diyAB Amp - The "Honey Badger"

[ostripper]

By Tsip - if Q5-Q6 can be replaced by BC560 i guess that BC550C would be ok as well what about Q9? Can we replace with BC560 or BC550?
thanks

Q9 sees nearly the full rail voltage , might want to keep a high Vce device (2sa992) here.

OS

[keantoken]

Hfe matching won't help. Vbe matching is what's important for a current mirror. Most BC5x0C I've measured are matched to 300uV. That's a just over a 1% mismatch. 1% at 3mA is 30uA. And this is without degeneration! Now lets consider temperature to make sure.

Q5 has about 1V Vce whereas Q6 has .66V. This corresponds to 3mW and 2mW respectively. The thermal resistance from junction to ambient of the average TO92 is 200, so this calculates to a temperature difference of .2C. The tempco of silicon is about 2.1mV/C, so we have a 420uV mismatch. So this leaves us with a 720uV total mismatch. This is 2.8%, corresponding to a mismatch of 84uA. If the cases are thermally coupled, differential thermal resistance is about 160 corresponding to a mismatch of 336uV.

And this is still without any degeneration.

Assuming an Hfe of 500 for the BC560C, mismatch caused by Hfe will be roughly 12uA. So improving matching by a factor of 8 with equal Vce and degeneration is the point of diminishing returns. BUT there is 220R degeneration in the original design. So lets factor this in. 720uV mismatch across 440R total degeneration means our Vbe-induced mismatch is reduced to 1.6uA - .005%! Despite mismatch and temperature variations, the biggest flaw with our current mirror is not a transistor mismatch. Of course, the degeneration tolerances will directly affect matching, so they are in fact the weakest link. Two 5% tolerance resistors have a potential mismatch of 10% - almost 4 times the max mismatch without any degeneration at all!

So, match R20/R21 first and only then worry about Vbe mismatching with the BC5x0C. You can check Vbe matching with your multimeter at the diode setting by placing the probes across the B-E junction and making sure they both match to 1mV (4%) or so. Let them sit a while and don't touch them with your fingers when you do this.

If you have matched R20/R21, you can verify transistor matching by probing across the emitters of Q5/6 with your DMM. It should be less than 1mV; short your DMM leads to make sure it can read DC below 1mV. Don't have a speaker connected while probing around in the amp. Even if you don't slip or drop a probe, touching the sensitive signal nodes may cause oscillation which could be dangerous for your speakers. A good idea is to use 10k-100k resistors in series with both probe leads, as this will block the shunt parasitics of the DMM.

[wangyungpi]

Thanks OS for clarifying about this amp, some members exibit blown transistors, maybe because of faulty parts. This amp is real God like..!!.. if built properly and nicely...

[naf]

wow thank you kean, your expalanation will take time for me to fully understand.
but when i build this honey badger i try to match as i could following part
q1/q2
q7/q8
q5/q6
r3/r6
r8/r9
r15/r16
r20/r21

[niss_man]

Originally Posted by ostripper

Member Keentoken also has a point about the current mirrors.

Replacing Q 5/6 would be simple , only the pair would be back-back instead of face-face. This improved current balance would indeed reflect on the sonics .. even if only by a small amount. This is explained in Hugh Dean's AKSA threads.

This is wrong information. Q5/6 are installed facing the same way on the board. Also KSA992 are ECB layout and BC560 CBE layout so it is not as simple to mirror them around as one has Collector center pin and other has a base center pin.

[naf]

Originally Posted by niss_man

This is wrong information. Q5/6 are installed facing the same way on the board. Also KSA992 are ECB layout and BC560 CBE layout so it is not as simple to mirror them around as one has Collector center pin and other has a base center pin.

correct niss_man, i have to bend my bc560 leg and isolate using heatshrink tube.

[ostripper]

Sorry for wrong info.
To redeem myself , here is a pin for pin equivalent:
http://www.mouser.com/ds/2/149/KSA1015-61942.pdf
KSA1015.

These match the BCxxx's typical characteristics almost exactly. http://www.mouser.com/ds/2/149/BC556-193550.pdf
Use the GR gain group. Buy extra's to match as Keen suggests (.10c at mouser).

OS

[keantoken]

That's a great find! They are probably even better, as you can see in the Fairchild BC560 datasheet there is a droop at 1V/50uS, but not for the japanese part. Only drawback is the lower Hfe. I will have to try these in my Kmultipliers whenever I have the chance.

Unfortunately the NPN part, KSC1815 doesn't show Vcesat comparable to the BC550. Possibly, the BC327-40 and KSC1015 are the ultimate BJT pair for low-Vce linearity. If Ft match is desired, BC550/KSC1015 both should still work great. However I want to test this to make sure.

[ostripper]

I was a little rusty(wrong info). But since I was going to order some (ksa1015's) , I thought it should be "shared".

OS

[danielwritesbac]

Can you see if this is correct?

R7+R8 constrained the dial to block overcurrent settings (did I do it right?)
R29+R30 constrained the dial to block overcurrent settings (did I do it right?)
R27 changed per information in this thread (and from there I just guessed)
Q5,Q6 changed per information in this thread (does it affect Q3, Q4?)
Q14,Q15 changed to avoid severe mismatch (onsemi's low HFE new China fab NPN)
Q16,Q17,Q19,Q20 per information in this thread, 50,50v rails, 8 ohm speaker
Marked all higher current resistors
C4, changed to a reasonable value for higher resolution bass with non-ringing treble
D3 indecipherable, "different compensation options" indecipherable and therefore these need checked out too.