[sdiy] Re: bread & butter sequencer+quantizer revisited

[harry bissell]

RIGHT NOW the beast I am working on is the quantizer. I started with 
Scott Stite's cool design for a quantizer using a counter and a 12-bit 
DAC and some LF398 S&H circuits, but naturally I didn't just have an 
AD7245 lying around & no 4516's either. But I did have 2 4520's and 2 
AD7524's and 2 AD7528's. I haven't quite wrapped my brain around the 
necessity for a 12-bit DAC when not all of the bits are used. Something 
about the precision?

Yes... if you use the high orser bits... it will be better than an 8 bit 
resolution. That said a lot of better
8 bit DACs would probably be good enough for civilians...

 I also have not wrapped my brain around how to hook 2 8-bit AD7524's up 
to get a 16-bit DAC. Would that get me the precision I need? Do I just 
connect the RFB pin on the first DAC to the REF pin on the second DAC?

That would be wrapping your head around a tree.  You will ~not~ get real 
sixteen bit performance from
cascading DACs like that.  You may get sixteen bits of resolution... if 
you are lucky it might even be monotonic
(ever increasing) but accuracy ??? forget it.


I thought about trying just a 4520 and only one AD7524 and see how that 
worked for me. Also would it be pointless to use synchronous 4520 
sections cascaded in ripple-mode (ie would I really be that much better 
off using 4516's?). It's really pointless to worry about that at this 
point because I am going to use them nonetheless. They do seem like a 
good candidate even though they're up-only - I only need one unit as 
well while with the 4516 I'd need 2. The datasheet shows a way cascade 
them synchronously, negative-edge triggered, but the parts count would 
be increased by 2: a 4012A of all things and a 4071. I have several 
4071's and I suppose I could make half (all I need) a 4012 with a 4011 I 
think. I have the room on the protoboard that I'm using so if it'd make 
that much diffrence. But really it seems a bit of a hassle.

If you want to use a ripple counter... you should change the circuit a 
bit. Have the comparator output
go to the data of a 4013 D flip flop... and clock from the edge of the 
counter clock that does ~not~
increment the counter (use an inverter if necessary).  This way you 
increment the counter, allow everything
to settle...and 1/2 clock later you latch and strobe the S/H. It works 
WAY better.

H^) harry
>