Oh thanks, what a nice and great way to say it!
I was just thinking, that if a transformer alone could not do the job, why not combine several solutions in one circuit
What you just suggest, to combine a number of transfer elements in a circuit that in total can handle the full frequency spectrum, is possible. It will hardly be simple, but possible.
Then for the use: Someone wants DC-10KHz (off-line SMPS, today an opto coupler); someone DC-50Hz (oscilloscope isolation, today often floating operation); someone wants 100Hz-10KHz for a control purpose etc. Some need precision, some not. All may like parts of your circuit but do not want all because they have no need for all and all is to big and expensive. The idea of covering DC-1GHz is ambitious but outside the scope of most.
The wish to isolate two circuits where you have analog information transferred from one to the other has existed for many decades. Look at the isolation amplifiers from Analog Devices and Burr Brown. Today, a number of solutions exist that cover different frequency ranges because most needs are for more limited frequency ranges. My fear is that an all-in solution will be too complex and expensive.
Some inspiration:
Isolated Oscilloscope Probe | Elektor Magazine
Isolated Current Probes & Differential Voltage Probes | Powertek
Last edited:
ref the scheme in original post: Am I being dumb?
The transformer has ac to the primary and supplies ac from the secondary - power in, power out.
The optical isolator takes a signal on the input side and switches the output side accordingly - it needs a supply on the secondary side to switch. It does not pass a voltage, it passes a signal. It would provide isolation of a dc voltage - complete isolation as nothing would come out.
The transformer has ac to the primary and supplies ac from the secondary - power in, power out.
The optical isolator takes a signal on the input side and switches the output side accordingly - it needs a supply on the secondary side to switch. It does not pass a voltage, it passes a signal. It would provide isolation of a dc voltage - complete isolation as nothing would come out.
I have taken several looks on diferent chips, transformers and what not, but to me do almost all of them ave som major downsides, except maybe the transformers... until I ded here about them not being to great for that purpose. If we say 20Hz to 350MHz, there is almost no problem in buying chips that, combined, can handle that range, but only up to maybe 0.5V. The transformers can handle fare more, about 500V but again, they should also be bad.
So maybe the most important question, how do I attack the problem?
You are right, but it's a digital isolator
How it exatly work is a bit abowe my level right now, the schematic for the isolator is from the datasheet, I did only put the transformer in parallel to illustrate my idea. 
Some loose ideas in bulk:
The concept of combining a LF/DC channel with a fast, pure AC one is not new: one of the first examples that comes to mind is the Tektronix current probe: it used a Hall sensor (DC/LF) and a current transformer (HF AC) combined in a single element to provide a DC to 100MHz bandwidth.
It is just a current sensor, but with a bit of imagination and some more turns than a single passage in the loop it could become the ideal device you strive for.
100MHz is minuscule by today's standards, but it was developed in the seventies, and it is certainly possible to do much better now using the same principles.
There are snags, because combining the two channels is not as straightforward as it seems, because of relative phase, but the problem has been treated in composite amplifiers (no galvanic isolation, but identical problem), for example here:
https://www.analog.com/media/en/technical-documentation/application-notes/an21f.pdf
You can find many other examples using similar search terms.
Note that a digital optocoupler is certainly not a solution for a linear transfer: you would need some kind of 1-bit/PWM modulation requiring significantly more bandwidth.
Fortunately, there are lasers and photodiodes having many tens of GHz of bandwidth.
Most of them are oriented towards digital applications, and that's how I used them when I worked in the telecom field, but I was aware that fast, linear devices were also available, mainly for cable TV applications.
Today, you can find fiber-based acquisition systems for testing ultra-high voltage systems: there is no other way to test GigaVolt power systems with DC and sub-ns capabilities combined.
Probably too expensive for a DIYer, but you should look to Chinese sources: they have to develop energy systems on a large scale, and can probably offer very cheap, reasonably effective acquisition systems sufficient for DIY.
It is possible to combine the output of two transformers, see an example here:
Splitting the audio band into two transformers
but it is not going to pass DC, and to my knowledge, the phase inversion in the Xover zone is unavoidable.
Note that transformers can be built to work for 4~5 decades, if the isolation voltage is not too high.
I have built such transformers for measurement applications, working in a 10kHz to 1GHz range.
They were not flat in this range, but the imperfections were correctable by a VNA calibration, and for 4 decades, they were directly usable for oscilloscopes or TDR's.
The signal handling capability was also severely limited at low frequencies, because of saturation, and the construction was fiendishly complicated: no such thing was available as standard from North Hills or Microcircuits.
Isolation amplifiers are a possible solution when you don't need a too large bandwidth:
Selection Table for Isolation Amplifiers | Parametric Search | Analog Devices
Some even include a supply for the isolated side
One of the most attractive solutions are differential probes: they do not provide a complete isolation, but the CM resistance of many megohms is generally not a problem.
They may look like overkill, because in fact you don't need the differential capability, but you simply use it for ground translation (and for true differential measurements if you need them).
Here is a very simple, cheap and low performance example, but I have used the same principles for a 3kV/30MHz probe (slightly more complex, of course):
A safe and inexpensive probe for direct mains measurements
The concept of combining a LF/DC channel with a fast, pure AC one is not new: one of the first examples that comes to mind is the Tektronix current probe: it used a Hall sensor (DC/LF) and a current transformer (HF AC) combined in a single element to provide a DC to 100MHz bandwidth.
It is just a current sensor, but with a bit of imagination and some more turns than a single passage in the loop it could become the ideal device you strive for.
100MHz is minuscule by today's standards, but it was developed in the seventies, and it is certainly possible to do much better now using the same principles.
There are snags, because combining the two channels is not as straightforward as it seems, because of relative phase, but the problem has been treated in composite amplifiers (no galvanic isolation, but identical problem), for example here:
https://www.analog.com/media/en/technical-documentation/application-notes/an21f.pdf
You can find many other examples using similar search terms.
Note that a digital optocoupler is certainly not a solution for a linear transfer: you would need some kind of 1-bit/PWM modulation requiring significantly more bandwidth.
Fortunately, there are lasers and photodiodes having many tens of GHz of bandwidth.
Most of them are oriented towards digital applications, and that's how I used them when I worked in the telecom field, but I was aware that fast, linear devices were also available, mainly for cable TV applications.
Today, you can find fiber-based acquisition systems for testing ultra-high voltage systems: there is no other way to test GigaVolt power systems with DC and sub-ns capabilities combined.
Probably too expensive for a DIYer, but you should look to Chinese sources: they have to develop energy systems on a large scale, and can probably offer very cheap, reasonably effective acquisition systems sufficient for DIY.
It is possible to combine the output of two transformers, see an example here:
Splitting the audio band into two transformers
but it is not going to pass DC, and to my knowledge, the phase inversion in the Xover zone is unavoidable.
Note that transformers can be built to work for 4~5 decades, if the isolation voltage is not too high.
I have built such transformers for measurement applications, working in a 10kHz to 1GHz range.
They were not flat in this range, but the imperfections were correctable by a VNA calibration, and for 4 decades, they were directly usable for oscilloscopes or TDR's.
The signal handling capability was also severely limited at low frequencies, because of saturation, and the construction was fiendishly complicated: no such thing was available as standard from North Hills or Microcircuits.
Isolation amplifiers are a possible solution when you don't need a too large bandwidth:
Selection Table for Isolation Amplifiers | Parametric Search | Analog Devices
Some even include a supply for the isolated side
One of the most attractive solutions are differential probes: they do not provide a complete isolation, but the CM resistance of many megohms is generally not a problem.
They may look like overkill, because in fact you don't need the differential capability, but you simply use it for ground translation (and for true differential measurements if you need them).
Here is a very simple, cheap and low performance example, but I have used the same principles for a 3kV/30MHz probe (slightly more complex, of course):
A safe and inexpensive probe for direct mains measurements
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.