Crossover tuning
I had a ready-made crossover design now, and I was all ready to see the finished speaker singing for me in a week's time. I had thought that's the amount of time it would take to get parts from Corrson -- I'd dealt with them before, they are prompt and professional.
I sent them the list of parts. They said they didn't have some of the resistor values I wanted, so I decided to go with local Lamington Road white-coffin resistors instead. They also said they couldn't make microHenry coils -- I'd asked for a 220uH. So I went looking for someone in Lamington Road who could make a coil for me. I just didn't have the enthusiasm to wind my own coil if I could pay someone to make it for me.
So, bidding goodbye to my dream of seeing the Asawari sing in a week, I went trudging through the crowd and noise of Lamington Road, to this firm on the second floor of one of those ancient wooden-staircase buildings. This firm was Disco Winding Works; they make transformers to order, besides other things. I tried telling them about making a pair of coils for me. They had all the enamelled copper cable in various gauges, but didn't have suitable jigs to set up a bobbin and make a coil. (When all you Westerners blithely order coils from Parts Express or Madisound, I want you to remember this story. :) ) So I persuaded them to think of some way out, and the owner said he probably could rustle up something for me. Would a plastic bobbin do? He showed me the bobbin: it was two inches wide along its axis, made of white plastic, and had the words "Johnson & Johnson" clearly marked on it -- it had held bandages from the medical company at some point. I said it would do. The owner then asked me, very politely, could I tell him how many turns this 220uF would take? I went home, went online, used some coil calculators, calculated the number of turns for a 2" coil, and emailed it to him. Four days later, the coils were ready. They cost me Rs.700 (about USD 15.00). They were made out of 16 SWG copper, which is like 14 AWG for Americans. Angshu says I am fool to have got coils made by someone at this price -- he is probably right. I was simply tired.
Angshu suggested that I check whether Corrson can make 0.22mH coils; I shouldn't mention micro-Henry. I found the idea ridiculous -- does this mean Corrson does not know the relationship between milli and micro? Well, I did ask. And guess what? Corrson said "Sure, we can make 0.22mH coils." I felt like such a fool for not having checked with them before. And Gooroo Angshoo had won once again. :)
The rest of the components (large metallised polypropylene capacitors mostly) arrived from Corrson bang on time, and I fished out an unused pair of 1.5mH coils I had lying around. 10 W resistors of the requisite value came from Visha Electronics on Lamington Road. I put the crossover together, cursing the soldering headache of the huge, fat copper junction points where half a dozen 10 AWG wires met at the negative points of the crossovers. These are the points where one wire goes to the negative terminal of the driver, one to the negative terminal of the speaker input, and one each of the terminals of assorted coils, resistors and what have you. They can get as fat as an average-sized thumb. I hadn't seen Troels Gravesen's site by that time, or else I'd have picked up his idea of using lovely solder tags for crossover soldering.
Finally, the crossovers were ready, one Sunday morning. I hooked up my amp, turned the volume low, and played something, checking for screeches, smoke or anything else. Nothing. The sound was "normal". It was working. I spent the whole of that day just playing CD after CD, at fairly high volumes. There was no instability, crackling, or strange noise anywhere. The speakers were behaving themselves, on the whole. The enclosure was as dead as I had hoped it would be. My gamble of plywood over MDF had paid off well.
But over the next few days, I began to feel that the sound was way too sharp. Angshu came over on one of his visits from Delhi, and heard the speakers. He admired the appearance and was impressed by the deadness of the enclosures, but he too finally shook his head about the brightness and harshness. Everything was sounding way too analytical, cold, and bright. When I say everything, I really mean everything, including voices of Harry Bellafonte or Mark Knopfler, not just guitars and cymbals.
Angshu was worried. He told me to take SPL readings as soon as I could. He said that if I could find any anomaly in the SPL curve, it could be fixed by either debugging the crossover or tweaking it. But if the SPL curve showed up a straight, predictable curve, and the harshness persisted, I might be in deep trouble, because there would be no obvious problem to fix. He also wondered whether this harshness was because of my using plywood instead of MDF -- maybe there were cabinet vibrations at frequencies we couldn't detect with our finger tips.
I took SPL readings. Nothing wrong; it matched the modelled SPL curve within half a dB. We were at a loss.
Totally clueless, I decided to try some messing around instead of twiddling my thumbs. In the crossover schematic, there was a 15 Ohm shunt resistor. I wondered what would happen if I reduced that to a lower value, so that the tweeter level would be cut? I modelled the change in SW, and I could see how the band from 3KHz and beyond would be pulled down by a dB or two, without affecting the phase coherence at Fc. So I decided to add a 33 Ohm resistor in parallel with the 15 Ohm. It was magic: the sound cleaned out and vocals became very listenable. I knew I was onto something.
I replaced the 33 Ohm resistor with a 22 Ohm. The sharpness became even more subdued as expected, but it began to appear a bit dull now -- I didn't like it. So I now had a range to play in. I pulled out the 22 Ohm and put back the 33 Ohm. Back to the old sparkling but listenable sound. I left the system this way for a week, playing music for an hour or two, ranging from Pink Floyd to Dire Straits to Diana Krall to Patricia Barber to B B King to Chet Atkins to Rashid Khan to Shahid Parvez to Kishore Kumar. I just soaked in the sound. More on the sound in the next section -- let me complete the crossover story first. :)
After a week, I decided that some bits of some songs were still sounding really bright. The James Bond 007 album had two lovely songs by Shirley Bassey: they sounded too bright. In "Private Investigations" from "Love over Gold", there were a couple of points where the guitar and drums broke out into a crescendo, with very sharp transients. Those passages seemed over-sharp, almost as if glass was shattering. I decided that a good speaker should keep these bits within the limits of listenability. I decided to add a 150 Ohm resistor in parallel with the 33 Ohm which was in parallel with the original 15 Ohm.
It helped. I continued to listen for a couple of weeks more, and I was happy. Even voices seemed to become just that bit smoother and cleaner. My friends, who had wanted to buy the Asawari from me, came and listened for a couple of hours. They were thrilled. My wife gave the sound her stamp of approval. And that's where the crossover has remained till now, and probably where it'll remain till the Asawari moves to her new home in Goregaon East, where my friends live.
I took measurements of the speakers, on axis, first with just the original 15 Ohm, then with the 33 Ohm added, and then with the 22 Ohm in place of the 33 Ohm:
The graphs clearly showed the effects of the resistors. But some of my speaker-building friends were skeptical about the effects of such small tweaks: each step was a drop of just 0.5dB to 1.0dB in the tweeter level. One friend in fact flatly refused to believe anyone could hear such small differences -- he stated, "You can't hear differences less than 3dB, it's been scientifically proven." I didn't want to argue with him, but I could see how that 3dB figure from those lab tests was out of context here. If you generate, say, a 400Hz test tone in a lab environment, and ask listeners to identify a change in volume, they will probably notice volume changes only in steps of about 3dB. But if you change the level of one part of the spectrum with respect to another part, and play full-spectrum music on the system, then this relative tilt in the spectral balance is detected as a change in the tonal characteristics of the sound, not a change in volume. And such tonal changes can be detected very easily even with a 1dB tilt -- you don't have to go up to 3dB for this. I didn't get into an argument on this issue, but I retained my shunt resistors. :)
So now I know that tweeter level tweaking alone can change the sound of a speaker from intolerably bright to lovely, in some cases. My original crossover schematic had a 15 Ohm resistor across the tweeter; my fine-tuned one has something like 10 Ohms there. If I were to build from scratch again today, this is what I'd start with. I wrote to Roman about this -- it's his crossover design after all. He said my observations made sense, though he was a bit skeptical about the audibility of that last 150 Ohm tweaking I'd done.
I took measurements of the speakers, on axis after the tweaks. The two graphs are for the two enclosures, and the two lines in each graph are the modelled and measured SPL.
Previous: Crossover | Next: From passive to active |