The diodes do track.
There is some doubt about which diodes are integrated into the various devices and whether some diodes were changed (by accident or deliberately) without changing the device code.
There is some doubt about which diodes are integrated into the various devices and whether some diodes were changed (by accident or deliberately) without changing the device code.
RIP ThermalTrak :-(
--Major amp designers have already completed lifetime buys.
DIY peons need a group hug :-|
--The JFET is DEAD! The ThermalTrak is DEAD!
--Oh Lord, Help Us rise above..... rise above.
...OK....OK...now... NOW we need a group think!! 🙂
1) select an epoxy package diode like the 1N4004 with a (guess) 4 C/watt "physically direct attach" thermal conductivity
2) slip teflon insulation tubing tightly on the diode interconnect leads
3) put high-conductivity epoxy glue around the diode
4) Wrap the bipolar output transistor collector output metal lead around the diode_body+epoxy (Clamp until hard)
5) use low electrical resistance, low'ish thermal conductance tin wire to attach/thermally_isolate this collector+diode lead to the PCB.
At least 3 genius audio engineers invested months analyzing ThermalTrak!
Where are they now? Where are they now after this market diaster!
--Major amp designers have already completed lifetime buys.
DIY peons need a group hug :-|
--The JFET is DEAD! The ThermalTrak is DEAD!
--Oh Lord, Help Us rise above..... rise above.
...OK....OK...now... NOW we need a group think!! 🙂
1) select an epoxy package diode like the 1N4004 with a (guess) 4 C/watt "physically direct attach" thermal conductivity
2) slip teflon insulation tubing tightly on the diode interconnect leads
3) put high-conductivity epoxy glue around the diode
4) Wrap the bipolar output transistor collector output metal lead around the diode_body+epoxy (Clamp until hard)
5) use low electrical resistance, low'ish thermal conductance tin wire to attach/thermally_isolate this collector+diode lead to the PCB.
At least 3 genius audio engineers invested months analyzing ThermalTrak!
Where are they now? Where are they now after this market diaster!
Don't panic. You don't need Thermaltrak or JFET's to build world class amplifiers.
And, you don't need to be an audio genius either.
🙂
And, you don't need to be an audio genius either.
🙂
Two types still listed as available.
The NJL3281DG NPN ThermalTrak, and the NJL1302DG PNP complement are still listed as active items on the OnSemi site, although Mouser and DigiKey both list them as backordered.
The NJL3281DG NPN ThermalTrak, and the NJL1302DG PNP complement are still listed as active items on the OnSemi site, although Mouser and DigiKey both list them as backordered.
Using diodes, what is the best way to mount them next to the transistors, and supply the correcting offset to the bias to maintain safety?
By gluing a diode connected sot23 transistor to the Collector lead of the output device, right next to where it exits the package.
I have not researched the matter to know if the ThermalTrak is really discontinued. As many of you know, I researched the ThermalTraks a great deal and wrote about them in my book, "Designing Audio Power Amplifiers". I had great success with them and felt that they made a great contribution to making BJT output stages competitive with MOSFET output stages in terms of thermal stability. I spent many hours characterizing the devices themselves and a prototype amplifier in which I used them.
It is true that some had difficulty applying them, including McIntosh. Also, the OnSemi App note was terrible and misleading. However, once one understood the devices and how to manipulate the design of the required bias spreaders (e.g., usually variants of the Vbe multiplier), very good designs were possible.
The mismatch in tempco of the Vbe to the diode was an annoyance, but readily dealt with by those who understood the needed techniques.
Having the diode monolithically implemented on the same die would have been great, but breakdown and isolation problems intervene. In practice, the sensing diode in the ThermalTrak is mounted on the same copper header as the power transistor die. This still gets it much closer thermally to where it wants to be. Attaching diodes to the case of the power transistor is better than to the heatsink, but a far cry from the thermal tracking performance of the ThermalTrak diode.
I don't recall having had trouble with any of the devices, but my sample size was fairly small. I do understand that there were some instances of reliability or defect concerns. If OnSemi was unable to such problems, it would be understandable for them to discontinue them. I seem to recall paying about $3 for them, which I thought was a great price. The 5-pin TO-264 package may have been a PIA for OnSemi.
Overall, I am very sad to see their demise, if it is true.
The Sanken devices, being Darlingtons, appear to lack the ability to have strong suck-out current for the output device. I have not had any experience with the Sanken devices, so I do not know to what extent this limitation interferes with good high-frequency, high-amplitude performance. This matter relates directly to dynamic crossover distortion.
Cheers,
Bob
It is true that some had difficulty applying them, including McIntosh. Also, the OnSemi App note was terrible and misleading. However, once one understood the devices and how to manipulate the design of the required bias spreaders (e.g., usually variants of the Vbe multiplier), very good designs were possible.
The mismatch in tempco of the Vbe to the diode was an annoyance, but readily dealt with by those who understood the needed techniques.
Having the diode monolithically implemented on the same die would have been great, but breakdown and isolation problems intervene. In practice, the sensing diode in the ThermalTrak is mounted on the same copper header as the power transistor die. This still gets it much closer thermally to where it wants to be. Attaching diodes to the case of the power transistor is better than to the heatsink, but a far cry from the thermal tracking performance of the ThermalTrak diode.
I don't recall having had trouble with any of the devices, but my sample size was fairly small. I do understand that there were some instances of reliability or defect concerns. If OnSemi was unable to such problems, it would be understandable for them to discontinue them. I seem to recall paying about $3 for them, which I thought was a great price. The 5-pin TO-264 package may have been a PIA for OnSemi.
Overall, I am very sad to see their demise, if it is true.
The Sanken devices, being Darlingtons, appear to lack the ability to have strong suck-out current for the output device. I have not had any experience with the Sanken devices, so I do not know to what extent this limitation interferes with good high-frequency, high-amplitude performance. This matter relates directly to dynamic crossover distortion.
Cheers,
Bob
Don't be so sad, bob..
May be sanken just want to say that sanken bjt driver + sanken bjt op will have good sound, and it is their advantage since long time ago.
Most very good bjt amps (soulution710, dan'agustino, etc)
using the arrangement with its secret touch. I also try this touch to other bjt and it also work with some fast sanyo bjt.
Bob Cordell,
.................
It is true that some had difficulty applying them, including McIntosh. Also, the OnSemi App note was terrible and misleading. However, once one understood the devices and how to manipulate the design of the required bias spreaders (e.g., usually variants of the Vbe multiplier), very good designs were possible.
....................
Are you saying that a Vbe multiplier is still required? The application note shows only the diodes and electrolytic capacitor as the entire bias circuit.
.................
It is true that some had difficulty applying them, including McIntosh. Also, the OnSemi App note was terrible and misleading. However, once one understood the devices and how to manipulate the design of the required bias spreaders (e.g., usually variants of the Vbe multiplier), very good designs were possible.
....................
Are you saying that a Vbe multiplier is still required? The application note shows only the diodes and electrolytic capacitor as the entire bias circuit.
Yes, a Vbe multiplier arrangement, or something equivalent, is generally required for proper biasing of ThermalTrak output stages. I discussed this at length in my book.
Cheers,
Bob
Cheers,
Bob
Thank you, Bob. Since asking that, I have read much of your book (and Selfs book) and have a better understanding of what the bias spreader needs to accomplish.
Sorry guys, I revived this thread but deleted my comment when minutes later I found a bunch of answers 😛 It must have stayed up the list of active threads though.
check out this thread:
https://www.diyaudio.com/community/...odels-with-sankens-sap-std-series-why.317340/
https://www.diyaudio.com/community/...odels-with-sankens-sap-std-series-why.317340/
I have them in my Leach Amp and have been thinking of changing them to 5200/1943 as they are not sounding good
gajanan phadte
gajanan phadte
I'm not sure what the status of ThermalTrak was when this thread was started, but as of August 2024 they are alive and well. The NJL3281DG and the NJL1302DG are both marked status Active on the OnSemi web site and are available for sampling or purchase direct, or through distributors such as Mouser who have thousands in stock. It was the suffix D devices that were discontinued when the suffix DG lead free variety superceded them.
This is good news, assuming that they addressed the sensing diode problem that caused reliability problems in the past. In those cases, the forward drop of the diode would change suddenly and unpredictably on occasion, throwing off the bias on the output stage. It probably had to do with some kind of thermal stress having to do with the fact that the diode and transistor die were mounted on the same header.
Cheers,
Bob
Cheers,
Bob
I've asked OnSemi and a company that was using them about any changes in manufacture, but no reply yet. I'll shout if I hear anything. So the diode and transistor are separate Si dies, but mounted on the same header, yes?. I can see how that could set up fatigue but it sounds like a rare event, especially with a crystal as perfect as a semiconductor and with a technique so commonly used. I'm tempted to buy a small batch and put them on a thermal stress cycle test