What are the tradeoffs? Simplicity and cheap vs. higher accuracy and more expensive?
I guess an Led and a transistor will do? Or maybe just JFET in some cases?
Scott there is a potential problem with your circuit, the lt6656 need at least 1uf at the output terminal for stability.
Using the lt3092 in this case is preferable as this is a true floating current source with very good specifications and relatively inexpensive.
look at the datasheet of lt3092 and see how they create the floating current source , is also ingenious and interesting.
Scott question on that large cap from B to Gnd, wouldn't that cause a delay in the temp tracking between the input Vbe and the output trio? Even assuming all is isothermal?
Jan
Somwhat related are current mirrors - maybe a little out of place at the moment; regarding accuracy - nothing to do with TC;
View attachment Current Mirrors -BJT.pdf
How do your's compare with other topologies?
Thx-RNMarsh
View attachment Current Mirrors -BJT.pdf
How do your's compare with other topologies?
Thx-RNMarsh
low accurate current sources can also be very noisy. And unfortunately there is a lot of bad examples.With all your experience, do you know any jfet or similar device that the zero tempco is similar in diferents units. That is something I never think of or experiment in the past.
I also want to buy some 2n4391 to measure the zero tempco. As I also find the simplest solutions more elegant.
I am still under moderation and can't see my own post, but I suppose that you must have seen this. Ed Oxner worked at Siliconix for decades and knows more about FETs that almost anybody else outside of Japan. He has written four different books and I have them all. They are excellent resources.
By the way he derived the zero tempco point for JFET's (what I referred to in my earlier post) theoretically. But it is very easy to verify. Simply measure the Idss at room temperature -- typically 20°C or 25°C. Then immerse the FET into a glass of ice water, which will be 0°C±0.1°C. This is a large enough spread to verify the theoretical calculations.
The output impedance of the current source will depend on the "flatness" of the curves. (The Tek 576 curve tracer I bought on eBay was the best $1000 I ever spent. You can find the older 575's for $100 or $200 but they are full of tubes and electrolytic capacitors that are going bad. Still it is better than nothing.) The P-channel Toshiba JFETs were very flat.
But the other approach is to simply cascode the first JFET with another JFET. This will increase output impedance by a factor of between 10x and 100x. The cascoding FET needs to have a Vgs pinchoff of at least a volt or two higher than the actual current source FET. So in the old days, you would just cascode a J74 with a J103. Unfortunately the J74s were discontinued five years ago and all you can easily find are fakes. The J103s were discontinued two years ago and you can probably find some stock on those.
If you are in the US, the only reputable places that won't sell fakes that I know of are B&D Enterprises (B+D Enterprises - Home page) and Solid State Inc. (Solid State Inc. | Electronic Components Suppliers). I bought some real NEC J44's and K163's from B&D Enterprises just a few years ago even though they were discontinued in the early '90s. The only problem was that they were of the lowest current grade (R?) and the Idss ranged from 1.5 to 3.0 mA, which wouldn't work for my application.
Fairchild still makes some FETs, and then of course there is Linear Systems. I bet that a LSK389 cascoded with a LSK489 would be very nice and quiet if you need a twin matched current source. I think they have singles of the 389s and are working on singles of the 489.
Have fun,
Charles Hansen
Ayre Acoustics, Inc.
Most attempts to heavily filter the noise can be problematic, I'n not sure how thermal tracking at frequencies over 10Hz is much of a problem. If you want high output impedance in a floating source at audio frequencies the FET probably wins over ones using lots of feedback.
I think the original derivation from Siliconix was J. Hoerni in 1963 who BTW was one of John Hall's mentors. So you see these old guys are still the only ones carrying the torch for JFET's.
Since the TC of the mobility of silicon is non-linear the TC is zero at only one temperature. Cobbold has some graphs in his book, but an IC reference is far better than the best. The FET works fine in many places but it is not a precision reference.
B&D is great but don't seem to restock when these things run out.
I was going to ask this, how necessary is this for audio projects and why if there's still a demand for Jfets, why were they taken out of production..
Last edited:
There is not much demand or new designs for most of the discrete ones anymore. OTOH there are millions of JFET op-amps shipped every month.
In audio it's a boutique business frankly a little too tied to fashion. Do you think anyone could put together a compelling argument that would create a large demand from the consuming public?
Scott is only telling you one half the story:
Complementary jfets are still regarded as one of the finest and most elegant ways to make a linear amplifier. However, IC implementation never came to being, because the IC companies decided that the cost/effort to MAKE true complementary jfets was not worth it.
Now, WHY did the Japanese make so many very high quality complementary jfets over a 30 year period? Yes, they started in about 1978, and gave up perhaps 5 years ago. It had to do with professional competition between various discrete component manufacturers, who competed with each other for their various 'clients', most of whom were in competition with each other as audio product manufacturers.
Therefore, Sony made certain spectacular products, Toshiba made theirs, Hitachi made complementary jfets, NEC made complementary Vfets for awhile, etc, etc.
Each semiconductor company would show their products in Japan at engineering shows, showing their technical capabilities. Most were very much at the same level, even in 1978, but EACH company made a slightly different part. Some were higher voltage, but somewhat higher noise, others might be lower input capacitance, but with lower Gm, etc. An independent, like me, could pick and choose among them, but their sister companies attached to the specific semiconductor manufacturer had to use what their host semiconductor manufacturer provided to them. The need to compete between the audio products companies kept the semiconductor companies attached to them, on their toes, so to speak.
Well, what happened? Many Japanese audio manufacturing companies went bankrupt due to the high cost of the yen, indifference in the audio marketplace, etc. so the primary semiconductor manufacturers found that they could make more money making something else, like IC's, for example, rather than exotic complementary jfets. AND after a time, they closed down their production facilities for something more profitable.
Still, the BEST jfets were made in Japan, and there is no real substitute quite yet. However, Linear Integrated Systems is trying, and is succeeding in replacing the Toshiba, etc., jfets in the marketplace, unfortunately at many times the original price.
So that is where we are at, today.
Complementary jfets are still regarded as one of the finest and most elegant ways to make a linear amplifier. However, IC implementation never came to being, because the IC companies decided that the cost/effort to MAKE true complementary jfets was not worth it.
Now, WHY did the Japanese make so many very high quality complementary jfets over a 30 year period? Yes, they started in about 1978, and gave up perhaps 5 years ago. It had to do with professional competition between various discrete component manufacturers, who competed with each other for their various 'clients', most of whom were in competition with each other as audio product manufacturers.
Therefore, Sony made certain spectacular products, Toshiba made theirs, Hitachi made complementary jfets, NEC made complementary Vfets for awhile, etc, etc.
Each semiconductor company would show their products in Japan at engineering shows, showing their technical capabilities. Most were very much at the same level, even in 1978, but EACH company made a slightly different part. Some were higher voltage, but somewhat higher noise, others might be lower input capacitance, but with lower Gm, etc. An independent, like me, could pick and choose among them, but their sister companies attached to the specific semiconductor manufacturer had to use what their host semiconductor manufacturer provided to them. The need to compete between the audio products companies kept the semiconductor companies attached to them, on their toes, so to speak.
Well, what happened? Many Japanese audio manufacturing companies went bankrupt due to the high cost of the yen, indifference in the audio marketplace, etc. so the primary semiconductor manufacturers found that they could make more money making something else, like IC's, for example, rather than exotic complementary jfets. AND after a time, they closed down their production facilities for something more profitable.
Still, the BEST jfets were made in Japan, and there is no real substitute quite yet. However, Linear Integrated Systems is trying, and is succeeding in replacing the Toshiba, etc., jfets in the marketplace, unfortunately at many times the original price.
So that is where we are at, today.
- Status
- Not open for further replies.