EEVblog #909 - Defibrillator Teardown

[mcinque]

Maybe it's a noob question but... why 3 diodes in series after the transformer? 1 diode doesn't do the job for a half wave rectifier as intended?

[b_force]

Maybe it's a noob question but... why 3 diodes in series after the transformer? 1 diode doesn't do the job for a half wave rectifier as intended?

Probably because the huge voltage spikes from the switching.
Bear in mind that we are talking about 1500-2000V, which means that switching spikes will be even higher.

[mikeselectricstuff]

Maybe it's a noob question but... why 3 diodes in series after the transformer? 1 diode doesn't do the job for a half wave rectifier as intended?

Probably because the huge voltage spikes from the switching.
Bear in mind that we are talking about 1500-2000V, which means that switching spikes will be even higher.

could also be  a safety thing in case a diode fails short - if it dumped all the cap energy into the transformer secondary it would probably explode.

[chris_leyson]

This is somewhat off topic but about 20 years ago the company I was working for had an inverter designed by a third party to charge up a bank of capacitors fairly quickly. You could hear the magnetics in the converter changing frequency slightly as the capacitors charged up and a colleauge of mine who had worked on defibrillators in the past said that when they had self oscillating converters running at tens of kHz you could "hear" the converter running and knew when the capacitors were charged. As the electronics improved and the switching frequencies got higher you couldn't hear the converter anymore, some customers complained and they had to add additional electronics just to generate the classic defibrillator sound.

[AF6LJ]

This is going to be Good.....
I have to wait until my work is done...

[max666]

Couple other things about them: a +10%/-10% spec is a little unusual for an electrolytic, but obviously required for this thing since you don't want too much energy available & can't overload that smaller SCR.  Also there doesn't appear to be balancing resistors so perhaps they're matched for leakage.  Maybe that's what the dots are indicating.  Throw in a form factor requirement and you might be married to one manufacturer.

The diodes appear to be straight up 1000v PIV so they're not pulling anything fancy there.

Huh, that's somewhat disappoint, I was hoping those diodes could be TVS (transient-voltage-suppression) diodes, that way you would have reverse polarity and over voltage protection. Or are TVS diodes a bad substitute for a balancing circuit?

[ChunkyPastaSauce]

The RTC is for the built in event data recorder, which records not only time stamp data of events but also records an ECG during each stage intervention. The software can also calculate CPR information from the data.

The data recorder can be read out using their software, see http://heartsinelive.s3.amazonaws.com/uploads/2015/06/H013-001-400-2-US-ENGLISH-SAVER-EVO-1.4.0-MANUAL.pdf

[Fenichel]

This particular (as most AEDs) will actually detect CPR being performed and provide visual/audible feedback. Every compression on the chest is detected and measured by the device. The AED will only allow a shock when it detects a certain irregular heartbeat. It won't help anyone who has no cardiac output.

  This is almost right, but "cardiac output" is a plumbing term, not an electric one.  All patients on whom defibrillators are used have no cardiac output.  What they must have are signs of cardiac electrical output.

  When the plumbing and wiring are all working well, the heart pushes out a few liters of blood every minute.  That is cardiac output.  Sometimes all the wiring is OK, but there are leaks, obstructions, or nonstandard paths for blood to take, other than the proper one out to the body.  After heart attacks, parts of the cardiac muscle are dead, so the pump can't properly respond to electrical signals, however well organized those signals are.  In any of these situations, the electrocardiogram may be almost normal, but the cardiac output may be very low.  Defibrillation wouldn't be relevant, and an AED would refuse to fire.

  Sometimes the plumbing is fine, and the muscle is healthy, but the wiring is shot.  Instead of getting organized signals, the muscle gets useless chaotic instructions, or none.  Cardiac output may be low or zero. 

• Looking at the electrocardiogram, one may see a pattern of electrical activity that has a fair chance of responding to defibrillation.  If the AED thinks that's what's happening, it will fire.  It's analogous (very loosely) to rebooting the computer. 
• Other electrical activity seen on the ECG may be responsible for poor, even life-threatening, pump performance, but not likely to benefit from defibrillation.  There may be other appropriate interventions, depending on what is seen.
• The wiring may be intact, but the pulse generator may be working only intermittently or too slowly.  One can see that on the electrocardiogram, too.  That's when one wants a pacemaker.

[Cerebus]

This is somewhat off topic but about 20 years ago the company I was working for had an inverter designed by a third party to charge up a bank of capacitors fairly quickly. You could hear the magnetics in the converter changing frequency slightly as the capacitors charged up and a colleauge of mine who had worked on defibrillators in the past said that when they had self oscillating converters running at tens of kHz you could "hear" the converter running and knew when the capacitors were charged. As the electronics improved and the switching frequencies got higher you couldn't hear the converter anymore, some customers complained and they had to add additional electronics just to generate the classic defibrillator sound.

Older professional photoflash packs are the same. After a while working with a particular pack you learned to judge the state of charge of the pack from the pitch and volume of the charging whine.

What I felt was missing from this teardown was an analysis of what had been done to combine circuitry that whacks out several hundred joules at 2kV with circuitry that amplifies around 60 mV peak nerve pulses - some interesting input protection challenges there.

[Brumby]

The RTC is for the built in event data recorder, which records not only time stamp data of events but also records an ECG during each stage intervention. The software can also calculate CPR information from the data.

The data recorder can be read out using their software, see http://heartsinelive.s3.amazonaws.com/uploads/2015/06/H013-001-400-2-US-ENGLISH-SAVER-EVO-1.4.0-MANUAL.pdf

I was just thinking .... the RTC clock doesn't need to be second perfect.

A unit can lay idle for years - and even if the clock is out by hours in absolute terms, it will still record fairly accurate timestamps, relatively speaking.  If you want accurate absolute time, simply connect the device to a computer with an accurate clock, determine the difference and apply that as an offset to get sub-second accuracy.

[trophosphere]

Generally speaking, efibrillation protection is a lot more straight-forward than the rest of the circuitry given that a couple things are addressed.

• The first is that the protection circuitry must not shunt an excessive amount of energy through itself and therefore "taking away" energy (energy reduction) that can otherwise used to depolarize the heart muscles: This can be fulfilled by placing current limiting resistors before the protection circuitry.
• The protection circuitry needs to have sufficient spacing/insulin/power rating to withstand multiple pulses.
• The high-pass filter used to remove DC offset/baseline wander needs to recover from saturation (after being subjected to defibrillation) within a specified amount of time.

An example pathway would be current limiting resistor->GDT/MOV->resistor->TVS/Zener/Capacitor*->resistor->diode (need to specify adequate dv/dt rating)->resistor->Input to In-Amp
*Capacitor acts as low pass filter and can be used for common mode/differential mode filtering

Note that with this protection circuitry in place, there is a trade-off between protection and SNR.

[Fungus]

Maybe it's a noob question but... why 3 diodes in series after the transformer? 1 diode doesn't do the job for a half wave rectifier as intended?

Safety.

(belt + braces + a piece of string)

[AF6LJ]

The USB to Serial cable is most likely for firmware updates.
I did a firmware update on an American Made AED for a client about eight years ago. A cable was supplied with the unit, Firmware was supplied from the manufacturer's website and was required to maintain certification. The firmware update app required Internet access in order to update the AED, not only for downloading the software but to confirm that AED had been updated.

[tatus1969]

I am wondering about one thing: they use two SCRs in series, where one of them would not be able to widthstand the working voltage. I know that this would be safe when using MOSFETs, because their overvoltage breakdown mechanism is of the Avalanche type and acts like a zener diode, so we only would need to care about thermal limits. Of course one would choose a particular part that is actually specified for going to that limit.

Does anyone know more about SCRs and how they behave during overvoltage? I looked in the datasheets but couldn't find anything. Also, voltage distribution between the two devices will depend on their leakage currents, and these aren't defined as well, and they will certainly spread widely. Unless you select devices, you have to expect that one of the two may see most of the bus voltage, and the other, more leaky one, will see only a little. Will the first one then break down because of overvoltage?

[SeanB]

SCR overvoltage behaviour is very simple, they turn on. Either the avalanche leakage current as it approaches the breakdown voltage turns the junctions on, or the rate of rise of voltage induces a capacitive current that does the same thing. Using 2 SCR devices in series means they will be able to still function even if one is leaky or failed short, as the other is a back up for the off state guarantee.

Turn off is done with the relays, which are energised to hold the contacts closed during the pulse, but they are released when the pulse is at a lower level, using the breaking ability of the relay to interrupt the current pulse. Then, after a short interval, they close again, so the reverse polarity pulse can be applied, with the massive high voltage SCR using it's turn on immunity to not trigger during the positive pulse and short the power capacitor bank, with the snubber assisting there as well. The reverse pulse is lower energy, and needs a better control of turn off, thus the IGBT devices to ensure a fster turn off than the relay can provide.

Then the relays are released, and the monitoring circuit can then read the heart again with minimal interference.  Relay life, even for those Matsushita units, will be very poor, around 500 cycles, but for a device that in it's entire life will only have 50 cycles of test firing in production, and perhaps another 20 cycles over the entire standby time, that is still a very reliable lifetime.

Failure mode of a SCR depends entirely on the energy available, it will always turn on as a protective mechanism, and then fail as a short if the energy is moderately past the maximum it can handle, which overheats the die and destroys it, or it will blow open if the switched energy is too large and at the level of fusing the bond wires.

[tatus1969]

SCR overvoltage behaviour is very simple, they turn on. Either the avalanche leakage current as it approaches the breakdown voltage turns the junctions on, or the rate of rise of voltage induces a capacitive current that does the same thing.

Sounds good. So they will effectively balance out when sharing the voltage, and cannot be destroyed when excessively unbalanced.

[SeanB]

SCR overvoltage behaviour is very simple, they turn on. Either the avalanche leakage current as it approaches the breakdown voltage turns the junctions on, or the rate of rise of voltage induces a capacitive current that does the same thing.

Sounds good. So they will effectively balance out when sharing the voltage, and cannot be destroyed when excessively unbalanced.

No, they are still effectively limited to the current of a single device, and the voltage of a single one as well. All you get is redundancy, unless you add balancing resistors and can live with this extra off state current.

In this application it is so that there is a much lower chance of a false triggering on when the other side is fired, as the turning on of a single device is not going to turn the other on easily, and if both turn on at once there will be a very big bang as the capacitors all discharge through this and blow them apart.

[bktemp]

Turn off is done with the relays, which are energised to hold the contacts closed during the pulse, but they are released when the pulse is at a lower level, using the breaking ability of the relay to interrupt the current pulse.

They look much too small to me to be able to interrupt something like 1kV @ >10A DC safely. At that power levels you typically need much larger relays with blowout horns or other arc quenching devices.
The relays used are RTE24012, rated for 250VAC 10A or 250VDC 0.2A. It looks like they have connected both contacts in series on each relay, but it is still way out of specs if it is used for interrupting any current at 1kV.

[SeanB]

The relays only have to interrupt current long enough for the SCR's to turn off, thus they will not have high voltage applied to them, and only have to break a current for a short time.

[bktemp]

That does not work:
If you disconnect 1kV DC at several amps, it will form an arc between the contacts shorting the interruption and dropping only some 10 volts. So the SCRs will not turn off.

[max666]

SeanB, I know you are a very smart cookie and I won't pretend to know what is going on, but I can not believe those relays are used to break the current. Like others have mentioned, I don't think that relay (TE Connectivity/Schrack RTE24012) is up to the task of breaking up that discharge. And even if it is, it would be right at the absolute maximum ratings and I don't think a medical device would be designed that way, especially since others keep throwing around the term "belt + brace" as reasons why things are done the way they are, in this unit.
To me the function of those relays is to isolate the high voltage capacitor bank from the helper handling the pads (SCR's could break down, or leak, or how about ESD from touching the pads).

But even ignoring that, the relay has a max. release time of 6ms, the positive pulse is 4ms after which you should break it. How could you possibly operate a relay reliably in the ms range?

[tatus1969]

SeanB, I know you are a very smart cookie and I won't pretend to know what is going on, but I can not believe those relays are used to break the current. Like others have mentioned, I don't think that relay (TE Connectivity/Schrack RTE24012) is up to the task of breaking up that discharge. And even if it is, it would be right at the absolute maximum ratings and I don't think a medical device would be designed that way, especially since others keep throwing around the term "belt + brace" as reasons why things are done the way they are, in this unit.
To me the function of those relays is to isolate the high voltage capacitor bank from the helper handling the pads (SCR's could break down, or leak, or how about ESD from touching the pads).

But even ignoring that, the relay has a max. release time of 6ms, the positive pulse is 4ms after which you should break it. How could you possibly operate a relay reliably in the ms range?

When talking about this being a safety appliance, my first guess would be that they let the SCRs do all the high v/i switching work, and put the relays in series with them in order to have a reliable air gap between the dangerous energies and the "PUT".

[bktemp]

When talking about this being a safety appliance, my first guess would be that they let the SCRs do all the high v/i switching work, and put the relays in series with them in order to have a reliable air gap between the dangerous energies and the "PUT".

Yes, the relays are most likely only an additional safety feature in case one of the SCRs or IGBTs fails. They may help when sensing the heartbeat because otherwise the leakage current from the SCRs or their parasitic capacitance could disturb the very low level signals.

[Brumby]

Yes, the relays are most likely only an additional safety feature in case one of the SCRs or IGBTs fails.

Indeed.

I don't think the relay operating speed has any real impact on the timing of the shock pulses - other than to be closed before the pulses are delivered and open after they are finished.  100ms either side would do the job and this is easily within the capabilities of a relay.

They may help when sensing the heartbeat because otherwise the leakage current from the SCRs or their parasitic capacitance could disturb the very low level signals.

This sounds like a far more likely explanation.

[mal_uk]

I'm trying to work out the thyristor commutation on this thing
I can see how it might work by momentarily using the igbts to divert the current from the bottom scr and turn it off
but...
going back to the description of the pcb, Dave you suggested there are three igbts and 2 1200v SCRs so is the bottom SCR on the Davecad diagram actually an IGBT?
If thats the case it would be a deal less clunky and you could actually redraw the circuit as a bridge with SCRs as the top legs, IGBTs as the bottom and the PUT (load?) across the middle