NJW0281/0302 output devices: I read many times in this forum, that this devices are factory selected for 10% gain difference. So I have 140 pcs. of NPN, and same number of PNP devices. No fakes (all ordered from Mouser, which is good source, I believe).
So I want to select pairs for my amplifiers. I measured all devices with Vce=5V, and Ic=100mA. This current is close to the bias, I want to use.
Just for fun, I tried with Ic=10mA, and Ic=1A as well, but there was no significant difference, so the the gain is quite flat.
The NPNs have really bad results on this test. The Hfe disperse from 84 to 122. The low gain is more typical, there are only few high gain (over 110) device found. But more than 50% of the transistors are below 90
The PNPs are very good: the minimum gain is 97, and the maximum is 118. But only 10 found over 110, and only one less than 98. So I can tell that the 10% is true for PNP devices!
The final result is 56 pairs from Hfe= 98 to Hfe=110. The pairs are within 1%.
Sajti
Very useful, thanks!
How many pieces do you suggest me to order to have 6 matched NPN or PNP (i'm thinking quasi) I guess more than 100 (each)...
NJW0281/0302 output devices: I read many times in this forum, that this devices are factory selected for 10% gain difference. So I have 140 pcs. of NPN, and same number of PNP devices. No fakes (all ordered from Mouser, which is good source, I believe).
So I want to select pairs for my amplifiers. I measured all devices with Vce=5V, and Ic=100mA. This current is close to the bias, I want to use.
Just for fun, I tried with Ic=10mA, and Ic=1A as well, but there was no significant difference, so the the gain is quite flat.
The NPNs have really bad results on this test. The Hfe disperse from 84 to 122. The low gain is more typical, there are only few high gain (over 110) device found. But more than 50% of the transistors are below 90
The PNPs are very good: the minimum gain is 97, and the maximum is 118. But only 10 found over 110, and only one less than 98. So I can tell that the 10% is true for PNP devices!
The final result is 56 pairs from Hfe= 98 to Hfe=110. The pairs are within 1%.
Sajti
The Hfe disperse from 84 to 122 is also a good result.
In my view the variation of the individual Ube voltage is more critical regarded the THD results (not possible to investigate by simulation)
This Ube voltage variation causes different idle currents through each output devices.
Please correct if i'm wrong - this is a problem also in class A biasing, right?
If You use NPN only, 25-30pcs would be enough. If I have to make NPN pairs only, I can make 50-55 pairs with the 140 pcs of NPN devices. If You remember, 50% of the devices are between 84 and 90. You have about 35pairs if this gain is enough.
Sajti
It's interesting, because I measured absolutely same Ube on the output transistors of the prototype. And I did it without any selection.
But I will check the Ube!
Sajti
I measured the bias within 15%, without transistor matching, with Re=0.22R, in the prototype
Sajti
Sajti,
there is a huge advantage of using MOSFETs as a driver stage - they do not transfer load variations to VAS. You may get a progress of one order when you use them as drivers for 8 pairs. This is a well meant suggestion ...
Best,
Sajti,
there is a huge advantage of using MOSFETs as a driver stage - they do not transfer load variations to VAS. You may get a progress of one order when you use them as drivers for 8 pairs. This is a well meant suggestion ...
Best,
Thanks PMA!
So I use tube VAS for this amplifier. But would You recommend any MOSFET I can try? (And costs less than 5Euro each
)
Sajti,
there is a huge advantage of using MOSFETs as a driver stage - they do not transfer load variations to VAS. You may get a progress of one order when you use them as drivers for 8 pairs. This is a well meant suggestion ...
Best,
But is this really a problem ?
If the VAS is loaded then it will push harder from the LTP to move the output voltage.
If it is getting to the point where the VAS has run out of current then you need a larger current through the CCS.
If You load the VAS, the loop gain will fall. This means less feedback, and higher distortion.
Sajti
The feedback is always the same Rfeedback/ Rtoground.
I am pretty sure the LTP gain plus the VAS gain will always be sufficient for the feedback to stay the same.
Most BJT amps use a driver stage anyway which would get rid of the problem.
My estimate is yes; this could be even the reason, why some guys claim, the best way it is, the paralleling of a bunch complete amplifiers (i. e. include the front end) instead only the output devices. This is often to observe by IC amps like TDA7293/7294.
I start for this question the follow thread:
http://www.diyaudio.com/forums/soli...voltages-parallel-power-bjts.html#post2147084
No. This is the closed loop gain, if the open loop gain is high enough.
In properly designed amplifier yes. But all amplifier has higher distortion with 4ohm load compared to 8ohm.
Can't eliminate the problem. Only try to reduce
Sajti
I want to try also all PNP (with today technology is a possibility) and the classic complementary. The gain is no issue but distortion is (no global feedback). Maximum number of devices to be used would be 6 pairs (i mispelled before, sorry), so 12 NPN, 12 PNP matched with themselves and 6+6 also matched with each other.
Will i go fine with 100+100 pieces?
I think that 100+100 must be enough. But for first I would by 50+50. And if it not enough, You can buy more.
This is not easy. Currently I thinking to build some SE class A amplifier, with NPN output devices, to use the remaining parts.
Sajti
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.