very narrow.
It is a fast transient peak value.
But the output transistors have to pass that current and the Base current will be determined by the gain at that peak transient value.
Then the driver has to pass that transient base current and it too will be into beta droop for the few tens, or hundreds, of nano seconds.
This is current clipping taking place. R.Cordell is one of the few audio authors that mentions current clipping
It is a fast transient peak value.
But the output transistors have to pass that current and the Base current will be determined by the gain at that peak transient value.
Then the driver has to pass that transient base current and it too will be into beta droop for the few tens, or hundreds, of nano seconds.
This is current clipping taking place. R.Cordell is one of the few audio authors that mentions current clipping
is that even remotely audible ?
i think if i remeber right lm3886 can do a repetative maximum corrent past 10 A.
please explain why would that no be sufficient, and what would be the sonic diference ?
what kind of psu would be needed to supply those huge current spikes ?
like larger than 5 times was mentioned,
so given a 20 volt swing, a 4 ohm resistive load would need 5 A current, and some say even 25 would be needed, for some nanosecund length.
mmm.dunno man, sounds alien.
i think if i remeber right lm3886 can do a repetative maximum corrent past 10 A.
please explain why would that no be sufficient, and what would be the sonic diference ?
what kind of psu would be needed to supply those huge current spikes ?
like larger than 5 times was mentioned,
so given a 20 volt swing, a 4 ohm resistive load would need 5 A current, and some say even 25 would be needed, for some nanosecund length.
mmm.dunno man, sounds alien.
Yes.
IMD (distortion at VHF) becomes audible.
Supply Rail Decoupling is a VHF technique and it is very audible, if the required decoupling is omitted, or done incorrectly.
No.
Typical is stated as 11A, when Tc is at 25°C AND the loading is non reactive. When we reproduce music, we use a reactive load and Tc is above 25°C. Those two factors reduce the actual peak output current to somewhat less than the typical, even when your actual chipamp meets that part of the specification.
However, the minimum spec for non reactive loading and Tc=25°C is 7A and that too is a peak value. Actual peak value will reduce when a reactive speaker and higher Tc is an operational condition.
If you were to select your chipamps so that all were able to meet say 10Apk @ Tc=25°C, you could design for that spec. But how much does that selected value for peak output current actually reduce for a warm, or very warm chipamp? National do not tell us !!!! But we can safely assume that by keeping the chipamp cooler, we can get closer to the Tc=25°C values.
the combination of remote smoothing capacitors and local MF+HF supply rail decoupling supply these "huge" spikes in demand. But they are not "huge".
A 10nF ceramic can supply 15Apk for a short duration.
I have posted many times my belief that a chipamp is not a suitable driver for 4ohms speakers and certainly not for a target 50W into 4r0.
If I were designing for 50W into 4ohms (requiring 20Vpk output) I would set my amplifier a target of 15Apk and use a two pair output stage to minimise the risk of beta droop at that peak output.
That cannot be matched by what National fit inside their chipamp package. As far as I know no other manufacturer tries to better National's minimum spec of 7Apk @ 25°C
Last edited:
the possible current peaks are lots and lots of ns long - as mentioned they come from audio frequency resonance in XO and electromechanical resonance in the drivers being pumped and then driven anti-phase so the peaks are good fractions the audio band resonant frequencies involved ~1/8 wave or so
Last edited:
Supply rail decoupling is a good engineering/design practice, in particular for high-current op-amps, which is what an LM3886 is.
You can find my analysis and optimization of the LM3886 decoupling network here: LM3886 chip amp supply decoupling.
All correct. I do happen to be rather intimately familiar with the design practices at National, though. The fact that a parameter is specified at 25 ºC should not be taken as an indicator that the performance falls apart at higher temperatures. It just means that the IC was tested in production at 25 ºC and found to meet the specs listed in the spec table under the conditions listed in the spec table. Testing at the temperature extremes is more expensive and only done if necessary.
There are many ways to design an IC to provide consistent performance across the full temperature range of operation. I am sure those ways were followed in the LM3886 design. That said, the performance of the part will vary some with temperature (damn physics!) and the output current will likely be lower at higher die temperatures. I'd be surprised if the peak output current drops by more than 10~20 %, though. If your amplifier design falls apart under those conditions, I suggest you use two LM3886es in parallel for higher output current.
~Tom
Well... The SOA limits are characterized for pulsed loads. Depending on the thermal impedance from the semiconductor junction to the case, there can be a significant temperature difference between the two. Thermal inductance reacts to an instantaneous increase in dissipated power very much like an electrical inductance reacts to an instantaneous increase in current.
~Tom
~Tom
That I get, but to get a 225 degree deltaT is impressive.
Now AndrewT said "It is a fast transient peak value." which is to me comparable with a pulsed load, as well as " But how much does that selected value for peak output current actually reduce for a warm, or very warm chipamp? National do not tell us !!!!"
Now happy to be corrected, but this suggests that the Chips IF implemented properly, can do the peaks when hot? Not huge headroom, but enough for many use cases.
Now AndrewT said "It is a fast transient peak value." which is to me comparable with a pulsed load, as well as " But how much does that selected value for peak output current actually reduce for a warm, or very warm chipamp? National do not tell us !!!!"
Now happy to be corrected, but this suggests that the Chips IF implemented properly, can do the peaks when hot? Not huge headroom, but enough for many use cases.
i would suppose it can deal with a 4 ohm speaker, maybe at full blast those nasty nanosec peaks can give it some hard time. can't tell, i would usually not use an amp at full blast anyways.
if i need parallel amps, i might not choose the 3886.
TDA chips offer a nice solution , why torture the 3886 with it.
if i need parallel amps, i might not choose the 3886.
TDA chips offer a nice solution , why torture the 3886 with it.
Physics at work...
See the last two paragraphs in Post #30.
Also note that in an actual experiment, I had no trouble getting the advertised output power from an LM3886 even under extremely hot conditions (heat sink temperature above 100 deg C). You can find the experiment write-up and data here: http://www.diyaudio.com/forums/chip-amps/265771-lm3886-thermal-experiment-data.html
~Tom
wrong
the range is milliseconds for woofer resonance, and proportionately shorter for higher frequency resonances that may be driver mass-spring fundamental resonance or XO Audio Frequency Complex Z
the higher Audio Frequency Complex Z effects still are in 10s-100s of microseconds for typical multiway loudspeakers - quite long enough to need to look at dynamic SOA curves, or expect current limit protection in chip amps to kick in
why aren't amps without the current reserve immediate fails? - because even in the millisecond range errors near in time to dynamic peaks can hard to hear - temporal masking is strong too, in addition to the critical band frequency masking that may be more familiar
Last edited:
interesting.
not lm3886, but makes no difference in this regard..
http://bolthely.hu/kepek/mikroshop/00537.jpg
that was a quite popular kit around here for disco use.
+-43 volt rails, driving 4 ohm at high levels, yet none did burn up or die.
nor did the protection circuitry kick in.
do you have any mesured data available on this subject?
if peak current draw could exceed ~5x what one would suspect vs a resistive 4 ohm load, then accrording to this, one would need an amplifier to drive 4 ohm rated complex load to be stable at 0,8 ohm resistive load (are You kidding?? for a 20 volt swing that would require 25A --> wtf... 500 watt... )
or consider in case of lm3886 peak pulsed current to be around no more than 5-7A, so use 3 of them parallel.
i would really like to see where would that benefit, and if it does any audible/mesureable perfomance increase at all whitin reasonable use.
not lm3886, but makes no difference in this regard..
http://bolthely.hu/kepek/mikroshop/00537.jpg
that was a quite popular kit around here for disco use.
+-43 volt rails, driving 4 ohm at high levels, yet none did burn up or die.
nor did the protection circuitry kick in.
do you have any mesured data available on this subject?
if peak current draw could exceed ~5x what one would suspect vs a resistive 4 ohm load, then accrording to this, one would need an amplifier to drive 4 ohm rated complex load to be stable at 0,8 ohm resistive load (are You kidding?? for a 20 volt swing that would require 25A --> wtf... 500 watt... )
or consider in case of lm3886 peak pulsed current to be around no more than 5-7A, so use 3 of them parallel.
i would really like to see where would that benefit, and if it does any audible/mesureable perfomance increase at all whitin reasonable use.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.