I again challenge Syn08 to come up with equivalent quality complementary matched fets that are available today, except for Linear Systems, which are relatively expensive at the moment, and hard to get, as well. It is not professional to criticize without any proof, but it is done with impunity, here, that's for sure.
Output is unable to get to rails. Look once more at the circuit and operating point voltages.
Interesting, but first everyone please realize that the CTC Blowtorch circuit, and similarly PMA's approximation, have a differential gain of about 16 dB, or about 8 dB on each side. This is not a unity gain buffer, as it would be pointless to put this topology in the circuit, if it were.
Please correct me if I am wrong, but isn't Syn08 presuming that this circuit has negative or at best, unity gain?
Please correct me if I am wrong, but isn't Syn08 presuming that this circuit has negative or at best, unity gain?
Thanks PMA, I simply approached it from the input angle, but everyone please realize that INTERNALLY this circuit operates from +/- 15V and the input devices can't even swing that far, unless you hit them with lightning, or some such, as they are clamped at 15 volts on their drains, no matter what the output swing.
Oh, now I see what Syn08 is saying: He thinks that the input noise does not matter, because the cascode connection is 'compromised with the 220 ohm resistors? Well, it is partially true, especially at very low frequencies, but I still need complementary matched pairs, the highest transconductance I can get in the input pairs, and they have to be quiet, in order not to add to the second stage.
Now, for everyone else: It might be a good time to show what is happening.
First, a cascode connection is always composed of 2 active devices.
The classic cascode is usually the same polarity devices (for example, 2N channel fets) and the top fet, is biased at a voltage between the input (usually zero volts) and the power supply. This connection gives almost all the noise contribution to the FIRST or input device. Now, what happens when we add a parasitic resistor (for some reason) of 220 ohms at the junction between the first and second fets and capacitively couple it with a large cap, to ground?
Well, this changes things, somewhat. Now the top fet can have actual voltage gain of its own, and this noise can inject itself into the output.
I think that this is what Syn08 is hinting at. Of course, I make the resistor as large as possible to reduce this noise, is why we use 30V instead of 18V or so. Make more sense, now?
Now, for everyone else: It might be a good time to show what is happening.
First, a cascode connection is always composed of 2 active devices.
The classic cascode is usually the same polarity devices (for example, 2N channel fets) and the top fet, is biased at a voltage between the input (usually zero volts) and the power supply. This connection gives almost all the noise contribution to the FIRST or input device. Now, what happens when we add a parasitic resistor (for some reason) of 220 ohms at the junction between the first and second fets and capacitively couple it with a large cap, to ground?
Well, this changes things, somewhat. Now the top fet can have actual voltage gain of its own, and this noise can inject itself into the output.
I think that this is what Syn08 is hinting at. Of course, I make the resistor as large as possible to reduce this noise, is why we use 30V instead of 18V or so. Make more sense, now?
From this perspective, you finally got it almost right. Add that, within the schematic that was posted as BT gain stage, the input JFET transconductance doesn't matter a iota. In the first approximation it is the ratio between the drain resistor and the source degeneration equivalent resistor that gives the input stage voltage gain. Good bye very large JFET transconductance requirement.
And no, as long as the impedance the first stage sees in the MOSFET source is very large, the input stage is NOT a transconductance stage and the MOSFET is NOT a I/V converter. The situation would be completely different if you would use bipolars in the output stage. Which, in fact, could be a better solution here.
Gotta run...
Times up. It is about 5 ohms. Now, 5 ohms is a bit less than 200-220 ohms, does this change the situation?
After all, lets say that there is NO resistor but an ideal current source or nothing at all. Then what is the 'voltage gain of the first stage? Does anyone see the complexity of the situation? How could a cascode work in the first place, if you thought of it as 2 separate gain stages?
After all, lets say that there is NO resistor but an ideal current source or nothing at all. Then what is the 'voltage gain of the first stage? Does anyone see the complexity of the situation? How could a cascode work in the first place, if you thought of it as 2 separate gain stages?
It makes almost no difference, if 220R resistors, or current sources.
And output mosfets have no current gain, output voltage is created on 1k load resitors.
It behaves like voltage output with 1k output impedance, so I do not know why syn08 complains to "gain dependent on load" - same as many common output circuits. Frankly speaking, I would rather see 100 ohm output impedance, 1k is a way too much to me.
And output mosfets have no current gain, output voltage is created on 1k load resitors.
It behaves like voltage output with 1k output impedance, so I do not know why syn08 complains to "gain dependent on load" - same as many common output circuits. Frankly speaking, I would rather see 100 ohm output impedance, 1k is a way too much to me.
Look closer. Try hard.
In fact IT IS.
The LT1363 has, of course, only one polarity (the +side) to the gain stage and output. And it has not a folded cascode like the BT but current mirrors.
Regarding gain, nothing changes. You can mirror the differential currents into the high impedance node with a mirror as well as with a folded cascode.
Same gain (with same load at the high impedance node), as long as the current mirror has no current gain. (Which it doesn't have in the LT1363.)
If you would analyze a PASS LABS gain stage (eg. those in the power amps) you'd be surprised that they too have a very very very similar 4 quadrant JFET input stage, the gain stage being MOSFET's in common source and not folded cascodes. And they have feedback (SUSY patent).
Ayre is similar I think, without feedback and current mirrors, not folded cascodes.
Seems like these top notch products converged to what John did.
Not too much to learn for John I fear.
Tino
Thanks Jam.
Well.............I have the dual FET's since some time, and wanted to use the THAT transistor arrays. But they can only stand 10mA (or 20, still too few, if i'm going to parallel 4 Fets). I think I'll look for 2SA1085/2SC2547.
Our summers here are short, so not much to stay behind the welding station. I'll keep you informed Jam, as soon as I'm going to make progress.
Tino
Ok. If to look closer and to try hard, and if to use an imagination, IT IS. But...
But if to see from a different prospective: bridged diamond buffers VS complementary diffstages each in tail of the other, it is not.
The main difference is, which devices are closer coupled. Vertical transistors, or horizontal transistors.
To say IT IS you have to modify BOTH. But it will be the 3'rd case, that is neither the 1'st one, nor the 2'nd one. However, all 3 of them, different thingies, can be made of THE SAME transistors.
It’s just another multi-tanh circuit for discussion.
I’m sorry if this is OT, but it’s about linearization so I hope that Netlist doesn’t Bin me.
Edmond Stuart and Scott Wurcer discussed the multi-tanh circuit by Nathan (Patent US 6963244) earlier. There has also been discussed other multi-tanh circuits, but as I see it the circuit by Nathan and other discussed circuits are related to IC’s and they are not very easy to implement in a discrete circuit, because they depend on the effective emitter area, Aeff.
I have looked at another multi-tanh circuit fig 16 in Barrie Gilberts’s IEEE journal “The Multi-tanh Principle: A tutorial Overview”
“A practical series connected doublet”
I have attached a schematic as an example.
As I understand it, this circuit doesn’t make the Aeff by changing the effective area of the emitter, but you can use the same transistors and produce the Aeff by external resistors, in the attached schematic RB1 and RB2.
I have also attached the calculations that I have used.
As you can see in the calculations, to keep the Aeff = 4 with various temp, the Ib should change.
So my simple questions are; is it possible to make a discrete Multi-tanh circuit this way?
Is it necessary to change the Ib with varying temp?
Cheers
stinius
I’m sorry if this is OT, but it’s about linearization so I hope that Netlist doesn’t Bin me.
Edmond Stuart and Scott Wurcer discussed the multi-tanh circuit by Nathan (Patent US 6963244) earlier. There has also been discussed other multi-tanh circuits, but as I see it the circuit by Nathan and other discussed circuits are related to IC’s and they are not very easy to implement in a discrete circuit, because they depend on the effective emitter area, Aeff.
I have looked at another multi-tanh circuit fig 16 in Barrie Gilberts’s IEEE journal “The Multi-tanh Principle: A tutorial Overview”
“A practical series connected doublet”
I have attached a schematic as an example.
As I understand it, this circuit doesn’t make the Aeff by changing the effective area of the emitter, but you can use the same transistors and produce the Aeff by external resistors, in the attached schematic RB1 and RB2.
I have also attached the calculations that I have used.
As you can see in the calculations, to keep the Aeff = 4 with various temp, the Ib should change.
So my simple questions are; is it possible to make a discrete Multi-tanh circuit this way?
Is it necessary to change the Ib with varying temp?
Cheers
stinius
Why don't you open your new thread?
There may be lots of variations using multiple diffstages, and all of them give very different results.
I have a break (just back from Shanghai), yes there are ways to make the offsets necessary for the muti-tanh circuit with currents and resistors. You will probably find PTAT currents are needed (easily generated). You also need to be careful on the noise budget, but throwing away current on a no compromise circuit is not much of a problem.
You can also do a fully complimentay version pretty easily, go for it!
- Status
- Not open for further replies.