To construct Voltage-Controlled Multivibrator (VCMV) for circuits in my Down-Conversion Frequency Synthesis concept, I've been moving closer and closer to a FET switch to direct the charging currents. One of things I want to do with the VCMV is switch charging directions across a capacitor to set the half-periods of a square-wave, but not use any sort of current mirror, as in a traditional triangle wave generator. I want to use exactly one voltage-controlled current source to supply both the up-ramp and down-ramp current, by switching the capacitor terminals. The idea behind this is that the square-wave period produced always has precisely a 50% duty cycle, because there is only one current source, presuming of course that switching itself incurs extremely small offsets.
An electronic switch seems in order, hence FET or MOSFET. I don't want to use ICs (e.g. a device like the CD4066), I was looking for a discrete semiconductor device. Back in January 2018, when I first started to deal with this engineering problem, MOSFETS seemed in order. After investigating several devices, I ordered a batch of 2N7000 devices from Digi-Key, because (i) they seemed capable of doing the job (switching a current source for a capacitor), and they are still available as TO-92 packaged parts (the Model III EMS will not use surface-mount technology -- it's not maintainable). As it happens, this device has 2x larger sustaining voltage than I need (60V vs. 30V), and it's wicked fast.
Today, I was able to test some basic circuits, before attempting to construct my VCMV in full.
First circuit was just the simple switch application, and it worked extremely well! Using a +5V square wave to switch currents sourced from the +15V Vcc supply was trivially easy. Here's the workbook notes of all of today's experiments:
An electronic switch seems in order, hence FET or MOSFET. I don't want to use ICs (e.g. a device like the CD4066), I was looking for a discrete semiconductor device. Back in January 2018, when I first started to deal with this engineering problem, MOSFETS seemed in order. After investigating several devices, I ordered a batch of 2N7000 devices from Digi-Key, because (i) they seemed capable of doing the job (switching a current source for a capacitor), and they are still available as TO-92 packaged parts (the Model III EMS will not use surface-mount technology -- it's not maintainable). As it happens, this device has 2x larger sustaining voltage than I need (60V vs. 30V), and it's wicked fast.
Today, I was able to test some basic circuits, before attempting to construct my VCMV in full.
First circuit was just the simple switch application, and it worked extremely well! Using a +5V square wave to switch currents sourced from the +15V Vcc supply was trivially easy. Here's the workbook notes of all of today's experiments:
2N7000 Testing Circuits
This MOSFET as a switch is very fast! The 2N7000 was able to pull down 15 mA for ≈25 nS TPHL, whereas the 1KΩ pull-up limited TPLH to more like 75 nS. Still, for audio frequencies, and for supersonic frequencies 20 KHz < f < 100 KHz, this is extremely small equivalent phase delay. It's also clear that the driving the gate input for switching is highly-capacitive. The HP33120A test generator was terminated in 50Ω, but the input was in fact still perturbed by the gate capacitance.
2N7000 MOSFET as Switch
The second circuit was a traditional FET source follower. This also worked extremely well, but the offset voltage between input and output is VGS. This is is far greater than for an emitter follower, or a complementary symmetry emitter follower. As tested, the VOS was about -2V. Otherwise the circuit is virtually perfect: it offers very high-impedance input, has a low-impedance output, and the bandwidth is very large. Dynamic response was extremely clean. It was difficult to see any appreciable signal delay between input and output! Of course, viewing this response is limited to the 60 MHz bandwidth of my Tektronix TBS1064 oscilloscope. With a higher-bandwidth 'scope, delays in the hundreds of picoseconds would probably be discerned. What an incredible circuit, VOS aside!
Source Follower Response at ≈500 KHz, ±5VPP
input output risetime
input-output falltime
The last circuit tried was an adaption of from the Art of Electronics, by Horowitz and Hill, to try to eliminate the VOS of the source follower. Unfortunately, this circuit was not successful. The circuit attempts to remove the VOS by using a second similar FET (though, I'm using MOSFETs instead), to act as a offset voltage, as well as to buffer the pull-down resistance as a constant current source for the voltage follower. While the VOS offset does in fact get eliminated, there was an apparent gain drop of -3dB. More particularly though, the resulting circuit has such high-impedance that 60 Hz hum was everywhere in the output, and could not be eliminated. In addition, looking into the drain of Q2 is very highly capacitive, the square-wave response had huge pulse tilt, for frequencies of only 5 KHz.
Comments
Post a Comment