LM3886 parallel

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The delta T remains the same. Thanks to improved heatsinking the average temperature will be lower, so the peak temperature as a result of average temperature plus delta T must also be lower.

So do you think that thermal coupling is a function of peak temp rather than delta T? If not then we agree here.

The IC designer's meaning of reasonable. I quote from the datasheet.

In my estimation the suggested applications given in a datasheet are not intended as an exhaustive list. What leads you to think they are?
 
this seems to confirm that the protections/limiters are monitoring Tj and adjusting the limiting thresholds to suit the level and time of the transient.
One of the several protection circuits does. And it does so at an elevated dissipation level.

On transients, I think that what I hear is the early onset of current limited transients...
Did you try to confirm what you hear with an oscilloscope. You could compare the waveforms with the pictures in AN-898 and see which of the several protection circuits is the culprit, if any.

I chose to ignore your other two questions,
They were also rather directed at panson_hk.

So do you think that thermal coupling is a function of peak temp rather than delta T? If not then we agree here.

I can only assume that you refer to delta T as the temperature difference between two media as in junction is at 140 °C, heatsink is at 120 °C, delta T = 20 K. If that is so, I agree that heat transfer is faster when delta T is higher.

Delta T in my post above refers to temperature rise due to the short term thermal time constants you mentioned as in x W of power due to a transient produce a temperature rise of delta T = 25 K. I made no relation to thermal coupling.

In my estimation the suggested applications given in a datasheet are not intended as an exhaustive list. What leads you to think they are?

What makes you think that National's marketing people would not take the chance to add all possible applications to the list when every single item increases the number of possible customers for the product?

You may find more uses for those amps, use different schematics around them, bridge them, parallel them, whatever you like. I just find it unfair to blame them for not delivering extreme currents, not being able to deal with high amounts of reactance or having limited thermal coupling by people who ignore the manufacturer's recommendations on intended applications. I also find it inadequate to blame the protection circuits for doing their job. If that has an audible influence, it is a broad hint to use the amp right, not a reason to complain.
 
I can only assume that you refer to delta T as the temperature difference between two media as in junction is at 140 °C, heatsink is at 120 °C, delta T = 20 K.

Nope - the context was thermal cross-coupling. I took it that delta T was the short term temperature rise over part of a cycle of the audio waveform.

What makes you think that ...<snip>

I note you did not answer my question.

You may find more uses for those amps, use different schematics around them, bridge them, parallel them, whatever you like. I just find it unfair to blame them for not delivering extreme currents, not being able to deal with high amounts of reactance or having limited thermal coupling by people who ignore the manufacturer's recommendations on intended applications.

Oh I agree - who is blaming them here? Cites would be useful as I haven't noticed blaming going on.

I also find it inadequate to blame the protection circuits for doing their job. If that has an audible influence, it is a broad hint to use the amp right, not a reason to complain.

Ditto the above remark - cites for who is complaining please.
 
Measured data

Some data are available from the test board. We can see below the THD vs power for single, two and three 3886s. The rails is +/- 25 V, no load.

From left to right: 1k, 4Ohm; 1k, 8Ohm; 20k 4Ohm; 20k, 8Ohm. We can see the clipping level is higher for parallel in particular from one to two 3886s and 8 Ohms load. We get lower THD by using an additional 3886. Will the fourth lead to better performance? We will know after I stacked the boards.
 

Attachments

  • THD vs power 1k 4Ohm.jpg
    THD vs power 1k 4Ohm.jpg
    81.2 KB · Views: 347
  • THD vs power 1k 8Ohm.jpg
    THD vs power 1k 8Ohm.jpg
    85.1 KB · Views: 337
  • THD vs power 20k 4Ohm.jpg
    THD vs power 20k 4Ohm.jpg
    76.9 KB · Views: 334
  • THD vs power 20k 8Ohm.jpg
    THD vs power 20k 8Ohm.jpg
    82.4 KB · Views: 343
We get lower THD by using an additional 3886. Will the fourth lead to better performance?
that applies to vitually all amplifiers not just chipamps.
Reducing the loading improves the amp performance. 8ohm speakers will always get a better performance out of a 4ohm capable amplifier when used with 4ohms speakers.
Similarly 16ohm speakers should allow better performance again.

Increasing the number of 3886 in parallel is a similar effect to increasing the load impedance. National datasheet confirms the 4 to 8ohm changes.

I remind users that the 3886 and others are crippled for peak current output. Any help to reduce current demand on any individual chipamp is likely to improve performance. But why bother? Just accept that the 3886 has limitations and adopt another amp more suited to higher current duty.
 
I remind users that the 3886 and others are crippled for peak current output.

If any users are interested, there's no evidence at all that the 3886 and others do not meet their specs. When I for one see such evidence presented I'll eagerly devour it. So far, there's been nothing. I recommend potential users of 3886 and similar not to hold their breath for it. :D
 
Paralleling more than three 3886 may not be necessary aimed for higher current demand. If this brings improvement, we may get better sound even the speaker impedance dip is about 4 Ohms. Using 3886 or other single-chip chipamp has the advantage of eliminating protection circuit design. I never doubt LME498xx with discrete output is much better. We just need to take care the protection.

Left is THD vs output power for 20 kHz, 4 Ohms load. 80kHz filter was employed. We almost get 0.01 % before clipping.

Right is THD vs freq for 44 W into 4 Ohms load. 80 kHz filter was employed.
 

Attachments

  • THD vs power 20k 4Ohm Up To Four.jpg
    THD vs power 20k 4Ohm Up To Four.jpg
    77.5 KB · Views: 270
  • THD vs freq 4Ohm 44W 80k.jpg
    THD vs freq 4Ohm 44W 80k.jpg
    72.5 KB · Views: 79
Last edited:
When one of the four 3886 gain is set different from the other three (18.8 vs 19.4), distortion is higher due to this discrepancy. 1% resistors with multi-turn trim pot are employed in the test board for gain setting. The figure below shows the THD vs power for board with and without gain error among the chips.
 

Attachments

  • THD vs freq 4Ohm 44W 80k with gain error.jpg
    THD vs freq 4Ohm 44W 80k with gain error.jpg
    74.1 KB · Views: 91
Last edited:
Anyone seen more than three 3886 in parallel? Make sense?

This is an interesting thread for someone like me, who depends on chip amps for his active loudspeaker projects.

Not to hijack this thread or anything, but have you considered using the TDA7293 for parallel applications? The datasheet shows that it can be used in a Master + Slave(s) configuration, where only the output stage of the slaves is paralleled. This should make it possible to scale current without having slightly different input stages competing against each other. Also, you do not need the current distributing output resistors that are required for the LM3886.

I would love for you (Panson_HK) to whip up a board using 3 or 4 of these in parallel, or something that the user could build as 2,3, or 4 in parallel by adding components as needed. I have a design for a dual board with some tips on selection of components to keep the TDA7293 doing what they should, which can be tricky. PM me if you want the details.

-Charlie
 
This is some really great info Thanks for taking the time to do this and present the data.
It has answered alot of things that I was wondering my self.
I have been Contemplating a 4xLm4780 to drive my ESL panels that are in the 1 ohm load range.
Right now I still have 5 chips and I will start with those and if it proves to perform well I can always get more of them to finish the project. :)

Thanks again. jer :)
 
have you considered using the TDA7293 for parallel applications?

Thank you for your suggestion. I was thinking to include TDA7294. TDA7293 is definitely a better choice for parallel. I will test it after 3886.

I would love for you (Panson_HK) to whip up a board using 3 or 4 of these in parallel, or something that the user could build as 2,3, or 4 in parallel by adding components as needed. I have a design for a dual board with some tips on selection of components to keep the TDA7293 doing what they should, which can be tricky. PM me if you want the details.

Yes, this is being considered. I will either make a board for, say, 4 x parallel or a 2 x parallel board easing for 4 x, 6 x ... parallel.

I will send you a PM. Thank you.
 
I have been Contemplating a 4xLm4780 to drive my ESL panels that are in the 1 ohm load range.
Right now I still have 5 chips and I will start with those and if it proves to perform well I can always get more of them to finish the project. :)

Thanks again. jer :)

You are welcome. All the channels (4 x 2) will be configured in x 4 parallel stereo or x 8 parallel mono?
 
ESL panel C often resonate with step up Xfmr leakage L to give big peak just beyond 20KHz
I've seen impedance plots that suggest a few Ohms series R in the amp would be a good thing

you can make this up with the current sharing R - damps the resonance, protects the amp chips from destablizing cable Cload - but the R need to be big to handle the power loss at the top of the audio range into the transformed C of the panel
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.