It is difficult to see from the pic, but drivers and VAS tranies seem to be 2SB649/2SD669. These are good parts, and I would keep them. Not so low Cob as some, but still good overall.
For 1815, I found this http://www.mouser.com/ds/2/149/KSC1815-59515.pdf Spec is just about the same except for the middle lead is base instead of collector. I suppose I can just bend it to fit the board. 😕
Try KSC1845 instead of 1815 🙂
Sajti
I thought there was something odd about that datasheet. look at this, also from Fairchild......KSC1815 Pinouts , KSC1815 Pin out,KSC1815 pin diagram 😕For 1815, I found this http://www.mouser.com/ds/2/149/KSC1815-59515.pdf Spec is just about the same except for the middle lead is base instead of collector. I suppose I can just bend it to fit the board. 😕
Also, I should have posted the complement to 2SA970, ie. 2SC2240 and that is almost the same as satji's suggestion, KSC1845. These are all listed in Greg's chart, #39
I thought there was something odd about that datasheet. look at this, also from Fairchild......KSC1815 Pinouts , KSC1815 Pin out,KSC1815 pin diagram 😕
Also, I should have posted the complement to 2SA970, ie. 2SC2240 and that is almost the same as satji's suggestion, KSC1845. These are all listed in Greg's chart, #39
The 2SC2240 2SA970 is not manufactured any more, unfortunately. Only smd version is available. It's not easy to get original Toshiba version now.
KSC1845/KSA992 is available from Mouser for peanuts. I have ordered 1000pcs each.
Sajti
Thanks, sajti. I checked this earlier this year and the "not suitable for new manufacture" was still showing. Well, it seems to happen to every good thru-hole part. 

Well, it seems to happen to every good thru-hole part.![]()
Unfortunately 🙁 I was the bad guy, who ordered the whole 2SC3423/2SA1360 stock from Mouser (300pcs each), and get the last stock of 2240/970 from Newark. It looks, that we have to make our own stock for the future, to keep our hobby up, and running...
Sajti
I noticed that after posting my finding yesterday. I was going to check the pinouts with DMM before installing anyway. 🙂I thought there was something odd about that datasheet. look at this, also from Fairchild......KSC1815 Pinouts , KSC1815 Pin out,KSC1815 pin diagram 😕
Thanks Ian and sajti.
Bump.
Also replaced 5401 and 1815. Now DC offset is 2 mV. It was a cheap (they are 10 - 15 cents each) and fun learning experience. 🙂
Also replaced 5401 and 1815. Now DC offset is 2 mV. It was a cheap (they are 10 - 15 cents each) and fun learning experience. 🙂
Good, I honestly thought you would have more difficulty getting a match there. Was it the current mirror pair that made the difference?
It was 5401 pair (Q15 & Q16) that made the main difference. When I replaced 1815 pair (Q11 & Q13, Q17 & Q19), DC offset still remained at 14 mV.
So here's the DC offset track record:
1) -24 mV out of the box stock.
2) 15 mV when I replaced 1015 pair (Q6 & Q8).
3) 14 mV when I replaced 1815 pair (Q11 & Q13, Q17 & Q19)
4) 2 mV when I replaced 5401 pair (Q15 & Q16)
So here's the DC offset track record:
1) -24 mV out of the box stock.
2) 15 mV when I replaced 1015 pair (Q6 & Q8).
3) 14 mV when I replaced 1815 pair (Q11 & Q13, Q17 & Q19)
4) 2 mV when I replaced 5401 pair (Q15 & Q16)

Last edited:
Ah..so it was the cascodes, then. (Q15,16) Well, I wouldn't like to try to design and prove that input stage - what a beast! 'nice to be satified with the effort though.
My latest modules turned up yesterday and I noticed some very new looking 2SD669/B649 drivers. Sadly, these parts, last made by Renesas, have been out of production for more than a decade. They have Hitachi marks but the plastic moulding is not as originals I have. I won't state the obvious - they may work fine anyway if that type was also used to prototype the L12 design in the first place - a factor a lot of DIYs ignore as we are often preoccupied with building older designs based on OEM components, when ranting over Ebay traders peddling counterfeit parts.
There are still a few reasonable quality T0126 drivers available such as Fairchild KSC2690/A1220m so there are OEM alternatives if max. rail voltages and heavy duty are envisaged. Average home duty is probably not likely to raise problems even with substandard parts so some DIYs will never know the difference, which is what I guess some sellers are banking on.
Oh well... 🙁
My latest modules turned up yesterday and I noticed some very new looking 2SD669/B649 drivers. Sadly, these parts, last made by Renesas, have been out of production for more than a decade. They have Hitachi marks but the plastic moulding is not as originals I have. I won't state the obvious - they may work fine anyway if that type was also used to prototype the L12 design in the first place - a factor a lot of DIYs ignore as we are often preoccupied with building older designs based on OEM components, when ranting over Ebay traders peddling counterfeit parts.
There are still a few reasonable quality T0126 drivers available such as Fairchild KSC2690/A1220m so there are OEM alternatives if max. rail voltages and heavy duty are envisaged. Average home duty is probably not likely to raise problems even with substandard parts so some DIYs will never know the difference, which is what I guess some sellers are banking on.
Oh well... 🙁
Yeah, my next project (if I find time & money) will be a bare board with my own shopping of parts from reputable supplier.
Yeah, my next project (if I find time & money) will be a bare board with my own shopping of parts from reputable supplier.
My latest modules turned up yesterday and I noticed some very new looking 2SD669/B649 drivers. Sadly, these parts, last made by Renesas, have been out of production for more than a decade. They have Hitachi marks but the plastic moulding is not as originals I have. I won't state the obvious - they may work fine anyway if that type was also used to prototype the L12 design in the first place - a factor a lot of DIYs ignore as we are often preoccupied with building older designs based on OEM components, when ranting over Ebay traders peddling counterfeit parts.
I found 669/649 at Profusion plc, made by Unisonic of Taiwan. They claim that these are of very good quality.
In fact Unisonic parts are 669AL C and 649AL C.
A=higher collector emitter rating (160V)
C=hfe rank (mid rank 100-200)
A=higher collector emitter rating (160V)
C=hfe rank (mid rank 100-200)
Yes, Profusion have a small but really useful selection of standard thru-hole audio parts.I found 669/649 at Profusion plc, made by Unisonic of Taiwan. They claim that these are of very good quality
I've never used Unisonic semis but there is little to argue about from their data sheets and users seem satisfied with results in the field - that's what counts if they are pro. audio users who likely purchase large quantities of parts and use them in many different designs, so they will more likely be giving components a real workout before endorsing them.
There are still a few reasonable quality T0126 drivers available such as Fairchild KSC2690/A1220m so there are OEM alternatives if max. rail voltages and heavy duty are envisaged. Average home duty is probably not likely to raise problems even with substandard parts so some DIYs will never know the difference, which is what I guess some sellers are banking on. Oh well... 🙁
I have some KSA2690A/KSA1220A in stock (about 150 each.), but their sound quality not as good as I expected

But now Mouser stocked some new (former Sanyo) drivers, which looks promising. I will get some 2sc4614/2sa1770, and 2sc3902/2sa1507 to try. Hopefully they will be good.
Another possibility to try the well known Toshibas as driver (2SC4793/2SA1837 or 2SC5171/2SA1930). They are still cheap, and available, the only disadvantage that we have to redesign the pcb.
Sajti
That's a pity, Sajti. I know these are a bit sluggish compared to other drivers but I have used them as substitutes for BD139/40, MJE243 MJE171,81 etc series to good effect. To qualify that, I should say that I often adjust frequency compensation of the amplifier to restore the sound quality, particularly this makes sense when used for the VAS transistor, but sometimes drivers too.I have some KSA2690A/KSA1220A in stock (about 150 each.), but their sound quality not as good as I expected....
It will be very interesting to see what On-Semi have done in cloning 2sc4614/2sa1770, and 2sc3902/2sa1507. Please, do keep us posted.
Since summer is coming, at last! I'm planning some new diy projects😀 I thought it would be nice if I could mount 4 of these boards into one case and bridge them. Is this possible or will they go up into smoke?
Duco
Duco
There's no compelling reason not to, provided they are first stable into a 4R or lower load. Each amplifier "sees" half the load impedance so it works at higher current to deliver the power increase. I would get them operating singly into a low impedance first, before adding a bridge adaptor and driving them hard. It's one problem to burn an amp on test and have to replace a few transistors but quite a big deal if you take ot four amplifiers at once due to some consistent error.
Yes, I have made that mistake and with someone else's equipment. too! 😱
As a caution though, I would make very sure there was no sign of instability (bursts of oscillation following signal transients, square wave ringing etc) that sometimes is present on CFP stages, depending on components choice even. You need a 'scope for this, but it doesn't need to be a particularly fast one for observing this. The bigger problem is making or obtaining a high power, low inductance dummy load of 4R/8R for the tests.
Note that the heatsinks some suppliers mount the wired modules on are really only for testing at low power. I doubt you could use these at more than 25W RMS, even if bias is lower with CFP designs (~13mA bias/output pair)
Yes, I have made that mistake and with someone else's equipment. too! 😱
As a caution though, I would make very sure there was no sign of instability (bursts of oscillation following signal transients, square wave ringing etc) that sometimes is present on CFP stages, depending on components choice even. You need a 'scope for this, but it doesn't need to be a particularly fast one for observing this. The bigger problem is making or obtaining a high power, low inductance dummy load of 4R/8R for the tests.
Note that the heatsinks some suppliers mount the wired modules on are really only for testing at low power. I doubt you could use these at more than 25W RMS, even if bias is lower with CFP designs (~13mA bias/output pair)
- Home
- Amplifiers
- Solid State
- L12-2 CFP Output amp 120W*2 8R