As TI now do cheap 4 channel 10 bit A/Ds which can sample at MS/s, what do you think about a SOA protection circuit that measures both voltage and current of both output devices. A 8 bit PIC or AVR cpu should be able to do the processing and control a mute and a speaker protection relay. The voltage signals would also be usable for tweeter protection and dc fault detection. I would use one A/D and one cpu per channel to avoid routing the input signal too far.
It should be possible to follow the manufacturers SOA curves much more accurately than the usual analogue circuits.
It should be possible to follow the manufacturers SOA curves much more accurately than the usual analogue circuits.
In case of Reactive loads...Muting isn't a viable solution..because there is still finite dissipation occurs in corresponding output devices due to V-I phase Lead/Lag....which could damage the devices...
It has been my experience that unless you have forced cooling methods, excellent thermal engineering, and very well defined waveforms, you should not operate a device anywhere near its SOA.
Max rated current / 3 - 4 will make things that don't melt.
😉
Max rated current / 3 - 4 will make things that don't melt.
😉
My experience is that bipolar amps either:
1) have very conservative and intrusive SOA that can get into real trouble with reactive loads
2) have overkill multiple output devices in parallel with the cost issues
3) blow output devices
Judging from the postings in this forum (3) is way to common
Checking against the VI curve and talking into account the time would allow you to be certain that you are safely inside the SOA limits, instead of the "rule of thumb" limits which make bipolars usable at only about half of the rated voltage.
1) have very conservative and intrusive SOA that can get into real trouble with reactive loads
2) have overkill multiple output devices in parallel with the cost issues
3) blow output devices
Judging from the postings in this forum (3) is way to common
Checking against the VI curve and talking into account the time would allow you to be certain that you are safely inside the SOA limits, instead of the "rule of thumb" limits which make bipolars usable at only about half of the rated voltage.
Guess you didn't really read my post...
It's takes a lot of engineering, and TESTING, to run devices, especially with wildly undefined waveforms at there maximum.
For a "one off" DIY project it is probably not worth the hassle. For a design to be built in the thousands it is worth the hassle.
And in the the end, when you have maximized device usage, what you have built is a modern TV set... 6-8 years and hurl it in the trash.
🙄
It's takes a lot of engineering, and TESTING, to run devices, especially with wildly undefined waveforms at there maximum.
For a "one off" DIY project it is probably not worth the hassle. For a design to be built in the thousands it is worth the hassle.
And in the the end, when you have maximized device usage, what you have built is a modern TV set... 6-8 years and hurl it in the trash.
🙄
It is best with audio to design via the DC SOA curve to be safe. The problem is that it is way to easy to get fake devices and because of the cheaper, smaller die, the SOA goes out the window usually as smoke.
BJT's generally can't handle much current with a high Vce. This makes them more likely to become toast with highly reactive loads then FET's.(with the exception of genuine MJ21194/5😀) Personally I like BJT's, but you have to observe the limitations.
BJT's generally can't handle much current with a high Vce. This makes them more likely to become toast with highly reactive loads then FET's.(with the exception of genuine MJ21194/5😀) Personally I like BJT's, but you have to observe the limitations.Hi,
on a recent thread the designer suggested using the 10mS SOAR as the limiting factor for music based loadings.
This invites trouble.
As an alternative, I have used DC or one second SOAR and then applied a cap to allow the passage of very short term peak currents that would normally exceed the long term SOAR.
This applies to both BJT and FET output stages into reactive loads (usually a theoretical 60degree phase angle).
My main difficulty is identifying the cap value that does not permit damage to the output stage.
on a recent thread the designer suggested using the 10mS SOAR as the limiting factor for music based loadings.
This invites trouble.
As an alternative, I have used DC or one second SOAR and then applied a cap to allow the passage of very short term peak currents that would normally exceed the long term SOAR.
This applies to both BJT and FET output stages into reactive loads (usually a theoretical 60degree phase angle).
My main difficulty is identifying the cap value that does not permit damage to the output stage.
Using the 10mS SOA for music is reckless, the lowest impedance of most reflex speakers is below 100Hz and the reflex tuning resonance leads to the the worst phase shifts below 100Hz too. Then play music with a lot of bass and you are going to get overload.
What I am trying to propose is using a digital method to constantly check that the devices are always inside the SOA curve by a safety margin of say 30%, not just guessing. So long as the devices are genuine, I trust reputable transistor makers to make a reliable product that meet their specs.
What I am trying to propose is using a digital method to constantly check that the devices are always inside the SOA curve by a safety margin of say 30%, not just guessing. So long as the devices are genuine, I trust reputable transistor makers to make a reliable product that meet their specs.
Hi all
Can any of you suggest a test reactive load (and the frequency
of the test signal ) to be used during simulation stage to verify SOA limits are not trespassed?
Federico
My two cents:
I think it is most important to define what the amp is going to be used for:
1) Active speakers (No passive X-over to worrie about)
2) 8 Ohm passive speaker
3) 4 Ohm passive speaker
4) 2 Ohm passive speaker
Case 1 and 2 are less of a problem, 3 and 4 are hard!
If you know the load it is possible to design the SOA protection to match this.
If you are unsure what the amp is going to be used for (This is often the case) you have to be conservative in your protection levle and maybe even expect case 4.
From Rods pages:
For the 3 ohm case, a reactive load (at 45° phase angle) can be simulated by using a 477uH inductor in series with the 3 ohm resistor (for a frequency of 1kHz)
\Jens
I think it is most important to define what the amp is going to be used for:
1) Active speakers (No passive X-over to worrie about)
2) 8 Ohm passive speaker
3) 4 Ohm passive speaker
4) 2 Ohm passive speaker
Case 1 and 2 are less of a problem, 3 and 4 are hard!
If you know the load it is possible to design the SOA protection to match this.
If you are unsure what the amp is going to be used for (This is often the case) you have to be conservative in your protection levle and maybe even expect case 4.
From Rods pages:
For the 3 ohm case, a reactive load (at 45° phase angle) can be simulated by using a 477uH inductor in series with the 3 ohm resistor (for a frequency of 1kHz)
\Jens
It seems to me that it would be easier and cheaper in the long run to just over design the amp for what you are going to listen too. This would take out the potential of the elaborate protection circuits to affect the quality of the audio signal and keep it back to basics.
If SOA is such a problem for you, and you have no problem with making some sort of complex protection circuit, then maybe you should investigate class G (or maybe H) designs. I like G better because the top rail follows the audio and doesn't introduce HF switching noise. Perhaps Fets for the top rail and BJT's for the bottom? 🙂
If SOA is such a problem for you, and you have no problem with making some sort of complex protection circuit, then maybe you should investigate class G (or maybe H) designs. I like G better because the top rail follows the audio and doesn't introduce HF switching noise. Perhaps Fets for the top rail and BJT's for the bottom? 🙂
I have seen some unusual speaker designs (ribbon tweeters?) which fell to about 1R at 20kHz. Try overdesigning for that one. It seems crazy but actually measuring the output transistor status in real time is cheaper than more robust ( and more quantity) output transistors.
- Status
- Not open for further replies.