So, I am referring over the two weeks old theme from PS section as it much more refers to measurements not a power.
The task was to design an optically isolated oscilloscope probe working at multiple GHz scale and giving more than 100 kV DC bias-caused security between 16 channels, and having an expectable small crosstalk between channels, means cross-capacitance. Thus only way to fulfil is fiber optics.
Now, few days after the Bremen Space Tech exhibition I may refer to the precised names of core-importance components: it must be Emcore 2651E receiver (4GHz) and Emcore 1933 or 1935 linear laser transmitter (4 GHz). In the input I went away from the FET with a tears in eyes, but only bjt have a cheap 90 GHz and 0,03 pF exemplare at 0,1 USD named BFU910FX. Thus, for now I meditate about the transmitter part circuitry.
Firstly, no doubt, it ought to incorporate the Wilson 4-transistor type of current mirror. Input then is between Vcc and upper shoulder bjt base, where the some 1,6 V a least voltage is unavoidable against Vcc. Therefore ought take one such mirror in input wire and other identical in braiding line. Thus the voltage between input terminals will be brilliant zero, or other wording - the DC measurement capability is set on.
But there I struck. Because i(c max) for those modern days wunderbars are 15 mA only. Therefore laser which needs a 150 mA have to be fed somehow rather tricky, simply not exist any other more amperage standing transistors except the diamond-semiconductor what are too slow in their way to shelves (invented about 2005).
One idea was ignored from very beginning, it is add the ten emitter followers working on common load. Ignored because of the slow reaction expected and need to apply the surplus emitter resistors of few Ohm tenths what makes a GHz scale PCB patch lengths problematic. However may be it was too fast so cut own tail??
Other idea is to duplicate all the right side of circuit (fig 4) at www.planetanalog.com/an-analog-tutorial-current-sources/?f_src=planetanalog_section_3065 somehow ten-fold in the raw like here in the second figure from the top https://en.wikibooks.org/wiki/Analogue_Electronics/Current_Mirrors/Wilson . Like soldiers in the death-battle. So adding a 20 surplus bjt~s. So, now I started to think what about the frequency here? Didnt I just killed a frequency response 10-fold?? At least with the mosfets in normal halfbridge output cascade it would happen for 1000% sure.
By the way, it may be possible to substitute the 90 GHz F(T) to 65 GHz and then the BFP620F; E; H; FH are available for 80 mA. Only two may carry the amperage. But if to put much different made bjts where it must be matched it will certainly give a problems of "random character". Or put all bjts of this kind will bring a problem of input current so big that may change the DUT.
Probably is worth to add the cacsode for amplification??
Your ideas how to overcome this problem without of killing the Gigahertzes are impatiently awaited.
P.S. More, at the max V(ce)=3V there are no much options as to put the pull-up resistors as 100 Ohm. Thus the input active resistance between coaxial braiding to central wire stands 200 Ohms and no way how to 50 or 75 Ohms. Yet my hope is that is rather unimportant as the physical distance between the transistor and point of DUT be measured may easily be some 2...3...5 mm and for 3 GHz or 100 mm lambda it makes a 1:20 or 1:40 of lambda. Thus to conjugating the impedances is no any need. Am I true?
The task was to design an optically isolated oscilloscope probe working at multiple GHz scale and giving more than 100 kV DC bias-caused security between 16 channels, and having an expectable small crosstalk between channels, means cross-capacitance. Thus only way to fulfil is fiber optics.
Now, few days after the Bremen Space Tech exhibition I may refer to the precised names of core-importance components: it must be Emcore 2651E receiver (4GHz) and Emcore 1933 or 1935 linear laser transmitter (4 GHz). In the input I went away from the FET with a tears in eyes, but only bjt have a cheap 90 GHz and 0,03 pF exemplare at 0,1 USD named BFU910FX. Thus, for now I meditate about the transmitter part circuitry.
Firstly, no doubt, it ought to incorporate the Wilson 4-transistor type of current mirror. Input then is between Vcc and upper shoulder bjt base, where the some 1,6 V a least voltage is unavoidable against Vcc. Therefore ought take one such mirror in input wire and other identical in braiding line. Thus the voltage between input terminals will be brilliant zero, or other wording - the DC measurement capability is set on.
But there I struck. Because i(c max) for those modern days wunderbars are 15 mA only. Therefore laser which needs a 150 mA have to be fed somehow rather tricky, simply not exist any other more amperage standing transistors except the diamond-semiconductor what are too slow in their way to shelves (invented about 2005).
One idea was ignored from very beginning, it is add the ten emitter followers working on common load. Ignored because of the slow reaction expected and need to apply the surplus emitter resistors of few Ohm tenths what makes a GHz scale PCB patch lengths problematic. However may be it was too fast so cut own tail??
Other idea is to duplicate all the right side of circuit (fig 4) at www.planetanalog.com/an-analog-tutorial-current-sources/?f_src=planetanalog_section_3065 somehow ten-fold in the raw like here in the second figure from the top https://en.wikibooks.org/wiki/Analogue_Electronics/Current_Mirrors/Wilson . Like soldiers in the death-battle. So adding a 20 surplus bjt~s. So, now I started to think what about the frequency here? Didnt I just killed a frequency response 10-fold?? At least with the mosfets in normal halfbridge output cascade it would happen for 1000% sure.
By the way, it may be possible to substitute the 90 GHz F(T) to 65 GHz and then the BFP620F; E; H; FH are available for 80 mA. Only two may carry the amperage. But if to put much different made bjts where it must be matched it will certainly give a problems of "random character". Or put all bjts of this kind will bring a problem of input current so big that may change the DUT.
Probably is worth to add the cacsode for amplification??
Your ideas how to overcome this problem without of killing the Gigahertzes are impatiently awaited.
P.S. More, at the max V(ce)=3V there are no much options as to put the pull-up resistors as 100 Ohm. Thus the input active resistance between coaxial braiding to central wire stands 200 Ohms and no way how to 50 or 75 Ohms. Yet my hope is that is rather unimportant as the physical distance between the transistor and point of DUT be measured may easily be some 2...3...5 mm and for 3 GHz or 100 mm lambda it makes a 1:20 or 1:40 of lambda. Thus to conjugating the impedances is no any need. Am I true?
Last edited: