AMC1203 is a deltamodulator that generates a 1 bit stream. you can digitally transfer that across the barrier using either opto/fiber or inductive methods.
for example a toroid sectorially wound with furakawa triple isolated wire. then a decimation filter or simple low pass filter. I built my first sigma delta modulator back in 1989 using a set of logic inverters as integrators. my boss had no idea what to do with it..
for example a toroid sectorially wound with furakawa triple isolated wire. then a decimation filter or simple low pass filter. I built my first sigma delta modulator back in 1989 using a set of logic inverters as integrators. my boss had no idea what to do with it..
I’ve seen optocouplers using fibre optic cable to get very high isolation voltages. Are you using this in the feedback loop Jan, the way they do in some SMPS regulators? I’ve also seen emitter and detectors coupled with a black plastic light pipe about 10 cm long.
Normally you have to slot the PCB to increase creepage distance.
Normally you have to slot the PCB to increase creepage distance.
I've built an isolating analogue optocoupler (using 2x6N137 and 1-bit switching) floating at around few hundred volts. However, I'm of the opinion that at 4000V, it maybe better to keep the dV/dt minimal.
https://www.ti.com/product/ISO121
https://www.ti.com/product/ISO121
Now for something completely different!
I can couple the audio input signal capacitively to the HV amp input.
6kV isolation caps are almost standard and I only need a fraction of a uF.
But there's a catch. The HV amp itself has the -4kV bus as reference, as local 'ground'.
And that -4kV is not regulated so has several V of ripple with respect to real gnd.
That has the effect as if the input signal contains the ripple too, and it gets amplified with the full HV amp gain.
A volt of ripple completely saturates the amp output...
A seemingly simple solution suggests itself: take a sample of the ripple, add it to the input signal in the proper phase to null the ripple on the output signal.
But that doesn't null to zero; I can only null it for the amp gain but the actual ripple voltage still appears on the HV amp output with a 'gain' of one.
That amp has a gain of more than 1000 and apparently that gain is compensated for but not the actual ripple magnitude.
Long story, I hope it's somewhat intelligible.
Jan
I can couple the audio input signal capacitively to the HV amp input.
6kV isolation caps are almost standard and I only need a fraction of a uF.
But there's a catch. The HV amp itself has the -4kV bus as reference, as local 'ground'.
And that -4kV is not regulated so has several V of ripple with respect to real gnd.
That has the effect as if the input signal contains the ripple too, and it gets amplified with the full HV amp gain.
A volt of ripple completely saturates the amp output...
A seemingly simple solution suggests itself: take a sample of the ripple, add it to the input signal in the proper phase to null the ripple on the output signal.
But that doesn't null to zero; I can only null it for the amp gain but the actual ripple voltage still appears on the HV amp output with a 'gain' of one.
That amp has a gain of more than 1000 and apparently that gain is compensated for but not the actual ripple magnitude.
Long story, I hope it's somewhat intelligible.
Jan
Bingo! The ripple occurs as common mode between the stators so causes no output.
I do need to cancel the ripple in the input stage as mentioned before to avoid saturating the amp with it.
Then the CM ripple level on the amp output is limited to the ripple level on the -4kV supply, but not amplified.
Cancellation can be done with a simple diff input stage, the ref of which is the AC ripple I can couple in with another HV cap.
I just need per channel two 10nF 6kV WIMA's that Mouser sells for a couple of bucks each https://eu.mouser.com/datasheet/2/440/e_WIMA_FKP_1-1139839.pdf
No longer need ticking optocoupler time bombs.
Happy camper!
Jan
I do need to cancel the ripple in the input stage as mentioned before to avoid saturating the amp with it.
Then the CM ripple level on the amp output is limited to the ripple level on the -4kV supply, but not amplified.
Cancellation can be done with a simple diff input stage, the ref of which is the AC ripple I can couple in with another HV cap.
I just need per channel two 10nF 6kV WIMA's that Mouser sells for a couple of bucks each https://eu.mouser.com/datasheet/2/440/e_WIMA_FKP_1-1139839.pdf
No longer need ticking optocoupler time bombs.
Happy camper!
Jan
It may be a good idea to choose the coupling time constant very much greater than 3 ms or to use bootstrapping to ensure capacitor tolerances don't mess up the suppression. You can easily get 0.1 % tolerance resistors nowadays and trim away any remaining error with a trimmer potmeter (*), but those capacitors have 5 % to 20 % tolerances.
(*): That may actually not be trivial when the trimmer potmeter is at a potential of -4 kV and you want to survive the trimming procedure.
(*): That may actually not be trivial when the trimmer potmeter is at a potential of -4 kV and you want to survive the trimming procedure.
Last edited:
I once spanned 10kV with an analog isolator. It used a fiberoptic cable to connect the transmitter and receiver. Transmitter used two leds in series, one for local feedback of light emission. An overall DC feedback loop was closed around the optical isolator system using a frequency compensated HV resistor array back to ground reference level.
EDIT: Looks like a solution for Jan has already been found.
EDIT: Looks like a solution for Jan has already been found.
Last edited:
Marcel, yes I can trim it at the ground level, at the input.
But because there's all that gain coming after that, the trim minimum is extremely sensitive.
In LTspice it needs adjustments of a factor 0.0001 or something like that. Not practical.
Probably best to take large caps; I don't mind if the cancellation deteriorates below 100Hz as the ESL will be crossed over very sharply at around 200Hz.
Jan
It's pretty expensive, though. Maybe actually going TOSLINK RX/TX modules with a very short cable is more promising.
You need a lot of photons for a low noise signal though, which is hard to push through a thin fibre. Electrons in a wire are vastly more numerous and move in concerted fashion!
For instance this commercial unit only gets 55dB SNR: https://www.analogmodules.com/admincenter/datasheets/732tr.pdf
For instance this commercial unit only gets 55dB SNR: https://www.analogmodules.com/admincenter/datasheets/732tr.pdf