B1 Active Crossover

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Just another Moderator
Joined 2003
Paid Member
Thanks for the link d to the g :) It's funny I was originally thinking I didn't want to do sallen key and would prefer state variable, because I thought it performed better than sallen key (but had decided it required too many opamps) as it turns out sallen key is probably better anyway. I think using a sallen key for the low pass and SY's Gyrator circuit for the high pass looks like a very good alternative, and it doesn't really add much complexity either. I like the fact that it is basically a shunt type arrangement as well. The only thing I'm uncertain of is it looks like SY's one is configured as a third order?? not second... (I'm assuming that it works just like it's passive equivalent, therefore C3 turns it from 2nd to 3rd order, and that C3 could be omitted for a second order electrical slope??).

Tony.
 
I agree that the low pass is far less demanding, unless you're using it to bandpass your midrange. For a dedicated woofer a sallen key seems more than fine.

The gyrator circuit at the bottom of page 2 ends up as a resistor and inductor in parallel, which shunts after the capacitor in the 2nd order filter.

The last circuit diagram on page 3 is a full implementation, and I'm pretty sure C3 is the capacitor for a cascoded filter. Hopefully SY jumps in here eventually, as I'm just a hack.

I had been planning on using on of the SE borbely jfet circuits for discrete simplicity, but if someone did boards up with a B1 based gyrator circuit on it I'd be all over them.
 
Just another Moderator
Joined 2003
Paid Member
Thanks SY :) I'm a bit challenged electrically ;) I think I should invest in some of the books you mention on your site... though the calculus reference scares me a bit ;) The fact that the two caps were both 100nf made me think c3 was part of the filter :)

Tony.
 
Topology... topography is part of a map...

the map is not the terrain, the map is not the terrain. 5 points for whoever knows which [huge hint]De Niro[/huge hint] movie this is from....

I agree with SY about avoiding SK filters, i've not been happy with any i've tried, led me to what i'm using now, PLLXO (buffered or unbuffered)

dave

That's great for a first order electrical... but when you need a second order you still run into the giant inductor issue. Which is why SY's gyrator circuit is just that good. It also fits with the high bandwidth simplicity that the B1 circuit does.
 
Hi Everyone,

I've been thinking a little about a version based on the BF862, mostly SMD although maybe not entirely so. I really haven't got a clear idea yet of what I'd like to build, but I've prepared a schematic so that I can ask a couple of questions. I'm thinking about doing a 24db/octave version, so following the basic ideas mentioned above (and from Rod Elliot;s ESP site) I have repeated the filter "blocks", and have the same output buffer as before, so (unless I have missed something) the circuit is still unity gain. I have drawn it with the same filter point as Jacques used, although I will probably change this if/when I build it.

Here are some questions.

1. The gate stopper resistors are shown as 220R, as Jacques used. Other circuits using the BF862 use 470R (including juma's BF862 preamp), whereas somewhere Scott Wurcer talks about using just a ferrite bead. How does one decide the best value to use?

2. Would the lower BF862 ( the current source) in each buffer also need a gate stopper? Again, juma's preamp has one, and if I have understood other threads about the BF862, it is a little prone to oscillations, so maybe the answer is yes? (These aren't drawn in the schematic)

3. I presume that by repeating the filter to get 24db/octave there is no longer any phase inversion? (Or rather, it has been inverted twice?)

4. If I want gain, say of 2, can I just replace the final output buffer (after the trimpots) by juma's preamp circuit? If I do, this means four copies of it, which is OK, but increases the complexity of things a bit. Is this sonically preferable to using just a stereo pair on the input of the ciruit? (Jacques mentioned a few dsays ago that he thought putting one on the input would increase distortion - would putting them on the output be preferable?)

5. Did I miss anything? (Probably... :confused:)

Thanks in advance for any answers, and for any suggestions you might have.

Cheers

Nigel
 

Attachments

  • LR_BF862.jpg
    LR_BF862.jpg
    60 KB · Views: 1,425
I've made a few of B1-like buffers with BF862 (I like them more than k170).
This is the way it works best for me:
- R3 increases Zout a bit (to about 35 Ohms) but it gives you the possibility to adjust the DC offset in wider range, and eliminates the need to match the JFETs - although they come in pretty tight range of Idss right out of the factory (I bought 500 pcs. and they all have Idss ranging from 12.5mA to 14.5mA with 80% of them being in 13ma to 13.5mA range)
- P1 sets the Id of Q1 (CCS) and consequentially the DC offset on output
- The gate stoppers (R6 and R1) are there to add some resistance to JFET's input capacitance in order to prevent possible oscillations. With bf862 their value is critical only in low-noise applications (MC & MM phono preamps) or in radio circutis. Anything from 100R to 1K will do the job without affecting the bandwith we need (because the input capacitance of BF862 is even lower than one found in k170). I just happen to have a lot of 470R 1/16W resistors so I used them on SMD adapter boards.
 

Attachments

  • bf862-adj-b1.gif
    bf862-adj-b1.gif
    2.9 KB · Views: 1,392
Hi juma,

Your circuit is still unity gain, right? (If I understand your BF862 circuit it's the first jfet that gives the gain of 2, which isn't present here... Sorry if it's a stupid question...)

Also, in this application it won't be very practical to have a trimpot on each of the six buffers, since I can't see how you would do the adjustments. Any reason not to use 10R in place of P1, like R3 and as you do in the BF862 preamp? Maybe keep the trimpots on the last buffers to handle the DC offset, I guess...

Cheers

Nigel
 
Hi juma,
Your circuit is still unity gain, right? (If I understand your BF862 circuit it's the first jfet that gives the gain of 2, which isn't present here... Sorry if it's a stupid question...)
Right on all accounts. The circuit in post #69 of this thread is still a unity gain buffer modified to allow DC offset adjustment.

Also, in this application it won't be very practical to have a trimpot on each of the six buffers...
You put adjustable vesion of the buffer only there where it's needed - on the output of the circuit - and even that is not neccessary if you use coupling caps afterwards.

Any reason not to use 10R in place of P1, like R3 and as you do in the BF862 preamp?
You could do that but it would work only for perfectly matched JFETs (such JFETs would not need any resistors in their Sources to maintain 0V DC offset).
In my preamp I use single polarity PSU for the output buffer and the coupling cap because output point sits on +12V DC.
 
Hi Everyone,

So, if I have understood juma's posts correctly, we would need a schematic something like the one posted below, assuming we want unity gain. I have no particular reason for choosing 220R gate stoppers - I just kept what was in Jacques' schematic.

juma - Did I understand the business about the trimpots correctly?

I'd still like to understand where would be the best point to put a gain stage. If I build this it will be for using with an F5 on the high-pass, so it would be nice to have a gain of 2, so that (as juma notes on his thread) I can use the F5 at full capability, if I wanted to. (On the other hand, I don't have a major problem with volume from the F5 as it is, so this is a little theoretical, perhaps...) Jacques posted above that putting a gain stage before the crossover would increase distortion. Would putting a gain stage in the last buffer (say, as in juma's BF862 preamp) be better in this regard?

Cheers

Nigel
 

Attachments

  • LR_BF862_ver2.jpg
    LR_BF862_ver2.jpg
    69 KB · Views: 1,432
3. I presume that by repeating the filter to get 24db/octave there is no longer any phase inversion? (Or rather, it has been inverted twice?)

That is correct.

Before you finalize your crossover point please have a look at active crossover section of Elliott's and/or Linkwitz's site. Note that the calculation of the crossover point is different for a 4th order LR than for a 2nd order. For the forth order LR the resistance is calculated by: R = 1 / (2 * PI * SQRT(2) * f * C). Also note that for the HP network the value of the resistor which is connected to ground is doubled, and for the LP network the value of the feedback cap is doubled.

Also, there are two buffers missing from the LP section that are necessary to isolate the equalization network of each section from the impedance of the following section.
 

Attachments

  • DCB1XO-LR4b.pdf
    9.1 KB · Views: 719
Hi Jacques,

I'm still quite a long way from finalising the crossover point, but thanks for pointing that out. (It means that the C and R values in my previous post are wrong, of course...)

...Also, there are two buffers missing from the LP section that are necessary to isolate the equalization network of each section from the impedance of the following section.

Just one buffer missing, right? (The last buffer was already present...) Or did I misunderstand something?

Cheers

Nigel
 
So following the help Jacques and juma posted above, here's the latest proposed schematic. It is (I hope) a 24db/octave 2-way filter, followed by juma's BF862 preamp circuit on both the high- and low-pass outputs. I've corrected the capacitor and resistor values, so right now it's set for a crossover point around 500Hz. Here are a list of practical design considerations - I'd be very interested in any comments.

1. I'm planning a single-box preamp, twin volume controls (left and right) no balance controls, toggle switch to choose betwen two inputs. (Like the "official" B1.) The idea is to produce a preamp which is as flexible as I can make it, without things getting too complicated. (I have a couple of nice wirewound 10-turn pots I'm going to use for volume controls, but I'll make the case big enough to include Lightspeeds later, if I feel like it.

2. A gain of two will let me drive the F5 to it's limit. May well not be necessary on the LP output, depending on what amp (and speakers) I use there, but it seems better to keep things as symmetrical as possible, so we'll put it on both.

3. B+ and B- will likely be 9-10V. One PSU should be enough to run all the filter parts in both channels, and a separate pair of PSUs can run the four output buffers. (At the voltages juma is using.)

4. Careful design of the PCBs (or protoboards :D ) should let me do everything in a modular form, so boards can be switched in and out depending on the set-up I want at the moment, using some kind of connectors inside the case. (OK, call me fickle if you want...) So, I imagine an initial stage which is the inputs and volume controls, with the first (common) buffer, then two separate boards which are the HP and LP filters, then a last pair of boards with one "gain" buffer each. These could then be connected to be used either as a normal preamp (gain of two) using just the input and control board, followed by one pair of the output buffers, or as a crossover with gain buffers on all outputs. Careful design using connectors inside the case should make this relatively easy, althoguh doing it switchable from outside the case seems unecessary.

5. Of course, I'd like to keep my options open as to crossover freq, so as massimo suggested above I'm thinking of putting the parts that control the frequency on a daughter board. Does anyone have any experience of using computer motherboard-style connectors for this? I mean like the PCI-express connectors so you could clip in a "frequency board" using pads etched into the PCB?

Well, still planning, but I'd be fascinated to hear an and all comments.

Cheers

Nigel
 

Attachments

  • LR_BF862_ver3.jpg
    LR_BF862_ver3.jpg
    93.7 KB · Views: 1,379
Hi Jacques,
Just one buffer missing, right? (The last buffer was already present...) Or did I misunderstand something?

Nigel,

The second buffer is needed to isolate the network of the second LP section from the 22k pot on the input of the gain block. Without it the 22k pot will upset the correct operation of the network.

5. Of course, I'd like to keep my options open as to crossover freq, so as massimo suggested above I'm thinking of putting the parts that control the frequency on a daughter board. Does anyone have any experience of using computer motherboard-style connectors for this? I mean like the PCI-express connectors so you could clip in a "frequency board" using pads etched into the PCB?

I think Marchand does something like this on their commercial offerings. I rather liked AndrewT's suggestion to use DIP headers and sockets for this.

I also like your idea to include everything on the board initially. That makes it easy to de-populate a section and run jumpers if you decide you don't want or need a particular feature or section in any particular usage.

Cheers,
 
5. Of course, I'd like to keep my options open as to crossover freq, so as massimo suggested above I'm thinking of putting the parts that control the frequency on a daughter board. Does anyone have any experience of using computer motherboard-style connectors for this? I mean like the PCI-express connectors so you could clip in a "frequency board" using pads etched into the PCB?

In my opinion the best way to get this are ordinary IC-sockets. You have two possibilities:

  1. Use 7,5mm capacitors (which is a restriction in Parts selection) and put the whole filter-selection (R and C in the Linwitz-Page) on one socket that is fastened into a soldered socket. Use one standard-value of R and C per module and get the 2R by series-connection of the Rs, the 2C by parallel connection of the Caps.

  2. Use a fixed Cap value and make the Rs interchangable via sockets (IC-sockets e.g.). With a B1 you are in the save region with R form 5k to 20k which leaves you two octaves for the selection of the crossover frequency. You could even get lower freqs with higher R. I am not shure if its wise to go below 5k.

Marchand uses IC-Sockets too. At least in some models.

Just my 2ct.

Floric
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.