[sdiy] Outputs grounded! Re: SSI2164 datasheet updated

KD KD pic24hj at gmail.com
Fri Aug 17 02:02:13 CEST 2018


2018-08-12 11:37 GMT+02:00, Mattias Rickardsson <mr at analogue.org>:

> I've always found good use for all VCA channels, there's always some
> additional feature to come up with - and if not, why not use them in
> parallel to win some extra SNR - so I have no experience in connecting
> unused VCAs in different ways.
>
> But *if* I would leave a VCA channel unused, I'd definitely connect the
> output to ground, since that's the potential it's supposed to sit at when
> used.

Thats would be a problem since you didnt GND the input.

 And it *is* still used - not by me, but by the IC. If someone left it
> open and didn't notice any problem, then fine - but it doesn't mean that
> I'm going to do it or recommend it to anyone. It just means that someone
> didn't notice any problem in the actual application. It doesn't mean that
> there wasn't any problem, it doesn't mean that the person made all the
> measurements needed to detect a problem, it doesn't mean that it gave
> optimum results even if there was no problem. And since I don't even know
> the application, I can't know that I have the suitable conditions to start
> with - so I'm not interested in taking risks using methods that might both
> work and fail for me when there are safer methods around.

>It wouldn't be a robust design choice.

See, you also dont know to 100% and thats the whole point i brought this
up in first place because just like me lots of people was and still are
lets say "surprised" by the GND'ing of in's and outs in the later datasheets..

> But I'm happy to hear that it seems to work well in all the existing
> products out there with unconnected VCAs, it would be really sweaty
> to find out it didn't! :-)

Dont indirectly put gasoline onto the just extinguished bush fire! ;-)

> When it comes to the recommendation of grounded inputs I'm a bit more
> confused though.

I think Neal and Roman finally gave a proper suggestion onto why
in's and out's to GND. I buy that and now ditch and dismiss Dixons
approach although it seams to work! :-D

 >The same applies to the "Figure 7: Ultra-Low Noise VCA"
> example in the v3.0 datasheet, where all four inputs are tied together
> after a common input resistor and compensation network.

I guess you get equal input bias and saves resistors.

>And also many VCF schematics I've seen through the years, where the
>control currents of OTAs are generated as one common current, and then
>divided by connecting it to all OTAs with their control current inputs together.

The attempt is to distribute even currents, current split to each OTA
to sink rater to introduce a resistor % in uneven current for each.
What you get is a better performing VCF.

> When a (virtual ground) VCA input is grounded, how can you know it doesn't
> get a substantial DC signal current due to imperfections?

Measure it! According to Roman the cell is perfectly ballanced.

> When many VCA inputs (or OTA control ports) are tied together,
> how can you know they get the same (current) signal?

You can measure it, polysillicon used  as resistors depending on
process have ugly tolerances but apparently good enough on the
2164. Lm13700 use two DMM's

> My gut feeling tells me that the unused VCA inputs should be left
> unconnected - or perhaps connected to ground via a typical input resistor
> value & compensation network - in order to be "provided no signal".

Neil/Roman suggest GND.

> Connecting it directly to ground makes me know nothing about what happens
> to the input signal. But I'm sure some of you can clarify this a bit?

Neil/Roman have.
> /mr


More information about the Synth-diy mailing list