Improving TPA3118 Class D Amplifier

[SanjeevM]

I doubt that a big cap bank will improve the bass response. What a big cap bank is supposed to do is provide instantaneous energy when the program material calls for it (like sudden dynamic peaks).

In general I find the bass weight of 3118 lesser than two other amps I compared it to (this is only in relative terms - it has enough bass heft in my system for my taste), but the deficiency in bass weight is in some way made up by the texture of the bass. It brings out the tonality of the bass much better, for example, than the LM3886 based MyRef Rev C amp (though MyRef has more bass weight).

Guess that sums it up nicely :sad:

 

[keith_correa]

Ok but I'm wondering how an amp costing 3.5 USD got escalated to 39 USD just with some mods here and there. The other day I was interacting with Joshua who seems to be convinced of the cost escalation but still, I'm wondering! Any comments?

It's not "just some mods here and there" as you put it. The $3.5 board uses "cheap" no-name parts vs. the Allo using good quality parts. The Allo also has a cap multiplier + 2 RCA + log pot on board. Also the Allo seems to be using a modified circuit [more parts!!!] as compared to the datasheet circuit which the cheap amp uses. Now I don't know if that justifies the price difference but then factor in the R&D cost + profit margin needed for Allo and I think that for $39, it is well priced.

 

[captrajesh]

It's not "just some mods here and there" as you put it. The $3.5 board uses "cheap" no-name parts vs. the Allo using good quality parts. The Allo also has a cap multiplier + 2 RCA + log pot on board. Also the Allo seems to be using a modified circuit [more parts!!!] as compared to the datasheet circuit which the cheap amp uses. Now I don't know if that justifies the price difference but then factor in the R&D cost + profit margin needed for Allo and I think that for $39, it is well priced.

Hmm! You guys do have a valid point. It would be an interesting experiment if a comparison between the basic amp that you have with an Allo. Wish someone has it in Mumbai (near you) for a one on one comparison.

 

[SanjeevM]

Hmm! You guys do have a valid point. It would be an interesting experiment if a comparison between the basic amp that you have with an Allo. Wish someone has it in Mumbai (near you) for a one on one comparison.

I am using a XH-M189 3116D2 board. I did a 1 on 1 comparison with Volt+.

The XH board sounded better on the 6R spkrs I used.


Sent from my ASUS_Z00AD using Tapatalk

 

[greenhorn]

As long as people are putting in requests, i wonder if anyone has tried using it for auto use. There are 150W DC to DC step up boards available for slightly more than 300 INR. theoretically, if you add that to a one of these boards and run it at 24V, you have something very competitive spec wise to an entry level auto amplifier at the fraction of its price!

 

[shaan]

I doubt that a big cap bank will improve the bass response. What a big cap bank is supposed to do is provide instantaneous energy when the program material calls for it (like sudden dynamic peaks).

In general I find the bass weight of 3118 lesser than two other amps I compared it to (this is only in relative terms - it has enough bass heft in my system for my taste), but the deficiency in bass weight is in some way made up by the texture of the bass. It brings out the tonality of the bass much better, for example, than the LM3886 based MyRef Rev C amp (though MyRef has more bass weight).

I have seen this comment about reduced bass response for a lot of T amps - even saw the same for my TDA7498E. wonder why

Can you measure C4? Looks like it's in series with the input and if too small it can cause a bass bottleneck.

 

[jls001]

Can you measure C4? Looks like it's in series with the input and if too small it can cause a bass bottleneck.

Board 1 measures 2.95 uF and board 2 measures 2.98 uF.

 

[shaan]

Can you replace them with 10uF polyester/bipolar and see if anything changes?

 

[jls001]

I could try mounting it off board with a long lead. C4 is a very tiny surface mount cap. It will be a rather delicate surgery because the leads of a 10 uF cap will nearly touch each other when soldered in place of C4. Alternately, I could try tracing board traces, scratch off some of mask and solder the 10 uF without touching the existing C4 so that the new cap and C4 are in parallel.

If using a bipolar, is electrolytic okay? And what polarity must be observed? A polyester of 10 uF will be quite bulky and heavy and will be a challenge to fit in the existing space. Plus the sheer weight could upset the rather thin traces of the board.

To add: I just checked the datasheet circuit and there are 1 uF caps on both the positive and negative input leads. So I guess I'll need two caps per board.

 

[greenhorn]

does it have to be replaced? as long as we are trying to increase the value of a cap, can't we just strap something in parallel?

Edit: or even easier, remove the cap, bridge the solder pads to short it, and put the coupling cap off board somewhere along the input ?

 

[greenhorn]

It's not "just some mods here and there" as you put it. The $3.5 board uses "cheap" no-name parts vs. the Allo using good quality parts. The Allo also has a cap multiplier + 2 RCA + log pot on board. Also the Allo seems to be using a modified circuit [more parts!!!] as compared to the datasheet circuit which the cheap amp uses. Now I don't know if that justifies the price difference but then factor in the R&D cost + profit margin needed for Allo and I think that for $39, it is well priced.

Eight audio also has a TPA 3116D2 board which should hopefully be a midway between the two - looks like it uses the reference design (no way to be sure) but lacks the extra bells and whistles on the allo board except the RCA's

 

[shaan]

I could try mounting it off board with a long lead. C4 is a very tiny surface mount cap. It will be a rather delicate surgery because the leads of a 10 uF cap will nearly touch each other when soldered in place of C4. Alternately, I could try tracing board traces, scratch off some of mask and solder the 10 uF without touching the existing C4 so that the new cap and C4 are in parallel.

If using a bipolar, is electrolytic okay? And what polarity must be observed? A polyester of 10 uF will be quite bulky and heavy and will be a challenge to fit in the existing space. Plus the sheer weight could upset the rather thin traces of the board.

To add: I just checked the datasheet circuit and there are 1 uF caps on both the positive and negative input leads. So I guess I'll need two caps per board.

What greenhorn suggested is a better idea to connect the capacitor.

Before desoldering the SMD capacitor turn the amp on and measure the DC across it and also check the polarity of the voltage. If it shows 2-3V (IIRC the datasheet indicated this voltage in this node), then install either a polyester or a polarized electrolytic. If less than 2V then install a nonpolar electrolytic.

I wonder whether increasing the value of this capacitor will cause a better bass response from it or not. So you don't have to go premium with the cap as LF is where our ears will be focused at in the listening.

 

[greenhorn]

Just realized that doing so would short the differential inputs together. Not sure if the chip will like that. Too lazy to check the datasheet now!

 

[shaan]

Just realized that doing so would short the differential inputs together. Not sure if the chip will like that. Too lazy to check the datasheet now!

How would shorting C4 short the differential inputs together?

 

[greenhorn]

i assumed that there was one capacitor each at the input side INPR and INNR, and both would need to be changed :/

anyway in bridge mode, they are not tied together I guess anyway. perils of trying to interpret datasheets while half asleep

 

[shaan]

It looks in the board like the -in (pin 5) is shunned to ground via C2 (Or maybe I should say C20, the markings are scrambled by the footprints. C4 is actually C40 something).

 

[keith_correa]

Gain on Joshua's board is set for 26 dB so input impedance is ~30 k?. Given this, it would be interesting to see what a change to 10 F will bring about.

Joshua, if using a 'lytic, the direction is positive towards the chip pin.

 

[jls001]

Joshua, if using a 'lytic, the direction is positive towards the chip pin.

I must be having a few Elna Silmic II 10 uF 'lytics. The last I searched locally I could only get only few values below 10 uF in polystyrenes. Polypropylene is unobtainium.

 

[greenhorn]

TPA3118 listening impressions


Later, I swapped out the 470 uF caps to 1000 uF caps (a brand called CapXon which I'm using for the first time). I didn't hear immediate gains in sound, but now that I've put in more hours on the amplifiers, I think it was a worthy investment.

Given that TI explicitly calls our that the bypass capacitors be low ESR types, its worth a shot swapping them out with some good known low ESR type caps ( Panasonic FC/FR?)

 

[keith_correa]

From the datasheet:

The input capacitors used should be a type with low leakage, like quality electrolytic, tantalum or ceramic.