TDA8947J, Opamp filters and virtual GND

[jeroen94704]

I'm a noob in the diy electronics field, so this may be elementary, but I just can't seem to figure this out:

In order to make a 2.1 amplifier, I've disassembled a Logitech PC speaker set containing two TDA8947J's, which is a 4-channel amplifier chip. I have successfully built the reference circuit from the datasheet in a simple 2-channel (2.0) configuration using channels 1 and 2. This works like a charm.

Now I want to add some filters, so I can send the highs to channels 1+2, and the lows to 3+4, which can be wired in bridge mode to drive a single subwoofer driver.

Since the opamp-based filter circuits need a split supply (+, - and GND), and since I only have a single supply (Vcc and GND), I've created another opamp circuit that outputs Vcc/2.

I've verified that the filter works as expected using a scope and signal generator.

However, when I pass the input signal through the filter (using the virtual GND and signal ground), the filtered signal is offset by Vcc/2 compared to the supply ground, which is the ground used for the TDA. Apparently the TDA doesn't like this, and an incredibly loud noise, completely drowning out any music, is the result.

So what I'm wondering is: Is the principle of what I try to do sound? I.e: Use a Vcc/2 as virtual ground for the opamp filter and input signal, while still using the power ground as ground for the TDA? If not, how should this be wired?

If you need more info, please let me know.

Thanks in advance,

Jeroen

 

[AndrewT]

Create a dual polarity supply by using a pair of half wave rectifiers connected to the transformer secondary. Use these only for the low current opamp power.

 

[jeroen94704]

Doesn't that do the same as a simple voltage divider? In either case, I have a signal ground that sits halfway between my power supply voltage, which seems to be a problem for the TDA.

 

[AndrewT]

Why do you bother asking?
Just to argue rather than go and research half wave rectifier?

 

[jcx]

active supply splitting is fine for the filter circuits

likely the TDA internal DC signal ref doesn't match your active gnd

the TDA shows inputs AC gnded to the negative rail, you should C bypass your Vcc/2 ref to Vneg and keep the blocking caps between your filters and the TDA signal inputs

 

[jan.didden]

However, when I pass the input signal through the filter (using the virtual GND and signal ground), Jeroen

What do you mean by this Jeroen? You are not connecting signal gnd to virtual gnd are you?
The filter input needs to be coupled to the input (which presumably sits at Vcc/2) capacitively.
Then the output of the filter (which sits at Vcc/2) is capacitively coupled to the power chip input (which also sits at around Vcc/2 I presume).
In all these 'transactions', signal ground is the only reference.
If indeed the filter output AND the power input sit at Vcc/2 you can just DC couple filter output to power input, of course.

jan didden

 

[jeroen94704]

Thanks for the replies everyone. I'm trying to wrap my head around it all, and tried some more things based on your feedback, but no luck so far.

I thought a diagram of how I have things wired up right now might help. It probably something obvious. I hope the diagram makes sense at all . One important thing to note is that the diagram shows a TDA7375, because I can't find the TDA8947J in Eagle. The two are not pin-compatible, but they are functionally close. At least I hope to get the idea across:

An externally hosted image should be here but it was not working when we last tested it.

To answer Janneman's questions:

- By signal ground I mean the ground of the input signal. The SGND pin of the TDA is connected to power GND.
- The TDA is not connected to VCC/2, but is powered by the full VCC and power GND voltage, mainly because there is no way the opamp used to obtain VCC/2 can provide the current needed by the TDA.

 

[jcx]

the op amp may not like driving the rail Caps on the active gnd Vcc/2 output

I'd remove them, maybe move one to the input R divider center to power gnd

 

[jeroen94704]

The "rail caps", that would be C3 and C4, right? So remove those, and move one of them to "the input R divider center to power gnd". But where is that?

 

[jcx]

parallel with R4

 

[Gaftbart]

Hi there, right at the moment, i am building an amp just the way you do jeroen. I use tl07x for the preamp and filter stage, with single supply referred to vcc/2 as virtual ground. As power amp there is an TDA 7381. It has 4 seperate inputs and outputs as BTL configuration. Everything is mounted and grounded on one board. supplied by the same power supply with 16v dc. Both units work perfectly on their own as for you jeroen. now as i connect both stages with no load at the output, i can read a clean output signal (comming from a sine wave signal generator). but as soon i connect a speaker at the output, large dc voltages from max to min supply voltage, changing every half second. this results in that the speaker is moving back and forth very unhealthy. but still a bit of the original signal comes though. as well i decoupled the signal the same way on the diagram above. tried to without c10/c11. now there is a constant direct current pull on the speaker. i didnt use C3/C4 either and put one where jcx recommended. but still no difference. i am very close now to finish this project, but still stuck at this point. i wonder if its even possible to make it work. but as both stages work fine on their own... i would love to hear some more opinions on that matter