[sdiy] Outputs grounded! Re: SSI2164 datasheet updated

[Neil Johnson]

Hi David,

> a) The 2164 datasheet gives no guidance at all as to what to do with unused
> VCAs.

That is definitely true of the original SSM2164 datasheet (and poorly
copied by CoolAudio).  There are some real howlers in that datasheet
and I understand the designers of the SSM2164 were rather annoyed with
Analog Devices when it came out.

> b) The control input is already grounded (across an internal impedance -- 5k
> originally, 10k in the new ones) so this is not an issue.

Indeed.

> c) The 2164 VCA amplifies current, so zero current in equals zero current
> out, except for the input bias (up to 10 nA) and output offset (up to 50 nA)
> currents.

Bias currents are the killer if not treated properly.

> This would suggest that one probably should supply paths to ground for these
> small bias and offset currents.  I suppose I will start doing that in my
> designs now.  However, looking at the inner 2164 circuit in the datasheet,
> it would appear as if all of these currents have alternative paths to the
> negative rail through the various transistors.  Presumably, when the inputs
> and outputs are unconnected, then either these small currents simply don't
> flow, or they get out of the device by some other means.

The currents have to go somewhere.  By connecting the output to ground
you provide a simple safe output path, and that keeps the steering
transistors happy.
Connecting the input to ground appears to be a valid thing to do, and
saves three components per unused input pin.  There will be bias
currents flowing into or out of the input pins, but in this case we
just want to park the pins at a safe level.  It is also guaranteed to
be stable, since the noise gain is now some huge number and now we
have gobs of phase margin where it counts.  And since the inputs are
virtual earth nodes anyway it is an easy option to use ground.

> The Dixie II+ involves three unconnected 2164 VCAs.  The Rubicon2 also
> involves three unconnected 2164 VCAs.  Both of these VCOs track beautifully
> and otherwise work exactly as designed.  Hence, while I'm willing to concede
> that we may have done it "wrong", I'm also somewhat dubious that it makes
> any difference whatsoever.

It depends on where the unused cells are.  Unused outputs may well peg
to one output voltage due to output offset currents.  What effect this
has is hard to say, although I can imagine it maybe causing unwanted
excess to currents to flow in the VCA core, which may have a heating
effect on the die.  In a VCA or VCF application this is probably not
going to be much of an issue, but in an expo converter in the VCO it
might require extra thermal compensation.

> I know that I am a relative neophyte to
> electronic design, and have no formal background in electronics (my day job
> is as a metallurgical engineering professor) other than 10 years of
> experience learning by doing.  However, my designs represent millions of
> dollars of revenue for Intellijel, and many hundreds of happy customers, so
> I figure that counts for something.

It certainly does!  The most important aspect though, and is reflected
in pretty much any profession, whether it is electronic engineering or
metallurgical engineering, is continual learning.  Certainly the IET
(through which I get my chartered status) mandates continuing
professional development - keeping your technical and professional
knowledge and skills alive through continuous learning.  To bring this
thought home: when I first started playing with the SSM2164 it is was
only available from Analog Devices, and tech support was thin on the
ground.  Then ADI EOLed it, so goodbye any official support.
CoolAudio somehow picked up the line, but good luck getting any
technical support from those guys.  It is only in the last year or so
that the emergence of Sound Semiconductor Inc has brought out of the
woodwork the original designers of the SSM2164, who took the
opportunity to fix some of the issues in the original design, make
improvements, and are more than happy to support these new parts.  And
all through this I have been learning more and more about the internal
workings of the 2164 (not to mention the technical archeology of
digging through the SSM2014 and SSM2018 chips, reading through Doug
Frey's patents and AES articles, and so on).

Learning... it's fun!

Cheers,
Neil






-- 
http://www.njohnson.co.uk