B1 Active Crossover

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Since it is basically a B1, I'm sure it can drive a F5, but you may want some gain.
I have a crazy sandwich mother daughter board system in mind, but it might get too complicated. :D

But this is not my design, not my thread or anything, so I'll leave it to some others to take those decisions.
 
I have a crazy sandwich mother daughter board system in mind, but it might get too complicated. :D

I am putting one of these together for use with an F5, and while I almost certainly won't buy a board I'd like to make a couple of suggestions, if I may.

Why not extend your "mother daughter board" idea to a fully modular approach? I am using small PCBs which slot vertically into a larger board which in my case is p2p, but in yours could be a "mother" pcb. If you follow a modular design, builders can choose from a list of options; whether to use SMD or through the hole devices, final output buffers with gain or without, and how many stages to use, to give 12db/octave or 24, and so forth.

If you'd like to see photos of my approach they are on http://www.diyaudio.com/forums/pass-labs/146310-bf862-preamp-10.html, but I'm sure you'll be able to do better than that. (And avoid the stupid mistakes...)

Cheers

Nigel
 
I have a crazy sandwich mother daughter board system in mind, but it might get too complicated. :D
Me too, with the possibility of using input/output buffers capable of both SE and balanced input/output.

My thoughts include one mother board for each output. I'm looking to make both a 2-way and a 4-way.

I have some conceptual work done for this, but not with this actual crossover.
 
Hi Guys,

I now have a working BF862 preamp, with four output buffers as shown in post #82 here. (At least, two are working - the others probably have something trivial wrong... There are details and photos at http://www.diyaudio.com/forums/pass-labs/146310-bf862-preamp.html if you'd like to hear more.) Time has come to plan practical details about the filter section. Following the experience with plugin boards on these buffers, the evident shortage of space, and juma's comments on the other thread about the size of the boards, I am thinking of making the buffers as small boards to plug into DIP-8 sockets. This isn't a new idea, of course, there's a fascinating thread at http://www.diyaudio.com/forums/pass-labs/120445-pass-discrete-opamp-dip-8-package.html which describes something physically similar, although electrically different. The same idea could be applied to the LP section, using DIP-16 sockets to allow building the buffers as pairs, which is pretty natural given the schematic. My concern is heat; juma will doubtless say I am worried about nothing (again), but there really isn't much space on the top of a DIP-8, and I'm not sure if I fancy the idea of fixing a heatsink on top like EUVL did for his opamp. (I feel it shouldn't be necessary in this case, anyhow...)

If this idea works it might be helpful to cviller with the PCB idea... Use DIP-8 sockets and either plug in a small SMD board, or plug in the 2SK170s directly...

I think I might do a test run with a pair of buffers and wire up a passive buffer with some sort of symmetrical PSU and see how it works, before doing 14 of them...

Anyone have thoughts on this?

Cheers

Nigel
 
Nigel,
Your construction sure looks nice. Hope it all works. Looking at your construction and the comments by cviller, a mother/daughter board setup will be good. The mother board can route output of one as input for the next one. One can always cut traces and jumper it if output from one has to be routed to 2 or 3 stages.
 
No problem. I found it when looking for a good schematic program. For layout I usually use Layout50 and wish I could link you but the link is so buried in google I cant find it. Illegal copies are abundant but at $40 its pretty nice to be legal. It generates useful gerbers and is super easy to use. Usually have to build your own components if its anything other than the norm, but thats easy to.
Uriah
 
diyAudio Editor
Joined 2001
Paid Member
Somewhere in and about the MOX crossover threads there is a design to make discrete opamps on a pcb board that would plug into a socket. I actually got a PCB boards from the buy- all discrete components, based on Pass opamps, that plug into a socket. If we can find it we could avoid re-inventing the wheel..

I think that Jens Rasmussen was involved.

Mark


Hi Guys,

I am thinking of making the buffers as small boards to plug into DIP-8 sockets. This isn't a new idea, of course, there's a fascinating thread at http://www.diyaudio.com/forums/pass-labs/120445-pass-discrete-opamp-dip-8-package.html which describes something physically similar, although electrically different. The same idea could be applied to the LP section, using DIP-16 sockets to allow building the buffers as pairs, which is pretty natural given the schematic.
If this idea works it might be helpful to cviller with the PCB idea... Use DIP-8 sockets and either plug in a small SMD board, or plug in the 2SK170s directly...

I think I might do a test run with a pair of buffers and wire up a passive buffer with some sort of symmetrical PSU and see how it works, before doing 14 of them...

Anyone have thoughts on this?

Cheers

Nigel
 
Well, this thread is about using the B1 as unity gain buffer instead of a complete opamp, but you are right that is also a viable path...

dviswa, the reason I haven't been saying anything is because I'm thinking about the number of active stages we need for the low pass. It would be really great if we didn't have to have twice as many buffers, but I can see that the filter is messed up when it is loaded...

I can't really figure out how it was done in the XVR1 (I don't have the schematics, only some high res pictures), it looks like there the signal travels single ended through the filters and there is a buffer for each 2 poles. But there is no double up on the low pass compared to the high pass (at least as far as I can see).
 
Somewhere in and about the MOX crossover threads there is a design to make discrete opamps on a pcb board that would plug into a socket. I actually got a PCB boards from the buy- all discrete components, based on Pass opamps, that plug into a socket. If we can find it we could avoid re-inventing the wheel..

I think that Jens Rasmussen was involved.

Mark

Is that maybe the same thread I referred to? If not it is simlar in principle.

I've spent a very fun afternoon and evening burning PCBs and soldering up a couple of buffers to slot in a DIP-8 socket, and making a test board to see how they perform. (I mentioned above I was thinking of doing this.) If all goes well then I plan on t\doing all the buffers in the filtere this way. (DIP-8 for HP, DIP-16 for LP). I failed to get a good photo of the little boards themselves; the first photo posted is the best that I could manage with a digital camera and a maginfiying glass... The second shows a mostly complete preamp board I put together to test the SMD thingies. I ran out of time and energy before finishing, but hopefully tomorrow I'll be able to tell you all how it sounds.

The little DIP-8 boards are each one buffer as in the high-pass part of the circuit posted above in post #82, except that the resistor to ground on the input is 330k instead of 20.4; there is no filtering necessary in this test case. (I would have used the 1M that juma uses in his circuit, except I don't have any). If this idea is to be used in the crossover then this resistor won't be on the board anyway; it'll be with the capacitors in the switchable DIP-** that AndrewT proposed. I found it all less technically difficult than I expected - the boards were just drawn by hand, and there is plenty of space for two jfets and three resistors.

The preamp board is really intended as a test device, not a version to be used long-term. I deliberately omitted any trimpot to control DC offset, since if these are to be used in the crossover there won't be any trimpot except at the end. Hence the caps on the output; I reckon these should be sufficient to listen to the thing for long enough to check things out.

I am really hoping to find out two things. The first is how much heat is generated, and whether any heatsinking is necessary. (juma was certainly right that my output buffers are on boards uncessarily big). The second is to experiment with power supply variations; as I mentioned above I am uncertain how to deal with the current draw that 14 of these buffers will need, if I use (some version of) Salas' regulator. So I'll try first just with 7809/7909 and decoupling right next to the SMD "IC", and try more complicated solutions after.

I don't have any sophisticated equipment to measure things, but I'd be grateful for any suggestions for things that could be added to the experiment.

Cheers

Nigel
 

Attachments

  • IMG_5251.small.jpg
    IMG_5251.small.jpg
    246.2 KB · Views: 1,056
  • IMG_5257.small.jpg
    IMG_5257.small.jpg
    414.6 KB · Views: 1,034
Hi Mark, Nigel and all,

One of the major problems with the all purpose designs is that the pcbs just grow in size and complexity and the tracks just get longer and effect the signal to the point that the whole potential of the xovers are severely compromised - it happened with the MOX with all those resistors, links and things.
The discrete IC worked really well, but the original big board setup was intended as a testbed for experimenting and not the final design - Marchant's moster is another disaster if you're talking about lost quality - they actually measure well, until you start looking at signal compression and "stuff" that is most unwelcome.

So, I wonder if we can approach this design with 2 boards in mind - a simple no frills 12dB with optional Salen Key &/or Passive filters with short tracks, buffers built into the power track layout (especially the earthing) and avoid all the plug in connections/ mother board disasters, despite it's attraction. (These things are FILTERS, not multiple choice exercises)

The other one could be a complete variation optional masterpeice, if required, but can simply go back to the MOX optional layout scheme and maybe not even require different components - you may need to design another pcb for the buffer stage, maybe add some rail bipasses, etc, etc but the mother board had every conceivable option required, including a large track count!

It would still be better to have the buffers built into any mother board and optomise the power tracks, earth retuns, signal earthing (not the same) etc
and have the filter components off plugin connectors (still pretty yuck!)

At least it would be a good universal testing board for Xover experiments.

... my 2 cents.
 
dviswa, the reason I haven't been saying anything is because I'm thinking about the number of active stages we need for the low pass. It would be really great if we didn't have to have twice as many buffers, but I can see that the filter is messed up when it is loaded...

I can't really figure out how it was done in the XVR1 (I don't have the schematics, only some high res pictures), it looks like there the signal travels single ended through the filters and there is a buffer for each 2 poles. But there is no double up on the low pass compared to the high pass (at least as far as I can see).

cviller,

Can you please explain what you meant. My EE skills are not that high. How is a low pass sallen-key filter any different from a high pass sallen-key filter, except for the fact that R and C are interchanged. Why would low pass need twice as many buffers?

If you are concerned about the number of stages in general, as I see it, one might need a minimum of 2 stages or as much as 5 stages. 4 stages if you need if you need Volume control, Cross over and "LPAD". LPAD in quotes, I mean conceptually, if you need to adjust for difference of efficiency among the drivers.

Best case:
Buf->Filter

Worst Case:
Buf->Attenuator->Buf->Filter->LPAD.

The Filter can be 2 stages for a band pass leading to 5 stages. If you choose the drivers right, you can avoid the LPAD stage. If you do not need the Volume control, you can save one stage.

I am only counting how long the chain is and not how wide.

Regards,
Dinesh
 
Hi Guys,

I have the test watchamacallit up and running; as described above, powered by a regulated +/-15V power supply that I had lying around feeding a 7809/7909 pair. For the record, the circuit of the buffer-IC being tested is posted below - I've spared you the caps on the output and so forth since these aren't really relevant. (Observant readers will note that it's really juma's circuit attachment from above, altered in gimp...)

One of the questions I wanted answered already is. I'm listening to music through the buffer now, and it doesn't get hot enough to worry about; I can hold my finger gently on the IC for well over 10 secs - generally I take it off because of crackling in the speakers, not because it's hurting too much. The little PCB is too small for me to try and measure temperature more accurately with a probe (or at least, with the probe I have....) so that will have to do. Once in a closed box the temp may be a little higher, but overall I think it's OK.

Next up is a list of experiments with regulator options...

Dinesh: the doubling up of buffers in the LP was discussed above; if you take a look at the schematic in post #82, which is what I'm planning to build, you can see the buffer count in the LP is higher. Explanations of why this is necessary have been posted before #82 by people who understand all this much better than I do.

On a related note, I'm still trying to absorb jameshillj's comments above. For my own needs I'm not concerned about the final PCB design, since I almost certainly won't be buying or using one. However the issues raised (longer signal paths and so forth) will definitely affect my p2p version. I'm hoping that these DIP-8 (and DIP-16) ICs will help in this regard, but I'd be most interested to hear more details about "signal compression and "stuff" ".

Meanwhile, the test buffer sounds surprisingly good - I expected the 7809/7909 to give a sub-par performance but ths sound is still pretty good. Here's hoping a discrete reg gives better!

Cheers

Nigel
 

Attachments

  • BufferTestCircuit.gif
    BufferTestCircuit.gif
    1.9 KB · Views: 966
Nigel, cviller,

I went back and reread bits of this thread and the article that Jacques Merde had linked. The graph in the article showed the frequency response rising after about 2 decades past the cut-off and from about 60db down. I do not have scope to measure it. If treble is being fed into the woofer, I am not able to hear it. What is the woofer's response to this treble is also a good question.

For now, if we can get a building block that is sufficiently cheap, small, and easy to handle, we can add the additoinal buffer either behind the LP filter or within the feedback loop as shown in the article.

Regarding what Jameshillj, said, yes, it is a concern, careful PCB design can mitigate most of it. The basic building block will contains 2 FETs, 2 bypass caps, 2 protection resistors and one bias resistor. The filters will replace 2 resistors while adding 4 new components. I think we should still try and design a single board with 2 blocks which can be used either as filters or buffers. My wishlist includes enough space for the 4 filter components so that we can double up on components instead of using components of different values. For instance, in my high pass I am using 10K, 33nf, 10K+10K and 33nf. I do the same for LP, parallel the 10K resistors. Perhaps this is where what jameshill says might come to pass, but we should still try.
 
As you see on Juma's BF862 buffer, the o/p cap(s) have all got a DC bias applied to them (16v and about 5v) - now, on the Sallen Key filters, the caps don't - it's the way they work.
There's usually much confusion in selecting the type of filter cap as they all have different sound characteristics when passing AC signal with no (or little) DC bias, and the "better" ones are often physically larger.

Now, I'm NOT saying that bigger caps give better sound by any means, but suggest you keep it in mind (with all the other things!) when choosing the pcb layout &/or plug-in systems - known as a "design protocol".

... and I'd avoid the "cheap" thing - go for the best quality you can do and trim it back later - too many cheap things already.

.... up to 4 cents now!
 
.... up to 4 cents now!

That's OK! Keep it coming, can be 40 cents, 4 dollars or more... Interesting stuff :)

Actually I'm still trying to absorb your earlier comments, while experimenting with the test buffers. (Salas-style power supply is next up...) Let me repeat, I will not be buying or using a pcb, but your remarks help with designing my version. Presently I have in mind a p2p main board with the buffers all done in these plug-in DIP-8 and DIP-16 boards. They have no caps on them; all the filter caps and resistors will be on other DIP sockets. I hope doing it this way I can avoid the worst of the problems you indicate, since I don't have to do a "one-size-fits all" main board, but still have the flexibility to change crossover point if the rest of the system changes at a later date.

Nonetheless I am starting to feel that the 24db/octave version may be impractical. May be necessary to scale back to 12db/octave as the OP proposed. (Speaking of the OP: Jacques, have you lost interest in this? Or are you way ahead of us?)

Cheers

Nigel
 
Sanity check...

Ok, I have done a quick sketch and I would like to hear what you guys think - is this completely insane or something that could end up being really cool?

There is a volume control for each channel - I would probably use double pots for stereo.

The filters are going to have dip switches so that the cross frequency can be set without too much trouble.

The headers J1-J6 are intended to be expansion slots where additional filters can be added (to get more than 12db slopes). I thought I would use jumpers to short them out when not in use.

As you can probably see, I have been very much inspired by B4 and XVR1... ;)
 

Attachments

  • xo_mboard.pdf
    19 KB · Views: 493
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.