yet another electromagnetic-harp update!

[oolfur]

hey again,

for reference, see the two older threads ::

so - the design of the new electromagnetic harp has come a long way since my last post - i am almost ready to send PCB drawings to the production plant to finish the first 4 instruments !

one of the new features im super excited about allows you to pass any audio signal through the strings. here is Chet Baker being played through only ONE string! imagine what it will sound like when I hook up the remaining 25 coils…!!

https://soundcloud.com/ulfurh/one-string-chet-baker

anyway - i have a few problems i'm trying to solve during the weekend, and i thought you guys might be of help

so, the first stage of the circuit is a simple guitar preamp / phase flipping circuit that allows me to:

• amplify a signal from a single-string pickup (its around 2.5kohms)
• flip the phase of the signal using a single switch
• bypass the signal from the output, or rather cutting it off using a single switch

the reason why i want to be able to flip the phase is that switching between the two different phases makes the string in question settle on a different harmonic (it's just like the two different "harmonic modes" on commercial ebows...)

here is my design;

(first switch to the right flips the phase, second is the bypass switch)

the circuit works great when i do the switching manually with physical spdt switches, however i need to be able to do the switching electronically - preferably using the quad CMOS 4066 switch and the 4049 HEX inverter (so that by default, the switches are ON).

so my problem is that if i power the 4066 with a single ended supply (let's say, 0 - 15v) it wont pass any negative voltages through the switches, and effectively clips the negative swing of the input signal off. it sounds pretty nasty, and looks like this on the oscilloscope.

appearantly the 4066 can handle dual ended power supplies(let's say -5 to +5), and then it will pass the signal unaltered(no clipping) - but my understanding is that this also means the incoming logic signal has to be pulled to the VSS (-5v) to turn the switches off. so a 0v logic signal will still be seen as "high"... right? i think thats what the datasheet says...

4066 datasheet is here::http://www.ti.com/lit/ds/symlink/cd4066b.pdf

maybe all i need is a capacitor between the signal and CMOS, im not sure... there seems to be a plethora of circuits online that use the 4066 in a single ended configuration to switch audio with no problems at all... maybe i can configure the 4049 inverter to swing the logic signal to negative, even though the input logic still is 0v - 5v?

maybe you have better ideas on how to approach this?

any thoughts on this would be greatly appreciated!

thank you ever so much for your input,

-úlfur

[curtisa]

Regarding the switching there are two ways to deal with this. You essentially want the output wobbling either side of a non-zero reference to lift the negative swings of the signal away from the 0V cutoff point imposed by the CMOS gates:

1. Introduce a DC offset of half your intended signal swing to the signals as they pass through each switching element - can be as simple as a couple of resistors and capacitors per switch.
2. Run your CMOS chips from +/-5V and use a PNP transistor as a level translator to interface your logic 0/5V signal to -5V/+5V - very easy to do, but will require one translator for each switching element.

Both solutions will require some more engineering on your part. CMOS switches aren't very robust and introduce obvious foibles with regards to switching voltages and the types of signals they can carry (as you've now found). Are small relays out of the question? Then you can mix and match switching supplies and audio signals as much as you like without having to worry about the interfacing problems.

There's some useful info regarding using CMOS switches to control signals here.

Your schematic shows the output being switched via a 4066 in series to achieve mute. A slightly more elegant way of achieving mute is to move this switching element to the input side of the opamp and use it to short (shunt) the signal to ground.

Will this opamp being actually driving a speaker as is drawn?

[oolfur]

curtisa,

thank you! that link is very insightful, probably the most helpful reading i have found so far on CMOS audio switching...

it looks like the DC offset method would be easiest at this point, not sure how to implement a level translator that swings to negative... how will the offset affect the audio signal output of the CMOS?

the reason why im not using relays is im worried about audible clicks - since this will be a high end musical instrument... this is only one of 25 channels that need similar switching. maybe there is a super silent relay IC out there somewhere - do you have any ideas?

Will this opamp being actually driving a speaker as is drawn?

nope - from there it goes to a VCA, and then to a power amplifier that drives the actuators. it's gonna be hot & heavy!

thank you so much for your help, now i'll go back to the lab and experiment...!

all the best,

úlfur

[curtisa]

it looks like the DC offset method would be easiest at this point, not sure how to implement a level translator that swings to negative... how will the offset affect the audio signal output of the CMOS?

A level translator needn't be complicated. This is about as simple as it gets:

DC offset won't affect the audio signal as such. As far as the audio signal is concerned it is merely shifted up. Instead of going positive and negative around 0V it will now go "more positive" and "less postive" around a point midway between the two extremes. You will have to remove the offset after you've shifted it before passing the signal on to other processing stages. A series blocking capacitor on the output of the last shifted stage is the simplest way to do it. The value of this cap may have an impact on the signal and is dependent on the impedance of the surrounding circuitry, so leave room for experimentation.

the reason why im not using relays is im worried about audible clicks - since this will be a high end musical instrument... this is only one of 25 channels that need similar switching. maybe there is a super silent relay IC out there somewhere - do you have any ideas?

Dedicated switching ICs exist, but theyre not easy to come by and usually expensive as well. Don't forget that relays are used all the time for channel switching in guitar amps, mixers and the like, where small signals need to be handled gently to avoid excess noise. Simple techniques exist to minimise switching noise from relays and mechanical switches.

[Prostheta]

Same here. I'd be tempted to use a balanced supply and differential buffering/amplification of the signal to reduce noise whilst I was at it.

[oolfur]

it looks like the DC offset method would be easiest at this point, not sure how to implement a level translator that swings to negative... how will the offset affect the audio signal output of the CMOS?

A level translator needn't be complicated. This is about as simple as it gets:

DC offset won't affect the audio signal as such. As far as the audio signal is concerned it is merely shifted up. Instead of going positive and negative around 0V it will now go "more positive" and "less postive" around a point midway between the two extremes. You will have to remove the offset after you've shifted it before passing the signal on to other processing stages. A series blocking capacitor on the output of the last shifted stage is the simplest way to do it. The value of this cap may have an impact on the signal and is dependent on the impedance of the surrounding circuitry, so leave room for experimentation.

the reason why im not using relays is im worried about audible clicks - since this will be a high end musical instrument... this is only one of 25 channels that need similar switching. maybe there is a super silent relay IC out there somewhere - do you have any ideas?

Dedicated switching ICs exist, but theyre not easy to come by and usually expensive as well. Don't forget that relays are used all the time for channel switching in guitar amps, mixers and the like, where small signals need to be handled gently to avoid excess noise. Simple techniques exist to minimise switching noise from relays and mechanical switches.

thank you for this - this transistor circuit makes it a lot easier for me to deal with the 4066 - i have a lot of transistors laying around.

your schematic kinda looks like a transistor "not" gate, - so im wondering.... since im already using a 4049 hex inverter, could i simply supply the inverter with +-5v to get similar results?

your help is greatly appreciated! thank you so much!!

[curtisa]

Yes, I probably should've mentioned that this will invert the output. Logic 1 (+5V) at the input will translate as logic 0 (-5V) at the output, so you can do away with your 4049 inverters.

[oolfur]

Yes, I probably should've mentioned that this will invert the output. Logic 1 (+5V) at the input will translate as logic 0 (-5V) at the output, so you can do away with your 4049 inverters.

hmm... and logic 0 (0v) at the input will translate as logic 1 (+5) also? that would be perfect!

[curtisa]

Correct.

By running opamps and 4066s on +/-5V supplies and substituting one transistor level shifter for every 4049 inverter you should be able to achieve the required switching without having to resort to DC offsets for each 4066 switching element.

Out of curiosity what are you using to control all these switching functions? Some kind of microcontroller?

I forgot to mention earlier, I'm intrigued by your electroharp, I'm liking where it's going and the sounds you're getting from it - very innovative!

[oolfur]

Correct.

By running opamps and 4066s on +/-5V supplies and substituting one transistor level shifter for every 4049 inverter you should be able to achieve the required switching without having to resort to DC offsets for each 4066 switching element.

Out of curiosity what are you using to control all these switching functions? Some kind of microcontroller?

I forgot to mention earlier, I'm intrigued by your electroharp, I'm liking where it's going and the sounds you're getting from it - very innovative!

thank you - that means a lot - i've put a lot of work into it.

to control the switches i am using optocouplers with capacitive touch switch IC's from pcbgadgets.com. the whole electrical interface will be a touch "keyboard" made with copper inlay on the walnut soundboard.

there will be superbright LED's just beneath the surface of the walnut, so each note or setting will be indicated as on/off, with glowing amber light coming through the woodgrain.

today i will hopefully be sending the last protoboard to the PCB lab, and if it works flawlessly, i will draw the 25 channel PCB's by the end of the week. im planning on having the instruments completed this summer.

will keep you posted, and thank you so much for all your help!

-úlfur

[oolfur]

hey again,

i have fixed all switching issues thanks to your help!

now i have run into another problem however, and i thought i'd ask the forum and see if anyone has any experience with this sort of thing.

so - as previously demonstrated - one of the main features of the instrument will allow you to route any type of audio signal into each and every string.

now - for the signal output, i need to sum together all of the 26 strings to a single output.

before i buy opamps in bulk - id like to ask the forum about noise accumulation and gainstaging. can i safely sum together 26 pickup coils using TL07x's like this?

one opamp on each and every coil operating at unity gain, then all fed into a single unity gain TL071 through 10k resistors... does that make sense?

im also wondering if i should sum the coil signals before or after the pre-amp stages...

any thoughts would be greatly appreciated!!

thanks again,

úlfur

[curtisa]

Have a look at this article, particularly section 3. What you want to build is a mixer. With your diagram as shown you are mixing the pickup signals, but doing so passively. The main drawback with your scheme is that each pickup signal is in parallel with 25 other signals, loading each other down, and with 26 stages each pickup will only be capable of delivering 1/26th of their full output at the opamp, requiring 26x more gain at the opamp to make up for the loss.

The example at section 3 is a better solution, where the mixing occurs at the inverting input of the opamp. You can still use your 10K resistors, but the adjacent-signal loading that would occur with passive summation is non-existant, and each input maintains its level through the opamp.

You may want to consider creating several sub-mixer circuits of say 6-8 pickup signals, which then get passed in groups to a master mixer circuit. Mixing in groups will be less noisy than mixing one batch of 26 signals all at once. If your prototype board is big enough try experimenting with it before committing your design to PCB.

[oolfur]

very cool article... many thanks.

since i want to sum, and not mix - are the pots necessary? could i do something like this?

thank you so much for the info, im going print it out for future reference!

[Prostheta]

Yes. It makes sense, however do you really need to be buffering them all prior to mixing? I think that even super high impedance sources would be fine going straight to a mixer. Don't TL07x amps have an input impedance in the order of teraOhms?

[curtisa]

I think the "buffers" are simplified representations of his per-pickup preamps (first post up top).

If you don't want to do volume for each string you can leave out the pots - just connect each pickup signal to the left-hand point of the 10K mixing resistors.