I received the 2 pairs of 21193-4 for $15 with 10 day free shipment. They are heavier than the counterfeit ones with perfect finish. 3 of them measure the same Vbe voltage with diode position on multimeter 461mv, the other 93 measured 477. The simulator model was 60% difference, here is 2.8%. This is important to know that the base resistors 100 ohm 1% is good enough to be symmetrical for ripple rejection due to base to base capacitor.
If I have perfect Vbe matched pair, it is not the case with Hfe. The npn 94s measured both 38_39 but the pnp 93s are 90 _121.
I have also ordered a pair of counterfeit C5359 A1987, they are Hfe matched pair of 168_164. They have Vbe 550_561.
The counterfeit 21193_94 have Vbe522_532, Hfe 161_128.
If I have perfect Vbe matched pair, it is not the case with Hfe. The npn 94s measured both 38_39 but the pnp 93s are 90 _121.
I have also ordered a pair of counterfeit C5359 A1987, they are Hfe matched pair of 168_164. They have Vbe 550_561.
The counterfeit 21193_94 have Vbe522_532, Hfe 161_128.
It is normal for real NPNs like 21194 to measure lower hFE at very low current, and for counterfeits to measure high (at low current). They use a smaller, lower power die in the fakes.
One more test - Take a pair of slip joint pliers over the top hat and give it a gentle squeeze. Real ones won’t even dent. The tops will pop right off most fakes without even really trying.
Indeed, I measured the npn 94 at 100ma, I got Hfe of 68. The pnp which was 90, at 100ma remained 92.
The 2SD669 has 250 vs 2SB649 200. Then the ratio of positive to negative product is 1.69.
Now, there is the early phenomenon which will act inversly and reduce the ratio. Unlike the simulation, the second harmonic will have the same phase as the VAS, adding instead of subtracting.
The DS of njw says they are sold matched paired. I looked the main serious suppliers, no one sells matched paired. My source of A1987,C5359 who also have A1941,C5191 matched paired counterfeit but on Ali, there is another vendor has the njw 0302, 0281 matched paired for about a dollar.
https://www.aliexpress.com/item/1005004148811514.html
Maybe using counterfeit matched paired could result better then original unmatched.
The 2SD669 has 250 vs 2SB649 200. Then the ratio of positive to negative product is 1.69.
Now, there is the early phenomenon which will act inversly and reduce the ratio. Unlike the simulation, the second harmonic will have the same phase as the VAS, adding instead of subtracting.
The DS of njw says they are sold matched paired. I looked the main serious suppliers, no one sells matched paired. My source of A1987,C5359 who also have A1941,C5191 matched paired counterfeit but on Ali, there is another vendor has the njw 0302, 0281 matched paired for about a dollar.
https://www.aliexpress.com/item/1005004148811514.html
Maybe using counterfeit matched paired could result better then original unmatched.
Last edited:
Drive it from a high impedance source, not a low impedance. Which is more realistic of what happens in the complete amplifier. Then you’ll get some even harmonics.
This is also why high gain in the whole output stage is good, and why some of us really like EF3. When the current gain goes to infinity (or just extremely high) distortion gets less sensitive to the actual hFE. Mismatches don't bother it as much. CFPs also have some of this masking effect for output transistor gain (But it still matters in the driver).
This is also why high gain in the whole output stage is good, and why some of us really like EF3. When the current gain goes to infinity (or just extremely high) distortion gets less sensitive to the actual hFE. Mismatches don't bother it as much. CFPs also have some of this masking effect for output transistor gain (But it still matters in the driver).
Hayk,
I think even harmonics are suppressed because each half's transfer function is predominately determined by D1 and R3 (top half) and D2.R4 (bottom half). These 4 components are perfectly matched in your sim.
So the differences between CCS1xCCS3 and CCS2xCCS4 get squashed down by the high loop gains of the CFP.
If you want to model the transistors more realistically then set the 'IS' for D1 to the driver tr models 'IS' parameter divided by the Hfe. Likewise D2 for the other driver. Repeat for the power transistor diodes.
I think even harmonics are suppressed because each half's transfer function is predominately determined by D1 and R3 (top half) and D2.R4 (bottom half). These 4 components are perfectly matched in your sim.
So the differences between CCS1xCCS3 and CCS2xCCS4 get squashed down by the high loop gains of the CFP.
If you want to model the transistors more realistically then set the 'IS' for D1 to the driver tr models 'IS' parameter divided by the Hfe. Likewise D2 for the other driver. Repeat for the power transistor diodes.
You are right, the feedback is reducing considerably the dissymetrical Hfe. I simulated for 24 ohms and 14k sources with equal and unequal Hfe shown bellow. Indeed, the even harmonics get increased but remains unsignifacant compared to odd ones. It is not the case with 14k.
I received $0.6 njw1302 and njw0281, if the finishing of 1302 is excellent the later looks cheap quality. This means there are several counterfeit sources. Strange that the njl1302 3281 are not counterfeited.
My nearby vendor ordered for me 2sc2690. They are labeled NEC, but instead of having the pins ECB, I have BEC. The tap is the emitter!!. This what I call fake component.
I received $0.6 njw1302 and njw0281, if the finishing of 1302 is excellent the later looks cheap quality. This means there are several counterfeit sources. Strange that the njl1302 3281 are not counterfeited.
My nearby vendor ordered for me 2sc2690. They are labeled NEC, but instead of having the pins ECB, I have BEC. The tap is the emitter!!. This what I call fake component.
Attachments
Last edited:
Have a look at this post and bellow #94. The CFP is transformed into diamond buffer.
Post in thread 'Output BJTs for buffer' https://www.diyaudio.com/community/threads/output-bjts-for-buffer.386324/post-7030024
It had to perform according to Sim with Cordell models 0.0015%, in real world he gets @8W 0.048%. According to Cordell models the 21193_94 pair has 4 times higher distortion than the 1302 3281. The source impedance should be the input 100 ohms.
As the Hfe mismatch doesn't effect much, the only character to watch is the early phenomenon and the 21193_94 showed to be the best followed by the njw0302 njw0281 which have identical early which is probably what the DS refers to be more linear than the njl1302 3208.
The mighty mjl21193_94 has economic version costing $4 instead of $5.70 from Mouser.
Post in thread 'Output BJTs for buffer' https://www.diyaudio.com/community/threads/output-bjts-for-buffer.386324/post-7030024
It had to perform according to Sim with Cordell models 0.0015%, in real world he gets @8W 0.048%. According to Cordell models the 21193_94 pair has 4 times higher distortion than the 1302 3281. The source impedance should be the input 100 ohms.
As the Hfe mismatch doesn't effect much, the only character to watch is the early phenomenon and the 21193_94 showed to be the best followed by the njw0302 njw0281 which have identical early which is probably what the DS refers to be more linear than the njl1302 3208.
The mighty mjl21193_94 has economic version costing $4 instead of $5.70 from Mouser.
Last edited:
I found on Ali a shop specialized selling used transistors.
I bought a pair of njw21193_94 for $4.5.
https://www.aliexpress.com/item/1005005575597217.html
For trials and measurements it is good choice for sacrificial tests as SOA protection circuit test.
I bought a pair of njw21193_94 for $4.5.
https://www.aliexpress.com/item/1005005575597217.html
For trials and measurements it is good choice for sacrificial tests as SOA protection circuit test.
I have looked at a minimal changes version of the circuit same components with different values and a few additions.
These can be changed to suit - some of mine are arbritary I am as peckled potato (spectator) on the sideline so I will let the main players get on with the game. This should be capable of handling capacitive loads up to 2uF THD is reasonable without C8 in circuit but the FFT pattern might not be pleasing to all.. As a passing thought I had considered making this a full compound emitter follower by increasing R24 and R25 additions to 100R as this would make setting the standing current easier to do.
These can be changed to suit - some of mine are arbritary I am as peckled potato (spectator) on the sideline so I will let the main players get on with the game. This should be capable of handling capacitive loads up to 2uF THD is reasonable without C8 in circuit but the FFT pattern might not be pleasing to all.. As a passing thought I had considered making this a full compound emitter follower by increasing R24 and R25 additions to 100R as this would make setting the standing current easier to do.