alone wasn't nearly enough; several iterations of changes led to a design that maintained the height at 40 inches, increased the depth from 21 to 24 inches (slightly deeper than the 801 which was 22 inches back to front) and kept about the same width (17"). The new bass enclosure will weigh in at 110 liters before bracing. (It will also be fairly heavy)

Still baffling

To fix the destructive interference issue, I building a new baffle. It will also be almost twice as thick, partly to reduce resonance, and partly as a way of loosing all the ugly screws on the front but still allowing access to the inside. The star of the piece is the new circle cutting jig.

(It fits only some Craftsman routers - the plunge router has a larger base, but it does fit the fixed base model). Depending on which way you put in the center locating pin slider thing, you can cute really small circles (<2 inches in diameter) to fairly large ones.

The router jig enabled me to rebate the drivers which I hadn't done on the first baffle. Although I don't understand the physics, there is a strong school of thought that the tweeter in particular needs to be flush with the baffle.  Presumably something to do with diffraction from the sharp edges?

This shows the double layers and the neat rebate the router cuts. The lower section of the baffle will be screwed in close intervals to the cabinet, and the upper section will be attached to the lower plate with the driver mounting  bolts and possibly four screw, one at each corner. 

The AC 250F1's are now in so I want to get this (Phase IV) completed before building he new cabinets.Once the new baffle is built I will likely line the cabinet with SAE-13 or SAE-15 1/4" felt. I didn't do this first time round since I wasn't sure I wanted to blow $80on felt if the sound really sucked. (I also wanted to test the reflective diffusing properties of the cabinet design without any absorbing material). 

Phase V

Current plans are forming around the AC-250F1, an AC-130F1 or MkII, probably the latter (I like the idea of using carbon fiber somewhere) and Hiquphon OW1. The two driver idea will have to wait. Some testing of the implications of the phase change created with any second order or higher filters suggest using only first order. I am assuming the trade-off is with harmonic distortion when significant signal gets to the drivers outside of their ideal operating range. This remains, as they say, an empirical question.  What bothered me was the real mess second order filters seem to make of a square wave.

Above is the simulated result of passing a square wave though the second order crossover designed around the AC-250, AC130 and OW1. Speaker Workshop generates the phase shift data in the total frequency simulation. Using these, I simulated a square wave from odd harmonics, and then applied the phase shifts from the Speaker Workshop output which I had dumped to a file. This (above) was the result and it looks terrible - nothing like a square wave.

This, by way of comparison, is the simulated for the first order crossover I designed for the same drivers. While obviously not square, it seems much better than the signal passed by the second order network.

If the rendition of detail has a lot to do with the accurate reproduction of transients, then the first order network should do much better at this than the second order.

My current plan is to build the speakers (when the wretched drivers get here in March or April) and two pairs of cross overs, one first order, the other second order, and see how they sound.

Doh!

You may have noticed that the mid range and the tweeter are too far apart. I completely ignored interference fringing around the crossover frequency. The units need to be separated by less than the wavelength of the crossover frequency - and they aren't. So as you move vertically (perpendicular to the axis), the waves from the two units interfere constructively and then destructively. Doh! This is something I really should have thought about. So I will be building a new baffle with the two units mounted much nearer together. Good job it's only screwed (sic) rather than glued. Oh, and I need to get a new amp or repair my old SUV6 which is cutting in and out on the left channel. A cheap valve amp would be nice but probably out of reach. 

Winding the coils

Since the exact values for the inductors I needed weren't available, I decided to buy some 16 gauge magnet wire and wind my own. After looking at a number of U-tube videos, I built a simple winding jig:  It wasn't quite as easy as the video made it look but eventually I had four coils. One was about 20 turns light because I ran out of wire. I think I was short changed on the spool I bough on Amazon - I won't be  buying from TechFix again. The online calculator seemed way off for two of them. Instead of 2.25 mH, the large coil was 1.6mH; not even close. So it was back to the shed with a hack saw and some nails. The addition of a steel core (should be iron but steel was the best I could improvise) boosted the impedance to 6.8mH. That meant I could unwind more than half the turns on both coils. The two home made cores are on the right - on the 2.25mH coil you can just see the nails under the blue insulating tape.  On the right is the network schematic. 

While I stated with a set of theoretical values, must succumbed to trial-and-error fine tuning in Speaker Workshop as I matched the network to the particular measured characteristics of the mounted drivers.

My wife's comment when I showed her the finished crossover was "it looks like something a 5-year old would have made". I was pleased; that something that looks so rough produces such nice results evoked the same feeling I had when I first visited the 8th floor of the Blackett laboratory. In display cabinets in the elevator lobby were devices that had been instrumental in a number of major experimental discoveries in particle physics. They too looked like "something a 5-year old would have made". It doesn't need to look good to work well, something Prof. Ken Bignell also taught me

Phase IV - measurements

Here are the far field frequency response graphs for my Phase IV prototype and my old B&W DM6s. Both have a trough at 110 Hz which I am assuming is a function of reflection in the room. The Phase IV has a trough at 215 which I think is internal reflection in the cabinet which currently has no damping whatsoever - so I need to buy the SAE 13 felt but at $80 I had been holding off. I  don't imagine that it will make a huge difference when I put it in but I may have to design the crossovers. From 700 Hz and up the response is very flat and looks rather better than the DM6 which has a biggish hole from about 550 to 1.5k. I'm pretty much done with this experiment now and in March will be beginning a new design based on a similar approach to the cabinet but with better quality bass and high frequency drivers.

Three goals were accomplished with Phase IV: the first was to test the general approach to the design of the cabinet; the second was to test the design process from beginning to end, in particular the design of the crossover based on the measured 'in-situ' response of the drivers; the third was to build a speaker comparable in quality to my DM6.