Problem with TDA7295 oscillating

[Adrian Tuddenham]

I am using a TDA7295 to drive a nominal 12 - 15 ohm load. On the test
bench the load is a chunk of electric fire element with connections made
by adjustable straps, so it is slightly inductive. The load in the
final version will be a loudspeaker voice coil on the end of some
unavoidably long wiring.

The TDA7295 data sheet claims "No Boucherot cell" but does not elaborate
any further. I am using the recommended circuit and the gain is set by
a divider of 22k and 680 ohms, exactly as specified.

An oscilloscope shows bursts of H.F. oscillation at two points on the
output waveform as it approaches +Vcc and -Vcc. These only occur on
load and they disappear if the signal amplitude is reduced below
approximately half the maximum peak-to-peak voltage. The power supply
decoupling is as recommended by the manufacturers and is mounted within
0.25 " of the chip pins.

To eliminate the possibility of earth loops or inductive coupling from
the output currents by temporarily connecting the load across different
signal and earth points in the layout. I have connected the signal
generator direct to the chip input. Neither of these tests made any
difference. The effect is independent of signal frequency below 25
Kc/s; but above that frequency it rapidly fades out.

The only thing which has removed the oscillation is the addition of a
Boucherot/Zobel network across the output terminals.

1) Have I missed something?

2) Is this one peculiar faulty chip?

3) Is the manufacturer making a false claim about needing no Boucherot
Cell (or a claim which is only true under exceptional circumstances)?

 

[Klaus Bahner]

The TDA7295 data sheet claims "No Boucherot cell" but does not elaborate
any further. I am using the recommended circuit and the gain is set by
a divider of 22k and 680 ohms, exactly as specified.

Where have you found that statement? The datasheet says "The Boucherot
cell R6, C10 is normally not necessary for a stable operation. It could
be needed in presence of particular load impedances at Vs < +/- 25V"

An oscilloscope shows bursts of H.F. oscillation at two points on the
output waveform as it approaches +Vcc and -Vcc. These only occur on

Sounds like an artifact from the MOSFET output stage. Perhaps your
bootstrap is not working properly, starving the MOSFETs for sufficient
gate drive, causing the typical MOSFET bursts of oscillations due to
their non-linear gate capacitance.

Klaus

 

[Phil Allison]

"Adrian Tuddenham"

I am using a TDA7295 to drive a nominal 12 - 15 ohm load. On the test
bench the load is a chunk of electric fire element with connections made
by adjustable straps, so it is slightly inductive.

** SFA - in reality.

The load in the
final version will be a loudspeaker voice coil on the end of some
unavoidably long wiring.

** Have you tried that ??

The TDA7295 data sheet claims "No Boucherot cell" but does not elaborate
any further.

** Might mean with typical loudspeaker loads.

Or it might mean that there are no damaging CONTINUOUS oscillations.

What you have described is hardly likely to be audible on any programme - at
all !!

I have used the TDA7295 and added the standard 0.1 uF plus 8.2 ohms as a
matter of course.

The only thing which has removed the oscillation is the addition of a
Boucherot/Zobel network across the output terminals.

** You a man or a mouse ?

1) Have I missed something?

** A bushel of common sense maybe?



.... Phil

 

[Adrian Tuddenham]

On a sunny day (Thu, 14 Mar 2013 12:01:31 +0000) it happened
[email protected] (Adrian Tuddenham) wrote in
<1kzq0bp.4xngrzszv7weN%[email protected]>:
[...]

An oscilloscope shows bursts of H.F. oscillation at two points on the
output waveform as it approaches +Vcc and -Vcc. These only occur on
load and they disappear if the signal amplitude is reduced below
approximately half the maximum peak-to-peak voltage. The power supply
decoupling is as recommended by the manufacturers and is mounted within
0.25 " of the chip pins.

[...]

I use that other chip (TDA7294 IIRC) and I have never seen any oscillations.
Are you sure your power decoupling caps are of good quality?
Perhaps bypass those with a few hundred nF to test?

There are two 63v 220nf capacitors mounted within 6mm of the chip.
Further away there are two good-quality 220uf electrolytic capacitors.

No, at least in my case, with this chip, it clean as it can get without thet
RC to ground.
Wiring, layout?

The layout is 'difficult' becaue of the circumstances. The IC is
screwed to the inside of the large die-cast box which houses the
equipment. It has a small PCB mounted directly on its pins which
carries the H.F. decoupling, the bootstrap capacitor, the input D.C.
blocking capacitor, the input bias resistor and the two feedback
resistors.

There is then a long hank of cabling which connects this to the main PCB
by means of a set of header pins and sockets. I was worried that
inductive or capacitive cross-talk in this cable might be part of the
problem, but I have eliminated that by connecting the load and a low
impedance source directly to the chip pins.

 

[Adrian Tuddenham]

Where have you found that statement? The datasheet says "The Boucherot
cell R6, C10 is normally not necessary for a stable operation. It could
be needed in presence of particular load impedances at Vs < +/- 25V"

We must have different data sheets. Mine is by STMIcroelectronics; the
reference to Boucherot cells is in the list of bullet points on the
first page and there is no other mention of them anywhere else in the
document. There are no components R6 or C10 shown on any of the
circuits.

It seems as though you have found the answer. My supply voltage is +/-
30v and I have no idea what the load impedance will be when the circuit
is in use. it looks as though I ought to incorporate a Boucherot/Zobel
network

Sounds like an artifact from the MOSFET output stage. Perhaps your
bootstrap is not working properly, starving the MOSFETs for sufficient
gate drive, causing the typical MOSFET bursts of oscillations due to
their non-linear gate capacitance.

I'm not aware of anything wrong with the bootstrap, but I will check it.

 

[Klaus Bahner]

We must have different data sheets. Mine is by STMIcroelectronics; the
reference to Boucherot cells is in the list of bullet points on the
first page and there is no other mention of them anywhere else in the
document. There are no components R6 or C10 shown on any of the
circuits.

Mine is from ST, too. Dated April 2003. The Zobel network and quoted
text are included in figure 1 "Typical Application and Test Circuit".

In fact the bullet point regarding the Boucherot cell is also in my
datasheet - I seldom read these feature lists, because they are way too
often "marketing highlighting", which is proofed again by this datasheet

Klaus

 

[Adrian Tuddenham]

Mine is from ST, too. Dated April 2003. The Zobel network and quoted
text are included in figure 1 "Typical Application and Test Circuit".

The date on mine is January 2003 and there is definitely no Zobel
network shown in Fig,1. It looks as though they had discovered their
mistake by April (the Zobel components on your circuit diagram are
numbered at the end of the numbering range, which suggests that they
were added last).

In fact the bullet point regarding the Boucherot cell is also in my
datasheet - I seldom read these feature lists, because they are way too
often "marketing highlighting", which is proofed again by this datasheet

Exactly - but I searched for further information and couldn't find any.

 

[Adrian Tuddenham]

The TDA7294 only mentions 'NO BOUCHEROT CELL' on page 1, and seems from
1998. It is possible they changed production processes since then, renamed
the chip to TDA7295, and then found out later it needed the RC network. I
have done pretty strange things to the TDA7294, and usualy with a scope
connected, never a problem. 'Strange things' include driving capacitive
loads (lots of piezos parallel at > 20 kHz), and power transformers (step
up) at 60 Hz.


Just add that RC, and be done with it

That seems to be working. I've got it on a thermal soak test at the
moment, but will try an aasortment of loads and signals tomorrow.