[sdiy] Digi-Mod schematic

[Seb Francis]

Hi Neil,

Thanks for your detailed comments.  Your input is very much 
appreciated.  My answers below ...

>
> - page 1: where are the slave-select signals for the SPI interfaces? 

Hmm, I thought it's possible to operate SPI in 2 or 3 pin modes.  But 
reading the dsPIC datasheet again it's not clear whether the SS pin can 
be disabled when in master mode.  Anyone know this for sure?

If the SS pin can't be disabled in master mode then I guess I should 
remove access to SPI1 (no way to use SS1 pin) and just have SPI2, 
including the SS2 pin.  This is also needed for the EEPROM as you 
pointed out.  This does mean no external SPI if the EEPROM option is 
included, but there's always I2C.


> - page 1: do LEDs _really_ need to run from 3V3 rail?

No, I suppose not really.  I guess I just preferred not to have high 
current switching on the clean +15V supply.  I will change it so they 
run from digital +5V (i.e. the raw +5V power input), and the builder 
will have to ensure the correct resistor values are used in the case 
where there is no +5V supply and +15V is used instead.


> - page 2: do something with the enable input of X1 (don't leave it to 
> float) 

According to the datasheet of the oscillator I was looking at, this pin 
can be left NC or high to enable the oscillator.  But perhaps this isn't 
the case for all oscillators and it's no problem to connect it high, so 
I'll do this.


> - page 2: where is the supply decoupling capacitor for X1? 

Yep, got to be a good idea!  I'll pop in a 100nF.


> - page 6: if you're going to use the 7805 to drive the LM2937 I 
> suggest a low-value resistor (a few ohms) between the two, otherwise 
> the active load of the LM2937 can upset the 7805, causing possible 
> oscillation.

You misread it.  "+5V_" is the raw 5V power input.  (This can also be 
connected to 15V if 5V is not available.)
And since 5V power in most modular systems is used to drive all the 
digital stuff, I didn't want to use it for the analog stuff here, so I'm 
getting a clean 5V by regulating the 15V with the 78L05.  This clean 5V 
is what label "+5V" refers to.  The current requirements are very low 
for this (just the CODEC VA).

I'll add a little note to the schematic as it's not clear.

You think I should have a low value resistor on the input of the LM2937 
anyway for the benefit of the external 5V power?  Or maybe a low 
inductance ferrite bead and a bigger capacitor?


> - page 7: you need to control the CS\ pin, otherwise you won't be able 
> to talk to the chip.  Read section 7.2 of the Winbond datasheet: 

Ah, yeah.  Good point!  That'll teach me not to read the datasheet 
properly.  Would best be driven by the SPI !SS pin.


> - page 8: given the critical nature of the 220p capacitor I would put 
> the BAT54S on the left-side of C47.  Or, an alternative approach would 
> be to put the BAT54S on the input side of R85, and put, say, 100R on 
> the output of the 4580 inside the FB loop to protect its o/p stage 
> when going -ve.  Also simplifies layout as you don't need to protect 
> up to +15V rail then (one less track to worry about).

I'm not sure about putting the protection on the left of C47 as the DC 
bias levels are quite different: 0V to the left and 2.5V to the right.  
Can't see a neat way to do it other than how I have it.

The BAT54S datasheet quotes a capacitance of 10pF max, presumably per 
diode.  Do you really think this will make much of a difference?


> - general: the 4580 is a bipolar op-amp so you really need to think 
> about input bias currents, otherwise you're going to get unwanted 
> offsets all over the place.

This is a good point and I will include the resistor from +ve to ground 
for U14:B/U16:B.  But I don't know how to handle this for the VCA output 
into U14A/U16A since the VCA has a current output, not fixed output 
impedance.  Presumably though any DC offset here won't matter because it 
will be blocked by C47/C55.

For the output stages U17:A/U17:B I will adjust the parallel combination 
of the 5K1 and 1K resistors to match the combined impedance of the CODEC 
and 470R resistor.



> - page 12: not sure what you're trying to achieve here with your funky 
> dual feedback circuit.  This looks like an output buffer, for driving 
> long capacitive lines.  Are you trying to make a LPF? 

As Tom posted, I was trying to provide optimum drive for the 
time-multiplexed ADC inputs of the dsPIC (see "Compensated buffer 
topology" - http://www.audiodesignline.com/210605596?printableArticle=true )

And yes I'm also happy that it will be a LPF as this should help CV 
stability.

Perhaps this is OTT for dsPIC ADCs, but I'd like to get the ADC inputs 
accurate enough for pitch CV so thought I'd try it.  I can also solder 
in a more basic buffer configuration and compare if it makes a 
difference.  On the final PCBs I may simplify this if it proves unnecessary.



>
>
> 1. If you can avoid the external +5V input then you can make your 
> module supply-compatible with MOTM and Oakley (amongst others). 

+15V can be used instead, it will simply need a heatsink on U4 for the 
extra power dissipation.


> 2. For the high-current 3V3 rail I would seriously consider a small 
> switching converter, and keep the lower-current analogue supplies 
> separate. 

I did think about this, but I was a bit wary of the HF switching noise 
getting into the analog stuff.  Particularly I wanted a cleanish digital 
supply for the CODEC.

I decided it was going to add too many components to clean up the power 
from a switching regulator, and anyway when there is +5V power available 
the LDO reg is not so inefficient.

But yes, with only +15V available there's a fair amount of wastage 
(~3W).  A modest heatsink (<=20C/W) will take care of the heat from this 
however.


> 3. Took me a while to work out what you meant by "V ptp".  This is 
> more commonly written "Vpp". 

Can you tell I didn't have any formal electronics education ;)


> 4. Any out-bound supply pins on connectors I'd protect with a 
> low-value series resistor, to protect against shorts.

Is this really necessary?  The voltage regulators are protected anyway.  
And what if the power is to be used to for some external device like an 
SD card or something.  Isn't it better to have low impedance power?

But this does remind me to specify that break-before-make type rotary 
switches should be used for sample rate select and other multi-bit inputs.


Seb