Are my Chinese 2SA970 and 2SC2240s really low noise?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
To fix a Marantz amp I needed high voltage low noise transistors. I could use the standard 2N5401/5551 transistors but they're not exactly appreciated for their noise performance. To put this to the test I built a simple test jig. It emulates a noise figure meter, only it's not calibrated so I can only do comparisons. But that's OK. The jig (schematic on its way) has an input for a reference signal (10 kHz), an attenuator, the socket for the DUT (Device Under Test) and an opamp. The output goes to a spectrum analyzer, although a scope should be possible too.

With the analyzer I can look down to almost DC. A transistor has 1/f noise, which on my analyzer (0-10 MHz) peaks at around 2 kHz. Hence the 10 kHz reference frequency. After all, I must compensate for the amplification of the DUT. So, with my Chinese loot having arrived, I did the tests with six types, 2SA970, BC560C, 2N5401 (PNP) and 2SC2240, BC550C and 2N5551 (NPN). The results are the difference betwen the arbitrary reference signal level and the noise level for my analyzer's bandwidth:
DUT Noise [dBm] Ref [dBm] diff [dB]
2SA970#1 -41 -30 11
2SA970#2 -41 -30 11
BC560C -38 -29 9
2N5401 -39 -35 4
2SC2240 -38 -27 11
BC550C -38 -24 14
2N5551 -38 -29 9

Apologies for the crap formatting but I can't seem to be able to use HTML code... Anyway, the difference means the higher the better. These results are consistent with my expectations and the Chinese transistors are indeed low noise. Now I have to test whether they're also high voltage...

Oh, the amp (PM80 Mkii) just suffered from dry joints.
 
Last edited:
Jig

Attached is the test jig I used. The emitter current is set to 0.1 mA, V(CE) is 6 V. These values are identical to the ones in the datasheets. Because the DUTs have different gain the results must be normalized. I compare to a known good transistor, in this case a BC550C. Turns out it has a better performance than the 2SC2240, which in turn is much better than the 2N5551. But that is no surprise. Funny enough is the result for the PNP counterparts reversed. The Japanese transistor now is better than the European. The 2N5401 is much worse than its NPN equivalent.

Also attached is a screenshot from my analyzer, showing the peak of the 1/f noise and the reference signal at 10 kHz. If this is replaced with a square wave (50 Hz obtained from the power transformer would be perfect!) the display can also be a scope. In this case timebase must be locked to the mains (LINE) and the square wave controlled with an accurate attenuator calibrated in [dB] to be able to compare. This exercise is left to the reader, as they say...
 

Attachments

  • test_jig2.jpg
    test_jig2.jpg
    44.4 KB · Views: 1,478
  • IMGP3518_BC550C.jpg
    IMGP3518_BC550C.jpg
    119.1 KB · Views: 1,445
Last edited:
Not junk

Only one thing is certain: it is junk.
Not so fast. They test excellent, on par with my European low noise devices. If they're really Chinese clones of the Toshiba devices they did a good job. My goal was to test whether or not they really are low noise. The answer is yes. There is one thing left to verify, and that is if they can handle the 120 V V(CEO) as spec'd in the Tosh datasheet. And that is easy.
 
  • Like
Reactions: 1 user
There is one thing left to verify, and that is if they can handle the 120 V V(CEO) as spec'd in the Tosh datasheet. And that is easy.

I got some lovely counterfeit 2SC2240/2SA970. On the picture, the fake ones are the 2 laser-marked on the right. The one on the left is the one off the commercial phono preamp by Azur.

Here are the fake's qualities:

- physically they are significantly thinner than the spec. Originals measure close to 3.8mm or more, fake are about 3.6mm or less, measured with a Japanese and Chinese caliper :) This is something where the Japanese and Chinese agree.

- laser marked, see picture. The link above is the cropped full resolution pic.

- significantly lower break down voltage and HFE, tested with a Tektronix 577 curve tracer

- noisier by at least 20dB than the original part, tested on a commercial phono preamp that uses 2SA970 in its front end. You can hear the higher background noise and also see the higher noise floor on an FFT analyzer

Post a picture of your Chinese 2SC2240s and 2SA970s, they might turn out to be Japanese :D
 

Attachments

  • 5 legit 2 fake 1024.jpg
    5 legit 2 fake 1024.jpg
    269.3 KB · Views: 1,648
No, they're laser marked like yours but test excellent. I'll post piccies tomorrow. They're prolly produced by yet another party and cloned better than your devices. Also TODO: V(CEO) test. Or rather V (BR) CEO. Too bad I don't have that Marantz amp anymore, I'd have borrowed the transistors from it...
 
It's high on my TODO list but I'm pretty busy with other fun stuff. Like tube audio. :D

Hope to have those values by tomorrow. It indeed is clear they're of different manufacture. The leads have more burrs on them too. There's even case material between the leads, your devices are clean.
 
Last edited:
Wow, just wow!

Yeah, I have results! They exceed my expectations. Spec from the Tosh datasheet is V (BR) CE >120 V @ I(c)=1.0 mA. So I hooked up my high voltage supply to a 10k resistor, measure the voltage across that, and connect the other end to my DUT with a second voltmeter across it. Turn up the voltage and stop when the voltage across the resistor becomes 10 V. Then read the second value. Snag was that it takes some time to settle, especially with the 2SC2240. Eventually I got for the NPN 250 and 242 V and for the 2SA970 223, 217 and 221 V. Almost double that of the Tosh spec! This is an Amazing Discovery! (Tell Sell ad). It doesn't happen often, but I'm gobsmacked!
 
Lucky you! That's the problem with unknown transistors: they certainly have unknown performance. Sometimes, it seems, serendipity hits. (Maybe you got pulls, but then again, you'd see solder on the legs).

Smart guys following this thread: what would be a quick/dirty way of getting a noise spectrum from a transistor? Synchronous sampling of a common emitter-connected transistor (with as much gain as you can handle) followed by a known-good transistor in common collector to buffer the input transistor?
 
Regenpak, may I ask what your source is? I need some 2SA970BL transistors with hFE>350, matched within 7 percent for a salas riaa build. I also need matched 2SK369BL and 2SK170BL transistors. If this source is good, I would like to buy a batch of each so that I can match them. Thanks! Andy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.