2SK170 on Ebay

I just bought a bunch of 2SK170 on Ebay. To hedge my position I purchased from 4 different vendors. Then I measured them to see if they were legit and what sort of Idss and Vp they are. Here are the results:



So they all look pretty legit near as I can tell. Just not much variety. Note that the hkutsource parts were NOT cut leads as suggested by the "used" condition.

UPDATE:

I created a simple test circuit to try to evaluate the authenticity, noise and performance of these different batches of "2SK170". It makes ~40dB more or less depending on the transistor but I did adjust the input level so that the output was ~-2.0dBV on the QA400 so that the plots could be compared for distortion and such. I can post the cirucit go into the method details if necessary but here are the spectra. If you open these images in separate tabs you can switch between them such that the differences are clearer:

ceitron (GR part):


alweit:


giorgio11185:


hkutsource 1:


hkutsource 2:


I did two of the hkutsource transistors since that's the one that's the 50 for $9.50 deal (giorgio11185 are the same as hkutsource and even have the same '3F' code on the transistor).

I have no idea if any conclusions can be made about noise. After doing this it's not obvious to me how to go about measuring noise of a JFET.

Note that you can see what the differences in gain are by taking the difference of the Gen 1 values. For example, alweit is -3.3 and hkutsource is 3.0 (meaning I had to reduce input for alweit to match level of hkutsource) which is a difference of 6.3dB. Meaning alweit is 6.3dB higher output so even though hkutsource noise floor looks to be ~2dB lower the SNR at ~1kHz (if that makes sense) is 4.3dB better with alweit and ceitron parts.

The -BL and -GR  are graded for Idss.

Presumably the BL may be slightly quieter.

I used some 2SK117 toshiba parts in my last phono preamp front end (decades ago) and they were nice parts.

JR

2sk170's even from legit places like mouser, etc will be different. We spend hours matching them in pairs just make sure the circuits we use them in perform the same.

Well for $9.50 USD you can get a whole bag of 50 that are already within about 3%. They're all in the 10-11mA range but that's probably just great for a lot of things.

I used to sell matched pairs of N jfets for my old P-10 preamp...  I didn't match for same Idss at  0V Vgs, but matched for same Vgs in my specific preamp circuit.  I measured a couple hundred, then put the closest pairs together.

Using unmatched parts worked, but the THD baseline was measurably higher.

JR 

squarewave said:

Well for $9.50 USD you can get a whole bag of 50 that are already within about 3%. They're all in the 10-11mA range but that's probably just great for a lot of things.

Click to expand...

Where do I have to look for that? At that price I´d instantly grab a few bags...

jensenmann said:

Where do I have to look for that? At that price I´d instantly grab a few bags...

Click to expand...

The "Listing Desc" column is the exact text of the listing. Use that to search.

Note that that listed was Item Condition: Used. I expected salvaged parts (meaning cut leads) but they were not. They are all completely unused near as I can tell.

Thanks!
I´ve instantly ordered. Usually I´m a bit hesitant buying from chinese sellers because you never know what you really get. Fake products are all over the place. But in this case having a confirmation that these transistors seem to be the original part is great.

You still don't know what you get. There are plenty of JFETs with similar IDSS that could have been relabelled. Only real proof would be a noise test.

I got very good results (noise wise) with 2SK170 from Ampslab (2$ each).
For very low noise apps 2SK369 may be considered also.

volker said:

You still don't know what you get. There are plenty of JFETs with similar IDSS that could have been relabelled. Only real proof would be a noise test.

Click to expand...

Interesting! I will have to do a noise test now.

But a few things to consider:

1) The gain is right so it's not just IDSS but also Vp that is correct.
2) The seller has >100k transactions so it seems foolish to jeopardize those credentials for such a small profit.
3) Why go through the trouble of faking a transistor to sell it for 19 cents?

> Why go through the trouble of faking a transistor to sell it for 19 cents?

The major Nickel company got an order for an odd alloy. So odd, they called the Treasury. Or a paper-boy got a coin that didn't feel right. Agents descended. This guy had prepared the stampers to make his own US 5-cent nickels at home, and turned-out many thousands. It appears his costs were 6 cents each. Thrown in prison (again- he'd done $5 bills before) and vanished from history.

PRR said:

> Why go through the trouble of faking a transistor to sell it for 19 cents?

The major Nickel company got an order for an odd alloy. So odd, they called the Treasury. Or a paper-boy got a coin that didn't feel right. Agents descended. This guy had prepared the stampers to make his own US 5-cent nickels at home, and turned-out many thousands. It appears his costs were 6 cents each. Thrown in prison (again- he'd done $5 bills before) and vanished from history.

Click to expand...

Only the US government makes money that cost more to make than it's nominal value.

1 cent cost $0.015

1 nickel cost $0.08

the rest is worth more than it costs to mint.

JR

Updated original post with QA400 spectra of each batch.

Have followed this with interest. You may have run across this too, but I stumbled on this EE lab procedure for measuring JFET 1/f Noise and corner frequency for the purpose of comparing measured to theoretical values.  After a casual read I can't distill a simple procedure  that will suit your goal of comparing relative noise levels and spectral content, but maybe someone smarter than I am could.

https://leachlegacy.ece.gatech.edu/ece6416/f09/JFET_Noise.pdf

BT

> a simple procedure  that will suit your goal of comparing relative noise levels and spectral content

Simple basic JFET gain-stage, opamp booster (with gain fixed after initial setup), headphone amp. Or JFET gain-stage to a high-gain mixer with phone jack. Leach's circuit is fancy but good.

95% of a batch will be all the same. A few will have more hiss or deep rumble.

I'm not sure moving everything up another +40dB with an op amp is going to reveal anything new. The noise floor of the QA400 in loopback is lower than the test circuit and that's all that really matters. The QA400 is perfectly cabable of measurements at -130dB.

After thinking about this a little I think the hkutsource / giorgio11185 parts (which are the same exact parts) are clearly "different" from the alweit and ceitron parts which have spectra that are nearly identical and yet their IDSS/Vp is very different and they come from completely different sources. Even though the noise floor of the hkutsource / giorgio11185 parts is technically lower by ~2.5dB, the gain is also 6.3dB less. In other words if the gain were boosted 6.3dB to match the alweit and ceitron parts, the SNR would end up being 4.3dB less. So more gain without a corresponding increase in noise means the alweit and ceitron parts have significantly better noise performance (about 4.3dB better).

Having said that, they're not bad transistors. Their gain is pretty high compared to just about anything commercially available (2N3819 only in the 2-6.5mS range). It would be difficult to tell the difference between a genuine 2SK170 and one of these hkutsource / giorgio11185 parts.

squarewave: Did you have any genuine ones tested too? I got some, if you want to test the real deal as well