GB NP matched Toshiba 2SK1530/2SK201 for F5X

[pchw]

OK, I will help to get this kicks started to see there will be any traction.
For the time being, there is no organizer yet. The silver anniversary with my love one is coming up, we will be doing some traveling in May. Therefore, I will not be the one in near future.

A first things to start, here are the names expressed interest to this GB.
So, I copied them over.
For the F5X, the most desirable is NNNNPPPP matched, next will be
NNNN + PPPP or NNPP + NNPP (??)
I am including the most desirable matching to everyone, but feel free to
edit your name to add your next choice. If ,let say, there is not enough
NNNNPPPP or none at all, that will be the default.

NicMac 8 pairs (2 stereo F5X) (NNNNPPPP x 2)
CanAm Man 4 pairs (1 stereo FX5) (NNNNPPPP)
ramallo 4 pairs (1 stereo FX5) (NNNNPPPP)
oneaclock 4 pairs (1 stereo FX5) (NNNNPPPP)
andrewt 4 pairs (1 stereo FX5) (NNNNPPPP)
Studley 4 pairs (1 stereo FX5) (NNNNPPPP)
ashaw 8 stereo sets? (NNNNPPPP x 8 ??)
triode4 4 pairs (1 stereo F5X) (NNNNPPPP)
heinz1 4 pairs (1 stereo F5X) (NNNNPPPP)
Mravinsky 4 pairs (1 stereo FX5) (NNNNPPPP)
JBdV 8 pairs for 2 F5X (NNNNPPPP x 2)
smyslow 4 pairs (1 stereo F5X) (NNNNPPPP)
APK 4 pairs (1 stereo F5X) (NNNNPPPP)
Horio 4 pairs (1 stereo F5X) (NNNNPPPP)
Westley 4 pairs (1 stereo F5X) (NNNNPPPP)
alazira 4 pairs (1 stereo F5X) (NNNNPPPP)
pmchoong 4 pairs (1 stereo F5X) (NNNNPPPP)
CeeVee 4 pairs (1 stereo F5X) (NNNNPPPP)

There are a few factors to consider:
1. How to source the genuine Toshiba? Maybe Davide can help sourcing
from Japan. If so, this must be decided, start, and closed pretty soon
because he is moving to Europe.
2. How big the pool is enough? Just the above list will require 26 sets of
NNNNPPPP. If you look at ths picture from Zhou's thread
He probably bought 1000 each, that means total of 2,000 devices.
Yet, he got very bad yield from his last batch and raised his price.
Notice that his matching method is less stringent that what Patrick's
3 point match, not mention the curve tracer match. If NP match is this
difficult to obtain, NNNNPPPP will be likely quite a few times harder.
3. This maybe the most important to consider, what if there is not enough
NNNNPPPP match? In another thread, I posted that I bought 10 random
pairs from a US seller, the N and the P were not even close. Therefore,
I am not sure there can be any guarantee of such quantities of match
sets. The group must agree upon a "what if" strategy as this will be
a lot of money up front.

I hope this will help get the GB started.

Express your preference and acceptable alternatives here.

Cheers,

[ashaw]

For the PPPP and NNNN vs PPNN x2. you want the latter as distortion and DC offset depends on the match within one channel. The main benifits between matching PPPPNNNN is balance between amplifiers, mainly matching gain.

[buzzforb]

FWIW, I have looked and are currently looking for these. They are not widely available through most channels. I would be very careful, for without access to factory boxes, the risk of counterfeit skyrockets.

[EUVL]

Here are my personal opinions. So just take it as that :

1) Davide has bought these devices in JP for XEN Audio (so not from Zhou). They come in original Toshiba boxes.
2) To be practical, you should aim for NNNN PPPP. Distortion cancellation is still possible within the X.
3) A good matching pool would be 200~500 of each device type.
4) Allow up to 50% matching reject rates if you only want quads like NNNN. Pairs are easier.

Some additional background info & discussions :

F5X -- the EUVL Approach
F5X -- the EUVL Approach
F5X -- the EUVL Approach
F5X -- the EUVL Approach
F5X -- the EUVL Approach


There will be different levels of match naturally, not only within the Ns & Ps, but especially N to P.
So you need to agree upfront on different price levels (e.g. 1% match, 5%, 10%, etc.).
And who is prepared to paid premium price for best match, who not.
I think this will be the one single point that will invite a lot of debate.

Someone has to do the matching.
And if you want to match properly (at current and bias under heatsink temperature), even single point match is a huge task.
The last time I did this with a friend was in 2005 when we built the Aleph-X. We matched 200 2SK1529 curve traced.
The measurement alone was 8 hours non-stop (2 persons). The data processing another day.
No easy task and not very entertaining.


Patrick

[ashaw]

Though I am in a non-ideal location to do matching, I am happy to offer my services to do data processing. Someone else will have to do the measurements though.

Regards

[pchw]

First, update the list:
NicMac 8 pairs (2 stereo F5X) (NNNNPPPP x 2)
CanAm Man 4 pairs (1 stereo FX5) (NNNNPPPP)
ramallo 4 pairs (1 stereo FX5) (NNNNPPPP)
oneaclock 4 pairs (1 stereo FX5) (NNNNPPPP)
andrewt 4 pairs (1 stereo FX5) (NNNNPPPP)
Studley 4 pairs (1 stereo FX5) (NNNNPPPP)
ashaw 8 stereo sets? (NNNNPPPP x 8 ??)
triode4 4 pairs (1 stereo F5X) (NNNNPPPP)
heinz1 4 pairs (1 stereo F5X) (NNNNPPPP)
Mravinsky 4 pairs (1 stereo FX5) (NNNNPPPP)
JBdV 8 pairs for 2 F5X (NNNNPPPP x 2)
smyslow 4 pairs (1 stereo F5X) (NNNNPPPP)
APK 4 pairs (1 stereo F5X) (NNNNPPPP)
Horio 4 pairs (1 stereo F5X) (NNNNPPPP)
Westley 4 pairs (1 stereo F5X) (NNNNPPPP)
alazira 4 pairs (1 stereo F5X) (NNNNPPPP)
pmchoong 4 pairs (1 stereo F5X) (NNNNPPPP)
CeeVee 4 pairs (1 stereo F5X) (NNNNPPPP)
lhquam 4 pairs (1 stereo F5X) (NNNNPPPP)

To all potential participants, you can copy the list and append your name
or edit your entry.

Second, I am not the organizer at the moment. There is no organizer yet.

I mentioned to have Davide to source from Japan at the beginning because that is the only reliable bulk source that I know. I honestly won't spend time to look for another one.

Patrick is right on the money and was to the exact point that I didn't manage to express. I am almost sure that 500 each type will not yield enough NNNNPPPP matches from the above list, not even close. So, tier pricing is almost a must, but how to set the price is going to be quite task. Just for reference:

Zhou's P/N match is $45 per N/P now. Therefore, NNNNPPPP will be $180.
Zhou's precision match of NNNN + PPPP is $64 ($16 x 4).

If anyone wants to go for only NNNN + PPPP, then Zhou's price is quite good, considering I paid $7.50 each before matching to the US reseller.

[EUVL]

I can confirm that 7.50 USD is as good a price you can get anywhere, including from JP.

But Zhou does not match at 16V Vgs and 2A Id at 60°C Tc.
Perhaps you could get a few sets from Zhou at $64 and then rematch yourselves using my 3-point method.

F5X -- the EUVL Approach


Patrick

[pchw]

Patrick,

LOL, I did mean to suggest to buy a few N's and P's and perform another match, but I forgot, a senior moment :-)

[EUVL]

One good solution is to see whether you can buy NNNNNNNN & PPPPPPPP from Zhou.
Then you have a better chance to find good NN-NNs & PP-PPs.




Patrick

[AndrewT]

I raised a concern about N to P match when we started out on the regulator mosFETs.

I was told we would have no difficulty getting the matches we need @ the 4A Id.
This turned out to be impossible. On average the N Vgs was ~200mV lower than the P Vgs.

Fortunately, These were for regulators. The solution is dead easy. Simply select the Zeners to have a 200mV mismatch to correct the error. The problem goes away.

For the N to P match in the Push Pull Amplifier, I reckon we have no hope of getting N to P matches in significant numbers and there is a chance we will get zero yield of N to P matches.

Forget that best case option. Decide what we need for good F5X without that best case option.

NN & PP for a single channel using 3 pt centred on 2A Id (2A +-1A) may be good enough.
precision matching between channels is worthless in my view. Get both channels working properly and within each channel good enough 3pt matching and fairly close selection of NN to NN and PP to PP between channels will in my view yield a good stereo performance, (note, I said "selection", not matching between channels). Euvl may completely disagree, but let this discussion proceed.

In summary for a single stereo F5x I think we need
NN, NN, PP, PP, i.e. 4pairs for the stereo amplifier. Each pair 3pt matched at 1A, 2A & 3A Id.