Tripath TAA4100A issues

nertog

Member
2008-01-27 1:41 pm
Hi...first post on the forums:)

I've been designing a 2-channel TAA4100 based amp in the last couple of months and I just finished the first prototype. It's almost interely based on the application schematic provided in the datasheet. The only changes are the input capacitors (1.5 µF foil instead of 0.47µF elco) and the input are clamped with 5.2V zeners. The power supply is a 20V 120W SMPS.

Upon first testing I discovered a few problems. First of all: the transition from mute to unmute is not quiet at all. I get a pretty large "thump" followed by +/- 0.5 seconds of white noise (oscillations?).

The little amp sounds very nice at low power, but once I ask for more than a few watts of output power I get some strange sounds: at a certain power output level you can hear a slight "ticking" sound through the speaker..when you raise the volume some more this "ticking" goes over in white noise (again: oscillations?). When the heatsink and chip is warm (let's say 45°C) this noise will be more prominent and sometimes won't stop even when you turn down the volume. If you play, for example, a pure 25hz sinus tone, there is a ticking or crackling sound even at lower power output (speakers are ok, I tested this with a few different pairs). At higher frequencies (80-100hz) I can get a low more power out of the amp. The problem seems to be concentrated at lower freqs....

I triple checked all soldering, schematic and board layout...and everything seems to be ok. I have the feeling this has something to do with a grounding issue...or maybe something at the inputs: when I touch the RCA jacks (metal, signal carrying part), the amp stays quiet, as it should. However, when I touch the input pins on the tripath chip (or anything after the input caps) there's a loud hum. Even stranger: if I touch input nr. 3: 50hz hum + white noise. The other inputs: just a 50hz hum.

FYI: Power ground (= - pole of the SMPS input jack) and signal ground are connected at 1 point on the PCB.

Any ideas what this could be? Maybe a faulty component? Or a major design fault?

Please have a look at the schematic, board layout and pics of the board.

Many thanks,
Wim


[IMGDEAD]http://img155.imageshack.us/img155/2163/schemadc4.jpg[/IMGDEAD]

[IMGDEAD]http://img155.imageshack.us/img155/8329/printlr4.jpg[/IMGDEAD]

[IMGDEAD]http://img155.imageshack.us/img155/2235/hpim5522mx8.jpg[/IMGDEAD]

[IMGDEAD]http://img100.imageshack.us/img100/2831/hpim5523tn1.jpg[/IMGDEAD]

[IMGDEAD]http://img352.imageshack.us/img352/9530/hpim5527lb9.jpg[/IMGDEAD]
 

wakibaki

Banned
2008-01-08 11:51 pm
What's going on with MUTEB here? How are you switching it? I'd expect to see a cap to ground here so that at switch-on the voltage on MUTEB would slowly rise and unmute the output. I don't know, maybe you have a micro driving it, a cap here wouldn't hurt to keep any digital noise out of the chip.

Otherwise nothing obvious on the diagram. I can't tell from the layout exactly how the diodes at the outputs are returned to VPP and PGND, but these want to go to VPP1, 2 etc. and PGND pins with as short a track length as possible.
 

Iain McNeill

diyAudio Moderator Emeritus
I wouldn't use zeners on the signal inputs personally. Zeners make excellent noise sources. This could be what you're hearing as you un mute, there's a brief conduction in those zeners. I prefer an RC for ESD protection - 100ohms/100pF won't impact audio too much.

You say you are connecting AGND and GND at one point. Is that a hard connection or is there a cap/resistor connecting them. Split ground designs can be very problematic. A resistive load for the signal inputs wouldn't hurt either, 47Kohms is usual.

I would be a bit concerned about the circulating currents in the output LC filters. There's a lot of HF energy in this circuit and your layout looks like it has rather large loop areas for these traces. Could be some EMI effects here at higher power levels.

Do you have a scope to look at the clicking sound and see where it appears?
 

nertog

Member
2008-01-27 1:41 pm
Thanks for the input on the zeners! I removed them and there's no more noise at start-up!

It doesn't solve the other issue though....I did some more testing (just with a DMM, will have acces to a scope somewhere next week): maximum power output before the 'ticking' appears, tested with a 50hz sinewave, is 15.6V at 2.32 amps = 36 watts. Speaker impedance at that frequency seems to be 6.7 ohms. There is a very clear point at which the distortion appears.

When lower the power supply voltage, the point at which the distortion appears also lowers....seems to be a case of amplifier clipping, but why?

I guess this rules out the inductors. If they were the cause, the distortion would be independant of the power supply voltage, right?

Concerning the layout: if you look at the pictures of the PCB itself (last pic), you'll see there's a ground plane on both sides of the board...I left this out of the Eagle screenshot. I tried to keep the length of the other high power traces to a minimum....
 

nertog

Member
2008-01-27 1:41 pm
wakibaki said:
What's going on with MUTEB here? How are you switching it? I'd expect to see a cap to ground here so that at switch-on the voltage on MUTEB would slowly rise and unmute the output. I don't know, maybe you have a micro driving it, a cap here wouldn't hurt to keep any digital noise out of the chip.

Otherwise nothing obvious on the diagram. I can't tell from the layout exactly how the diodes at the outputs are returned to VPP and PGND, but these want to go to VPP1, 2 etc. and PGND pins with as short a track length as possible.

I'm switching the amp on/off with the sleep pin...yeah, I know....my mistake. Muteb is pulled high to the +5V, all the time. This is probably the cause of the thump when switching the amp on.

Anyway, that's not the main problem as the amp is switched on 24/7.

The amp really struggles with low frequencies (20 - 30hz) even at low power...distortion and 'ticking sounds' everywhere...

The output diodes are connected to the ground plane and routed back to the correct pins on the chip. Maybe someone can check this on the schematic and layout, but it should be correct.

Thanks!
 

nertog

Member
2008-01-27 1:41 pm
Iain McNeill said:

You say you are connecting AGND and GND at one point. Is that a hard connection or is there a cap/resistor connecting them. Split ground designs can be very problematic. A resistive load for the signal inputs wouldn't hurt either, 47Kohms is usual.


AGND and GND are connected by a 2mm long pcb trace, no caps/resistor/inductors used. How else would you connect the AGND and signals?

What's the reason for a resistive load at the inputs? Wouldn't this change the gain of the amp (you're creating a voltage divider at the inputs....).

Thanks!
 

wakibaki

Banned
2008-01-08 11:51 pm
Sounds like the amp works, pretty much.

All the caps rated for the voltages they experience?

Is the ticking periodic or at random intervals? Faint or pronounced? How frequent? Coinciding with the music or unrelated apart from volume? Is the energy high frequency (hissy)
or low (thumpy) or both?

It sounds like you've got a lab-type PSU. It's not current limiting is it? Noise could be coming out of there, although it's not likely.

Have a look at the board and psu in the dark. I found some things by seeing a tiny, tiny spark.

None of the above really sound right for something that goes away when you touch the input gnd tho. The resistive load at the input sounds good anyway, 47k should be big enough to have no significant impact on the gain but will ensure that both terminals are at roughly the same DC potential under normal conditions.

Get somebody else who knows what they're doing to cross-check the components on the board against the BOM. Sometimes you can misread a value time and time again.

Iain could be right about the long traces, is that the smallest diode package that will do? It's hard to tell, of course, without trying to route the circuit yourself whether you could do any better, but the board does not really look densely packed. Nice looking low-profile inductors, but taller ones might have a smaller footprint Since these are not in series, I'd rather see them all in a single line. Better to try to eliminate any other causes before worrying about this tho...

w
 

wakibaki

Banned
2008-01-08 11:51 pm
Oh, GND and AGND should come together where the (decoupled) power comes in, this can be the terminal or a trace or board area but the basic idea is to keep circulating currents in GND out of AGND and vice versa, so you wouldn't connect power ground to one side of GND plane and connect AGND to the other side as all AGND returns would therefore pass through GND. Also you don't want them overlapping.
 

nertog

Member
2008-01-27 1:41 pm
wakibaki said:
Sounds like the amp works, pretty much.

All the caps rated for the voltages they experience?

Is the ticking periodic or at random intervals? Faint or pronounced? How frequent? Coinciding with the music or unrelated apart from volume? Is the energy high frequency (hissy)
or low (thumpy) or both?

It sounds like you've got a lab-type PSU. It's not current limiting is it? Noise could be coming out of there, although it's not likely.

Have a look at the board and psu in the dark. I found some things by seeing a tiny, tiny spark.

None of the above really sound right for something that goes away when you touch the input gnd tho. The resistive load at the input sounds good anyway, 47k should be big enough to have no significant impact on the gain but will ensure that both terminals are at roughly the same DC potential under normal conditions.

Get somebody else who knows what they're doing to cross-check the components on the board against the BOM. Sometimes you can misread a value time and time again.

Iain could be right about the long traces, is that the smallest diode package that will do? It's hard to tell, of course, without trying to route the circuit yourself whether you could do any better, but the board does not really look densely packed. Nice looking low-profile inductors, but taller ones might have a smaller footprint Since these are not in series, I'd rather see them all in a single line. Better to try to eliminate any other causes before worrying about this tho...

w

Yes, all caps are rated for 50+ volts (except the power supply caps are 35V).

I did some more testing: the lower the frequency, the sooner the ticking sound appears. If I try the amplify a 5Hz sinewave it starts ticking even at very low power. If I increase the input signal to a few watts the ticking sound becomes a steady hum...if I increase the input some more the amp starts to oscillate (white noise) and goes in protection mode.

The ticking itself is random but loud. There is a very pronounced threshold at which the ticking and other distortion start. The spectrum of the ticking sound is quite wide I think...it's probably a large spike on the speaker outputs. I'll try to measure it with a scope next week.

I can also hear this ticking sound (and the other distortion) on the amplifier board itself...weird!?

The ticking and distortion is independent of the type of PSU used. However, the oscillations that start after the ticking are a loud hiss with the SMPS and a loud 50Hz hum with a normal transformer based supply.

Connecting 1 input instead of 2 gives about double the amount of power at the output before the ticking starts (that's the power output I measured in the previous post). You would think it's a power supply problem...but it isn't.

This is what Tripath writes about connection AGND and PGND : "the input and output ground should be externally connected
at a single point as close to the TAA4100A as possible". That's what I did: AGND is connected to the ground plane at pin 10 and 14 (both pin 10 and 14 are AGND). The PGND and AGND circuit is completely separate, except for 1 connection between pin 10 and 14....

Thanks for all the help!
 
better late then never

Hi there, your simple issue is that you have connected the 4 channels to 2 channels at the output, probably not knowing that the switching frequency of each channel is different to prevent high spikes on the EMI/EMC tests, this means that the outputs have different switch frequency's and start biting each other, effectively short circuiting at the times the switching patterns are opposites of each other.

With this IC you cannot change it, as Tripath have put the caps that decide the switching frequency inside the IC, as are all resistors. so i m afraid the circuit as is is never going to function.

Greetings,

Arjen Helder