Charles Hansen said:
Hello Charles,
Please call me Chuck (Mr. Hansen is my father).
🙂
'Chuck' is the name I try to use in the audio world, so I hope it solves the confusion problem. audioXpress sometimes uses Charles, my fault I suppose since I have that name in the boilerplate of my articles for the magazine. I'll have to change that. I also put a note on the back of my latest book "A Brief History of Bendix Red Bank Tubes" that I was not you or Charles Hansen the music publisher.
I chat with John Curl once a month or so, and I have lately been following a few threads he suggested such as this one and Mosfet vs. BJT. I am amazed at the depth of knowledge that all you DIYaudio guys possess.
Best Regards,
Chuck Hansen
chascode said:
Hi Chuck,
Thank you very much! I think that will go a long way to straightening out what has been an understandably confusing issue.
chascode said:
John is a great guy and in my opinion, the "godfather" of modern solid-state audio design. There were a few other people working with complementary circuits back then, but none were as original as John, and he single-handedly brought JFET's (especially complementary JFET's) into the consciousness of audio designers.
In a way it is lucky that people like Mark Levinson cheated Mr. Curl. John then printed the schematics as a way to get some small measure of revenge, and that is where I learned so much -- just by studying a handful of John's designs. I spent countless hours analyzing them and learning from them. All of my designs are just extensions of the work that he began. I am very gratified that it has come full circle, as I have shared some of my ideas with John and he has incorporated some of them into his designs as well.
Sigurd Ruschkow said:
I am afraid that Grey is correct. The 2SJ74 is being discontinued by Toshiba. I don't know whether it is the country or the different grade of part, but we have been informed by Arrow USA that March 31 is the last date to place orders for the 2SJ74BL.
This is very bad news indeed. We will have to place a large order, as there seems to be no other replacement parts available now or in the future.
Arrow USA does not normally carry this part, so don't waste your time trying purchase from them unless you plan on buying 10,000 pieces or more. For the hobbyist, I would recommend going through one of two places:
http://www.bdent.com/
http://www.solidstateinc.com/
These suppliers have 2SJ74BL in stock at reasonable prices and enough quantity for hobbyists. Both of them are very reputable and I can vouch for them. They are US-based companies that have been in business for at least 10 or 20 years. They do not sell fakes.
Charles,
good to know! That was for the BL versionI read. My date in Sweden was for the SJ74V.
I remember that Ayre bought a huge stock of 389/109 (or was it 3381/1349) and it is now time for the SJ74.
I am stocking SJ74 and also SK170.
No need to hunt down the 389/109 anymore...been there, done that.
Sigurd
good to know! That was for the BL versionI read. My date in Sweden was for the SJ74V.
I remember that Ayre bought a huge stock of 389/109 (or was it 3381/1349) and it is now time for the SJ74.
I am stocking SJ74 and also SK170.
No need to hunt down the 389/109 anymore...been there, done that.
Sigurd
Charles Hansen said:
I checked the Toshiba data base in Japan, and 2SK170 / 2SJ74 have not been listed as inactive YET.
Moreover, 2SK370 / 2Sj108 seem to be replacement devices. The datasheetsare dated 1st Nov 2007, and if I remember correctly, Nelson Pass bought a whole bunch and used them for the F4 power follower.
So maybe it is not doomday as yet afterall.
Patrick
PS The datasheets are too large but this is the link :
http://www.semicon.toshiba.co.jp/openb2b/websearch/start.jsp
Moreover, 2SK370 / 2Sj108 seem to be replacement devices. The datasheetsare dated 1st Nov 2007, and if I remember correctly, Nelson Pass bought a whole bunch and used them for the F4 power follower.
So maybe it is not doomday as yet afterall.
Patrick
PS The datasheets are too large but this is the link :
http://www.semicon.toshiba.co.jp/openb2b/websearch/start.jsp
EUVL said:
Patrick,
I'm sorry to be so pessimistic, but I don't agree. Th K370/J108 are identical dies to the K170/J74, but in a smaller ("Mini") package instead of the TO-92. The only difference is the max Pd, listed at 200 mW instead of 400 mW. I would be completely surprised if these are still in production.
Please refer to the link provided by Grey at the very first post in this page. This is the general catalog for 2008. There are only three P-channel parts still listed -- the J103 (complement to the K246) and the identical die in two different SMT packages -- the J106 and the J144. (This die is *not* the low-noise, high-transconductance part like the J74.)
I am afraid that Grey is correct, and that there are only a few weeks left to order the J74 parts. There appears to be no replacement in the pipeline. I have very little faith in the small US companies, InterFET, and Linear Devices. They have been trying for years to make parts that are true complements with low-noise and high-transconductance, and to the best of my knowledge, without success.
Vishay recently sold the JFET intellectual property of Siliconix, I believe to Linear Devices. But none of the Siliconix parts are suitable replacments either. This is a bad day for high-performance audio.
Nelson Pass said:
Hi Nelson,
Glad to hear from you, and I am curious as to your meaning. It is not doomsday because you know of a source for these parts, or it is not doomsday because you don't care to use them?
Thanks,
Charles Hansen
I am in the semiconductor industry and work for one of the big guys in the field. I've been reading this thread, especially the latter part with some interest.
Unfortunately, I do not see a bright future for JFETS other than in RF applications (canned tuners) and even then, its position is under threat from IC solutions.
Bottom line is many of the functions that in the past could only be done with discrete FET's, can now be done on IC's because IC technology is very mature and the solid state physicists and process specialists in the fabs are able to do some truly remarkable things. This of course does not help audio amplifier designers.
The biggest cost in discrete semi development is the process. For a mos FET process, figure on about $500k to $1m and about 18 months work for 15-20 people. Since a JFET process is much simpler, th e cost is lower, but still a t least $100k to $200k. A process is a 'recipe' that the production guys can use to manufacture saleable parts with a high yield (normally >98% for discretes) with very repeatable performance.
Once you have the process in place, you can then design actual products (= piece of silicon that meets a defined spec, packaged, with spec sheet and you can produce it in volume and get it right at least >98% yield). Critically, your part mus t pay back in 2 years or less. Even for th e process, you would need to get payback in 2 years or less.
There are no discrete JFET audio applications I am aware of that will deliver this kind of payback. Audio traditionally piggybacked on all sorts of other applications . . . . which have now migrated to IC based solutions, or, gone digital.
So, I expect because of this that we will gradually see the pool of available JFETS decline over the next few years as semi companies take the decision not to develop new JFET processes, because the functions performed by discrete devices are required less and less. Keep in mind that a semi fab is an extraordinarily expensive undertaking and the people that run these have to squeeze every last penny of profit out of them to keep them running.
I still see that there will be a few small niche vendors (like Linear) but for how long they can keep the processes alive I don't know.
Sadly, the sun is setting on the JFET, so stock up while you can or look for alternative solutions. I guess for the JFET aficcionados, it will probably be back to tubes which seem to have stayed alive - but I think this is because setting up and running a tube factory is much less expensive than a semiconductor fab. Just to close with some numbers: A fully depreciated 6 inch semi-fab will for for about $20m. A new 8 inch discrete fab will cost about $300 million. If we talk 12 inch state of the art, it about $1Billion.
🙂
Unfortunately, I do not see a bright future for JFETS other than in RF applications (canned tuners) and even then, its position is under threat from IC solutions.
Bottom line is many of the functions that in the past could only be done with discrete FET's, can now be done on IC's because IC technology is very mature and the solid state physicists and process specialists in the fabs are able to do some truly remarkable things. This of course does not help audio amplifier designers.
The biggest cost in discrete semi development is the process. For a mos FET process, figure on about $500k to $1m and about 18 months work for 15-20 people. Since a JFET process is much simpler, th e cost is lower, but still a t least $100k to $200k. A process is a 'recipe' that the production guys can use to manufacture saleable parts with a high yield (normally >98% for discretes) with very repeatable performance.
Once you have the process in place, you can then design actual products (= piece of silicon that meets a defined spec, packaged, with spec sheet and you can produce it in volume and get it right at least >98% yield). Critically, your part mus t pay back in 2 years or less. Even for th e process, you would need to get payback in 2 years or less.
There are no discrete JFET audio applications I am aware of that will deliver this kind of payback. Audio traditionally piggybacked on all sorts of other applications . . . . which have now migrated to IC based solutions, or, gone digital.
So, I expect because of this that we will gradually see the pool of available JFETS decline over the next few years as semi companies take the decision not to develop new JFET processes, because the functions performed by discrete devices are required less and less. Keep in mind that a semi fab is an extraordinarily expensive undertaking and the people that run these have to squeeze every last penny of profit out of them to keep them running.
I still see that there will be a few small niche vendors (like Linear) but for how long they can keep the processes alive I don't know.
Sadly, the sun is setting on the JFET, so stock up while you can or look for alternative solutions. I guess for the JFET aficcionados, it will probably be back to tubes which seem to have stayed alive - but I think this is because setting up and running a tube factory is much less expensive than a semiconductor fab. Just to close with some numbers: A fully depreciated 6 inch semi-fab will for for about $20m. A new 8 inch discrete fab will cost about $300 million. If we talk 12 inch state of the art, it about $1Billion.
🙂
Bonsai,
If you note my codename and where I am, then you also know I am in semicon, like you.
The processes for all current existing JFETs are well developed, so there is no real development cost required. And if there is any write-off, it has probably long been done. So for now, it is just production cost. But of course the equipment, and perhaps more importantly fab floor space, can be used to make something else that brings more revenue per die. And JFETs simply do not bring revenue. I guess the same can be asked about any discrete devices, like small signal bipolar transistor. But they probably find more applications than just audio.
The only devices that are "safe" in that sense are switching power MOSFETs. We would still have IRF9610's and IRFP240s or variants of them for a long time to come.
And there will be small companies serving a niche market, like Sanken, Semelab, .... etc. Most of them however do audio power devices, which again bring 10x the revenue for about the same effort than JFETs.
Perhaps Audiophile companies should start making their own transistors, or getting them make (there is already one example in Australia ?).
Patrick
If you note my codename and where I am, then you also know I am in semicon, like you.
The processes for all current existing JFETs are well developed, so there is no real development cost required. And if there is any write-off, it has probably long been done. So for now, it is just production cost. But of course the equipment, and perhaps more importantly fab floor space, can be used to make something else that brings more revenue per die. And JFETs simply do not bring revenue. I guess the same can be asked about any discrete devices, like small signal bipolar transistor. But they probably find more applications than just audio.
The only devices that are "safe" in that sense are switching power MOSFETs. We would still have IRF9610's and IRFP240s or variants of them for a long time to come.
And there will be small companies serving a niche market, like Sanken, Semelab, .... etc. Most of them however do audio power devices, which again bring 10x the revenue for about the same effort than JFETs.
Perhaps Audiophile companies should start making their own transistors, or getting them make (there is already one example in Australia ?).
Patrick
Bonsai said:
Thanks for the insider info. Presumably Toshiba has no further use for the process they have developed for the J74. What do you think the chances are that they would sell this process to another company? Is that something they would be happy to make some money on, or do you think they consider it to be totally proprietary.
Because if I understand your post correctly, presumably Linear Systems or InterFET could purchase the "obsolete" process from Toshiba and start making clones in very short order. Any insight would be appreciated.
EUVL said:
The Australian company is having Semelab (Magnatec) make a variant on their existing lateral MOSFET designs. I don't know what the minimum quantity is for such a variation.
Several years ago I requested that they make a scaled down version of their lateral MOSFET that would be comparable to the Hitachi (Renesas) TO-220 parts (2SJ77/2SK213). At that time they said the minimum order to do so would be $100,000 worth of parts. Currently it is easier for us to simply purchase the Hitachi parts, but if they cease production then $100,000 would be feasible investment for us.
Today I got a last time buy notification from Arrow Nordic saying that the 2SK170V has a last time buy date 2008.06.16
and that the suggested replacement is the 2SK170-V(F).
I see this as a good sign as Toshiba is thus making the SK170V into a PB-free variant and is not stopping the process. Instead they adapt the SK170V to be a more modern part that customers demand.
Sigurd
and that the suggested replacement is the 2SK170-V(F).
I see this as a good sign as Toshiba is thus making the SK170V into a PB-free variant and is not stopping the process. Instead they adapt the SK170V to be a more modern part that customers demand.
Sigurd
Attachments
Charles,
> At that time they said the minimum order to do so would be $100,000 worth of parts.
That is based on a part which they already make. Try to get them to do 2SJ109's for you for $100k ....
I cannot explain to you in public so please just take my word -- semicoductor companies will never sell their processes. They only start to collarborate these days for highest spec processes (e.g. 45nm process for making processors and memories) because no one company can afford the development on its own anymore, except maybe Intel and Samsung. And even then, they collarborate only between equals.
Why should Toshiba sell their knowhow to Linear ?
How much can Linear afford to pay ?
Sad, but true.
Patrick
PS I still hope that Nelson would answer your question.
> At that time they said the minimum order to do so would be $100,000 worth of parts.
That is based on a part which they already make. Try to get them to do 2SJ109's for you for $100k ....
I cannot explain to you in public so please just take my word -- semicoductor companies will never sell their processes. They only start to collarborate these days for highest spec processes (e.g. 45nm process for making processors and memories) because no one company can afford the development on its own anymore, except maybe Intel and Samsung. And even then, they collarborate only between equals.
Why should Toshiba sell their knowhow to Linear ?
How much can Linear afford to pay ?
Sad, but true.
Patrick
PS I still hope that Nelson would answer your question.
Charles Hansen said:
I am happy using these or their equivalents. Parts similar to
them will be around in the future, even if we have to get them
made ourselves.
In the meantime, we have to live off our humps, as we did during
the Mosfet famine in the 90's.
😎
Nelson Pass said:
Thats a great idea mr. Pass.
Why not make them ourselves?
This guy makes his own tubes:
http://dailymotion.alice.it/video/x3wrzo_fabrication-dune-lampe-triode_tech
This storm has probably nixed a trip to Akihabara (it killed my free day in front of customer visits) but I wondered if anyone knows the chance of stocking up on any of these obsolete devices here these days?
scott wurcer said:
Hi Scott,
It depends on which parts you are looking for. At this point the dual monolithic JFETs were discontinued 3 years ago. You can probably find a handful of parts, but expect to pay $10 to $20 each for them. There was a group buy recently for $6, but they turned out to be fakes.
Toshiba has announced they are discontinuing the J74 (complement to the K170). The last time to order is March. Purchasing information is in an earlier post:
http://www.diyaudio.com/forums/showthread.php?postid=1398409#post1398409
The K170 remains in production (for the time being), as do the K246/J103 complementary pairs.
- Status
- Not open for further replies.