Digital Communication across transformer allready powering a circuit

Hz

h speed to the transformer

00kHz for just a couple of periods. That can easily be detected.

r:

SI07.pdf

detection circuit on the primary side to the target frequency on the secon dary side (load side). That's done, since it would need to detect maybe sma ll current changes

al (to inject a current into the secondary side winding), and detecting tha t signal on the primary side. But, that is really not cheap anyway, at leas t I have not found a super easy way to do that yet

of the system, but I would really like to stay away from resonance

nothing?

there's a reason people use iso-optolators.

NT

Full duplex?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Is the transformer with 1:1 ratio or something else?

Are there rectifiers with capacitor-input filters behind the transformer?

Can you measure the primary, secondary and mutual inductances of the transformer?

If the transformer feeds a rectifier, you need to move the communication carriers pretty far from the main feed to avoid interference from the rectification. Another option could be a cleverly timed signaling synchronous with the feed, so that the rectification spikes do not interfere, but this need pretty involved processing.

--

-TV

It all depends on how high "fairly high" is :-)

Powerline interfacing has the advantage that you can buy chips for it but those add cost.

Probably the cheapest way is to force a carrier into the output side and detect that upstream. Depending on the situation ISM at 6.78MHz or

13.56MHz may be needed in order not to blow the radio cert. Depends on the market areas and where used. However, ISM down there realistically only offers a few kHz of BW.

As NT hinted an opto-coupler might cost less.

--
Regards, Joerg 

http://www.analogconsultants.com/

I cannot use any other component except the transformer

Cheers

Klaus

Simplex is ok

The transformer is 1 to 1 ratio.

The primary side is half bridge capacitive driven. Secondary side is full wave rectified by a bridge rectifier

Cheers

Klaus

I can measure the primary inductance, but that would be really difficult while feeding power

Cheers

Klaus

The upstream rate would be great to have minimum 100 kB

Power line communication with a standard Ic is way too expensive

Cheers

Klaus

Do you have a max cost target for this feature? What is available on either side in terms of micro controllers and stuff? Preferably ones that aren't maxed out for port pins and ROM space.

For example, the uC on one side could wiggle a port pin at a high frequency, that could be forced onto the transformer and is detected on the other side. Whether the uC can do the detection job on the cheap depends on its capabilities and its resources already used up for other purposes. For example, a PWM in the uC could move one side of its internal comparator to where it just doesn't trip over yet, making it very sensitive. Then when a signal is able to make it through a simple LC filter it starts to switch this comparator's output and this can be detected by the uC. Can be iffy if the drive voltage on the transformer is high.

If you meant 100kbit because modulating between 200kHz and 300kHz doesn't allow 100kByte: Another method is modulating the 200kHz of the transformer drive on both sides. This has to be synchronous and cycle by cycle. For example, one cycle is always normal and this sets the comparator threshold of the receiving uC. IOW a "neutral cycle". The next cycle is a "data cycle". This either has the transformer load modulating down which means a "1" or it doesn't and that means "0". The next one is a neutral cycle again. And so on. The same can be done from the upstream side. On the downstream side the uC has to self-synchronize. But as we hardware guys always say while ducking the flying paper wads and coffee mugs, that's just software :-)

--
Regards, Joerg 

http://www.analogconsultants.com/

Cycle skip? ...Jim Thompson

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

| 1962 |

Thinking outside the box... producing elegant solutions.

That works well for downstream. For upstream you could only skip current cycles but that requires that the downstream side represents a detectable load at all times. Meaning no sleep modes and stuff. One could have a FET that cuts the charging of the rectifier capacitor every other cycle for a "1" and not cut it for a "0". However, Klaus might object to the 10 cents that this FET costs :-)

--
Regards, Joerg 

http://www.analogconsultants.com/

{:->)

Actually have a client that inserts boards side-by-side and they communicate inductively thru the air... works just great... posted a problem a few weeks ago, turned out his new calibration scheme changed the frequency, screwing up the comm's, but we're now back in business. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| STV, Queen Creek, AZ 85142    Skype: skypeanalog |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

     Thinking outside the box... producing elegant solutions.

for upstream maybe you can phase modulate the rectifier.

--
This email has not been checked by half-arsed antivirus software

Max cost for additional circuits for the communication feature is something like 0.1 USD ish

I have a microcontroller on both sides of the transformer. The frequency ju mp on the primary side is done by a microcontroller and the load shift sens ing scheme was also supposed to be detected by that microcontroller, with s ome few spare parts

That is somewhat what I used for the first idea, but I am not very happy ab out something that is riding on an edge. It needs to be a very robust solut ion. For example, high external RF fields must not be able to disturb the f unction

Sounds like a good idea. The only problem is that modulating the primary si de half bridge doesn't couple straight through, since mutual and leakage in ductance of the transformer and non-zero currents makes it somewhat sensit ive to big changes (high Q effects)

Another idea could be to modulate the zero crossing just very minutely. So starting out with say 500kHz. Let the receiving microcontroller lock in to that frequency, and then just do FSK with very small percentage shift. That will result in a stable system, that doesn't jump up and down. Would howev er need a very good zero crossing detector and fast too

Cheers

Klaus

That is how I am doing it now. A small 2N7002 cost almost nothing, which sh orts out the secondary side, with a resistor in series so I can set the loa d jump

It is a little tricky, since the MOSFET body diode comes into play. Then I need two in series, of a JFET, but that is not very cheap either

I could place the MOSFET in series with the charging capacitor, like you sa id, cutting off the charging of the cap. That would be ok with respect to t he body diode. The modulating current will be small however. My target is b elow 10mW load on the secondary side

Cheers

Klaus

That sounds VERY simelar to what I am doing

Cheers

Klaus

With large qties 10 cents buys at least a handful of parts.

It would have to be a robust protocol with error correction and all that so that disturbances can be detected and blanked out.

What makes it robust is the cycle-to-cycle nature. There is always one neutral or reference cycle between all data cycles. Only every other cycle has data on it. So even in a high-Q situation your regulator loop would have to be blazingly fast (and kind of borderline stable) to disturb that.

But then it is tough to achieve your high data rates. For 100kbit/sec over a 200kHz or 300kHz transformer I think a frequency domain approach is like getting 25 people into a VW Beetle. Not that it hasn't been tried before ...

I usually prefer frequency domain but in this case my first choice would be time domain. In either case the amplitude would need to be high enough to overcome the worst anticipated RF disturbance. Error correction will always be needed if data integrity is critical because there can always be hard unexpected disturbances such as nearby lightning strikes.

--
Regards, Joerg 

http://www.analogconsultants.com/

Can't you use a regular diode in series with it to dodge the body diode? Those can be had for under a cent in qties.

That could be iffy WRT RF disturbance. However, that depends on what this circuit is going to be exposed to and how well it is EMI filtered.

--
Regards, Joerg 

http://www.analogconsultants.com/