I am trying to figure out a new way to protect my amplifiers. The old VI limiting always adds non linearities to the signal as it approaches the limit, or even much earlier in some cases, which is very undesirable in my mind, or ears. While I’m sure someone else has already tried this, Id really like to figure this out.
My solution works great at removing the non-linearities and still protects the amplifier (sort of) but the trip point is kind of slow (1 or 2 ms) and the current reaches 60 amps before the protection cuts off the signal. Mind you the test I did is all on Multisim not real world, but I’m building it now.
What I’d like to know is if this is a dead-end path? I just need to figure out some way to cut off the signal earlier, or I could double up the output devices. IDK any suggestions would be awesome.
My solution works great at removing the non-linearities and still protects the amplifier (sort of) but the trip point is kind of slow (1 or 2 ms) and the current reaches 60 amps before the protection cuts off the signal. Mind you the test I did is all on Multisim not real world, but I’m building it now.
What I’d like to know is if this is a dead-end path? I just need to figure out some way to cut off the signal earlier, or I could double up the output devices. IDK any suggestions would be awesome.
Deosn't that depend on the gains of the output darlingtons being fairly well known?
Doesn't Hfe increase with temperature (sometimes? It's a long time ago!), so you'd need to limit to a low value when cool?
Doesn't Hfe increase with temperature (sometimes? It's a long time ago!), so you'd need to limit to a low value when cool?
Very likely I’ll find other issues when I get this built, like heat and hfe issue., However I believe the question remains how to speed up the cut off? if i put the signal level at 10khz it cuts off in the nano second range but that’s not real-world stuff. when at 100hz it takes 250ms and the outputs trannies far exceed their SOA if only for a short period, 60 amps or more. but i could double up the output transistors and increase the ballast resistors and that would work but that's not what i want
and for those wondering how much a diode in the signal path effects the sound ( which i was concerend about) its way less than a standard vi protection method. i get .01% right up to the cut off limit which cuts off completely. This amp will do square waves cleanly right up to 20k
What are you trying to achieve?
Limit current to protect the output transistors? why not just have big transistors and limit the supply current?
As far as I understand, the more complex SOAR limiting is all about getting closer to the limits of what the transistor can do, compared to simple current limiting. It will allow a higher current with less volts across the transistor, it attempts to limit dissipation in the devices.
If you are getting audible distortion from mainstream limiting circuits, maybe you just want a bigger amp?
How many home hifi users habitually drive their amps into limiting?
By the time you've turned it up enough for the limiter to be adding distortion, won't the speaker be distorting
More complex/subtle limiting might be in order for PA/live music amps,where it's likely things will be used to their max power, but you don't want sudden cut-offs to affect the sound any worse than necessary.
Hard clipping may also feed the speakers with high-powered harmonics. Maybe that won't be good for listeners or tweeters?
Limit current to protect the output transistors? why not just have big transistors and limit the supply current?
As far as I understand, the more complex SOAR limiting is all about getting closer to the limits of what the transistor can do, compared to simple current limiting. It will allow a higher current with less volts across the transistor, it attempts to limit dissipation in the devices.
If you are getting audible distortion from mainstream limiting circuits, maybe you just want a bigger amp?
How many home hifi users habitually drive their amps into limiting?
By the time you've turned it up enough for the limiter to be adding distortion, won't the speaker be distorting
More complex/subtle limiting might be in order for PA/live music amps,where it's likely things will be used to their max power, but you don't want sudden cut-offs to affect the sound any worse than necessary.
Hard clipping may also feed the speakers with high-powered harmonics. Maybe that won't be good for listeners or tweeters?
It’s just an exercise in trying to find better way for an amp to protect itself from the real-world issues, speaker, idiots, over driven. I believe the old vi limiter is bad as it starts to mess up the sound way before the soa is reached . I want something that protects right at the limit of the SOA and never injects itself prior too.
I never actually built it, but for an amplifier with output MOSFETs without Spirito instability (sort of second breakdown), I once came up with a translinear circuit that produces a current proportional to the momentary power dissipation of an output MOSFET. Filter it with an RC circuit that is an electrical model of the thermal impedance and you get a voltage that indicates the temperature increase. See post #11 of https://www.diyaudio.com/community/...ether-it-works-in-a-or-ab.308636/post-5100509
Bipolar devices often fail due to secondary breakdown, so you have to monitor the voltage and current to protect against this whilst still getting the most out of the devices, ie a somewhat more complex criterion than straight IV-limiting. It ought to be possible to design a floating circuit that tracks this using a few comparators to trigger shutdown...
Monitoring a driver's base current cannot be as accurate as monitoring the output's emitter resistor. Otherwise, the circuit implements conventional latching protection.
The circuit has no control over the VAS idle current. C12 and C13 do not look like workable frequency compensation.
Ed
The circuit has no control over the VAS idle current. C12 and C13 do not look like workable frequency compensation.
Ed
i dont think (and im offend wrong) anyone is capturing what im trying to do. i supplired the scem. what do yoiu think
In the world of industrial semiconductors, work has been done to monitor the junction temperature in real time.
For instance in a push-pull circuit, you could perform some measurement during the half cycle that the device is idle.
People have tried techniques like injecting some HF and measuring some temperature-variant parameter of the device.
Or you could just embed another junction or temperature sensor in the package.
It takes time to overheat a power device, if you understand the physical structure of the device you might deduce how far you can go beyond simple VI limits.
If you've got some meaty great thyristors, their thermal time constant at the junction might be several switching cycles.
It's all interesting stuff, but how far do you want to go?
If you want your output stage to last many years, then you need a big safety margin anyway.
For instance in a push-pull circuit, you could perform some measurement during the half cycle that the device is idle.
People have tried techniques like injecting some HF and measuring some temperature-variant parameter of the device.
Or you could just embed another junction or temperature sensor in the package.
It takes time to overheat a power device, if you understand the physical structure of the device you might deduce how far you can go beyond simple VI limits.
If you've got some meaty great thyristors, their thermal time constant at the junction might be several switching cycles.
It's all interesting stuff, but how far do you want to go?
If you want your output stage to last many years, then you need a big safety margin anyway.
At the top level, AIUI, you are trying to sense the base current into a darlington, and use that information to protect the output devices from blowing up.
Others one here will be better at the detail.
I don't think it will work well as a conceptual level for a control strategy.
I see you have a latch that makes everything currentless when overcurrent is detected. In my experience, that works fine for an amplifier for domestic use that never goes into current limiting unless something is really wrong.
Unless I'm missing something, weak points are the fact that you sense a base rather than an emitter or collector current and that you only sense the current through the lower side of the output stage, not the upper side.
Regarding the delay, does the latch trigger too slowly or is it the rest of the amplifier responding too slowly to the latch?
By the way, the rectangular box around the latch in the schematic looks a lot like a shorting wire. It makes it a bit harder to understand what is and what isn't connected to what.
Unless I'm missing something, weak points are the fact that you sense a base rather than an emitter or collector current and that you only sense the current through the lower side of the output stage, not the upper side.
Regarding the delay, does the latch trigger too slowly or is it the rest of the amplifier responding too slowly to the latch?
By the way, the rectangular box around the latch in the schematic looks a lot like a shorting wire. It makes it a bit harder to understand what is and what isn't connected to what.
Last edited:
while this circuit works quite well, but when i short .01ohm the output whatever transistor is conducting at the moment of the short goes behond the soa of the transistors current rating. this might be ok for that short duration which would make this converstion moot. but im looking for the smart people to look at this and say ?????
That's what we are doing 🙂
With due respect, you are still detecting power transistor current, only in an indirect way: monitoring driver base current instead of voltage drop across an emitter resistor.
2 problems with that:
* It's an indirect approach and introduces 2 parameters which vary a lot in commercial components: each transistor Hfe.
You would need to replace the 100 ohm sensing resistor by a trimmer, drive amp to peak acceptable current into a load (some amps specify 1 or 2 ohm just for that purpose) and adjust triggering.
Much easier and repeatable is to just monitor ballast resistor voltage drop. It does NOT depend on individual transistor variations.
* It is not a VI limiter since you are not monitoring voltage across transistors, then it's not a complete protection.
If you don't like VI type, fine with me, just don't use them, but then either use robust transistors, more of them, or like many Asians do, monitor current only (often using a dedicated chip) but trigger a relay within milliseconds.
Why don't you connect Q16 across R26 and R25, with a voltage divider if needed to get the right trip point? Maybe you would need a diode or two to prevent interactions with Q29, but at least you would protect both the upper and the lower part and without dependence on hFE.
In your simulation, it should be easy to see where most of the delay is located.
The SOAR graphs normally show what kind of voltages and currents are allowable for a few microseconds. If you get outside those contours, the transistors may very well be damaged.
In your simulation, it should be easy to see where most of the delay is located.
The SOAR graphs normally show what kind of voltages and currents are allowable for a few microseconds. If you get outside those contours, the transistors may very well be damaged.
I think the current protection circuit in the Cambridge A3i in front of me is quite neat, sensing the current in the '0V' sides of the transformer secondaries.
No connections within the output stages, and operates close to 0V.
Switching off the speakers via relay may not always be fast enough of course
No connections within the output stages, and operates close to 0V.
Switching off the speakers via relay may not always be fast enough of course
Any kind of indicators planned to inform the user the protection is active, but more importantly, why?
The problem I see with most consumer protection circuits is knowing why they are active.
Even the service manuals do a poor job documenting this.
Look at how many threads appear here looking for help with devices in protection mode.
I know it's not easy to ID every possible failure, but if a protection circuit is monitoring three different conditions, like temperature, over current, and DC out, then it should be possible to have a simple led indicate which one is active.
The problem I see with most consumer protection circuits is knowing why they are active.
Even the service manuals do a poor job documenting this.
Look at how many threads appear here looking for help with devices in protection mode.
I know it's not easy to ID every possible failure, but if a protection circuit is monitoring three different conditions, like temperature, over current, and DC out, then it should be possible to have a simple led indicate which one is active.