Signal Input Protection Diodes

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
squarewave said:
Diverting into power supply rails doesn't make as much sense to me. I think it makes the most sense to place protection diodes right next to pins (between 2 and 1 and 3 and 1) and then use the XLR sockets that have the metal spike that connects pin 1 to the chassis so that shocks or whatever transients will be shunted directly into the chassis without getting that far inside the enclosure.

Clamping to the power supplies is actually a good thing. The power supply voltage easily defines the maximum input voltage that the active devices can handle, so why not use that as the clamp voltage. Sure, you can choose some random voltage and pick a TVS part that clamps to that voltage, but won't you then be clamping to some arbitrary voltage, not based upon what the circuit can handle, but which TVS part you bought? The amplifier data sheet will tell you what they can handle, usually 0.7V beyond the rails, or maybe 0.3V beyond the rails. Since that's the spec, isn't it simplest to clamp to those rails, whatever that voltage is?

Also consider the situation when your device is driven with an input signal, yet it is unpowered. In that state, your device's power supply rails will be at zero, so that's a good voltage to clamp an I/O jack to, and not some random TVS part's threshold.

Finally, in terms of distortion, it's best to put your clamp voltages as far away from the active signal region as possible, since the clamp diode capacitance modulation is something worth avoiding, and it gets worse when the reverse bias across a diode decreases, i.e. when your signal gets close to making the clamps conduct. Placing the diode clamps at the rails will optimize this.
 
Most supply rails are populated with bypass caps that can absorb transient charge easily. If you work that back to the regulators and the primary power source, you find that you can grind down all manner of over-voltage pretty easily, back to the power cable that feeds the device.

Dumping BS to  the rails is a good thing! :)

Mind you, if the device is powered up, most regulators and bypass caps can just "eat" the fault directly. If powered down, then the fault can still get handled by some pretty meaty power supply components on the way back to the power line cord.
 
Monte McGuire said:
Sure, you can choose some random voltage and pick a TVS part that clamps to that voltage, but won't you then be clamping to some arbitrary voltage, not based upon what the circuit can handle, but which TVS part you bought?
It's not random. I use 16.7V TVS diodes right now. ICs almost always already have protection diodes. The problem is that often times they can only handle low-energy transients. TVS diodes are designed to handle the high energy transients that can actually damage the IC.

And you assume all of the circuitry and traces and parts between the input pins and power supply capacitors can handle high energy transients. What happens if it's a 1kV spike - can the caps handle that? Could it cause an arc to another trace that is routed too close?

Capacitance of TVS diodes is an issue yes. I have made that mistake. But from checking Mouser it seems there are some parts with very low capacitance (3pF). And they're one tiny 3mm part per signal line.
 
squarewave said:
What happens if it's a 1kV spike - can the caps handle that? Could it cause an arc to another trace that is routed too close?
It all depends on the capacity of the EMC source to deliver energy. 1kV from a typical low capacitance EMC source is clamped by a mere 100pF cap.
You cannot protect against high-power surges; for example, if you connect the 120V mains to the input, something will and has to give up. A good designer will ensure that a low-cost resistor burns before a costly  and hard to replace IC, or before the PCB traces.
 
JohnRoberts said:
This is a very old and well inspected topic.

Arguing with an experienced designer like Abbey is not productive.  What he said again.

JR
It's not an argument if no one presents arguments. So far the only real argument was posited by Monty who argued that the voltage should be relative to the device that is being protected. Ok. That makes sense. It's really not that compelling IMO but it is a valid argument. Most of the other replies seem to lean on "this is a very old and well inspected topic" that apparently everyone should already know. So then what's the point in asking questions on a forum? And telling people to search through forums has always been a weak punt IMO. If someone provides a good detailed answer to a specific topic, then provide a link to that post. But forum discussions very rarely answer questions completely or directly and group wisdom evolves over time. New parts become available. Are low-capacitance TVS diodes new? If yes, then maybe they should be considered? Why shunt transients into the rails if you don't have to? Do low capacitance TVS diodes meet all of the requirements for protecting circuits? If no, please explain why because so far I have not heard any arguments about that. I actually don't know myself. I'm just going on what the datasheet says.
 
squarewave said:
It's not an argument if no one presents arguments.
semantics
So far the only real argument was posited by Monty who argued that the voltage should be relative to the device that is being protected. Ok. That makes sense.
In fact it needs to be clamped to less than device power terminal voltages to prevent unintended conduction paths (like intrinsic diodes.) For low to modest voltage applications it is not unusual to clamp directly to ground, but lots of hand wringing over stray capacitance and perhaps nonlinear impedances (with voltage) on high integrity inputs.
It's really not that compelling IMO but it is a valid argument. Most of the other replies seem to lean on "this is a very old and well inspected topic"
yup
that apparently everyone should already know.
that Abbey surely knows.
So then what's the point in asking questions on a forum? And telling people to search through forums has always been a weak punt IMO.
If you go back and look at my several answers I did not make that suggestion yet..

That said questioning good advice can be confusing to people who do not have a firm footing with the technology.
If someone provides a good detailed answer to a specific topic, then provide a link to that post. But forum discussions very rarely answer questions completely or directly and group wisdom evolves over time. New parts become available. Are low-capacitance TVS diodes new? If yes, then maybe they should be considered? Why shunt transients into the rails if you don't have to? Do low capacitance TVS diodes meet all of the requirements for protecting circuits? If no, please explain why because so far I have not heard any arguments about that. I actually don't know myself. I'm just going on what the datasheet says.
I have never had to use a specialized TVS diode in decades of (audio) design... in fact I had to google it to understand the acronym. I have seen them used on telephone lines, and perhaps cable tv boxes where transient hazards are more common  (but I never designed one of those either.)

The THAT corp white paper I suggested is a good reference about input*** protection for audio lines.

JR

*** output protection is a slightly different animal and I prefer to use some build out resistance between the very output and the clamp diodes... of course this involves a trade off against lowest output impedance.

 
 
Some devices fail but still pass signal. Large geometry input devices can do that, and just give you more distortion. If you don't mind testing constantly, then yes, that'd work ;-)
 
JohnRoberts said:
Yes back to back (anti-parallel) diodes to ground  from the - input is simple and effective with inverting amps, BUT you said non-inverting in your original post.

Also using JFET gate diodes as low leakage diodes is another old trick if the cost is acceptable.

JR

Yes but I don't think anyone has suggested anti parallel diodes for a non inverting config. The TVS diode mentioned is bidirectional and allows for symmetrical signal swing.
JFet as diode - yes not a new thing and I designed the circuit I mentioned back in around 2000. I had just forgotten how low the leakage could be.
I'm wondering how it compares to a BJT configured as Base to Collector diode. I don't have capability to make those sort of measurements atm.
 
squarewave said:
Ick! They have super high capacitance. I measured 625p. Strangely most of the SMBJ15CA datasheets don't mention anything about capacitance and I chose the part because it was used in the schematic for an eval board for some fancy audio chip. Only the OnSemi datasheet has a graph of capacitance vs reverse voltage. I guess modern designs assume everything is being driven by low Z. The source Z would have to be in the 5K ohms range to start seeing distortion though. I suppose I can keep using them for the moment. It seems there are TVS diodes with much lower capacitance (like CDSOD323-T15C which is only ~3pF).

Attaching diodes to supply rails is a PITA. It would be nice if I could find a simple 2 terminal part.

Thanks for pointing this out. I would like to interface with high Z gear.

As you've found out there are lower capacitance TVS devices. TVS diodes are used in power / data lines etc. and the capacitance can obviously be a problem with faster data lines. Hence lower capacitance parts manufactured. But linearity with signal ie Voltage Coefficient and other parameters eg Thermal Coefficient are not really very important in typical applications so I wouldn't expect them to be in the same league as a good film or NP0 cap ??? Although if you are starting at a nominal 3pF then I guess it can't be going far off from there  :)

Do be aware that specs for TVS (and MOV come to that) are never that simple. The 'headline' voltage may be the nominal clamping voltage but there are a host of other parameters such as the maximum voltage a given the signal / transient will reach before the clamp is effective and response time etc. And there's also a power rating which has to be interpreted if it's an issue.

I was really looking at low energy  / signal / ESD transient issues here. Maybe a power amp output being misapplied - bearing in mind that the input will be relatively high impedance so not huge currents drawn.
As others have discussed - if a mains power supply gets in then it probably won't end well.
A mains supply should be able to to handle 1kV spikes etc (see EMC regs etc) but those are applied to the power input itself and likely dealt with by some TVS / MOV combination. ( I'll say without looking that MOVs won't be great if stuck on a signal line).
 
squarewave said:
I think a bi-directional TVS diode is essentially like back-to-back zeners. They have a reverse breakdown voltage.

Yes - they are. Essentially they are Zener type diodes characterised for transient duty - ESD / Spikes etc. - as opposed to standard Zeners which are characterised in terms of voltage regulation etc. ( Quick summary - Google is your friend - search "TVS vs Zener" etc...)
The technical pedant in me wants to point out that Zener diodes are actually acting more as Avalanche diodes in most cases  ;)
 
squarewave said:
Diverting into power supply rails doesn't make as much sense to me. I think it makes the most sense to place protection diodes right next to pins (between 2 and 1 and 3 and 1) and then use the XLR sockets that have the metal spike that connects pin 1 to the chassis so that shocks or whatever transients will be shunted directly into the chassis without getting that far inside the enclosure.

Clamping to the voltage rails is basically fine.
We are talking about transient overvoltages etc . here. These should be 'rare' events (something seriously amiss if not) and likely to occur when not passing audio - on plugging in / setup etc...
So all we want is for the protection to 'disappear' when not needed.
There's a slightly different argument wrt noise filtering and power rails.

'Pumping Up' the voltage rails is a 'thing' but not in typical small signal audio apps for reasons that have been outlined.

JR is right to point out that ideally clamp level would be below voltage rail (as opposed to a diode drop beyond it) to avoid intrinsic diode conduction but hopefully voltage rail clamping is enough and the device internal protection scheme can take care of what's left.
Rail clamping is 'self adjusting' as pointed out and so applies during power up/down and also if the voltage rails are changed (eg +/-17V supply replaced by +/-15V due to availability / current consumption...
 
bluebird said:
I say just socket the op amp... 8)

:) SMT ?
Actually I think SMT is fine - I'm really wanting to think about avoiding any issue that needs intervention.
As JR points out - it's not a new thing. Really I was wondering what people were currently using if anything.
I seem to have kicked something off !
Interestingly THAT 1510  /1512 data sheets favours rail clamping as back to back Zener scheme required larger series input resistors.
 
Back
Top