Discrete AD797 up and running

[nelsonvandal]

Originally posted by Sigurd Ruschkow
Nelson,

the full schematic I have of the AD797 is similar to what you have drawn but differs a bit.

I would add cascoding of the input stage JFETs.

I would also try some other BJTs for the output stage. Sanyo has some very good medium current BJTs that you might want to try.




Happy New Year!



Sigurd

Happy new year to you too! Without your help with SA1016/SC2362, I'd never be happy with the amp. I'm still very grateful. So what Sanyos are you proposing? I can't fit anyting larger than TO-126 on this board, but if I lay out a new one, I'll make room for TO-220. I've tried SA1538/SC3953, but they can't take even 20 mA of current.

Regarding cascoding the input, I have six fairly matched SK246 waiting for some experimenting. I just haven't tried it yet. It'll be very hard to improve on the sound of this amp since it sounds so very close to perfect. I think I'll experiment with the AD829 clone first since it sounds grainier, and improvement should be easier to hear. I'm also going to try CFP input and maybe the combination of CFP and cascodes.

[AndrewT]

Originally posted by nelsonvandal
I've tried SA1538/SC3953, but they can't take even 20 mA of current.

I don't understand. In what respect do they fall down when pushed to 20mA?

[nelsonvandal]

Originally posted by AndrewT
I don't understand. In what respect do they fall down when pushed to 20mA?

I can't understand it either. When biased to 20 mA the sound was colored. Like a glaze over everything. With heavy loads (10R), it started to clip at moderate levels. When I lowered the bias to something like 6 mA, the sound was less colored, but it was still clipping at moderate levels when loaded down.

Maybe it has to do with the low voltage supply, 10 V, but when using BD137/BD138 (ONSemi) there are no problems at all. I know the BD's sound slightly darker than neutral when I've used them in open loop diamond buffers, so I had great hopes for an improved sound with the Sanyos. I've used BC639/640 from ONSemi in this amp but they sound muddy in comparison. I don't know if I should try BC639/640 from NXP. I'm quite happy with the sound as it is at the moment.

[Lumba Ogir]

Nelson,
we have discussed this matter recently, you have a very short memory, don’t you? The Sanyo devices are inappropriate for that job, just like Mr. Carr insightfully pointed out. Use 2SC5171 or 2SC4793, if not possible 2SC3421 or 2SC2824.

[AndrewT]

Originally posted by nelsonvandal

I can't understand it either. When biased to 20 mA the sound was colored. Like a glaze over everything. With heavy loads (10R), it started to clip at moderate levels. When I lowered the bias to something like 6 mA, the sound was less colored, but it was still clipping at moderate levels when loaded down.

10ohm loading!!!
What happened to 300ohms?

6mA bias allows almost 12mA peak output for ClassA working.
That's just 120mVpk maximum output (0.7mW into 10r). Then it goes into ClassAB working and from the graphs starts to go off @ about 90mA. That's only 900mVpk and equivalent to just 40mW of ClassAB output into 10r.

[nelsonvandal]

Originally posted by AndrewT
10ohm loading!!!
What happened to 300ohms?

That was just for testing it's abilities. It should be able to handle any kind of headphones, including 16R IEMs. Listening tests is done with 300R Sennheiser HD650. There's no sonic benefits of using the Sanyos before the BD's even when driving a light load such as the HD650.

[AndrewT]

I don't know what signal level you use to drive the 300ohm headphones, but if one applies the usual worst case speaker loading to your headphone amplifier you will find that your +-5V supply could possibly be stretching the followers even into a 300ohm reactive load.

eg.
maximum output current ~Vpk/Rload/0.35.
for 4.5Vpk output into 300ohm speakers the maxmimum current could be ~= 4.5/300/0.35 ~= 43mApk. This is equivalent to ~33mW into 300r.
To remain in ClassA the bias would need to be set to >22mA.
Now the followers will be running warm to hot. Apply the temperature de-rating and you are really stretching these 100mA/200mA devices.

If your actual loading were 300r (resistive) then the operating conditions on the followers are in a completely different ballpark.

Decide what you're designing and then select devices appropriately.
Is this discrete opamp for 300r or 300ohm or 10r or 10ohm?

I note you kept headphone amplifier as the duty from us until post17, naughty!

[scott wurcer]

Originally posted by nelsonvandal

That was just for testing it's abilities. It should be able to handle any kind of headphones, including 16R IEMs. Listening tests is done with 300R Sennheiser HD650. There's no sonic benefits of using the Sanyos before the BD's even when driving a light load such as the HD650.

Looks good Nelson. I had the same idea when I started that discrete OPA627 thread which I am still working on. I was gifted with some HD600's.

You may look into cascoding T1 and connecting the collectors of the first drivers in the output stage back to the output. This works in IC's but some discretes don't like running that close to saturation.

With high bias you might not find a lot of benefit to the distortion neutralization.

[Sigurd Ruschkow]

I was thinking of some video Sanyo BJT:s as I thought you were making a line amp, but as you want to drive low Z headphones my suggestions no longer are valid. For low Z headphones I would try the 2SA1837/2SC4793 as Jonathan suggested.

You can use the SA1016/SC2362 for cascode work, too.



Sigurd

Originally posted by nelsonvandal

Happy new year to you too! Without your help with SA1016/SC2362, I'd never be happy with the amp. I'm still very grateful. So what Sanyos are you proposing? I can't fit anyting larger than TO-126 on this board, but if I lay out a new one, I'll make room for TO-220. I've tried SA1538/SC3953, but they can't take even 20 mA of current.

Regarding cascoding the input, I have six fairly matched SK246 waiting for some experimenting. I just haven't tried it yet. It'll be very hard to improve on the sound of this amp since it sounds so very close to perfect. I think I'll experiment with the AD829 clone first since it sounds grainier, and improvement should be easier to hear. I'm also going to try CFP input and maybe the combination of CFP and cascodes.

[maxpou]

Hi,
why not try 2SA1930/2SC5171, faster then 2SA1837, lowest cob and more current? Maxpou

Happy new year!