U1A keeps the FET current constant by feeding back cascode Q3's collector node, producing the output voltage in the process (not shown). G = R20/R(drain) = 1,000 if I'm not mistaken. That's rather ambitious for one stage.
Weirdness at Q3's base will affect the loop, & has to be ruled out. Q3 might also oscillate. I don't see that it should, but sometimes weird things happen (e.g., negative input impedance).
Indeed. Zin for the cascode Q3 emitter is 100/beta and beta = fT/f, which is 13 at 10MHz for FZT851, or Zin = 7.7 ohms at 10MHz, rising with f, acting as nasty 207nH inductor. Not to mention a 1.3nV.rt-Hz voltage-noise source. Axe that 100 ohms.
Yes, the 100R are just there to check the sensitivity. The bad thing is that even for a zero Ohm resistor, the S11 comes only a bit closer to the circle that encloses the positive resistance region, but it still takes a walk outside.
But I need at least a few Ohms so I can insert the buffered input voltage for the bootstrap.
The entire bias chain has no impact on the total noise, even including the Zener and with absolutely minimum decoupling.
Here is a stripped down version in LTspice:
screendump: <
formatting link
>
noise results, including some contributors:
It seems that above 1 MHz, the active loads quickly becomes ineffective in suppressing the VCC noise. But 1 MHz BW is enough.
SYMATTR SpiceLine Rser=0.1 SYMBOL res 336 128 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R14 SYMATTR Value 100 SYMBOL res 480 848 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R18 SYMATTR Value 10meg SYMBOL cap 416 -32 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C3 SYMATTR Value 47p SYMBOL zener -192 368 R180 WINDOW 0 24 64 Left 2 WINDOW 3 -46 -65 Left 2 SYMATTR InstName D3 SYMATTR Value BZX84-C2v7 SYMBOL pnp 112 -96 R180 SYMATTR InstName Q1 SYMATTR Value ZTX951 SYMBOL res 32 -336 R0 SYMATTR InstName R1005 SYMATTR Value 33 SYMBOL res 112 -128 R0 SYMATTR InstName R1006 SYMATTR Value 1k SYMBOL diode 112 -256 R0 WINDOW 0 46 -11 Left 2 WINDOW 3 40 48 Left 2 WINDOW 123 48 21 Left 2 SYMATTR InstName D1003 SYMATTR Value BAV99S SYMATTR Value2 n=5 SYMBOL res 480 -112 R90 WINDOW 0 0 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName R17 SYMATTR Value 400k SYMBOL res -176 112 R0 WINDOW 3 38 71 Left 2 SYMATTR InstName R1003 SYMATTR Value 240 SYMBOL cap 464 768 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C8 SYMATTR Value 2n2 SYMBOL cap -256 32 R0 WINDOW 3 42 43 Left 2
SYMATTR InstName C1001 SYMATTR SpiceLine Rser=0.1 SYMBOL res -64 224 R90 WINDOW 0 66 -8 VBottom 2 WINDOW 3 40 56 VTop 2 SYMATTR InstName R1414 SYMATTR Value 0.1 SYMBOL cap 320 528 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C6 SYMATTR Value 1p SYMBOL cap 240 -272 R0 SYMATTR InstName C1004
SYMBOL current -160 -224 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName I1 SYMATTR Value 10m SYMBOL Opamps\\opamp2 432 32 R0 SYMATTR InstName U1A SYMATTR Value ths4021 SYMBOL res -400 672 R0 SYMATTR InstName R1000 SYMATTR Value 0.01 SYMBOL cap -176 304 R0 WINDOW 3 42 43 Left 2 WINDOW 0 36 69 Left 2
SYMATTR InstName C1000 SYMATTR SpiceLine Rser=0.1 SYMBOL npn -16 192 R0 SYMATTR InstName Q3 SYMATTR Value ZTX851 TEXT 168 336 Left 2 !.noise V(op_out) V2 dec 1000 100 1E7 TEXT 168 304 Left 2 !;ac dec 1000 10m 100Meg TEXT -496 -80 Left 2 ;1nV/RtHz noise plus\njust non-zero offset TEXT -376 -352 Left 2 ;Vcc + E1 * 1n VRtHz noise TEXT 168 280 Left 2 !;tran 500s TEXT 168 248 Left 2 !.op TEXT 168 368 Left 2 !;.include d:\\lib\\spice\\ghf_spice_lib.txt TEXT -1528 -560 Left 2 !*******************************************************************************\n\n* THS4021 SUBCIRCUIT\n\n* HIGH SPEED MONLITHIC OPERATIONAL AMPLIFIER \n\n* WRITTEN 1/6/99\n\n* NULL PINS ARE NOT MODELED\n\n* TEMPLATE=X^@REFDES %IN+ %IN- %VCC+ %VCC- %OUT @MODEL\n\n* CONNECTIONS: NON-INVERTING INPUT\n\n* | INVERTING INPUT\n\n* | | POSITIVE POWER SUPPLY\n\n* | | | NEGATIVE POWER SUPPLY\n\n* | | | | OUTPUT\n\n* | | | | | \n\n* | | | | | \n\n* | | | | | \n\n.SUBCKT THS4021 1 2 3 4 5 \n\n*\n\n* INPUT *\n\nQ1 33
No, a very low noise amplifier with 60 dB gain, 100 mHz to 1 MHz. I'm aiming for 100pV/rtHz at least on the sweet spot.
The current source defines the bias current and the FET's gate voltage is adjusted by an integrator until the cascode collector equals the level of CENTER.
The signal gain from the gate to the output of U1A is determined by the gm of the FET array, which _is_ large and by the transimpedance amplifier. (assuming open loop)
The current source and the cascode transistor do not enter the equation other than second order effects like the ccs diverting some signal current or that the BJT has probably a much higher Early voltage and thus higher output impedance.
The integrator is also not in the signal path. It is as slow as molasse. It takes a minute until the window comparator thinks that the operating point has been reached. There may be some influence in the mHz region, but I have no problem there.
The current source is necessary to suppress the noise on VCC; without it it takes a supply better than 2nV/rtHz (for 8 FETs and twice the current). There also must be 5 diodes in series or the noise of the current source emitter resistor starts playing a role.
Gerhard
who got a 54754A differential TDR plug in this morning, still a month under service. That needs to be played with now. :-) :-)
If putting an inductor on the input makes an oscillator, this can only mean that energy is being fed from the output back to the input.
The two obvious paths for that are Cdg and Cgs.
You seem to have eliminated Cdg as a cause, by your report that the negative resistance is relatively insensitive to the cascode's base resistance.
Logically, this points suspicion at Cgs, a positive-feedback capacitance famous for making BJTs into oscillators.
I'd suspect your feedback loop, injecting energy into the FET array's source that couples through to its gate, except you've already reported that the problem persists even with the feedback loop disabled.
This leaves the FETs themselves.
You could also test the possibility that Cgs feedthrough is the cause, by temporarily lowering the operating current and disconnecting most of the JFET array.
If 4 x Cgs is a problem, 1 x Cgs should be one-fourth as bad.
If you can identify the source of the problem, you'll be able to devise a correction for it. (E.g., if this were r.f., neutralization.)
The frequency of oscillation for an input inductor L might give a hint as to the value of any capacitance at play here.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.