MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

R.I.P. LM3046: byebye affordable temp stabilized VCO's
MUFF WIGGLER Forum Index -> Music Tech DIY  
Author R.I.P. LM3046: byebye affordable temp stabilized VCO's
D' Naab 136
TI announced on Aug 7th. 2017 the discontinuation of a bunch of IC's. This due to the planned closure of the GFAB wafer fab in Scottland, which they acquired with their acquisition of National Semiconductor.

Among those is the LM3046 transistor array.

AFAIK this is the last general purpose monolithic transistor array that's widely available and affordable.

Without it, temperature stabilized (heater topology) exp and log-converters for VCO's and VCF's will quickly become a rarity as other usable arrays from AD,TT ,THAT,... come at prohibitive cost or are not easy to obtain in small quantities.

The same goes for decent current mirrors.

I think everyone using these should pester TI a bit with mails requesting moving production to one of their other fabs. Especially the bigger manufacturers here who use them in large quantities (wink wink Arturia) and have more leverage than a few DIY enthusiasts.


The relevant PCN can be found here:

https://tech.ebv.com/PCN/pcns_show.php?pcn=1502911127
jmagin
On the whole TI has been pretty good at not discontinuing products after their acquisition of National, though this is disappointing.

I'm not really sure the price of the AD or THAT parts is that unreasonable if you want really nice matched transistors. The THATcorp parts are $7-8 in single quantity from Small Bear, which if you're going to all the effort of building a good analog VCO doesn't sound so bad.

On the other hand, there are plenty of cheap dual transistor parts like BCM847DS, are these completely unsuitable?
guest
its a bummer, but i dont think its a show stopper. the LM3046 is really poorly matched compared to new matched pairs, which are much cheaper. for non-heater topologies you would be better off using BCM or PMP parts. having 5 transistors on the same die is really useful for a few niche applications, but for a lot of those the 4 matched pairs from AD or THAT would work better, but at a much higher cost (but not unreasonable for what youre getting). for a low cost option, there are still quad 3904 arrays, which although not gauranteed to match, will probably be about as good. finally, there are still a ton in stock, and 100$ would probably get you more than you could ever use personally. for companies its is a different issue.
D' Naab 136
The other monolithic arrays are simply overkill (matching doesn't matter when you keep the transistors at the same temperature) and $7-8 is more than 10x the price of single LM3046's.

dual transistors are of no use for this as they are not monolithic (on the same piece of silicon). SMD duals are just 2 separate transistors in a highly thermal insulating epoxy (they might as well be placed in another room) and worse is that their collectors pins are basically highly conductive heat pipes from the outside world directly to the dies. They are thus greatly affected by any difference in their operating points and temperature gradients on the pcb, drafts, convection,...

Monolithic arrays are vastly superior because they are on a single piece of silicon crystal which has low thermal resistance and it's also mounted on a copper paddle with even lower thermal resistance. This whole is mounted in an insulator, without low thermal resistance path to the outside.
PISS.EXE
They don't have a replacement lined up for it? d'oh!
guest
D' Naab 136 wrote:
The other monolithic arrays are simply overkill (matching doesn't matter when you keep the transistors at the same temperature) and $7-8 is more than 10x the price of single LM3046's.


yes, they may be overkill depending upon the application, but most LM3046 i buy are 0.80$ - 0.90$ @qty1, so the 7$ @qty1 THAT chips arent a real major expense unless im going into production.

Quote:
Monolithic arrays are vastly superior because they are on a single piece of silicon crystal which has low thermal resistance and it's also mounted on a copper paddle with even lower thermal resistance. This whole is mounted in an insulator, without low thermal resistance path to the outside.


again, for heater applications a dual array is of no use. but for other applications, i dont think the thermal conductivity matters too much, and the poor performance of the LM3046 doesnt make up for this benefit. i have done extensive thermal tests of the LM3046, PMP4201, THAT300, and MAT14 in exponential converter applications, and found that the high emitter resistance and low beta of the LM3046 made it the worst of the bunch, regardless of its monolithic process. the PMP4201 is so small, that another layer of insulation around the whole IC makes it pretty much the same temperature inside the plastic packaging. the PMP4201 and THAT300 performed identically, with the MAT14 being miles ahead of the rest.
D' Naab 136
It's lower beta and (seemingly) higher emitter resistance than the duals only introduces a static linear error which can be perfectly compensated for.

The duals only seem to have a much lower emitter resistance in exp convertors. But that's because one of them heats up much more than the other one at higher output currents. The net-effect of this is some sort of self compensation of the emitter resistance. Seems nice, but it's a dynamic process that lags.

All arguments aside of which is better, this is still a great loss.
guest
D' Naab 136 wrote:
But that's because one of them heats up much more than the other one at higher output currents. The net-effect of this is some sort of self compensation of the emitter resistance.


thats an interesting effect that i hadnt considered, although how big is it? the PMP4201 has a thermal resistance from junction to air of 625K/W, so a typical expo converter running at 100uA would produce 60uW of power and .037K of temperature rise (assuming no heat transfer between dice).

i think ill rip open one of the pmp4201s to see how the dice are layed out inside of it. im guessing the leadframe is continuous under the both of them, giving them a nice copper thermal conductor. but it also might just be a seperate die on each collector pad.
D' Naab 136
That 100µA doubles with each octave for the output transistor where the reference one stays at this resting current. It's also a bit optimistic to calculate power with collector at 0V. This is the case for example in opamp integrator based vco's. But could be anywhere between the supply rails. In case of the prodigy vco it's on average half the amplitude of the sawtooth.

The leadframe is not continuous as this would tie the collectors together as both transistors have no isolation layer or substrate at lower voltage that would keep them isolated. It's just 2 x BC547B dies together in a much smaller package.
widdly
Doesn't Intersil still make a 3046? The website says CA3046 is obsolete but they have an HFA3046 which could work. Not sure about prices.
guest
the HFA3046 is a different transistor, its a high frequency device. it has a really low breakdown voltage.

also, i ripped apart the PMP4201, and it is indeed two dice on opposite side of the packages (sitting on the collector paddles). the dice are ~0.3mm on a side.
widdly
Is the low breakdown voltage an issue for temparture ovens?
belltones
i have dozens of those! am i going to be rich now?
Paul Perry
ON Semiconductor has plants able to manufacture these, if they wanted to.
mskala
Looks like we're also losing the LM4250 programmable op amp. Do any programmable op amps remain in regular production? There's the place in Latvia with "revived" production of the old Soviet chip for Polivoks clones, but TI's recommended substitute is a digitally programmed instrumentation amplifier and not equivalent at all.
guest
widdly wrote:
Is the low breakdown voltage an issue for temparture ovens?


yes, its like 10V or something, if i recall correctly. you could work around it, but im betting there are other parameters that would make it less than ideal.
NewEnglandAnalog
I am bummed about the EOL announcement too, but it's not the end of the world. I have cloned lots of ARP gear using the MPQ3904 and MPQ3906 transistor arrays, and have had no issues. They aren't pin for pin identical, but they do the job just fine. Both of those chips are still made in abundance.
wackelpeter
Those MPQ390x arrays aren't matched Transistors, right? Had a look over 2 data Sheets and couldn't see any hints at this.
They only seem to be some kind of thermally coupled within the same package, is this correct?

Still have a few CA3046 and CA3086 but rarely use them in my builds.
Mostly i stick to some hand matched Transistors glued together then on the flat side.
NewEnglandAnalog
wackelpeter wrote:
Those MPQ390x arrays aren't matched Transistors, right? Had a look over 2 data Sheets and couldn't see any hints at this.
They only seem to be some kind of thermally coupled within the same package, is this correct?

Still have a few CA3046 and CA3086 but rarely use them in my builds.
Mostly i stick to some hand matched Transistors glued together then on the flat side.


Yeah, they aren't perfectly matched, but for the applications I have used these in, they sound great. I cloned the VCOs from the Odyssey, Axxe, and used them to clone the Moog VCF and the ARP 4035. I had no issues. To be clear, I am not using these for my exponential converters.
v8pete
This is probably going to cause a few headaches for the major synth builders too. I think that the VCOs in the UliMoog are LM3046-based, and I'm guessing that's what might be in the reissued SEM too?...

Maybe Mr Behringer will oblige and make us a nice replacement though his Cool Audio semiconductor company?... a V3046 perhaps! ;-)

If you just need a single matched npn pair in a package, then the BCM847DS is looking pretty good, as it's housed in the SC-74 package type, so the pins are 0.95 mm spacing, so are much more user-friendly for hand-soldering than many of the other SMT pairs. These are very cheap in quantity, and the basic specs look great at first glance.
strathTerrace
Greetings People,

Its been a while since there has been any activity on this topic, but I thought folks would be interested in Alfa Rpar's offerings. Most folks are aware that they are offering clones of famous synth chips and op-amps. But they also offer clones of the LM3046-CA3046 (Alfa's version is AS3046D) in both SMT and PDIP. They also offer several flavors of the LM194 (AS194) and LM346 (AS394) matched pairs in both SMT and through-hole in little cans.

URL here: www.alfarzpp.lv/eng/sc/transistors.php

I have seen AS194 and AS346 for sale, but not the AS3046 so far.

In Europa Erica Synths sells AS194 and AS346 , and I am sure other outfits do too. In the states cabintechglobal.com sells them along with a handful of other cool "rare" parts.

Cheers
ixtern
strathTerrace wrote:
Greetings People,

Its been a while since there has been any activity on this topic, but I thought folks would be interested in Alfa Rpar's offerings. Most folks are aware that they are offering clones of famous synth chips and op-amps. But they also offer clones of the LM3046-CA3046 (Alfa's version is AS3046D) in both SMT and PDIP. They also offer several flavors of the LM194 (AS194) and LM346 (AS394) matched pairs in both SMT and through-hole in little cans.

URL here: www.alfarzpp.lv/eng/sc/transistors.php

I have seen AS194 and AS346 for sale, but not the AS3046 so far.

In Europa Erica Synths sells AS194 and AS346 , and I am sure other outfits do too. In the states cabintechglobal.com sells them along with a handful of other cool "rare" parts.
Cheers

I've been buying AS3046 as well as AS194, AS3340, V3205D and other chips here:
https://syntaxis.com.pl/en/12-electronic-parts

Interesting thing is they still have the original RCA CA3046 in stock:
https://syntaxis.com.pl/en/home/46-ca3046.html
MUFF WIGGLER Forum Index -> Music Tech DIY  
Page 1 of 1
Powered by phpBB © phpBB Group