Son of Discrete OTA

[R.G.]

It is **possible** to lay out a discrete version of the CA3080 using the DMMT3904W and DMMT3906W on a PCB which has pins that will plug into the same socket as an integrated 3080.

Practicality may be something else, but a huge run of these things at an SMT assembly house is a possibility.

[lovekraft0]

You're a genius! Any idea what they'd cost, or how they'd compare in performance to the original chip? If enough people show interest, it might be worth doing a group buy or something, provided the logistics could be worked out.

[toneman]

how will U match all the trannys???
also, each will have different thermal considerations.
1 opamp and 3 trannys can make an OTA.
ARP used a similar configuration in their VCA.
T

[R.G.]

I don't know about the genius part. I get a lot of evidence to the contrary.  

Here's a picture:


Those traces, by the way, are as thin as 0.008" and they have 0.004" clearance the two places where they go between pads, so you have to have someone who knows what they're doing etch the boards. This is NOT a press-n-peel operation.

PCB is 0.4"/10mm square, so you get a lot of them out of the board blank, and they're efficient to print. They'll probably go for the nominal $0.25/sq in., so call it $0.04 per board. Mouser lists the following for the DMMT's:

1: $ 0.500
10: $ 0.400
100: $ 0.300
250: $ 0.230
500: $ 0.180
1,000: $ 0.160
2,000: $ 0.150
5,000: $ 0.140
10,000: $ 0.130

and you need 3 3904's and two 3906's per unit.
You also need pins. A suitable header pin strip is available from Digikey for about $5.00 for a 64 pin strip, which makes 8 units, so the per-unit pins cost is $0.63.

So we have
PCB................ - 0.04
DMMT3904W(3) - 0.54 (using 500 piece price)
DMMT3906W(2) - 0.36 (using 500 piece price)
Pins................. - 0.63

for a total of $1.57 in parts at 250 units.

Getting someone to tool up to populate and reflow these will be a giant pain, possibly expensive. I don't have a good guess for that. Maybe you could do something with etching a copper or brass sheet with the component placement and speed up manual placement so it would be feasible. Lay the template on the PCB, put all the components in the holes where they go, sticking them down with a dab of glue. Remove the template, put another template on which has the second component placement, and place all those. It works for only two different components on a board. I would etch the boards in 0.032 thick copper (half normal thickness) and leave them all on one giant board with breakaway routes in the board. That would make population easier than dealing with hundreds of tiny boards.

In large quantities - probably 10K, as they're so small - the assembly cost would tend to maybe a nickel a board, but that will not be true in quantities of less than several thousand at a time.

But you could probably get the whole thing under $2.50. Not a good thing with $0.50 LM3080's around, but possible if 3080's cost $5.00.

Performance? OK. Probably as good as a 3080 for most things, not as good for precision synth volt-per-octave stuff. Probably much better at noise.

[toneman]

hey RG,
can U link to schemo U R refering 2??
i C 12 trannys in on OTA, but your SMT layout only has 10(?)
(5 duals)?
thanx
tone

[Rob Strand]

Nicely done.

Haven't seen the schematic but if the pairs for the mirrors and diff-pairs are chosen correctly the temperature stability should be OK.

PS: probably worth emphasizing the W in  DMMT3906W.

[R.G.]

can U link to schemo U R refering 2??
i C 12 trannys in on OTA, but your SMT layout only has 10(?)
(5 duals)?

The schemo is my own. I used two-transistor current mirrors for all the mirrors. The 3080 uses three-transistor mirrors for the PNP mirrors because of the inherently poor gain of lateral PNPs when made on the same chip with NPNs. It's still a four-mirror OTA.

Nicely done.
Haven't seen the schematic but if the pairs for the mirrors and diff-pairs are chosen correctly the temperature stability should be OK.
PS: probably worth emphasizing the W in DMMT3906W.

Thanks.  I think epoxying a sheet of aluminum over the whole mess should contibute to equal temps across devices; I actually think that mirrors being what they are, temperature sensitivity from mirror to mirror is a minor consideration.  The critical pair is the input diffamp - I think.

And yes - HEY EVERYBODY!!! THE DMMT3904W IS WHAT YOU WANT!! THE DMMT3904 IS HARD WIRED DIFFERENTLY INTERNALLY.

[Rob Strand]

sheet of aluminum over

It would help but since the aluminum isn't connected to the sub-strate there can still be thermal gradients; the case -> sheet thermal resistance is pretty high.  Using two layer pcb with a copper area would be neater but possibly a little less effective depending on specifics.  The problem with putting metal areas down is they need to be grounded to prevent  noise picking up.

temperature sensitivity from mirror to mirror is a minor consideration

The trick is to put the each transistor pair for the mirror, and diff pair, in the same DMMT390x device.  That way the two devices are in thermal contact - fat tracks between the two can also help.  The same goes for the diff pair they should come from the same DMMT390x  package.      You may have already done this, and there's little point going further. There's only minor benefit  making the pairs of different mirrors the same temperature.

You might be interested that Philips have matched pairs available as well:  BCM847BS and BCM857BS; not to be confused with the unmatched BC847BS and BC857BS.  The vbe matching, which is more important, looks better on these, whereas the hfe matching looks worse.  These units have a different pinouts.  Prices?

[R.G.]

You might be interested that Philips have matched pairs available as well: BCM847BS and BCM857BS; not to be confused with the unmatched BC847BS and BC857BS. The vbe matching, which is more important, looks better on these, whereas the hfe matching looks worse. These units have a different pinouts. Prices?

I did find those, but didn't use them because they are not easily available that I can find - and I couldn't find any prices. Great minds seem to run in the same ruts.

[Rob Strand]

I did find those, but didn't use them because they are not easily available that I can find - and I couldn't find any prices. Great minds seem to run in the same ruts.

I guess we both go through the same motions.

Philips only started offering these at the end of last year.  It's a bit early for them to get onto general component sellers catalogues - no doubt Philips distributors will be able to get them.  I suppose no action will be taken by general suppliers until the next catalogue issue - presumably the BC devices will be easier to get for the European guys than the DMM stuff.  We'll just have to wait and see how things pan out.

[puretube]

now to spin the idea further, some small stone cloners might want an
additional darlington-"follower", to get a "son-of-3094".

to satisfy even more out there, there should be SIL version of the latter...

[mikeb]

Neat, though the miniscule clearances scare me.

Mike

[R.G.]

now to spin the idea further, some small stone cloners might want an
additional darlington-"follower", to get a "son-of-3094".

to satisfy even more out there, there should be SIL version of the latter...

Good ideas!

Now I gotta do more layout... bummer....          

[puretube]

everything smaller than the 1/10" grid annoys me - I`ll go for a 12AX7 solution...  :wink:  
(will take some re-design of the applications, though - e.g. the Dinocomp)

[puretube]

http://focus.ti.com/lit/ds/symlink/opa660.pdf  

[Arn C.]

What about some transistor arrays?  I purchased a bunch a while back, I can not remember the different ones I bought, but one of them was CA3046  it has two matching transistors, and I believe a total of 5 transistors.  I measured them all and they all had the same hfe.  5 matching transistors in one package.  

Something to look at anyway.....

Peace!
Arn C.

[gez]

What about some transistor arrays?  

There are a few schematics on synth sites which show how to use arrays wired up as OTAs.  Don't have the links though, sorry Arn.

[mojotron]

http://focus.ti.com/lit/ds/symlink/opa660.pdf  

This discussion is going to lead to some good things, I like the idea of using discrete components - which allows for some tinkering too.  

There's this one http://www.mouser.com/index.cfm?handler=displayproduct&lstdispproductid=347638&e_categoryid=131&e_pcodeid=52617 too... But, I have never used a BB part that I didn't love!! Thanks for the link.

[puretube]

HERE`s a nice one...
(VCA, actually)

[puretube]

http://arraydesign.com/downloads/index.html