I recently acquired an ancient philips dfa888 amplifier with builtin TDA1541A based dac. Disappointingly it didn't contain the S1 TDA1541A as I had read somewhere it would, just the plain TDA1541A. Nonetheless, it's a very nice dac with dc servo to maintain 0 volt at the output. It has an input circuit consisting of a lot of sanyo cmos switches , the LC7821 and LC4966, providing input and output selection and a "source direct" function, bypassing tone control and balance , but not loudness. The part of my query is the power amp section:
The audio signal from the inputy selectorcircuit enters at JE51 (from PV01 the selector circuit) with right the loudness circuit and left the 50 kOhm volume pot. The slider of the pot goes through the source-direct/tone-control selector ic lc4966. When source-direct is selected the ac-signal goes straight via the 680Ohm resistor RG02 to J701 input connector and there via C701 10uF cap. and 100Ohm resistor R705 to the inverting input of the NJM2068 opamp. As far as I am able of understanding the circuit the output of the opamp drives the emitters of the first stage of the fully balanced output circuit. The collectors of these drive the Vas , which drives a triple emitter follower power stage. The thus inverted output of the power stage is fed back via a 27kOhm resistor and 15pF cap to the the non-inverting input of the opamp , thus forming a shunt feedback gain factor determined by the 27k resistor divided by the sum of the 150Ohm R703 and the position of the 500Ohm pot RG19-2 (coupled with the the 50kOhm RG19-1 volume pot). The amp is spec'd for 85W in 8Ohm output. This equates 26V and the input sensitivity is spec'd as 150mV. The feedback arrangement has a gain of 27000 divided by 150 max, which is 180. The sensitivity of 150 mV requiers for 26V output a gain of 26 divided by 0.15 which is 173. That seems rather on top. My question is, did I interpret the circuit correctly and/or can someone more knowledgable shed a light on this power amp circuit. The used opamp NJM2068 gives a 35mV offset at the powerstage output. When replaced with a opa1656 the output offset voltage is reduced to less than 1mV and when replaced with an opa2189 the offset is ca. 1.8mV. The offset voltage of the NJM2068 is spec'd as 0.3mV , of the opa1656 it's 0.5mV and of the opa2189 it's 5uV , but the offset voltage at the power output seems more related to the input offset current,which is lowest for the opa1656 and highest for the njm2068. Is this a correct observation ?
Tha amp sounds wonderfull with the opa1656 by the way, as with the njm2068 , but the latter seems a tad less sparkly and spacious, anyway I like 1mV offset better than 35mV offset so the opa1656 is there to stay.
Thanks, systux
Tha amp sounds wonderfull with the opa1656 by the way, as with the njm2068 , but the latter seems a tad less sparkly and spacious, anyway I like 1mV offset better than 35mV offset so the opa1656 is there to stay.
Thanks, systux
Last edited:
The original bipolar op amp seems to want the input 220k resistor to be replaced with a 27k, to match the feedback resistor value.
Thought about that too, will match the input resistor later, for now I tried some more opamps I have lying around among which a Signetics NE5532 and a LM833 salvaged from old electronics along with a new LM4562 of recent production. The ne5532 produced 88 mV offset at the output, the LM833 43 mV and the LM4562 just as much as the opa1656 , ca. 1mV, so it seems bjt or fet input doesn't matter , but anciency. Maybe it's a matter of how well matched the input differential pair of the opamp is and newer production opamps have better matched input pair, be it fet or bjt. I pulled the LM4562 , the supply is 17.36V symmetric and the 4562 is 17V max. Maybe that's why old ne5532 are glorified over current production, they have less well matched input pair and thus produce more (pleasing ?) harmonics.
The soldering of the pcb's is horrendous. To get the dac circuit working I had to reflow a lot of broken solder joints. The disassembling of the power amp pcb will be quite an adventure, it contains the power supply for the other circuitry as well and has a lot of connectors, which are all prone to broken solder joints. The resistors are the very very small 0.125W type and space is cramped to fit new ones. Now first enjoying the music played from my laptop over an usb to spdif converter to the tda1541 dac.
The soldering of the pcb's is horrendous. To get the dac circuit working I had to reflow a lot of broken solder joints. The disassembling of the power amp pcb will be quite an adventure, it contains the power supply for the other circuitry as well and has a lot of connectors, which are all prone to broken solder joints. The resistors are the very very small 0.125W type and space is cramped to fit new ones. Now first enjoying the music played from my laptop over an usb to spdif converter to the tda1541 dac.
I just discovered I have been rambling in my first post. The power amp is just an opamp with an extra gain stage, so the inverting and non-inverting inputs of the actual opamp, the njm2068, switched function . Inverting and non-inverting is relative to the phase of the output of the total amp. Anyway, mystery solved how the nfb and signal input are connected to the inputs of the opamp. I allso discovered why the amp was so much cheaper than other offerings after disassembling the power amp pcb. Someone had allready been busy "repairing" what appears to have been a melt down of one channel and something what looks like a practice in using a drill around the solder joints of one of the power reservoir elco's, and filled up the holes in the pcb with solder afterwards. The replaced components are a strange collection of fake parts from aliexpress and what have been found in a junk box as resistors, matching in original value but not in quality or type. The speaker relay contacts were bridged , making the whole speaker protection circuitry pointless. And the quality of the pcb itself is such that any attempt to repair results in the dissolving of the copper traces, a typical costs very big objects production of that time strangely contradicting the quality of the original components in the circuitry , temperature stable resistors in the power stage, all elna audio caps, high quality solid state devices, than why the lousy quality of the pcb and underrated components in the power supply section. Coloration of the pcb where components got too hot and loosening the copper from the pertinax in the process. Is it worth the trouble trying to repair this, yes of course it is, see if you can find the faults in the relay circuit that made bridging the relay contacts necessary to have audio, replace the odd components with the original quality components , change the load resistor of the input to the value of the feedback resistor and see if that helps reducing the offset at the output of the power amp, that will teach me buying cheap stuff. Luckily I have a few donor parts to disassemble for the original components. Happy diy'ing .
It's rather simple: there are CB, CE and EF stages here; only the CE inverts phase, so there is a single instant of phase inversion after the opamp, hence the swapped inputs.
Did you renew C701/C713? Leakage in those will certainly have an effect upon the offset too.
Yes, I noticed the phase reversal with respect to the opamp inputs, but somehow in my head the inverting and non-inverting inputs stayed at the same legs as specified in the datasheet for the inverting and non-inverting inputs of those opamps, causing a big question mark what strange circuitry the power amp was, but actually it's a rather simple nothing special power stage. Anyway , I started "repairing" the power amp pcb. reflowing al the solder joints, repaired broken tracks and joints to the speaker protection circuitry, removing the shorts over the relay contacts for the speakers, change inappropriate resistor types in the one channel that has been repaired previously to match the types in the other, original channel. Changed the feedback resistors from 27k to 20k , making the amp a bit less sensitive and change the input load resistor from 220k to allso 20k, matching the feedback resistor in value. Replaced the opamp socket because there appeared to have been a break in leg 8 causing a less well connection. After all that, the speaker protection circuitry functioned again, offset at the output was less than 1mV for all tested opamps, jfet or bjt and the amp seems to be in good condition again.
I definitely don't need 85 watts/ch in my listening room, so I changed the power transformer , a 220V only EI type that becomes very, really very hot when in use , providing near + and - 60V DC to the amp, for a 230V toroidal transformer providing +- 35V DC , enough for allmost 50Watts/ch, which is more than enough for normal listening levels.
Up to the tone control and input selection circuitry, I don't like all that cmos switching in the signal path , going to figure how to make that simpler, don't need all those inputs and outputs. This is the tone conttrol :
Anybody has an idea how this works ? Input is at JE01 , output at JE03 via balance pot to power amp.
Thanks .Systux
I definitely don't need 85 watts/ch in my listening room, so I changed the power transformer , a 220V only EI type that becomes very, really very hot when in use , providing near + and - 60V DC to the amp, for a 230V toroidal transformer providing +- 35V DC , enough for allmost 50Watts/ch, which is more than enough for normal listening levels.
Up to the tone control and input selection circuitry, I don't like all that cmos switching in the signal path , going to figure how to make that simpler, don't need all those inputs and outputs. This is the tone conttrol :
Anybody has an idea how this works ? Input is at JE01 , output at JE03 via balance pot to power amp.
Thanks .Systux
Last edited:
It wasn't the input offset current, but the input bias current.
For example, suppose you have an op-amp with 500 nA of input bias current, 0 nA of input offset current and 0 mV of input offset voltage. Each input then draws 500 nA, causing a voltage drop of 110 mV across the 220 kohm input resistor. Both inputs therefore end up at -110 mV with respect to ground. The voltage drop across the 27 kohm feedback resistor, which will be 13.5 mV, brings the output DC voltage to -110 mV + 13.5 mV = -96.5 mV.
Apparently your NE5532 has an input bias current of about 450 nA. The spec is 200 nA typical, 800 nA maximum at 25 degrees Celsius.
The LM4562 has input bias current compensation, so its input bias current is quite small (and its input current noise large, but that's another topic).
Last edited:
I see , what rayma said at first. So with a bjt it's more important than with a jfet input opamp because the input bias current is so much higher with bjt than with jfet input.
I think I understand the tone control. It's a fixed nfb non-inverting opamp that gets additional parallel to the 6.8k R more or less frequency dependant feedback from QE02 and QE03 via the frequency dependant input filters at the inputs of QE02 and QE03, the 0.012uF , 56pF and 1.2k and the 3.3uF , 0.018uF and 1.2k. +-8dB at 100Hz and 10kHz according to the specs.
I think I understand the tone control. It's a fixed nfb non-inverting opamp that gets additional parallel to the 6.8k R more or less frequency dependant feedback from QE02 and QE03 via the frequency dependant input filters at the inputs of QE02 and QE03, the 0.012uF , 56pF and 1.2k and the 3.3uF , 0.018uF and 1.2k. +-8dB at 100Hz and 10kHz according to the specs.