The BA662 is a custom made DC controlled variable transconductance amplifier (or OTA) that Roland had manufactured by Rohm from 1978 and is now obsolete and very expensive. There are two types of chip with different suffices; the A version has been selected for lower offset than the B version. They are further classified by gain/gm. The BA662B can only replace another BA662B, whilst an A can be used in either application.
To quote Roland from the 100M Service Manual: “color has great importance in the circuits of some models”. However there is nothing mystical about the sound of the BA662, it is just a reasonable good OTA chip which is often carefully matched and selected.
Factory Selection The BA662 was used in places where low offset was important, such as VCA’s where a large offset causes a “thump” sound in the audio output. This is where low offset versions are used (BA662A), for example in the SH-09. Low offset is often marked on the top of the chip with a white stripe over Pin 1 or in the middle of the chip.
BA662’s were also selected for close matching of the transconductance (gm) for use in filters and portamento. Filters will not self-oscillate below 400Hz without each OTA being matched for transconductance. This is the gain of the OTA once a load resistor is in place. Roland measured each BA662 and categorised them into 9 grades, ranging from 1 (low gm) to 9 (high gm).
The transconductance(gm) for each chip is marked with a small dot of paint on the top of the chip (in the indent), the codes are:
- Brown
- Red
- Orange
- Yellow (gm =8080)
- Green
- Blue
- Dark green or violet
- Grey or black
- White
I am researching the exact gm value for each colour code. The gm range of the BA662 is not known but looking at the colour codes of BA662’s in synths its clear that the majority are Orange, Yellow, Green and Blue. not many reds and never seen dark green, violet or grey.
Roland required exact matching of colour dot in some applications (e.g. portamento and filters in the Jupiter 4), and within +/- 2 colours for others. This need for measurement and matching in the SH-2, SH-09, ProMars, JP-4 and JP-4 was time consuming. Roland replaced the BA662 in these applications with the precision quad OTA chip, the IR3109. This eliminated matching, and reduced PCB space from 1981 onwards.
The BA662 continued to feature in Roland’s 1980’s products, especially as a standalone VCA or integrated into a VCF and VCA chip (IR3R05), but by 1985 it was gone. It lived on in some Boss pedals and was ultimately retired around 2000, with probably over 250,000 manufactured.
BA6110 The more recent Rohm BA6110 OTA chip is similar to the BA662, albeit with a slightly different pin out, and it too is now obsolete. It is 3-7dB louder in the same circuit, has a low offset (max 3mV) and low distortion (below 0.2% typically) when the linearizing diodes are used.
We have used the BA6110 in many AMSynths circuits (often matched), as they are still available, work well and are low cost. Other designers have used the LM13700 or even dual transistor based clones of the BA662 (which are a lot more expensive than using the BA6110!).
The BA6110FS datasheet shows the transconductance range as 4800 – 12000, with a typical value of 8000. This is a wide range and probably explains my lack of success in using the BA6110 in low pass filter circuits, unless they are matched. The maximum input offset is better than a LM13700 (4mV) but much higher than the AS662D at 250uV.
BA662F The Roland 80017A chip array in the Juno 106 uses the IR3109 in SMD and 2x BA662F 14-pin SMD chips. The BA662F version has the same OTA and buffer as the SIP version but in a surface mount body with the following pin out:
- OTA Control Input
- No Connection
- Negative OTA Signal Input
- Positive OTA Signal Input
- No Connection
- Buffer Control Input
- Buffer Control Input
- Negative Power
- OTA Output
- No Connection
- Buffer Input
- Buffer Output
- No Connection
- Positive Power
BA662F Adapter PCB I have designed an adapter PCB which enables a BA66F SMD chip to be used as a BA662 SIP replacement. The PCB includes an SMD offset trimmer to enable the output offset to be trimmed. Typically we want this to be below 100uV.
The PCB is Open Source and shared on OSH Park here. The trimmer is a Bourns 3314G 50K and the SMD resistor is 1M 0805. The correct turned pin SIL headers can be bought from Rapid Online here. I found the trimmer did not do anything and can be omitted.
The PCB is 10mm x 22mm and will fit into narrow spaces, such as between JP-4 voice cards.
2022 Update Since I wrote this post two BA662 clones are due to hit the market in 2022. The first to arrive was the Rpar AS662 which is a dual OTA chip with the buffers, mounted into a 16 pin SOIC chip. More details and the datasheet are here. The chip came into production in April and I have used it in a variety of AMSynths products.
The second clone is the V662A from CoolAudio, which should arrive by the end of the year. I helped on the development of the chip by sourcing original BA662 chips to be analyzed in China, and writing the draft datasheet and schematic. More details and the datasheet are at CoolAudio here. It is a 16 pin SOIC chip with the same pinout as the AS662, the offset is 3mV max and the gm range is wide. This chip will certainly need selection for use in filter circuits.
I have created an adapter PCB for taking either chip and replacing a BA662 chip. This uses only one side of the new OTA’s, so it’s not suggested for new designs which can make full use of the dual OTA’s being on the same chip with close matching. The adapter PCB is shared on OSHPark here.
AS662D Experience I made extensive use of the low offset (250uV) AS662D chip in various AMSynths designs during the Spring of 2022. The chip works very well but it does require the transconductance(gm) to be measured, and chips selected with similar gm, when used in filters. The gm range is relatively narrow; 7700 to 9100 with a center point at 8500. The AS662 is therefore a higher tolerance OTA, and should be easier to find matches.
If they are not matched the filter will refuse to self-oscillate at lower frequencies, typically below 400Hz. In the SH-09 service manual Roland say that BA662 chips are ok if they can self-oscillate down to 50Hz. With close matching of gm the AS662D will actually go down to 20Hz.
BA662 Modes (added 2022) When using the BA662 (or AS662) it is important to know how to power the internal buffer, as it does not like being left floating or the input grounded. There are two modes for powering the buffer.
- Mode A – The OTA output is connected to a JFET or Op Amp buffer directly and the internal buffer is not used. Pin 4 is to -VE and pin 7 to +VE. This configuration is typical for a VCA, HPF and LFO.
- Mode B – The internal buffer is used and pin 4 to GND and pin 7 is connected to Pin 6. This is a typical VCA application.
However there are examples (JP-8 Controller IC16) where the buffer is unused and pin 7 is floating with Pin 4 to ground. The important configuration to avoid is Pin 4 to ground, when the buffer is not used.
AS662 and BA662 Test Harness The schematics are here.
I came across this site in a Google search. I’m interested in a JP04 oscillator and JP04 filter. Is any this actually available to buy?
Hi
The JP04 VCO and VCF are in development and will be launched around April time
Rob
Great thank you!
Remarked (or refurbished) BA6110 on your photo.
We made clones of the LM13700. I know you are going to say “well that’s easy”, but we did a lot of research and select the clones on our dedicated tester, just like Roland did. So, we can supply clones that fit at least on 90% of the positions the original BA662 needs to replaced.
https://m.facebook.com/story.php?story_fbid=2125381720877632&id=826416797440804
Thanks for the info, great idea! I may well need an accurate BA662 clone if the ones in my JP4 Voice Card have failed. Rob
Regarding the SMT BA662 pinout, there seems to be one extra functional pin compared to the SIP component. An extra buffer control input? If one were to adapt the SMT chip to replace the through hole part, how would this extra pin be best handled? Tied to the other control input? Tied to negative power rail? Left floating?
I was planning to leave floating and see how it performs
Thanks very much for this article, for your work on these chips and synthesizers in general, and for the 2022 updates regarding the new AS662 and V662A devices!
Thank you for this excellent info and the new SM chips.
My question:
How can i go about matching the forward transconductance of an AS662 adapted replacement in my SH-2 which has yellow paint?
Breaking down this question a bit:
Can you share a test circuit for measuring transconductance?
Do i need to order lots of chips and test them til i find a match or can i adjust the bias current to achieve a matched gm for my synth?
It is wise to match the gm on the SH-2 to ensure you get self oscillation down to 20Hz. You need to measure the gm of at least 20x AS662 chips, and match to the gm of a yellow dot BA662.
I have a test circuit and PCB layout, which I will post tomorrow. You cant adjust the bias current, as its going to all 4 chips.
Thank you
Hi, interesting article, thanks!
I replaced the 80017A chip arrays in my Juno 106. I’m thinking to have a go at using the old ones in a project, some are still operational.
Do you know control voltages are they expecting to receive on the control inputs?
Thanks,
I made the AM8106 module from 80017A chips. I can share the schematics and might even have a PCB.
Hello,
I don’t know if I can ask you a few questions, you’re probably more of a salesman than answering questions from strangers!
I’ll try anyway.
I wanted to recreate the Roland SH-1 filter with the alternative BA6110 chip. 4 tuned BA6110s are needed with the smallest possible offset, so I have a few questions:
1.) How and where on which pins do I measure on the BA6110?
2.) What is the maximum value that can be satisfactory?
3.) What is the maximum difference that can be between the other BA6110s (tolerance limit)?
4.) How many BA6110s do you think I should buy to get 4 identical ones?
…. Or would you advise against building the SH-1 and instead building the SH-2 with CA3080?
If you answer, I would be very pleased, thank you in advance!
Best regards
Maik
Hi Mark,
The SH-2 filter can be built from BA662’s, BA6110’s, AS662’s, LM13700 and V662’s. The key to an accurate SH-2 (or Jupiter 4 Mark 1) VCF is to ensure the resonance will self oscillate down to 20Hz. This is achieved by matching the transconductance (gm), of each chip to within a few percent. This means building a test rig with a precision voltage source and a digital multimeter that can read uA’s. Then each OTA can be measured and matched. This is easier with the chips that have one OTA like the BA6110, as the dual chips can have OTA’s that are never going to match, so you end up wasting one side of the chip. Typically you will need 25+ chips to get 4 that match, within 150 gm. This why Roland used color coding, so that large batches of chips could be measured and coded on their gm value. If you build an SH-1 or SH-7 filter you will need to match the CA3080 chips. WIthout matching the filter may or may not self oscillate at low frequencies, depends if you get lucky on their gm value! If you need the schematics for a tester I can send you one.
Best Regards
Rob
Thank you for your feedback!
I would be VERY happy to receive a circuit diagram from the tester!!!
I’m serious about building the SH2.
I have the “conventional” vcf’s as:
Diodes or transistors ladder filter or LM13700 or Ca3080 or CEM3320.!
But a BA6110 filter is not included yet!
I once needed a BA6110 to build a TR808 module and bought the chip on Aliexpress (China dealer), and everything worked.
Or would you rather advise against the Chinese chips for the SH2 (they certainly won’t be original)?
Otherwise I would buy 30 pieces from China (€35 including shipping). Which would be good price wise!
So 150gm is still a good value, ok.
Best regards
Mike
… Oh, I meant I wanted to build the SH2 and only if impossible the SH1 with ca3080 🙂
… I have another question: Is it also possible to buy 4 coordinated AS622F or BS6110 chips?
Hello Rob,
Happy New Year 🙂
I really don’t want to steal your time and come to the end of my questions, but could you please tell me exactly what the exercise should look like, from the original BA622 to BA6110, in the SH-2, preferably a small sketch or in words,
(… for example: Pin1 from BA622 to Pin3 BA6110… etc.)?
The fourth and last of the four OTA’s in the SH-2 also has a slightly different circuit! Can you help?
The translation I found on the internet for the TR 808, for example, is something else I also found for VCAs!
I wanted to create a circuit board, but I totally lack this basic knowledge. Unfortunately I can’t find any information about it on the internet.
You seem to be the only person in the world who knows about this!
You would help me incredibly!
Best regards
Mike
The BA6110 and BA662 pinouts are here:
https://obsoletetechnology.wordpress.com/repairs/roland-tr-808-repair/
The SH-2 VCF has Q15 as a buffer BJT on the 4th OTA cell. If you use BA6110 chips the gain will be a couple of dB higher than the BA662 but that wont be a problem. The AS3380EA chip is a potentially better matched OTA out of the box, but hard to find except via Rpar themselves. I will post the schematic of the transconductance test circuit here later this week Unfortunately I don’t have the time (or chips in stock) to hand match 50x BA6110 chips to get 4 that are close enough to use in an SH-2 VCF clone. AMSynths will be re-issuing the SH-2 VCF in 2015 probably with the AS3380EA. Cheers!
Hello Rob,
Here I am the pain in the ass again 🙂
Are you thinking about releasing the schematics for the tester when the time suits you? I would be very happy!
I read by chance that you also built the SH-5 filter and replaced the “ITS1276” with an LM394.
I built the System 100 VCF (SH5 VCF identical) and it works very well, I replaced my two adapted 2C945GR for ITS1276 with AS194 (new improved version of the Lm394), unfortunately when I inserted the AS194 the resonance disappeared completely ( E , B, C was aligned correctly).
Did you have to change anything so that the Lm394 could work correctly?
You have really great, interesting modules, I will buy from you as soon as you can supply the EU again!
Unfortunately some modules are sold out.
Greetings Maik
Hi, The test harness schematics have been added to the post. Rob
Hi,
If the pinout is correct (check with a transistor tester) then it’s the gain structure that’s wrong – best to measure the AS194.
What are you using at TR520 ? A high gain BC547C works well.
You can also play around with R544, as it sets the gain of the CA3080.
Hello Rob,
Thanks for your help, I got it working with the adapted transistor on the SH5 VCF 🙂
To use the BA6110 you referred to this link:
https://obsoletetechnology.wordpress.com/repairs/roland-tr-808-repair/
…. when I connected it 1 to 1 like this, I got a short circuit on the negative voltage (BA6110 got hot).
I then disconnected the line from pin 3 (BA6110) to pin 4 (Ba662).
So pin 3 of BA6110 is now exposed!
I didn’t try to connect pin 6 to 7, otherwise the external buffer would have no function.
The filter works quite well so far, I also have feedback, but not perfect yet.
Will try more BA6110.
I had thought that 5R6K and 2R2K were supposed to be inductions, but they are supposed to be resistors. I also used 470p instead of 470N.
Do the 2SK30A have to be adapted (4 identical)?
You may have time to reply again.
Mike
Yes 5R6K is 5K6 and 2R2K is 2K2 resistors.
The JFET’s do not have to be matched, a they are used as buffers.
The OTA’s do need matching.
Hi,
Hoping you wouldn’t mind answering one question for me? Is the BA662F closer in spec to the BA662A or BA662B? I just bought a BA662F for an RE-303 build and I’m hoping it will be a better choice than the clones that are out there.
Thanks
Nick
The BA662F is the SMD version and was used in the VCA part of the Roland D80017 hybrid chip. It is equivalent to the BA662 but with a different pinout.
They are NOT selected/color coded so they would need matching in a VCF circuit that uses 4 of them.
In a TB-303 clone the BA662A is used in the VCA and you can replace it with a BA662F as long as the pinout is adjusted.
The offset of the F chip maybe apparent and need a trimmer, the original circuit has no trimmer.
I would give it a go and see if you have any VCA “thumping”/offsets in the audio output.
An AS662D with 250uV of offset is probably going to be a better solution.