Class-D Output Filter Designs, SURE TDA7498 as example

[trevmar]

If you download the evaluation version of the MultiSim software I used then I posted the file which is already set-up to emulate the filter at:

http://trevormarshall.com/class-d-tu...ut_filter.ms11

All you have to do is flip the toggle switch on the GUI to start the emulation running I have set up the current and voltage plots The info is near the bottom of my detailed info page at:

Trevor Marshall - Class D Audio Amplifier Design - TDA7498 Output filters

When the MultiSim evaluation period runs out, there is a student version of the MultiSim software available for $40-$60, and the simulation runs just fine on that.

..Trevor..

[jyoung070848]

Thank you Trevor. It looks easier than I thought. I need to look over your whole web site. It is nicely designed.
James

[the7trumpets]

Trevor,

I'm trying to get a handle on all of the specific things you did to improve this amplifier. I'm actually considering getting the 6 ch. version of this and modding it as per your findings, and using three of them for a multi-channel amplifier build. I have attached Sure's schematic for their 6 channel version, and I have attempted to enumerate where your mods would be on the 6 channel version, so all component numbers below refer to the attached 6 channel schematic, with the exception of the added damping circuit which refers to your component names in your SPICE schematic. I have only listed the component numbers for the first TDA7498 chip, as I would duplicate on the other channels.

Could you go through this list and confirm that this is what you did and answer some questions?

Thanks!


Input section:

May: Replace Input Capacitors: MKS polyester (1uF)
C5, C11

How Important? Add 1nF bypass capacitors from signal input to ground:
Since Test Points are now on the input side of the input coupling capacitors (C5, C11) What is the optimal location of this? Does it need to go from ground to between C11 and INPA (pin 22), and from ground to between C5 and INPB (pin 32)?
Also, the reference design has the same configuration (1uF coupling, 1nF bypass) on pins 23 and 33. This is unnecessary since the - input is always tied to ground, right?

May: Decrease Clock Frequency:
R8 (33K) Replace with 30K
Why? Does this make a big difference, or affect the output filter for some reason?

Snubber circuit:
R14 / C41, R15 / C45
You say 'add' the snubber circuit, but it looks like it's already in the schematic at the listed component locations. Did you just swap the 0.1uF caps for 0.33uF caps that are the value recommended by ST, or am I missing something?


Output Filter:

Must: Replace Output Capacitor: MKS polyester (1uF)
C42, C46

Must: Replace Inductors: Bourns SRF1280-150M (15uH)
L1, L2, L3, L4

Must: Replace Existing Output Capacitors:
C43, C44, C47, C48
with MKS polyester 330nF/63V caps, soldered to banana plug pads and ground plane vias

Must: Add Damping Circuit:
On your SPICE circuit: C5 (220nF), R9 (22 Ohms, 5W)
from out+ to out- on binding post pads

[trevmar]

.
Replacing the input capacitors is explained in the application note which I linked on Ceramic input capacitors. It reduces low-frequency distortion, but loudspeaker distortion is always going to be dominant in this region. Makes the amplifier specs look better, though

Originally Posted by the7trumpets

How Important? Add 1nF bypass capacitors from signal input to ground:
Since Test Points are now on the input side of the input coupling capacitors (C5, C11) What is the optimal location of this? Does it need to go from ground to between C11 and INPA (pin 22), and from ground to between C5 and INPB (pin 32)? Also, the reference design has the same configuration (1uF coupling, 1nF bypass) on pins 23 and 33. This is unnecessary since the - input is always tied to ground, right?

It provides a high-frequency ground. I saw no change from doing this, I did it to match the reference design, and because I already had the capacitors lying around...

May: Decrease Clock Frequency: R8 (33K) Replace with 30K
Why?

Again, my goal was to get the best performance I could out of the TDA7498, and raising the clock frequency measurably reduced the distortion.

Snubber circuit: R14 / C41, R15 / C45
You say 'add' the snubber circuit, but it looks like it's already in the schematic at the listed component locations. Did you just swap the 0.1uF caps for 0.33uF caps that are the value recommended by ST, or am I missing something?

It is on the SURE schematic, but the parts are not on the boards they ship. And SURE's changing a 100,000pf capacitor to a 330pF is a very big deal indeed at that point in the circuit.

Output Filter: Must: Replace Output Capacitor: MKS polyester (1uF)
C42, C46
Must: Replace Existing Output Capacitors:
C43, C44, C47, C48
with MKS polyester 330nF/63V caps, soldered to banana plug pads and ground plane vias

Again, refer to the application note on distortion in ceramic capacitors. This made a huge difference to the distortion in the mid and high frequency ranges.

Must: Replace Inductors: Bourns SRF1280-150M (15uH)
L1, L2, L3, L4

This is particularly important at higher supply voltages (above 24V) and to provide a stable peak out current

Must: Add Damping Circuit: On your SPICE circuit: C5 (220nF), R9 (22 Ohms, 5W) from out+ to out- on binding post pads

This also provides damping to stabilize the feedback loop and reduce the peak current from the TDA7498 output switches to values less than their 6-8 amps maximum rating.

Hope that helps,
Trevor

[stujonesmd]

Fascinating. Looking back at various posts regarding the Sure TK2050 and TA2024 boards, it appears that Sure's suppliers iterate through design upgrades after initial release, with the V1 boards not quite living up to promised specs (or stability over time for that matter) and later revisions doing much better. It might be worth taking another look at the TDA7498 boards to see if they work better now. These are now provided in the USA by Parts Express, which in my (admittedly limited) experience probably means they are working better than your samples.
I wish I had time and energy to experiment with these boards. I have a number of TA2020 and TK2050 based amps as well as a ton of high quality recent and old class AB amps, and the class D amps all sound very to remarkably good. My best experiences have been with the Sure 2 and 4 channel TK2050 boards, which I have found powerful enough for all practical purposes with outstanding reproduction quality. I have also tried the Dayton DTA100, which is a TK2050 optimized for 8 ohm loads, also powerful enough, with slightly less but still very good sound quality. My AB amps include NAD2140s, a Hafler 220, and a couple of recent Dayton 150W class ABs that I am using as 150W monoblocs into rebuilt Dahlquist DQ10s, also excellent reproduction quality.

I find that all of these amps I have tried benefit greatly from some input signal gain. It turns out it is easy to provide this using a single op amp stage (only 6-10 dB is usually needed), easily built on perf board (use an IC socket--good chip insurance, especially if you are not a soldering wizard). A used control preamp will also work nicely. The DTA100s, in particular, benefit from this, and using one of these as a power amp means you can hide the unit (and its blinding blue LED power indicator) out of sight.