I have small circuit or "sub-circuit" that will be repeated several to many times as part of a larger analog active circuit that is to be implemented on a PCB. I would like to make something versatile and flexible because the overall circuit might use four or fourteen or whatever number of the sub-circuits plus a few other auxiliary signal processing circuits. I am thinking of perhaps a physical arrangement consisting of a motherboard plus edge connected daugtherboards. Or making the sub-circuits on small PCBs that can be stacked up on top of each other above places on the motherboard PCB. Once the PCB is built it would not need to be reconfigured. I am only interested in making a flexible implementation overall.
Are there any other techniques that might work well for audio?
Are there some problems with or concerns about small PCBs that stand off of the main board, e.g. regarding noise or physical staibility?
Are there any other techniques that might work well for audio?
Are there some problems with or concerns about small PCBs that stand off of the main board, e.g. regarding noise or physical staibility?
I've done something similar in the past.
The most compact is a stack of identical boards but that depends on the possibility that for instance the power supply lines
and/or i/o lines go through each board. Individually connecting i/o to the mother board gets complex this way.
Placing subcircuits vertically on a mother board is often easier for signal routings.
The advantage of vertical on motherboard is that it makes it easier to swap out defect boards without having to take down
the stack, but it probably takes a bit more space. A set of vertical boards probably isn't as physically stable as stack, but can be made good enough by taking a connector that is wide enough even if only a few pins are populated.
What are these boards?
Jan
The most compact is a stack of identical boards but that depends on the possibility that for instance the power supply lines
and/or i/o lines go through each board. Individually connecting i/o to the mother board gets complex this way.
Placing subcircuits vertically on a mother board is often easier for signal routings.
The advantage of vertical on motherboard is that it makes it easier to swap out defect boards without having to take down
the stack, but it probably takes a bit more space. A set of vertical boards probably isn't as physically stable as stack, but can be made good enough by taking a connector that is wide enough even if only a few pins are populated.
What are these boards?
Jan
Hi Jan, thanks for your thoughts. The sub-circuits are state-variable filters. Each will use a quad op-amp and there are about 8 resistors and 2 small capacitors, probably all surface mount components except possibly the op-amp package. The footprint for the sub-circuit will be relatively small and as-populated not very tall. There could be a "row" of boards connected edgewise to a motherboard. But stacking is also possible because there would be a max of four used together so even a 4-stack is probably still under 1U height or close to it.
I am still trying to figure out which configuration would provide the most flexible signal routing on the motherboard. The signal would run through N sub-boards in series. So the signal would run up the stack and then return to the motherboard either back through the stack or via a signal+ground pair of wires. My concern is that the wires dangling "in space" away from the main PCB in the interconnections might be susceptible to noise pickup. This would not be a problem with N edge-connected sub-boards since the signal would either be on the motherboard or on the sub-board PCBs.
Charlie
I am still trying to figure out which configuration would provide the most flexible signal routing on the motherboard. The signal would run through N sub-boards in series. So the signal would run up the stack and then return to the motherboard either back through the stack or via a signal+ground pair of wires. My concern is that the wires dangling "in space" away from the main PCB in the interconnections might be susceptible to noise pickup. This would not be a problem with N edge-connected sub-boards since the signal would either be on the motherboard or on the sub-board PCBs.
Charlie
I captures some nascent ideas on how to connect a series of vertical daughter boards on a motherboard in the attached image. A cartoon of the board is shown at top. Connections are blue for signal IO and green orange red for power supply connections and these are made on opposite sides of the daughter board. On the motherboard, traces run front to back next to each other and cards plug in along the way. When a card is not needed the signal IO connection is simply bridged by a jumper.
Attachments
If the signal goes through all boards in a set a stack is a good possibility.
I inderstand that ultimately the signal has to get back to the motherboiard (the 'dangling' wires), but they would be as long if they were tracks, so that wouldn't be an issue.
And you can always twist a signal wire with a ground wire from the top of the stack back to the motherboard in a quasi-balanced mode.
@rayma's way would also work but probably have longer wires/tracks in case the signal goes through a series of boards.
But in the end, with such a small board, it's probably not making a difference anyway and you can select what is easiest to handle for assembly/replacement.
Jan
I inderstand that ultimately the signal has to get back to the motherboiard (the 'dangling' wires), but they would be as long if they were tracks, so that wouldn't be an issue.
And you can always twist a signal wire with a ground wire from the top of the stack back to the motherboard in a quasi-balanced mode.
@rayma's way would also work but probably have longer wires/tracks in case the signal goes through a series of boards.
But in the end, with such a small board, it's probably not making a difference anyway and you can select what is easiest to handle for assembly/replacement.
Jan
If the board configuration is really only a one-time thing, then I'd have the subcircuits actually part of the main board,
and stuff or jumper them as needed.
Sounds like the boards will be small enough to not justify a complicated physical arrangement with sockets, wires, etc,
and risk noise pickup or contact problems.
If there will be a lot of main boards, it may be simplest to stuff everything, and just jumper the functionality as needed.
and stuff or jumper them as needed.
Sounds like the boards will be small enough to not justify a complicated physical arrangement with sockets, wires, etc,
and risk noise pickup or contact problems.
If there will be a lot of main boards, it may be simplest to stuff everything, and just jumper the functionality as needed.
Last edited:
Just for complete information, I would like the design to feature up to 16 "lanes" (assuming the daughter cards on edge concept) with each lane accommodating up to 4 cards in a "row". So in total that is 64 daughter cards. Each lane is a trace that runs more or less all the way from front to back of the motherboard PCB with 4 spots to either plug in a daughter board or jumper/byapss. Lanes could be connected to the overall input or to other lanes, to allow for various signal routing configurations. Not 100% sure how to lay all of that out, but I think it is doable on a board with a single layer on each side....
Next up, how to distribute V+, V-, and PS GND to all of those cards??? Seems like a lot of PS related tracks running around. Maybe use lots of local decoupling? Two sided PCB with one side having a signal track layer and a signal ground plane, and the other side having PS V+ and V- tracks and PS ground plane?
Next up, how to distribute V+, V-, and PS GND to all of those cards??? Seems like a lot of PS related tracks running around. Maybe use lots of local decoupling? Two sided PCB with one side having a signal track layer and a signal ground plane, and the other side having PS V+ and V- tracks and PS ground plane?
If this is a one time configuration, better to avoid sockets and use soldered pins, similar to these.
https://www.te.com/usa-en/product-2842114-5.html
https://www.te.com/usa-en/product-2842114-5.html
On the other hand... What if I could construct the "daughter cards" to be small enough so that the interfacing could be done via DIP pins and sockets. They are pretty reliable, eh?
I could use SMD components and then it might resemble something like a discrete op-amp. The larger components like caps would be off board. But that would be a way to make the "daughter card" very small indeed.
There is precedent for this: for example TI's UAF42. See circuit diagram, attached. It's a DIP-14 IC.
I could use SMD components and then it might resemble something like a discrete op-amp. The larger components like caps would be off board. But that would be a way to make the "daughter card" very small indeed.
There is precedent for this: for example TI's UAF42. See circuit diagram, attached. It's a DIP-14 IC.
Attachments
Last edited:
Sure, although dual in line DIP sockets have a larger footprint than vertical pcbs. Could use single in line SIP sockets instead.
Soldered connections would still be more reliable, especially with 64 sockets.
So a few years ago I designed a last gasp parametric speaker equalizer before I conceded that it was time for DSP. Mother board and modules all 4 layer PCB. Worked well and met its electrical design objectives, but proved unequal to the task.
Probably as a much a hint about what not to do as anything. I like the idea of plug in modules, but would recommend 2 rows of pins rather than the single row. Gold will be more reliable in environments like mine on the edge of the ocean. Had this continued in use rather than languishing with its companion in a dark corner of my listening room/lab I would have secured each board with a small puddle of hot melt glue. (Sorry for the dust filled pix.)
Soldering is more reliable but less flexible.
The cards are just basic SVF filters with options for LPF. HPF, BPF outputs jumper selectable as is the filter frequency setting which was inconvenient to say the least. (The cards were built to cover specific frequency ranges which could be adjusted over a limited range by moving jumpers.
There are 3 lanes each with two cards, the second and third lanes and one card slot in each of the 3 lanes can be bypassed by jumper setting, allowing for 1 - 6 cards as desired. The lanes are sequential (series connected).
Lots of dust. Mono, balanced I/O.
Probably as a much a hint about what not to do as anything. I like the idea of plug in modules, but would recommend 2 rows of pins rather than the single row. Gold will be more reliable in environments like mine on the edge of the ocean. Had this continued in use rather than languishing with its companion in a dark corner of my listening room/lab I would have secured each board with a small puddle of hot melt glue. (Sorry for the dust filled pix.)
Soldering is more reliable but less flexible.
The cards are just basic SVF filters with options for LPF. HPF, BPF outputs jumper selectable as is the filter frequency setting which was inconvenient to say the least. (The cards were built to cover specific frequency ranges which could be adjusted over a limited range by moving jumpers.
There are 3 lanes each with two cards, the second and third lanes and one card slot in each of the 3 lanes can be bypassed by jumper setting, allowing for 1 - 6 cards as desired. The lanes are sequential (series connected).
Lots of dust. Mono, balanced I/O.
Last edited:
I'd definitely go with connectors over soldered-on daughterboards, that's just setting yourself up for a hard time (*). For reliable connectors carrying signal levels use gold-plated pins and sockets. Vibration is an issue though - this could mean having extra mechanical support.
(*) how do you test daughter boards which don't have a connector? Solder them up, test, unsolder, then re-solder onto a mainboard - ugh!
(*) how do you test daughter boards which don't have a connector? Solder them up, test, unsolder, then re-solder onto a mainboard - ugh!
Some connectors have an extra part on the side with a hole to bolt it to the mainboard similar to the DB25 connector system..
Another option used in many test units with multiple boards on a main board is to place a bracket across the top of all daughter boards to keep them pressed to the mainboard. A couple of standoffs connect the bracket to the mainboard or chassis.
If you want to go fancy you can make a slot in the bracket to hold the daughter board so it is also fixed laterally.
Many audio analyzer use this technique for obvious reasons.
Jan
Another option used in many test units with multiple boards on a main board is to place a bracket across the top of all daughter boards to keep them pressed to the mainboard. A couple of standoffs connect the bracket to the mainboard or chassis.
If you want to go fancy you can make a slot in the bracket to hold the daughter board so it is also fixed laterally.
Many audio analyzer use this technique for obvious reasons.
Jan
This is similar to what I want to do. I like the idea of making the outputs jumper selectable. I will have to think about that.
After sleeping on the idea, I still think that a small board using SMD components and a DIP header would work well. If I use an SMD cap of around 15nF the corner frequency can be adjusted over a range of around 100-4kHz. The user could add two larger caps on the motherboard to access lower frequencies. There are three adjustment to the SVF, each controlled by a resistor: one of these controls the Q and two others set F. These would be on the motherboard, and I could set up the holes to accept a trimpot or fixed resistor. All other components can be on the tiny plug-in daughterboard. Because the daughterboard can be quite small, this will also shrink the potential size of the motherboard. I am a bit concerned about how much the PCB real estate of a large motherboard would cost, and this is something to think about.
Back in 2011 I designed and had fabbed a bunch of PCBs for analog active filters. I was thinking of using them for my own loudspeaker projects and maybe selling some to other DIYers. But that was about the same time that I discovered the miniDSP 2x4 and one I started using it I never really went back to analog. Here I am over 10 years later thinking that it would not be a bad idea to have some real-time zero-latency analog capabilities as well. Where this will go I am not sure, but it might be a good project to develop over the upcoming winter months when loudspaker building is stalled.
There is yet another approach that I have seen used in, for example, an LR4 crossover board by Marchand Electronics. Most of the circuit is on the "main" PCB and only the parts that are variable, e.g. resistors that set frequency, etc., are on a small DIP header. Marchand sold this as the XM-1. I built my first active crossover using several of these boards back in the 1990s. The XM-1 seems to be discontinued or NLA but the manual is still up on the company website:
https://www.marchandelec.com/ftp/xm1man.pdf
The circuit is basically two SVFs in series.
What was great about this setup was that Marchand sold the board and also various pre-configured header boards that set the crossover frequency at e.g. a few different frequencies between 100Hz and 6kHz IIRC. The user/owner could make up new headers as needed by soldering four resistors of the right value on to a forked header. It worked well and was pretty versatile.
https://www.marchandelec.com/ftp/xm1man.pdf
The circuit is basically two SVFs in series.
What was great about this setup was that Marchand sold the board and also various pre-configured header boards that set the crossover frequency at e.g. a few different frequencies between 100Hz and 6kHz IIRC. The user/owner could make up new headers as needed by soldering four resistors of the right value on to a forked header. It worked well and was pretty versatile.
HI Charlie,
These modules are just a quad op-amp and 2nd order SVF, the Q is set by that pot on top and frequency by 50 mil pin headers with jumpers. The rest of the circuitry was on the mother board.
Were I to do it again I would use a double row RA header which would be much more secure than the single row headers I used.
My focus at this point was addressing some room modes, and learning how to calibrate the individual filters and match them channel to channel.
These modules are just a quad op-amp and 2nd order SVF, the Q is set by that pot on top and frequency by 50 mil pin headers with jumpers. The rest of the circuitry was on the mother board.
Were I to do it again I would use a double row RA header which would be much more secure than the single row headers I used.
My focus at this point was addressing some room modes, and learning how to calibrate the individual filters and match them channel to channel.