Fast DC protection for HPamps.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
So here's an idea I've been thinking about , instead of the slow servo-like DC
protection circuits I seen here on the forum and elsewhere , a faster one that not
just detects DC on the output but other calamities too , like distortions when you
drive the output to the rails. This works not by comparing the average voltage of a signal (music) , but the difference between in and output. (adjusted for the amplification of course)

An instrumentation amp amplifies the differential voltage between input and output
by 20-50x and outputs this to a window comparator . Low offsets , like mV won't
make the threshold . Large ones will after a small RC filter ( < 100 us).
This could reduce the reaction time to less than a millisecond (with fast opamps/Inamps) + 5 ms for the output relay to switch off. This is an order or 2 better than the servo + filtering types and just might save those ridiculously expensive HP's.
My background in electronics is not that much and I haven't used an
instrumentation amp before , integrated or with 3 discrete opamps.
I'm sure I'm not re-inventing the wheel here , but I haven't seen this before.
What do you think , before I start soldering a test circuit. Is there something I didn't
think about , ... pitfalls ?

The 2 top opamps are a basic HP amp with a gain of 3. Ignore the part after the diodes , it's just a level shifter to go to a S/R flipflop and the relays switch.
I want the relay switched off after detection , not coming back after de DC has gone.
 

Attachments

  • DC protection.JPG
    DC protection.JPG
    44.8 KB · Views: 777
Your circuit is not applicable to D.C. coupled amplifier protection, where input D.C. may be present. There must be an D.C. blocking input cap., yes?

I guess you didn't notice the capacitor at the input that will keep all DC out from the source. All that is left is offset from the opamps.

I'm more concerned about phase shift between in and output that might be an issue when amplifying the difference between them.
 
I guess you didn't notice the capacitor at the input that will keep all DC out from the source. All that is left is offset from the opamps.
Sorry, that sentence of mine was not worded as clearly as it could have been. Actually, I had noticed the input cap. My observation was just that your circuit protects against internal D.C. faults. It is not applicable to protecting D.C. coupled amplifiers from D.C. input offsets.
 
So here's an idea I've been thinking about , instead of the slow servo-like DC
protection circuits I seen here on the forum and elsewhere , a faster one that not
just detects DC on the output but other calamities too , like distortions when you
drive the output to the rails. This works not by comparing the average voltage of a signal (music) , but the difference between in and output. (adjusted for the amplification of course)

An instrumentation amp amplifies the differential voltage between input and output
by 20-50x and outputs this to a window comparator . Low offsets , like mV won't
make the threshold . Large ones will after a small RC filter ( < 100 us).
This could reduce the reaction time to less than a millisecond (with fast opamps/Inamps) + 5 ms for the output relay to switch off. This is an order or 2 better than the servo + filtering types and just might save those ridiculously expensive HP's.
My background in electronics is not that much and I haven't used an
instrumentation amp before , integrated or with 3 discrete opamps.
I'm sure I'm not re-inventing the wheel here , but I haven't seen this before.
What do you think , before I start soldering a test circuit. Is there something I didn't
think about , ... pitfalls ?

The 2 top opamps are a basic HP amp with a gain of 3. Ignore the part after the diodes , it's just a level shifter to go to a S/R flipflop and the relays switch.
I want the relay switched off after detection , not coming back after de DC has gone.

Clever. I wonder if you could implement this idea with a 555 timer chip.

It's not a new idea, but I haven't seen it implemented. It could maybe be reconfigured to control a compressor on a power amplifier to prevent clipping.

I've designed and built DC protection circuits that used a comparator based circuit and referenced ground. They were "fast enough" and worked without glitches. Speed can be optimized by choosing the right resistors; the more current driving the base, the slower the transistor will come out of saturation when that current is removed.

It will also shut down on clipping or gross distortion, both of which shouldn't be happening on a headphone amp.
 
Last edited:
^
It triggers a flipflop , so triggering a 555 timer isn't a problem.

The meaning of the circuit is to protect the HP's not just against (too big a) DC offsets on the output but also in case of opamps that break down and give a short to either supply rail . It shouldn't come back on after a certain time. Servo based or the RC protection is far too slow for that . Since the goal is to switch off the output relay , the saturation of the transistor is irrelevant.
 
^ Yes xrk971 , I already saw that circuit researching solutions here on the forum , but this is the RC one that is too slow. It's ok for small DC's , switching off the relay or supply for the time a DC voltage is detected , but it offers little protection for real faults like opamps shorting to the supply rails.
I know there is little chance of that happening , but still , connecting a 500 $/€ HP (or even a 100-200 $/€) , you want to be absolutely sure . Maybe the 5 ms it takes for the relay to switch off, may be already to long . Why would an opamp break down ? Bad handling , ESD , bad/too long soldering, ...or even overdissipating like putting it in a 3 x 3 mm package (OPA 1622). Resistors aren't foolproof . If one breaks in the feedback or in the DC correction you use , it could force the output to extremes .

I shouldn't have added the transistor , because it's irrelevant what comes behind the window comparator. Some want a timer , others just want an off switch during DC detection , transistor or opto-coupler , it doesn't matter. For me it's a S/R flipflop. It's the comparing in and output by an instrumentation amp that is the essence here.
 
Last edited:
^
It triggers a flipflop , so triggering a 555 timer isn't a problem.

The meaning of the circuit is to protect the HP's not just against (too big a) DC offsets on the output but also in case of opamps that break down and give a short to either supply rail . It shouldn't come back on after a certain time.

Now I see a slight advantage to the circuit. But a circuit referenced to ground will still trigger, will it not?

Since the goal is to switch off the output relay , the saturation of the transistor is irrelevant.
Noted.

The reason I mentioned this is that I'm working on protection circuits that use the 555 and a handful of transistors. Some circuits cam be optimized with this knowledge.
 
Last edited:
Now I see a slight advantage to the circuit. But a circuit referenced to ground will still trigger, will it not?

Only a slight advantage ? :)

It doesn't matter what the circuit is referenced to . Use a levelshifter or an opto-coupler and anything goes. I'm not a fan of the old 555 or it's grandkids 7555/TLC555. For timers I prefer 4538 or HC4538. My whole protection circuit is in HCMOS , over and under voltage detection and now , trying to get the DC protection right.
 
Founder of XSA-Labs
Joined 2012
Paid Member
I don’t think it’s that simple. You have to look at maximum suspension movement and how many joules of energy the voice coil (VC) can take, plus how many mA can the power supply really deliver. Very few headphone IC amps can are capable of delivering more than 150mA and what you are saying is an internal failure in shorted to rail mode. 5ms of 13v at some typically limited current is probably not going to destroy a large headphone with typical 96dB sensitivity. 114dB per mW IEMs on the other hand will die if you even look at them wrong.

7w over 5ms is 35mJ of energy. Voicoil of a big over the ear can with 45mm dia VC can probably take 35mJ dissipation. Can the suspension handle it? Maybe.

Nevertheless, I agree that if your circuit is comparing out vs in modified by gain, and if that’s out of bounds then it triggers. This doesn’t prevent damage from perfect fidelity of out vs in but exceeding suspension Xmax or thermal energy max of the headphone driver VC.

Also, planar magnetics like Hifiman are very resistant to over voltage and can dissipate quite a bit of peak burst power.
 
Last edited:
I made a quick estimation on a real-world example, a 6mm dia. VC:

attachment.php


The gross volume is about 2mm³; let's say for convenience that the actual copper volume is 1.1mm³.
Since its density is ~9g/cm³, the total mass is ~10mg.
With a heat capacity of 380J/°K*kg, this means that in order not to exceed a temp rise of 150°C for a single pulse in adiabatic conditions, the energy must be < 0.01*150*0.38 = 0.57J, or 570mJ.

The estimation is conservative, and a number of heat sinks, like glue, varnish or mylar have not been taken into account.

With a 7W input, about 80ms is required to reach 570mJ.
This puts things in perpective (and with 80ms, the conditions will not be adiabatic anymore)

Mechanical stress could be more problematic, depending on the construction
 

Attachments

  • VC.jpg
    VC.jpg
    308.9 KB · Views: 494
^ So we're pretty safe with the 5 ms it takes the relay .

You also need to consider big caps. While the power supply may not deliver a big current , the caps can . Have you seen the spark when you short (or discharge with a few ohms) a 1000 uF cap ?
The tiny voice coils of those in-ear or even the slightly bigger ones like on LV's pic will not survive . To my shame I have to admit I made costly mistakes with HP's .

Of course if you're confident your HP can take a hit , the slow circuit like Jhofland will do just fine.
 
I presume you want to have protection for your headphone against a high power amp ?
Or you have IC based headphone amps driving IEM’s ?

A modern over-ear headphone will take 200mW.
So even at 32R, that is a continuous DC of 2.5V and the coil still survives.
For 300R, this goes up to 7.8V pure DC.
Any of the RC-based protection circuits set at say 100mV will trigger in milliseconds at that DC level.
Not to speak of full rail voltage of 15V or above.

Then you also want to protect against clipping.
Assume you have 9V rails and you are just below clipping.
At 32R, this is ~1.3W rms. Or 0.6W at 65R.
Even if your ears survive, your headphone coil will not, although there is no error between input and the attenuated output.

OK, what if you use IEM’s ?
I guess you are not going to use a 200mA Class A output stage for that.
Isn't it a lot easier to put a current limit on your power supply, to say 40mA, than to rely on 5 opamps to protect against 2 in your main amp circuit ?
I am sure you can find lots of those examples with 2 transistors per rail.
For me a lot more reliable than 5 opamps.

Or am I missing something ?


Cheers,
Patrick
 
I presume you want to have protection for your headphone against a high power amp ?
Or you have IC based headphone amps driving IEM’s ?

No , for protection against malfunction of the amp.
The HP amp is for 24 ohm HP's as well as 300 ohm ones and works on 2 x 15V.
The 24 ohm's will be much more at risk , because of their very small voice coil. But the 300 ohm's one are the most expensive and I don't want to risk them getting damaged.

Any of the RC-based protection circuits set at say 100mV will trigger in milliseconds at that DC level. Not to speak of full rail voltage of 15V or above.

Looking at the RC based (or servo based) circuits , I doubt it wil trigger in ms, more like 100's of ms.
See the pic below.

Then you also want to protect against clipping.

No I don't , but it will be inherent to the circuit when it compares in and output.

Isn't it a lot easier to put a current limit on your power supply, to say 40mA, than to rely on 5 opamps to protect against 2 in your main amp circuit ?
I am sure you can find lots of those examples with 2 transistors per rail.
For me a lot more reliable than 5 opamps.

A current limiter would surely f-up the low noise regulation , and regulation itself, and it wouldn't help against caps getting discharged through shorted opamps outputs.


A lot of why's I would want such protection and if it is useful ...I think it is.
I hoped this thread would be more about the instrumentation amp , how to do it right ,reliability, ... because I have no experience with it , not about the why , or what comes after the detection of an error .

Tomchr (neurochrome) was making another HP amp , with more output protection, I think it was with the OPA1656 or 1688 . I wonder how he would tackle the problem of watertight protection . But I haven't heard of him lately on the progress on that ... and then again , he does this for a living so he won't be forthcoming to show us how he does it.
 

Attachments

  • HPA-SSR-Protect-Schematic.jpg
    HPA-SSR-Protect-Schematic.jpg
    18.5 KB · Views: 219
Actually the likes of LT3045 and LT3094 have built-in programmable current limits (with 1 resistor).
And they do not need large output caps.
10µ will do.


Cheers,
Patrick

Yeah , I know them. Both are in a very tiny smd , which is a very big no-no for regulators.
LM317/337 limits current too , but at much higher currents. :)

But your idea could be a simple solution for limiting the current , a R between base and emitter , when it comes at a certain value , it limits or cut power to the relays.
Combined with the slow RC DC detector , can be a valid way to go , if my idea doesn't pan out.
Still the current limiting is different for 24 or 300 ohm HP's. And you can't have bigger caps behind that limiter.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.