[sdiy] THE LM13600/LM13700 STORY

[Ian Fritz]

At 04:23 PM 7/11/2007, JH. wrote:
> >Ah, I didn't know that.  I think most people find it superior to the 13600,
> >not only in linearity, but in S/N and -- especialy -- in CV rejection.
>
>Do you know what exactly they made wrong that leads to increased CV
>feedthru?

No, I don't know.  According to the Serge writeup (EN#113) the 3280 uses an 
interdigitated structure for the input transistors, which reduces noise and 
offset.


>I'm asking because _theoretically_ the 13600's topology should have an
>advantage:
>The common mode signal at the input transistors is smaller than in a 3280.
>And normally a change of Vce will also cause a little change in offset
>voltage ot the
>differential pair. (according to the data sheet of really good monolithic
>matched pairs.)

I understand what you are saying.  According to the RCA data sheet, the 
3280 has very high CM rejection.  I don't know why that is.  As Serge says, 
the 3280 has many non-obvious features and quirks.


>That's why I've chosen a similar topology in my discrete OTAs for this QVCO:
>http://jhaible.heim.at/matrix_fx/jh_matrix_fx_osc1_core_board_sch.pdf
>The rather small resistors R108 and R109 keeps the voltage swing at the
>bases of U6
>small (and J5 helps to to keep the two CE voltages similar).
>
>If I had chosen the CA3280's topology, the differential pair would have been
>shifted
>up and down in a common mode signal that's derived from the (differential)
>input
>signal, for which the MAT-02 specs would predict more CV feedthru.
>
>So (asuming I haven't drawn wrong conclusions, which is entirely possible),
>the 3280
>must contain much better transistors than the 13600, on order to compensate
>that effect
>that would give th e13600 a topology advantage, and be even better.
>
>Does this make sense?

Of course.  :-)

I guess what I would do to compare is to make a test setup (like a VCA) 
with the same I/O levels and compare the feedthrough by a direct measurement.

When I was working on discrete OTA's a couple of years ago I used a 
symmetrical 4-transistor Wilson mirror and took the output from the pair 
differentially.  This seemed to help performance.


I've been contacted by a guy who is working on making high-performance 
discrete OTA's using inexpensive transistor pairs.  He is getting 
performance similar to the 3280 and claims he can probably do 
better.  Since he is developing this for his job, he doesn't share any 
details, but I gather it involves using fairly strong degeneration (trading 
off headroom or something like that). He claims nobody has done things his 
way before.  He has sent me a lot of his data, which he takes with an 
automated Keithly system, and he seems to know what he is doing as far as 
characterization.  Maybe these will be commercially available some day!

   Ian