I understand that people want to stock JFETs if they have a production to secure. But for new designs, is it not better to move on and use parts that have not been discontinued for years?
John Curl, I also sense that Toshiba will discontinue the SJ74/SK170, so stocking up on those is a wise idea if one wants to play the "JFET game".
Still, in 10 years time maybe there will be no more JFETs for audio use, so what do we do then?
Maybe it is better to face the facts that there will be no more JFETs say in 10 years and already today start designing with other parts (and have a 10 year head starton the competitors).
Or hope that some company will have audio JFETs when Toshiba's are not available anymore?
Sigurd
John Curl, I also sense that Toshiba will discontinue the SJ74/SK170, so stocking up on those is a wise idea if one wants to play the "JFET game".
Still, in 10 years time maybe there will be no more JFETs for audio use, so what do we do then?
Maybe it is better to face the facts that there will be no more JFETs say in 10 years and already today start designing with other parts (and have a 10 year head starton the competitors).
Or hope that some company will have audio JFETs when Toshiba's are not available anymore?
Sigurd
MikeBettinger said:
In the current project I'm working on with Edmond we are using 2SA1407/2SC3601 in the VAS stage at about 6mA (about 300mW dissipation, running without a heatsink) and in the EC power stage, as differential amplifier, at about 10mA.
As VAS, I have experimented 2xparallel 2N5401/2N5501, 2SB649/2SD669, 2SA1381/2SC3503 and nothing came close (in terms of measured performance) to 2SA1407/2SC3601 at whatever current between 5 and 20mA.
In the class AB EC power stage they made a huge difference in terms of measured THD20+N, the current open loop THD20+N is under 100ppm. With any other trannies, the open loop THD20+N was not better than 170ppm, both measurements done after re-tuning the Hawksword balance.
Many here are working on projects that really go back decades. Bipolar transistors as an advanced current mirror can be very linear and interesting, but to replace fets with bipolars directly is a step backward. Of course, 30 years ago, we had no choice, and had to use bipolar transistors with input fets, but only Charles Hansen has had any real success in doing it lately.
For those of you who are worried about fets, just think of them as you would vacuum tubes. They have been obsolete for 40 years. That doesn't stop enthusiasts from using them.
Next, people will revert to IC's.
For those of you who are worried about fets, just think of them as you would vacuum tubes. They have been obsolete for 40 years. That doesn't stop enthusiasts from using them.
Next, people will revert to IC's.
jacco vermeulen said:
Yes, not the same distortion performance.
If you look in the datasheet, Cob stands in the same ratio as the max collector current, that is, about 2/3. 2SA1406/2SC3600 are only 100mA devices. Which means they are most likely a scaled down (in area) version of the same design. And as a result, if you look at the Ic-Vce curves, they have a lower Early voltage for the same beta. Which means higher distortions which is exactly what I was able to measure.
Which doesn't mean that 2SA1406/2SC3600 are not very good devices for a high performance VAS. It's just not the same... As usual, you get what you pay for.
AndrewT said:
I have limited experience with 2sa1360/c3423. Tried them once and they are indeed very good, an excellent bang for the buck. I guess it's a matter of requirements for your project... YMMV, but for the ultimate performance there are better options.
Strictly based on the datasheet, 2SA1540/2SC3955 look better (larger ICmax, larger VCEo, slightly larger Early voltage, higher Ft) but I haven't had a chance to experiment with. Do you have a reliable source for these?
My favourite for regular projects and best bang for the buck are the 2SA1381/2SC3503 or KSA1381/KSC3503 from Fairchild. Mouser sells these for dirt cheap, at 25 cents a pop in 10's. As far as I know, Fairchild is about to discontinue them, as they did with many other japanese clones, so it's probably time to buy... I got 100 of each recently.
Maybe the same story for the A1406-A1407 and counterparts applies to the A1540/C3955 duo and the higher number A1541/C3956.
The A1540/C3955 are currently around $1.20pc overhere, the 1 digit higher numbers are 50% more expensive.
A year ago i had to cross the border to Poland for a sufficient number of A1540s, global availability of the C3955 hasn't been an issue so far.
In the US it's likely a different story, maybe AudiolabGA still has some.
Syn08, thanks for the explanation and excellent info, i blacked-out on the Ic/Vce part in the datasheets.
The A1540/C3955 are currently around $1.20pc overhere, the 1 digit higher numbers are 50% more expensive.
A year ago i had to cross the border to Poland for a sufficient number of A1540s, global availability of the C3955 hasn't been an issue so far.
In the US it's likely a different story, maybe AudiolabGA still has some.
Syn08, thanks for the explanation and excellent info, i blacked-out on the Ic/Vce part in the datasheets.
Hi Syn,
thanks for the feedback.
I'm ordering 10 off 2sc3601 for premium projects.
I can order 20 off 2sa1540/c3955 or 2sa1381/c3503 at 50% more than 2sa1360/c3423.
I don't have the hFE rank codes for any of their stock.
Which would you recommend?
and of course I'm ordering yet more 2sb649/d669c for drivers.
All from Dalbani.co.uk
They don't stock 1407.
They also stock 2sk170/j74, 2sk246/j103, 2sk369 all in both bl and gr idss.
If buying 369 which idss is more useful?
they still stock 2sj109 but the price has risen to £3.99
thanks for the feedback.
I'm ordering 10 off 2sc3601 for premium projects.
I can order 20 off 2sa1540/c3955 or 2sa1381/c3503 at 50% more than 2sa1360/c3423.
I don't have the hFE rank codes for any of their stock.
Which would you recommend?
and of course I'm ordering yet more 2sb649/d669c for drivers.
All from Dalbani.co.uk
They don't stock 1407.
They also stock 2sk170/j74, 2sk246/j103, 2sk369 all in both bl and gr idss.
If buying 369 which idss is more useful?
they still stock 2sj109 but the price has risen to £3.99
AndrewT said:
Ah, I just noticed you are on the other side of the Atlantic... Mouser references are likely not doing much good for you.
I can't find any reliable source for 2SA1540 here.
As for the beta, I never use or recommend the highest rank. Reason is that for bipolars the Beta*Early voltage is (in the first approximation) a constant, kinda process merit factor. Higher beta means always lower Early voltage and you know what follows - higher distortions.
I have no experience with 2SK369. I am a fan of fully symmetrical designs and therefore I don't buy many parts without the matching complementary part.
What is important on this thread, is designing with fets. Both mosfet and jfet. It doesn't have any other reason for being. If everyone wants to talk about bipolar design, they should chose another thread, in my opinion.
An 'Ayre' thread that talks about Ayre's new mixed component topology would make sense, but I don't know if Charles would want to divulge as much as I have on this thread.
An 'Ayre' thread that talks about Ayre's new mixed component topology would make sense, but I don't know if Charles would want to divulge as much as I have on this thread.
Today, I would like to talk about the 'Blowtorch' philosophy.
Even when this thread first appeared, many misunderstandings of the overall design were expressed, and many aspects of the design might have even been laughed at. I started inputs to this thread, after some misunderstandings were expressed by others, in order to get things right.
First, the Blowtorch is an expensive to make design. However, it is not that way because we wanted to use rare metals or do a particular metal sculpture piece that would look elegant in a beautiful listening room.
The parts that we used to make the unit were both expensive, relatively large, with real force needed to turn any of the controls.
We did not design the moving parts, but they were designed for professional/military applications where no mistakes in position or transmission could be tolerated.
These parts had been selected by my former business partner Bob Crump and put into a working chassis. However, in Bob's first prototype, only a j-fet follower was used after the TKD pots. It was essentially a buffered passive preamp, with minimal input switching and dual volume controls, in order to NOT need a balance control as well. Someone, somewhere, still has this preamp and is probably very happy with it.
However, Bob saw a 'limited' need for a preamp like this, but with a few additions:
First, it should supply balanced output drive, because many power amps were being designed with balanced input in mind. This would make it easier, as well to separate the preamp from the power amp and not pick up too much interference from the cable itself.
Second, it should supply some forward gain, so that the pot has more real range. More than 6dB, but less than 20dB was decided upon.
Third, we would custom make each unit to meet the customer's real needs. We could make more inputs, add a phono stage, buffer the outputs for long cable runs, or even provide completely balanced input-output operation of the line inputs, WITHOUT any extra stages or other compromise. I.E. no added IC op amps in the audio signal.
Fourth, we wanted a no compromise, absolute polarity switch, that added virtually nothing to the signal but inverted absolute polarity.
We, of course, wanted one for ourselves, and that was part of the deal.
My part of the deal was to make the line amp module, and I decided to try to make it without global feedback. This was a first for me, except for j-fet followers in various circuits.
I decided to emulate what Charles Hansen had originally done, but to NOT use the output stage. I was also using the example of the Audible Illusions line stage which uses a single open loop tube as a line stage. Just think, I had to use 8 fets to emulate what 1 tube can do! Still, this offered some advantages. More simplicity than Charles' design and no coupling caps like in the AI design.
Since we were going to make a 'world class' design like an F1 race car, we chose to use Teflon circuit boards, polystyrene and teflon caps, and the best connectors, wire, solder, whatever, so there would be little or no compromise.
I know that many of you still laugh at the very idea that wire, solder, etc can make any difference, but they do, and we have known this for decades. We were not going to compromise the design, just because some young engineer working in another field primarily, doesn't have the experience with this, and thinks it is silly. I was 'silly' too, once, even when I developed many of the topologies that are in use today. It is in fact possible to be a good engineer, yet be a little wet behind the ears, in hard experience.
When we went forward with the project, we found a machinist who could make the chassis for us. It had to be aluminum, copper or silver. These were the only acceptable metals that would heat sink properly and not add extra distortion. We chose aluminum for obvious reasons, it didn't easily tarnish, it was cheaper than silver, and it apparently is easier to work. We originally tried welding the aluminum chassis together, and one prototype was made, but we decided that just hogging out a solid aluminum block was really the better, more consistent way, and we knew that the kind of customers that we would have, outside of ourselves, would worry over small imperfections in the chassis, etc. Of course, this all added to the price.
Perhaps I will say more, later on this project.
Even when this thread first appeared, many misunderstandings of the overall design were expressed, and many aspects of the design might have even been laughed at. I started inputs to this thread, after some misunderstandings were expressed by others, in order to get things right.
First, the Blowtorch is an expensive to make design. However, it is not that way because we wanted to use rare metals or do a particular metal sculpture piece that would look elegant in a beautiful listening room.
The parts that we used to make the unit were both expensive, relatively large, with real force needed to turn any of the controls.
We did not design the moving parts, but they were designed for professional/military applications where no mistakes in position or transmission could be tolerated.
These parts had been selected by my former business partner Bob Crump and put into a working chassis. However, in Bob's first prototype, only a j-fet follower was used after the TKD pots. It was essentially a buffered passive preamp, with minimal input switching and dual volume controls, in order to NOT need a balance control as well. Someone, somewhere, still has this preamp and is probably very happy with it.
However, Bob saw a 'limited' need for a preamp like this, but with a few additions:
First, it should supply balanced output drive, because many power amps were being designed with balanced input in mind. This would make it easier, as well to separate the preamp from the power amp and not pick up too much interference from the cable itself.
Second, it should supply some forward gain, so that the pot has more real range. More than 6dB, but less than 20dB was decided upon.
Third, we would custom make each unit to meet the customer's real needs. We could make more inputs, add a phono stage, buffer the outputs for long cable runs, or even provide completely balanced input-output operation of the line inputs, WITHOUT any extra stages or other compromise. I.E. no added IC op amps in the audio signal.
Fourth, we wanted a no compromise, absolute polarity switch, that added virtually nothing to the signal but inverted absolute polarity.
We, of course, wanted one for ourselves, and that was part of the deal.
My part of the deal was to make the line amp module, and I decided to try to make it without global feedback. This was a first for me, except for j-fet followers in various circuits.
I decided to emulate what Charles Hansen had originally done, but to NOT use the output stage. I was also using the example of the Audible Illusions line stage which uses a single open loop tube as a line stage. Just think, I had to use 8 fets to emulate what 1 tube can do! Still, this offered some advantages. More simplicity than Charles' design and no coupling caps like in the AI design.
Since we were going to make a 'world class' design like an F1 race car, we chose to use Teflon circuit boards, polystyrene and teflon caps, and the best connectors, wire, solder, whatever, so there would be little or no compromise.
I know that many of you still laugh at the very idea that wire, solder, etc can make any difference, but they do, and we have known this for decades. We were not going to compromise the design, just because some young engineer working in another field primarily, doesn't have the experience with this, and thinks it is silly. I was 'silly' too, once, even when I developed many of the topologies that are in use today. It is in fact possible to be a good engineer, yet be a little wet behind the ears, in hard experience.
When we went forward with the project, we found a machinist who could make the chassis for us. It had to be aluminum, copper or silver. These were the only acceptable metals that would heat sink properly and not add extra distortion. We chose aluminum for obvious reasons, it didn't easily tarnish, it was cheaper than silver, and it apparently is easier to work. We originally tried welding the aluminum chassis together, and one prototype was made, but we decided that just hogging out a solid aluminum block was really the better, more consistent way, and we knew that the kind of customers that we would have, outside of ourselves, would worry over small imperfections in the chassis, etc. Of course, this all added to the price.
Perhaps I will say more, later on this project.
- Status
- Not open for further replies.