DiY Discrete,Direct Coupled Crossover

I remember a tutorial Nelson Pass posted--one of many--and in this particular article on DIY op-amps (look for it), he went on to describe a simple circuit that consisted of little more than a differential amplifier. One could add another transistor to the differential's output to get more gain, or additional current, and there were the four basic types of amplifying devices to tinker with: bipolar, jfet, mosfet, and tubes--and, hey, he threw in a section on constant current sources as well. Yes, this article transformed my thinking. I mean, after all, there are many circuits that require op-amps, so why not build your own DIY op-amps, right? And if you can build your own op-amp, why not build something bigger--like an electronic crossover?!?

What a wonderful age we live in!

With all that being said, this is an ongoing saga of my quest to develop a basic, easy to fabricate, discrete electronic crossover with no interstage coupling capacitors (unless they're part of the crossover's filter network).  Nothing on this page is gospel (but I wouldn't post an untested idea), and I'm always open to fair minded criticism, thus will make corrections as the situation demands. 


As you will see, when i get the itch to do something, I just get the parts, toss them in the cauldron and stir like crazy! I find that if you spend too much time contemplating, you may run out of passion and never get the first breadboard stuffed. Things won't always be perfect, and you may never get it perfect, but if you don't try, YOU'LL NEVER KNOW.


?I begin with my first building block: the 135 hz low pass portion of my crossover:?


Want it larger? Click on the picture.

This extremely simple circuit is all differential, except for the lone output buffer on the right. Beginning from the left, the first of three differentials is based on the J310 jfet. It's extremely easy to work with, affordable, and plentiful--or at least it was, I see that Mouser only has 33,000 OnSemi devices in stock and that it's an end-of-life item. If you hurry, you can still get 100 pieces for $15.00.

FYI: HEY, don't fret if they sell out. There's plenty of modern jfets that will work in this circuit. Mouser has a truckload of PF5102 jfets in stock, and they're brutes that sell at $13.40 per hundred. They seem perfect as a constant current source. For the first differential pair, their Ciss is a bit higher than the j310, but still less than the ZVN-3310 mosfets (which make up the next two differential stages), so again, don't fret.

Moving on:

The Constant Current Source is made from the same j310 transistor. Truth be told, I used this transistor for all the Constant Current Sources in this circuit. In the schematic, you'll notice that I didn't add a cap/ground resistor to the input--it is advisable to add them when coupling to amplifiers that don't have output caps and/or have DC at the output.

The second and third differentials are based on the ZVN-3310 mosfet utilizing a fair amount of source degeneration (feedback). The beauty of this arrangement is that I can use the x-over filter from the prior stage to bias the mosfet stages, this way they're direct coupled. In fact, the only coupling cap is the 10uF fellow trailing the output buffer (which is a source follower ZVN-3310 direct coupled to the third differential). 

FYI: The source degeneration feedback resistor values can be altered to adjust overall gain, so if you want more gain, lower the values of the resistors. The simulator gives some pretty crazy predictions if you lower the values too much--as in the waveform distortion is ridiculous--so keep an eye on that oscilloscope.

It's worth mentioning that Rod Elliot (ESP) has a sweet little piece of software that does these calculations for you. Go to his website (http://sound.westhost.com) and grab this tool if you're interested in this kind of stuff. You can locate the program at the very bottom of this page:
 http://sound.westhost.com/project09.htm#download_esp_lr 

While you're at Rod's site, check out his amazing tutorials and projects. I've purchased pcbs from Rod and they're top notch. 

 

Above: Pretty simple. The top row of transistors are for signal processing. The lower row are for the CCS supplies to the differentials. The caps/resistors in the near field are part of the x-over filter--easier to swap out filter parts that way.

 

Above: Here's a screen capture from TrueRTA. Running Pink Noise, I'm measuring a little six inch Tang Band W6-1721 screwed to a small Parts Express double wall shipping container. The sonic signature is extremely smooth, nothing harsh, more than I expected really. There's low end roll off, something I expect to see with a 40 Hz speaker in a sealed enclosure.

Below: I didn't have a proper speaker/enclosure, so this was my attempt at a quick fix. Works pretty well, all things considered. I certainly can see positive results from my experiment. My little crossover paired to one channel of a 40 watt Phase Linear DRS 400 drove the driver to its Xmax, and then some.... I must be more careful. I must also build a better (plywood) box! ;^)

Moving on to the second building block: the 135 Hz to 500 Hz band pass portion of my crossover:

Having succeeded with the low frequency section of the crossover, I'm moving on to the band pass portion. I'm beginning with a 12db slope. What's interesting, to me, is that I can use the same number of components to create the band pass section as I would in the low pass. Simplicity is what I'm after and a simple circuit interests me. I prefer the sonic signature, the heightened details. Of course, it could all be in my mind.... ;^)

A day later, and I've set up the breadboard and run Pink Noise tests. I'm not too thrilled with the high end's roll off, although it may sound just fine.

There's no point running the circuit full tilt, so I'm using the CD player's headphone output with volume control. Once in the system, the pre-amp will control the crossover's input. 

Click on the picture to enlarge.

Below: with the 12 db band pass filters, there's a lot of information being passed above 500 hz in the Pink Noise measurement. I'll experiment with a 24 db low pass filter and see if there's much of an improvement.

Below: I coupled another low pass filter stage to the first, making it 24 db. This necessitated a reconfiguration in filtering resistors. 12db/octave filters do not use the same resistor value as a 24db/octave filter.

Click on the schematic to enlarge.

Below: I tacked another low pass stage to the end of the bread board--no desire to undo what's already resting comfortably on the board--and joined it with jumpers to the first low pass stage. The results are rather impressive.

There's a real, measurable decline in treble information. I listen to music and hear, unlike the 12 db/octave version, the absence of high frequency information. I'm extremely pleased.

Moving on to the third building block: 500 Hz high pass portion of my crossover:

Finally. The last section of my crossover awaits, and I'm eager to apply what I've learned from the first two experiments. You'll notice that I've learned that neatness is important, so important that I actually did a clean job of bread boarding this phase. By cutting off the longer resistor leads it's easier to trace out the layout against the schematic. If you wear high power glasses to see up close, like me, then you know that depth perception is nil and it's next to impossible to follow the circuity. Plus, being neat makes for a pretty picture ;^)


Here's the calculator's info:

Here's the schematic. Click to enlarge.

To be frank, I'm stunned that my little six inch woofer can reproduce music up to 10,000 hz. As you can see, the high pass circuitry works beautifully. 

This process has been extremely satisfying, and I'm rather pleased with myself for stretching my boundaries. I had plenty of mistakes to contend with (like forgotten resistors, jumper wires, or circuits that didn't perform to expectation), but I didn't blow anything up--ok, I did cook one resistor in the first experiment ;^)

The treble section has the same smooth, dynamic characteristics as the bass and mid-bass portions of this crossover design. Perhaps it's because I have a buffer on the tail end of the circuits, all I know is that the music sounds rather fantastic.

The next phase will be to design the pcbs, fabricate the chassis, lock down decent capacitors--boy is that going to be interesting. It takes thirty six 0.1uF capacitors to build this stereo, three way crossover. I see capacitor prices range anywhere from $1.69 to $350.00. I think I'll be hanging out in the sub $5.00 department. Another disappointment is the discovery that my favorite workhorse j310 jfet is no longer in production--at least I haven't seen them on Mouser or the like. Thank god I have a bag of five hundred to tide me over.

Oh, there's one more interesting tidbit: none of the transistors in this experiment have been matched. Beings that I'm running differential pairs with high source degeneration none of them are going to be current hogs. To be quite honest, everything sounds amazing. I'll match the future generations and see if there's a sonic improvement and report back.

With a bit of luck and a willing attitude, I managed to secure two lots of twenty four KSG-2 Soviet 0.1uF capacitors. These are absolutely HUGE considering they're only 0.1uF.

I've read many good reviews regarding these caps, and judging by their scarcity, they must be popular. Let's face it, a decent sounding crossover is only as good as the parts inside, and these caps seem to be just what the doctor ordered. 


 

Below: I decided to dust off "Dip Trace" (software to aid in the designing of pcbs), and lay out the high pass circuit and circuit board--the low pass version come next. This is a rough board, primarily intended as proof of concept. I'll order ten pieces each to begin with, and if it sounds as good as the  bread board phase, I'll clean the artwork up a bit and order a bunch more. I'll make them available to whoever wishes to build their own electronic crossover.

I ordered my pcbs from an overseas outfit going by the Internet name: 3pcb. As  you can observe in the graphic, below, you're allowed to follow the manufacturing progress of your boards. The entire manufacturing process should take less than a week--of course there's shipping and the delay incurred, but I think it's a fair trade for quality boards at a fraction of my local prices.

Their prices are extremely competitive and they communicate effectively. They let you know immediately what your upfront costs will be. I supplied the Gerber files (exported from DipTrace) which they examined and approved. All in all, it's been a very satisfactory business transaction.

Karen Lin is the representative with whom I'm transacting business; she seems professional, service oriented, and writes in excellent English.

So far, they have my recommendation. The boards aren't in my hands, yet, but I have little doubt I'll have anything to complain about. I'll post pictures when they arrive.


(click on the picture to enlarge)

Here's a close-up of the pcbs (beautiful):

I matched fifty ZVN-3310 mosfets (foreground) and fifty J310 jfets (background). The entire lot is not a match, rather, the idea is to get enough matched double and triples to cover all the various board placements: three jfets at each board's first differential (includes matched CCS) , second and third differentials require matched double jfets for CCS under matched mosfet differential. FYI, the mosfets were easier to match than the jfets.