Sound Gain board - design notes
I'm onto version 3 of the sound design board. This version has a pre amp with gain of 100 plus another stage of amplification of 1 to 255 gain (V2 only had gain of 32, also less grounding). The second stage is controlled by two 255 digital potentoimeters ( so that gain increases can be done with small and similar % increase/decrease). 20140422 - I've sent off the parts for this board, need to quantify the noise that I have from it and how low I can do with the signal. Note - if gain of 100 in pre amp is to much for high noise then could reduce this down, making sure that I have the full dynamic range.
Next Version Ideas
There are a couple of things that I have seen that would work for the next version. Although the current version (v3) works well with the get the signal process it and then adjust the gain settings. But there are much better ones out there. These would require redoing some of the code and dealing with the issues of design that come with something new (maybe this is less of an issue).
Conclusion 20140422
Its not worth changing at the moment. The design that I have for sculpture is simple and works. If I wanted something that didn't clip the signal and hence loss information when the gain was to high or low, then 24 bit way would work better. But I don't need this at the moment as I can just ignore the signal and adjust the gain, hence it is slower. Also I might be able to make something smaller and with less pins but take longer and more code changes. So lets see what happens. Interesting to see the technology and could be used in other places. Also I would need to more to another level of soldering (probably not a big issue, and might be something I want to do, i.e. use an oven and stencil).
One other thing to think about is focusing on getting a better signal. One design paper that I saw was talking about the ground plane and for the mic bit having it go though the steady voltage chip gnd pin, I didn't do this and this might help. But at what point is it good enough. Again lets see what happens and if the noise levels are acceptable. With the 24 bit design I think it can be made better noise levels, both by reducing the whats happening i.e. less changing of the gains and also by more knowledge of design and grounding and fluctuations. Interesting.
note 20140424 - with my V3 its all analog so I dont have to work for predefined sample rates. This is an advantage.
Note - There are some shields out there that have 6 input with better adc converters than arduino, I think 16 bit ( cannot find the reference at the moment).
Note 2 - Sencil service at http://smart-prototyping.com/Prototyping-Services/Electronic-Prototyping/PCB-Solder-Paste-Stencil.html, looks a really good price better than seeed, although not sure of the board quality. But it looks like I can just order the stencil and it only cost 20$ (seeed cost 70$). Note - does mean that I could do other boards.
- 24 Bit ADC - There are 24 bit adc converters (these are cool). At present I'm using 10 bit adc on the arduino, if I use the 24 bit one then the dynamic range of signal that I an deal with increases another 14bits (thats 16382 time more resolution, omg but they say that it really only goes to 19 bits so that 9 more bits which is 512 more gain, still twice as much as I have not and not really needed) , which is much more than I have with the 255 gain at present (although I cannot increase the gain more I can decrease the initial stage of gain, which would mean that if I can get the noise level down then I have better range, from the very strong signals down to the very small signals and I don't change gains.) Advantages - more dynamic range that won't be clipped, so less distortion happening with large changes in signal. Disadvantages - will need to rewrite some of the code so that I normalise the input signal for the fft to work. Also need to look at speed of processing and if this can be done. But very interesting didn't think about 24 bit processing and this looks like how they do CDs and there are chips around for this. Note 18bits looks like all I need ( but lets see what the noise is like in the next version). Note - with 512 gain means that I can decrease my initial gain and get more dynamic range. note 2 - someone has made one that interfaces with arduino, refer - http://forum.arduino.cc/index.php?topic=57873.0, they have some interesting stuff here and look at the design but the adc is expensive. They do talk about how to design it and that it needs an master clock as well as signal clock. With this design it looks like I could get the number of connecting wires down to 5 and the size might be able to be a bit smaller. Note 3 - 24 adc that a good prices, refer mouser.com http://nz.mouser.com/ProductDetail/Wolfson-Microelectronics/WM8783GED-V/?qs=sGAEpiMZZMvANcmvDb1WMdKoy6izcZEu.
- Single ADC with pre amp - OMG there is an all in one chip that would do everything. Wolfson looks like they are into this in a big way. and the WM8952 is the all in one. http://www.wolfsonmicro.com/documents/uploads/data_sheets/en/WM8952.pdf, they say it comes in QFN (quad flat no leads), but I cannot find this and its a BGA ( Ball gate array). advantages - all in one so should be better noise reduction if I follow there instructions. the BGA chip is only 2.5 mm sq now we are getting real small. Disadvantages - another level of soldering, even with the QFN this is not a hand solder (although it can be done), and I could use an oven. Or I could get someone to make it for me. Programming would be more difficult and require more work.
Conclusion 20140422
Its not worth changing at the moment. The design that I have for sculpture is simple and works. If I wanted something that didn't clip the signal and hence loss information when the gain was to high or low, then 24 bit way would work better. But I don't need this at the moment as I can just ignore the signal and adjust the gain, hence it is slower. Also I might be able to make something smaller and with less pins but take longer and more code changes. So lets see what happens. Interesting to see the technology and could be used in other places. Also I would need to more to another level of soldering (probably not a big issue, and might be something I want to do, i.e. use an oven and stencil).
One other thing to think about is focusing on getting a better signal. One design paper that I saw was talking about the ground plane and for the mic bit having it go though the steady voltage chip gnd pin, I didn't do this and this might help. But at what point is it good enough. Again lets see what happens and if the noise levels are acceptable. With the 24 bit design I think it can be made better noise levels, both by reducing the whats happening i.e. less changing of the gains and also by more knowledge of design and grounding and fluctuations. Interesting.
note 20140424 - with my V3 its all analog so I dont have to work for predefined sample rates. This is an advantage.
Note - There are some shields out there that have 6 input with better adc converters than arduino, I think 16 bit ( cannot find the reference at the moment).
Note 2 - Sencil service at http://smart-prototyping.com/Prototyping-Services/Electronic-Prototyping/PCB-Solder-Paste-Stencil.html, looks a really good price better than seeed, although not sure of the board quality. But it looks like I can just order the stencil and it only cost 20$ (seeed cost 70$). Note - does mean that I could do other boards.