AB100 Class AB Power Amplifier

Very nice. Raw aluminum. Is that the cheeks on the side? And the stock silver faceplate with some mods? Like some info how you added the HS? It didn't come with them correct?
Yes, that is it, on the side, behind the face plate. I had to add those because I needed the 350 mm internal depth but 350 x 80 x 40 heat sinks are nowhere to find. The face plate is from the stock, drilled by me for the IR window and the standby / on + volume indicator LEDs. The photos in #1531can give you an idea how the heat sinks are mounted. Basically, I drilled threaded blind holes in the the 10 mm thick base on all 4 sides.
 
The 2SD2390 seems to be currently out of stock at digikey but to my memories it is not obsolate, just in-and-out.

Matching the Sanken Darlingtons might be worth the effort. To get my 8 + 8 selection, I started out from a set of 24 pnp and 24 npn devices purchased from digikey. In the measurements I applied 0.5A compliance current between collector and emitter and measured the base current, after about 10 minutes of thermailzation on a massive HS. The I_B values scattered between 70-119 uA (npn) and 64-89 uA (npn). I could finally pick my 8 + 8 devices from the 78-85 uA (npn) and 81-85 uA (pnp) ranges.
 
I also bought some TIP142G / TIP147G Darlingtons, as originally suggested by NP. I measured only 8-8 devices which scattered in much narrower and perfectly overlapping ranges; 99-103 uA (npn) and 96-101 uA (pnp). But in the end I did not dare to use them at my +/-50V rails.
 
The Sanken's are in stock at Chip1Stop.com

2SD2390 = $1.38USD each for 10+
2SB1560 = $2.62 USD each for 10+

That is where I got mine, they offer great service ex Japan. I think it is an Arrow company.

I did not match mine - like all Sanken's out of the same lot, the OS current sharing was better than 5%.
 
The 2SD2390 seems to be currently out of stock at digikey but to my memories it is not obsolate, just in-and-out.

Matching the Sanken Darlingtons might be worth the effort. To get my 8 + 8 selection, I started out from a set of 24 pnp and 24 npn devices purchased from digikey. In the measurements I applied 0.5A compliance current between collector and emitter and measured the base current, after about 10 minutes of thermailzation on a massive HS. The I_B values scattered between 70-119 uA (npn) and 64-89 uA (npn). I could finally pick my 8 + 8 devices from the 78-85 uA (npn) and 81-85 uA (pnp) ranges.
Please post the measuring circuit
A draft circuit by hand is more than enough.😉
 
I usually use the attached sketch (which I would like to properly cite but I cannot find out anymore where it came from) for BJT characterization. Of course this schematics is optimized for milliamperes and normal gains, rather than Amperes and Darlingtons.

So in case of the higher current OS Darlingtons, I removed the Rc and Re resistors; used a voltage source with an 0.5A current compliance; applied a Vce voltage which brought the Ic current to this compliance regime; used R_base = 1k; found a combination of R2 and R3 values (few k and few 10k, I think) such that the resulting voltage divider supplies around 0.7 - 1V to R_base. Then I measured the voltage drop on R_base and calculated Ib. Since the Darlingtons are not that much different, I settled to the some resistor values after a short experimenting, which I of course kept the same for testing all devices. Don't forget to mount the Darlington on a massive HS and allow it to thermailze under constant 0.5A.

I arrived at this approach (constant Ic and measured Ib) because withuot supplying a constant Ic, the latter was rather unstable in the 0.1 -1 A regime. I focused on this relatively high current regime because of two reasons: (1) the Darlington gain is very high and for the sake of more precise measurements, I preferred to deal with 100-ish microamps for Ib rather than something much smaller; (2) I assumed that equal current sharing in the OS is more critical at higher currents than around the few 10 mA constant bias current.

Please note that this was just my naive approach. If it is very wrong somewhere, I hope that a more educated / experienced fellow will speak up.
 

Attachments

  • transistor_matching_bjt_hfe.jpg
    transistor_matching_bjt_hfe.jpg
    13.1 KB · Views: 19