Create project Need help?
joe2030

Discrete OTAs for Synth-DIY & Elektor-Formant-Upgrades

Status: Proposal
11425
28
votes
November 19, 2014
Vintage BA662-SIL9-OTA.jpg

This project is intended to give all interested readers some practical suggestions about possible replacements for obsolete OTA-ICs, like maybe the CA3080´s, or the Roland-BA662´s, as also how to get into some easy DIY-Prototyping - or Breadboarding -, to perform some High-Precision OTA-Parts for almost universal High-Quality purposes. Please fell free to contact me on jo2030_at_gmx_dot_net, for getting additional informations on variants, as also maybe all the eagle-files for the shown PCBs in the attached JPEGs (severeal hundrets possible and available!!) to the project. With todays available SMD-Parts, almost every obsolete OTA-IC is possible.

Like also my other projects here at the Elektor-Website, this starts again on a litte story...

 

I.) Introduction:        About 15-20 years ago, I´ve been in need to repair a broken 12dB-State-Variable-VCF, which was housed in a big Formant-Synth and owned by a guy in my near neighborhood.
This was a very good sounding Musical-Instrument, a real legendary Formant-Standard, which was also built as the "Oldschool-Gear" and therefore strictly upset with the recommended CA3080A OTA´s in Metal-Cans, where one of those was the bad thing, damaged and with quite urgent wish by the owner, to change in short time. On this very day it was next to Christmas Eve and also coming a long weekend ahead, so no chance to get quickly a replacement for that important part on local stores, and this shurely lasting for about 5 or 6 days.
Now, - with short words to say - the only possible but also little silly solution to this problem in the end on that, I recommended a discrete small DIY-OTA-Module, performed with some SOT23-SMDs and quickly placed on an Edding-Marker-Made-PCB, which got etched in the kitchen and then (not too complicated - see the attached Photo with BCX70/71´s, as dito) fixed onto the IC-Socket.
This was a direct clone from the schematics in the Harris-CA3080-Datasheet and with all diodes realized as transistors on collector-base-shorted terminals.
This beginning project for discrete OTAs was of course in that moment an experimental "open trial" and the first result was not perfect, but the thing worked, and this not too bad. It served much more offset-voltage, than the formerly used IC, but this lack was easily fixed with an additional trimpot and a little error in tracking was fixed by adjusting another resistors value.

By that time in the first third of the 90ies, there was no word or any worries on "Obsolescense", nor any Synth-Chips not available - the only IC which went really discontinued and sometimes dramatically missing, escpecially on the old Mini-Moog´s Series, was the expensive Fairchild-µA726 (Heated-Dualtransistor) for enhanced VCO-Usages, as also used on the first edition for the Formant-VCO Synthesizer-Modules, and the almost best industrial solution that ever existed for such temperature-sensitive cases in conjunction with Dual-Transistors.
For me, on this exciting repairing-job, there now grown more an more the idea, to construct OTAs as selfmade modules and therefore using available and valueable SMD-Parts, where the main push to such a project came almost on the problems of extensive matching OTA`s for good tracking results of VCFs and what had mostly always an expensive impact on any Synth-Projects.

Then in 2005 the "Big Accident" happend to the CA3080-Production-Line at the Intersil Manufacturing Management. They inadvertently "recycled" the whole manufacturing plant and also the whole facility around that machinery, and so they discontinued (by only some telephone calls) the popular CA3080-Series, also soon to follow the same happened to the CA3280-ICs, the one and only ever available Highend-OTAs for best Audio-Performance. This was quite an "Unlucky Strike" to the Customers Community, but also a somekind of funny part in electronics history, and shurely wasn´t really planned by anyone tangible (as you also may read the whole story on: http://www.till.com/blog/archives/2005/06/last_of_the_ota.html), but so there was also never a real replacement for that very important parts, to use in really much music-electronic devices!

By these days again, I decided clearly to implement my own OTAs on useful SMD-Modules and to get such parts in good conditions as directly or better replacements for that old and obsolete ICs.
The first demand for realizing such an idea is the availability of some good workin´Current-Mirrors, supplied by the industry on an acceptable price and quality, with an additional requirement for good temperature-coupling of therefore used transistors.
Back in the beginning 80´ies, there have been some TO92-Current-Mirrors (TL011/012/014/021) introduced and offered by Texas -Instruments, and this also with different available Current-Transfer-Ratios, what means the user could choose, how much current the device produced on its output, for a given input-current (0.5:1, 1:1, 1:2, and so on) to a pin.
These parts were discontinued after a few years, because of not too much customers orders and therefore not really a higher volume on selling them. Obviously nobody at that time needed some standardized Current-Mirrors!
By the end of the 80´ies, then came Siemens (todays Infineon) and Philips (todays NXP) with the series for the BCV61/BCV62 Current-Mirrors in optimized SOT143B-SMD-Packages and on three selectable Beta-Groups (A-B-C for hFE-group), where these offers are still present until up today, mostly needed for automotive applications.
Somewhere in the near past, the two polarities with NPN- and PNP-Mirrors were followed at NXP´s with some more of those parts, offered as the BCM61B/BCM62B-Devices, on closer matching-data and therefore on using optically selected transistor-chips (used selection-process on wafer-level - of course also some higher offering prices), were also Diodes/USA offers now some matched and compatible SOT143B-Parts in identical manner, with the DMMT847B/DMMT857B-Series, as directly arranged Current-Mirror-Configurations.
 
II.) Discrete OTAs in realisation as needed:        The first and also main problem for doing this with available parts, is a greater group of facts - as so to say - and what means that if one constructs monolithic chips, this is like working out a recipy for a special cake or pie, to bake in the oven (and when you got the right one - you just have to repeat it), where you may adjust almost anything what gets technical possible, going into the smallest details, - but not so if you are forced to use a restricted number of fixed transistor-data, out of some distributors-catalogs and also sometimes as a widespread of uncountable and changing data, and also a not imaginable and variable "Coming of Quality", at all.
In chips-production, everthing is changeable and only gets fixed in the end for a readymade series-production, and for practice this means a manufacturer has variable doping-levels, material-constants, optical stepping tracks on the wafers, gas-concentrations, implanting-energy and so on, for any needed and best condition of the desired product. On such predestinations the development of the CA3080-ICs took remarkably a very long time, from 1967 to 1969, - for a handful of good working transistors, but after that time they got it!
By that years at the first RCA-Manufacturingplant for the CA3080A´s, and at RCA-Labs at all, integrated Analogchip-Production was in its first start-ups, so no wonder for such very long development-times on a good product. At that times these important jobs were done on the chipfamily for the next 3-4 decades, the recipies worked out and they just needed to repeat it in production-charges. Now - Back to discrete parts - Todays...
A discrete OTA is of course upset in the same manner and almost similar architcture as the ICs, but with some restrictions and therefore with some unchangeable errors to compensate and then only using some expensive and good matched transistors, will not completely hit the goal!
On this experiences I spent hundrets of working-hours and much Euros for buying matched parts (as you may see herefore the attached Project-Photos, little report on almost thousand of units), without much useful results on my predestinations, or reaching the desired quality in larger numbers, so here are therefore some other explanations, - with just to upset some rules!

III.) Measuring OTA-Quality on a "Breadboarding-Standard":        If you respectively can get a CA3080A, connecting its Differential-Inputs by 100-120 Ohms (1%) to Ground and supplying 1 mA Iabc to the Amplifier´s Control-Input, with additionally the Output connected to a Loadresistor of 10 kOhms and this tied also to ground (see therefore the schematic to the project as Eagle-PDF), then you can measure on the output-node around 10-70 mV of an "Errorvoltage", where for the CA3080E-Variants this value could vary up to 100-140 mV for regular, and maybe more, but this measurement is almost varying on any ota-parts, seeming also to be a problem for manufacturing the chips, and almost due to varying geometries in chip-wafers-production.  
This voltage follows as a direct Product/Result of the Input-Offset-Voltage and then some additional "Impedance-Mismatch", produced by the used arrangement of the Current-Mirrors, where the active Impedances in the OTA are directly controlled by the amount of the applied currentflow, the so called Iabc. This means, if we see those stages as energy sources, where each value should set up to result as a "one" for the positive and the negative Parts of that System connected in Series, logically in the end this should ideally result as a zero-value, but this isn´t, because of unavoidable Tolerances!
Additionally, these Mismatches are poroportional or better say multiplicating with the Transistorsizes and for the Chipproductions once adjusted in several trials until best performance as possible (still varying!) - but not so, if one uses fixed transistors (very much more varying!) as SMDs.
This means the result in the same arrangement as described above, varies of course much more on that DC-Error and maybe arises up to 0,5 Volts, or even more, so the only way to get this fixed to an acceptable value, is to test one by one for the right constitution of Currentmirrors, to setup for an accetpable working arrangement, or quite easier lets say, to apply always a compensation - what means a trimpot - to adjust these offset levels in any possible circuit.
On accepting these facts, one can get very good electronics with usage of discrete OTAs and also getting into some real enhancements, but first I want to explain here again - as also done on my several other projects - the thing as it works properly, or better lets say acceptable. To get such conclusions to post here, I have tested combinations with around 400 parts of those 3-Pin-Currentmirror-Modules.
For the finally and here as PDF-Schematic presented result, I have tried also a lot of breadboard experiments, as also used a Simulation-Software and then in severeal tests done a comparision to the JRC13700-ICs (very good OTAs, but really identical to LM13700 ?? - still a question), as also sometimes longer ago tested the old CA3080A´s.
Please pardon, that I can not present any actual Oscilloscope-Photos, with some Sine- or Squarewaves, - shurely next to follow as soon as possible!

On the above described measuring method for OTAs, also in the Simulation-Software, the LM13700-Spice-Models exhibit around the above mentioned Offset-Voltage of 120 mV for 1 mA of Iabc and so do the JRC13700-ICs in the same manner again on breadboard units testing.
Both OTAs are quite good ones, but if you can compare them with some CA3080A´s, they have almost no chance - and to get those Dual-OTAs intended for higher quality performances, you need to select the best parts out of a "larger bunch" of them...
With the above told possibilty of trying also hundrets of Current-Mirrors together with matched KC811 Dual-Transistors (Czech Product), I got High-Precision-OTAs (unavoidable costs per item around 7-11 Euros, if collected alltogether!!) with less than 1 mV errror on this value, what is excellent and where no additional offset-trimpot was needed!
As another information on the possible performance for that OTAs with those "Balanced-Current-Mirrors", they will perform the same as the Datasheets report for regular known ICs, so to say with possible Frequencies as up to 1-2 MHz and this only with maybe some restrictions, caused by the PCB-Layout, and so on. The only difference to the CA3080-Series was on the driven "GM-Spread", where this product-value was upset for the 3080´s to 1:1,6 (1:2 for the E-Types) and the selfmade parts upset on this value for almost 1:1. This value stands for the amount of Current for getting out of the Amplifiers on a given amount of Input-Iabc, - the "Amplifiers-Bias-Current". By using selectable Current-Mirrors (as shown on the attached photos) in that told SOT143´s, one can get quite easy to matched parts, where the bought ICs - not matter if A- or E-Types - always differed!!


The next step was to test the basic construction by the usage of LM3046N DIL14-ICs - Transistor-Arrays, which provided also a properly matched Low-Noise Input-Stage and where three of the five Transistors got used for a Wilson-Current-Mirror, performing and serving the Iabc-Control.
If you choose this configuration and if you can get chips out of one similar production-charge, selecting them, then you can take for secure, that your OTAs will match on the Signal-Control-Products very well and therefore being highly recommended for cascaded and best-tuned VCF-Circuits! The same I tried dito with the UL1111´s from Manufactr. Unitra/Poland, where the selection got only a little, but smaller number of good matching devices, so conclusive to say the LM3046 will stay more "stable" on matching data.
The only restriction to the usage of these Transistor-Arrays is a reduced Powersupply Voltage, which should for that parts not exceed 20 Volts (60V Breakdown-Voltage, as typical for LM3046/CA3046/UL1111).
I tried those IC-Transistors in +/-15 Volt-Supplied-Circuits without any damage, but who knows how long this would work, so anyway better to implement for this usage an additionally Supply modification, lowering the voltages to +/-8 or 9V...
(The LM3046 Transistor-Arrays went for discontinued some years ago by formerly Manufactr. National Semiconductor, but since that Texas Instruments has bought the whole National Semiconductor´s Company, the ICs are now produced again and still available in volume as SO14-SMD-Parts!!)

If you want to perform your own OTAs, it should be highly recommended to use some of those parts in optimized packages, where the competitors directly telling the usage "For Current-Mirrors" in their datasheets, because these products present almost the best data for the required temperature-behavior, and like also recommended for the use of Differential-Amps inside of a common package for the two Input-Transistors. For example, this could get best realized on matched products as the BCM846/847-Series (really smart smd-parts by NXP, Infineon - Europe), or alt. also the DMMT3904 (Lower hFE recommended possible also as SOT23-6 - Diodes-Semic. - USA).
!! The smaller on these temp-facts a transistors package is choosen, the faster it will react on any temperature-dynamics !!
On this conclusion, you can also design and use your OTA as an air-flow-meter...

Selecting and testing a Quad-Set of OTAs for matched VCF-Usages:

This is quite easy, but should get carefully performed and with good measuring equipment and all used resistors with tolerances at 0.1%-1%. Please see therefore the "Breadboarding Schematic-PDF", attached to the project, where also an additionally needed Referenvce-Voltage with 2,45 Volts (ZN458B - alt. LM385/LT1004) is shown.
To test for matched OTAs, the Amplifiers need to perform several different Signal-Inputs to exactly identical Output-Values, and this again also at different values for driven Iabc. The best to control for final statements, this may get performed at three, or four values in different situations.
For a first startup with an Iabc @1 mA (with no Input-Voltage applied and only active the two 120 Ohms-Resistors on Input-Bases) adjust the Output (with TP1) on the 10 kOhms at the Output for zero Volts. Then applying the 2,45 Volts Reference to both Inputs in sequence, should read the same values for positive and negative Voltages (mostly dependent on the hFE of the Input-Transistors and in a second manner on the GM-Product of the used Current-Mirrors, hereby respectively on the one for the Current-Control, which is tied to the Emitters of the Signal-Input-Transistors).
Then change the amount of Iabc to 500 µA (100 µA), and doing the same, should read again the same pos./neg. Values. Now change R1/R2 from 33 kOhm to 10 kOhm and measuring with the same procedure, should again give the results in identical values for each OTA.
The more of such different measurements you may drive, the better control you get for your expected OTAs, and of course this possible test doing also for matching the LM13700/JRC13700-ICs!

Conclusively, on the possible performances for discrete OTAs, important to say is that the herein described project is generally not realizeable for just a few cents, like maybe some standarized Opamp-ICs, what means this will cost money and this tends for rising up, as quality is desired and different products are expected.
If you buy some SOT23-Transistors on a discount-offer, then you just spent those few cents, but that´s maybe ok for a stompbox and shurely will work, but not to say this works quite good (OTA-Overdrive with Discrete-OTAs?), so if you want some real good parts for musical Instruments, or Highend-Audio-Applications, this will of course "cost a little more", and if you buy therefore transistors supplied from suppliers like THAT-Semiconductor, Linear-Systems, or Analog-Devices, for maybe the best performances possible, and using those parts with some experiences from a needed run-up and some works to do, you can really "hit the goals" in all manner -  and hopefully also feedbackin´ on Elektor-Projects...!

Enhancments for OTA-Architectures - JFET-Inputs

Assuming that the todays available OTA-ICs are products on old technologies out of the long ago 70ies, then why not take another step of change on the architectures of that parts to some todays possibilties? In the past there maybe has been no need for such experiments, or maybe such ideas were just too expensive to realize, but however, so why not implementing an OTA with JFET-Inputs, or again a little more crazy - with Germanium-Transistors ?
As an IC-Solution, this would get quite difficult (nothing is impossible!), better say cost-intensive, because here you really need matched parts on chip-level - with biasing-currents around of one milliampere, where the Chip-Technologies for JFET-Opamps cannot easily fix this need, because they are driven only on a few microampere, or less.
To get matched pairs of JFETs, is again now a job on its own, but maybe can get quite easy performed, if one got such an discrete OTA-System with some well-selected Current-Mirrors, and when this System can get into an output-product adjusted to zero.

To be continued soon...

Many Thanks to Todor and Manuela from Bulgaria for providing me with some needed parts on very high quality and prepared since 2013, and therefore overall making this little project possible!

A small number of the above described Parts will soon be available as complete Kits (also well prepared for lots of breadboarding-purposes) and on some good-priced offers on eBay.

The herein described and for the Project used "Doubled-Current-Mirrors" are also useful for my other Projects, like that little "Noise-Generator", "Synthacon-VCF" or the "Ladder-Filters", here on this Website.

Read the full post
Show less
Images
Using Chip-Pin-Strips (SIL9-Leadframes) for BA662-OTAs
Replacing CA3080-Parts with some SOT23-Transistors on handmade Module
High Performance OTA with Dual-Transistor and SIL-Current-Mirrors - 2014
High Performance OTA with LM3046N for lower Supply-Voltages
High Performance OTA with UL1111-Transistor-Array, orig. from Poland
Dozens of SIL-Current-Mirrors prepared for Matched-OTAs
Homemade 3080-OTAs for sale...
DIL08-OTA-Multiple, manufactured by Gebauer Electronic - Filderstadt
Widely differing Offsets, but still getting matched Parts (2012/13)
Parameter-Test with Vero-PCB, CA3080A and 1994 made Module
2´nd Generation of handmade OTAs (2010) - not too bad...
Schematics

Loading comments...