I'm trying to diagnose an issue with a Sony TAE-F20 preamp. The preamp has a very noticeable hiss even at zero volume, and volume increases far quicker than it should when the volume is turned up from zero. The balance, volume and tone pots generate scratching. All of this points towards a problem with DC bias, which is visible on the output lines.
I downloaded the service manual, and have taken measurements at the points specified. A few things stand out:
I've tested the resistances across the transistor Q401 and both show around 800 ohms, which suggests that the transistor is not broken.
I've checked the resistances of nearby resistors R404, R405 and R406 and they are within a couple of percent of spec.
I've checked D402 and it is a 1.2V Zener diode which my component tester says is behaving correctly.
I've checked C402 and my component tester reads a capacitance of 110uF in situ.
Q401 is a 2SC1364 transistor, which the spec sheet says is NPN. The collector on Q401 is connected to the base of four output transistors (2SC1636, also NPN), and to a 1uF capacitor to the negative output terminals. I've checked all four output transistors and they're behaving correctly too (same as my test for Q401).
What could be causing this incorrect voltage? There are no obvious visible faults on the PCBs. The tools I have at my disposal are a multimeter; an SMD tester which can check capacitances, resistances, and the behaviour of diodes including Zener diodes; and a basic digital oscilloscope. I have the full service manual with circuit diagrams and PCB layouts. Any ideas?
I downloaded the service manual, and have taken measurements at the points specified. A few things stand out:
- The voltages on the main opamp ICs are out of spec, probably due to power supply issues
- the +31V and -31V lines are running over spec, at 32.7V. This is probably because the preamp is rated at 220V and the power here is 230-240V - but it should be able to cope with that.
- One particular voltage is way off spec - the collector on Q401 is reading -21.3V when it should be -24.5V. My best guess is that this is the issue that's causing the preamp to misbehave.
I've tested the resistances across the transistor Q401 and both show around 800 ohms, which suggests that the transistor is not broken.
I've checked the resistances of nearby resistors R404, R405 and R406 and they are within a couple of percent of spec.
I've checked D402 and it is a 1.2V Zener diode which my component tester says is behaving correctly.
I've checked C402 and my component tester reads a capacitance of 110uF in situ.
Q401 is a 2SC1364 transistor, which the spec sheet says is NPN. The collector on Q401 is connected to the base of four output transistors (2SC1636, also NPN), and to a 1uF capacitor to the negative output terminals. I've checked all four output transistors and they're behaving correctly too (same as my test for Q401).
What could be causing this incorrect voltage? There are no obvious visible faults on the PCBs. The tools I have at my disposal are a multimeter; an SMD tester which can check capacitances, resistances, and the behaviour of diodes including Zener diodes; and a basic digital oscilloscope. I have the full service manual with circuit diagrams and PCB layouts. Any ideas?
Those voltages are all derived from non regulated rails and so will vary a lot depending on the supply voltage. I think your issues lie elsewhere tbh. The -/+22 volt lines are the only regulated rails on that diagram.
Thanks. Where would be the next place to look? The noise appears with zero volume, which suggests that the problem must be after the volume control. If I turn the treble control down, the hiss decreases - which suggests that the problem must be before the treble control.
That takes us to this section of the board - where, again, the voltages are a little out of spec. I've checked the other components as far as I can, and the numbers on the resistors and electrolytic capacitors are where they should be. If I've read the circuit diagram correctly, with the loudness and low filter switches both set to OFF, that should take all the components on that board out of the circuit, so there's no need to look there. Do I need to go around checking the film caps, or is there some better path to investigate?
That takes us to this section of the board - where, again, the voltages are a little out of spec. I've checked the other components as far as I can, and the numbers on the resistors and electrolytic capacitors are where they should be. If I've read the circuit diagram correctly, with the loudness and low filter switches both set to OFF, that should take all the components on that board out of the circuit, so there's no need to look there. Do I need to go around checking the film caps, or is there some better path to investigate?
Having looked at the rest of the circuit and the specifications I think the main problem is simply that this is a design from many many years ago and our expectations have changed.
This happens because the unit is designed for the kind of signal sources of the time, cassette decks and tuners with output levels far lower than modern digital sources like CD. Look at the specifications. 150mv is the quoted line sensitivity.
That problem can be fixed easily by adding input attenuation in the form of a simple divider.
The hiss is going to be a combination of the older opamps (CX550, which have a unique pinout) and the basic circuit design which is all high impedance. That generates white noise. The tone control stage is wrapped around that final opamp as well.
We would do it all differently today.
This happens because the unit is designed for the kind of signal sources of the time, cassette decks and tuners with output levels far lower than modern digital sources like CD. Look at the specifications. 150mv is the quoted line sensitivity.
That problem can be fixed easily by adding input attenuation in the form of a simple divider.
The hiss is going to be a combination of the older opamps (CX550, which have a unique pinout) and the basic circuit design which is all high impedance. That generates white noise. The tone control stage is wrapped around that final opamp as well.
We would do it all differently today.
I've adjusted the input level (trivial, it's coming from a DAC with a volume control) but the hiss is very audible (input or no input) when put through the accompanying TAN-15F power amp, to the point where I'm pretty sure it wouldn't have left the factory like that. It'd also be a surprise to me given that the TAN-15F is incredibly silent when playing directly to it from the DAC. I don't know about your comments about the design, but to my untrained and inexperienced mind I'm perplexed as to why the volume control is before the tone amp. After all, if the volume control came last, I probably wouldn't even have noticed this issue.
One thing that's puzzling me is why the hiss isn't audible when it's driving headphones, only through the line output when connected to a power amplifier. I guess the different load changes the behaviour, and that a pair of headphones as a passive load will tolerate the fact that the output is resting around 10mV. But could that be a clue as to how to address the issue?
One thing that's puzzling me is why the hiss isn't audible when it's driving headphones, only through the line output when connected to a power amplifier. I guess the different load changes the behaviour, and that a pair of headphones as a passive load will tolerate the fact that the output is resting around 10mV. But could that be a clue as to how to address the issue?
The volume control is before the opamp so that a large input signal will not clip the op amp.
As Mooly said the tone controls are in the feedback loop of the opamp. To reduce the noise you could reduce the gain of the tone control stage. This would involve replacing several capacitors and resistors on the bass and treble controls. Before simulators I used to calculate by hand the values for tone circuits, not too difficult.
If you dont use tone controls you could remove them and just use a resistor divider network around the opamp to set the gain lower. I dont know the opamp and I dont know if it is stable at lower gains.
You could try a voltage divider after the opamp between the opamp and the output sockets ie 10K and 1 to 5K.
There are not too many modern opamps with low noise that can handle +/- 22volt rails.
As Mooly said the tone controls are in the feedback loop of the opamp. To reduce the noise you could reduce the gain of the tone control stage. This would involve replacing several capacitors and resistors on the bass and treble controls. Before simulators I used to calculate by hand the values for tone circuits, not too difficult.
If you dont use tone controls you could remove them and just use a resistor divider network around the opamp to set the gain lower. I dont know the opamp and I dont know if it is stable at lower gains.
You could try a voltage divider after the opamp between the opamp and the output sockets ie 10K and 1 to 5K.
There are not too many modern opamps with low noise that can handle +/- 22volt rails.
Thanks for the ideas - I was thinking about how to bypass the tone controls partly to test whether they're introducing any issues, as well as to provide a cleaner signal path.
There isn't another gain stage after IC201. There are muting transistors (controlled by the Q401 transistor with the wonky voltage) and some basic resistors and capacitors around them, but no other opamps.
There isn't another gain stage after IC201. There are muting transistors (controlled by the Q401 transistor with the wonky voltage) and some basic resistors and capacitors around them, but no other opamps.
That's OK doing that but you might compromise the DAC's performance. It all depends how the volume control operates and you may lose resolution. It's usually best to run the DAC at full output.
The TAN 15 adds another x18 numerically (or 25db if you prefer) of voltage gain and so that is why the hiss is suddenly noticeable via the power amp but not with headphones. If you have very efficient speakers then that will also make it seem worse.
Hiss is caused by thermal noise
That is why I say we would do things differently today.
You can be 99% sure that if the hiss is the same for both channels then there is not going to be a fault as such, it is I'm afraid just the way it is.
A keen and experienced constructor could replace the opamp with a more modern type (it would require a small custom made PCB to do it neatly) and you could also look at rebuilding the tone stage by reducing the overall impedance by a factor of 10 (so decreasing resistors and pot values and increasing the caps) but its a major exercise.
What multisync suggests would work, an attenuator at the output would reduce noise as well as the signal. From a modern design perspective it breaks the rules of amplifying and then attenuating... but it would cut the hiss down.
Thanks again. I think I will study and probe the circuits more, so I can understand more about exactly where the noise is coming from. It would be trivial to just attenuate the signal at the end of the path, but it would be nicer to cut it out earlier in the process. From some initial investigation it looks like the noise is also there before the opamps, which suggests that something else is not quite right. But then again I am a beginner at this, so I know my hunches will often turn out to be wrong and lead me to some learning points 😉
If you link this point to ground (R206) on both channels you will hear just the noise contribution of that stage. You won't get it any lower than that without fairly major alterations.
I wasn't sure if the test lead I was using to link the point to ground was working correctly, but then I realised that it's easy enough to test one channel at a time just by using the balance control. That confirms that the output stage has noise of roughly 6mV peak-to-peak, 1mV rms. The service manual says residual noise should be under 40uV, which means either I'm measuring it wrong or something's not right.
A couple of other things seem weird: there's a considerable ground hum if I turn the volume up with nothing connected; and connecting my DAC (which has almost inaudible noise) introduces an odd buzz. My guess is that it shouldn't do those things, but I have no idea if they're connected to the output noise issue. The only thing that all the parts have in common is the ground connection...
A couple of other things seem weird: there's a considerable ground hum if I turn the volume up with nothing connected; and connecting my DAC (which has almost inaudible noise) introduces an odd buzz. My guess is that it shouldn't do those things, but I have no idea if they're connected to the output noise issue. The only thing that all the parts have in common is the ground connection...
Easy one first. Always test for hum with the inputs shorted (a shorting plug is easy to make). Inputs that are floating can and do pick up all sorts of noise and hash. The pre amp and your power amp should initially be run with nothing else apart from speakers connected to prove that all is OK. Ground loops can be created via the RCA leads back to other equipment and then via mains grounding of all the equipment.
How are you measuring the noise? Scope?
How are you measuring the noise? Scope?
Yes, to eliminate other issues I've tried with the inputs shorted too, and it's still showing the 6mV figure - whether with headphones (low impedance output), connected to the power amp with speakers attached (high impedance), or with nothing connected to the output.
I have a cheap digital scope I'm using to measure the noise and that's where the numbers came from.
I still can't help thinking that the voltages around the EQ section being out of spec (see post #3) are either the cause or a symptom of whatever's wrong here. The +/-21.2v rails (which, as you said earlier, are regulated) are showing +21.9 and -22.3; the 21.2v rails going to the phono stage are showing +21.2 and -21.6. The EQ stage and phono stage are both connected to the voltage regulation circuit by separate 100ohm resistors, which I guess is what gives them the potential to be different. I guess that means the voltage regulation circuit is working correctly (enough), and something in the EQ stage is not working the way it should. Half board is covered by a metal shield which, irritatingly, is soldered on to the board at the edges, so I can't see the components without removing it - and that solder looks shinier than the other parts, which makes me think someone might have been there before me to try to figure out what's going on. I think I'll have to take it off to take a closer look...
I have a cheap digital scope I'm using to measure the noise and that's where the numbers came from.
I still can't help thinking that the voltages around the EQ section being out of spec (see post #3) are either the cause or a symptom of whatever's wrong here. The +/-21.2v rails (which, as you said earlier, are regulated) are showing +21.9 and -22.3; the 21.2v rails going to the phono stage are showing +21.2 and -21.6. The EQ stage and phono stage are both connected to the voltage regulation circuit by separate 100ohm resistors, which I guess is what gives them the potential to be different. I guess that means the voltage regulation circuit is working correctly (enough), and something in the EQ stage is not working the way it should. Half board is covered by a metal shield which, irritatingly, is soldered on to the board at the edges, so I can't see the components without removing it - and that solder looks shinier than the other parts, which makes me think someone might have been there before me to try to figure out what's going on. I think I'll have to take it off to take a closer look...
The exact voltages are not critical and the opamp would work just fine with lower and/or unbalanced voltages such as -12 and +25 or +20 and -10 and so on. The supply voltages are chosen to determine the maximum output voltage swing available.
Its really difficult to judge this without actually hearing it for real, but at face value 6mv noise is extremely high.
Just for interest can you try to replicate the 6mv result again and this time (and I know this sounds crazy) connect the scope probe input to the scope probe ground. That might show if there is a measurement anomaly. You should of course se 0.00 volts noise doing that.
And remember power supply problems would not really cause hiss.
If the amp might have been worked on is it possible someone might have tried modifying something such as trying to get more gain.
There might be things we can try in an effort to prove where the problem is but you are going to have to be prepared to experiment. I'm thinking as a test we could tag in a different value feedback network to retain the same gain but lower the noise. It might just prove something.
Its really difficult to judge this without actually hearing it for real, but at face value 6mv noise is extremely high.
Just for interest can you try to replicate the 6mv result again and this time (and I know this sounds crazy) connect the scope probe input to the scope probe ground. That might show if there is a measurement anomaly. You should of course se 0.00 volts noise doing that.
And remember power supply problems would not really cause hiss.
If the amp might have been worked on is it possible someone might have tried modifying something such as trying to get more gain.
There might be things we can try in an effort to prove where the problem is but you are going to have to be prepared to experiment. I'm thinking as a test we could tag in a different value feedback network to retain the same gain but lower the noise. It might just prove something.
Yes, 6mV sounds high too when it's put through a power amp - it's far more than "put your ear near the speaker and you can hear it", it's "I can hear there are speakers turned on in this room" territory.
With the scope input connected to the scope ground, I see zero - both when the probe is touching the output and when it's only connected to itself. There's maybe one pixel of fluctuation at the highest zoom level. So I think we can trust the 6mV reading.
I don't know the history of this amp, so I'm prepared for things not being quite right under that metal cover when I get the chance to take a look tomorrow. It looks like the pins on the bass control may have been resoldered as they look shiny, whereas most of the others look dull. I'm quite open to experimenting if there's nothing obviously wrong with any part of it, because I've got to learn somehow!
With the scope input connected to the scope ground, I see zero - both when the probe is touching the output and when it's only connected to itself. There's maybe one pixel of fluctuation at the highest zoom level. So I think we can trust the 6mV reading.
I don't know the history of this amp, so I'm prepared for things not being quite right under that metal cover when I get the chance to take a look tomorrow. It looks like the pins on the bass control may have been resoldered as they look shiny, whereas most of the others look dull. I'm quite open to experimenting if there's nothing obviously wrong with any part of it, because I've got to learn somehow!
OK, interesting. Lets work on the theory that either it has been modified or something has failed and that would fail the same way for both channels.
As you have a scope we should really look at the rails using that, and while I don't think for a moment there is a problem there we should still check. We know the voltages are good, lets make sure the rails are clean.
Look at the circuit. You need to look at the end of R311 and R312 that feed the chip. Use AC coupling on the scope and look at the rail. It should be clean, nothing more than a millivolt or two of noise and that noise depends greatly where you ground the scope.
The bass and treble controls are part of the feedback loop and so are critical to all of this. Do they actually work as expected?
Look at the circuit. C205 is the point where the signal is fed back into the inverting input of the opamp. You need to check carefully that all the parts in that area are the correct value (not been changed by someone). Check the pots are the correct value. Check that with them centred you read around 50k from the centre wiper pin to each end. You can isolate one end of R216 and R217 to isolate the pot and get a true result. The treble pot should be isolated by the caps and you should be able to read that one in circuit. Check R216 and R217 are correct values. These set the gain (with the pot resistance).
Check R214 and R215 are correct and that the caps C211/212/213/214 are correct values.
C205 seems an outsider for a fault like this but as it is in the feedback loop it needs checking. You can just bridge it with another.
Remember whatever is going on is affecting both channels and so someone altering something is a possibility.
As you have a scope we should really look at the rails using that, and while I don't think for a moment there is a problem there we should still check. We know the voltages are good, lets make sure the rails are clean.
Look at the circuit. You need to look at the end of R311 and R312 that feed the chip. Use AC coupling on the scope and look at the rail. It should be clean, nothing more than a millivolt or two of noise and that noise depends greatly where you ground the scope.
The bass and treble controls are part of the feedback loop and so are critical to all of this. Do they actually work as expected?
Look at the circuit. C205 is the point where the signal is fed back into the inverting input of the opamp. You need to check carefully that all the parts in that area are the correct value (not been changed by someone). Check the pots are the correct value. Check that with them centred you read around 50k from the centre wiper pin to each end. You can isolate one end of R216 and R217 to isolate the pot and get a true result. The treble pot should be isolated by the caps and you should be able to read that one in circuit. Check R216 and R217 are correct values. These set the gain (with the pot resistance).
Check R214 and R215 are correct and that the caps C211/212/213/214 are correct values.
C205 seems an outsider for a fault like this but as it is in the feedback loop it needs checking. You can just bridge it with another.
Remember whatever is going on is affecting both channels and so someone altering something is a possibility.
If you are really interested you could take a frequency sweep (20 to 20kHz) in the form of an MP3 file or burn a test CDR and look at the output of the preamp on the scope. The output amplitude should be flat across the band with the controls centred.
If you want an MP3 file I can post one here.
If you want an MP3 file I can post one here.
Thanks for the detailed instructions! They were very easy to follow.
I measured the voltage rails, and they both have around 2mV of noise measuring back to the main ground connection where everything meets.
I took the cover off and got to the tone amp board, and aside from the big film cap between the bottom two pots having a bit of a bumpy shape, everything looks normal and original:
The pots work more-or-less as expected - they're not linear variable resistors, they're stepped, and whilst the resistances go from zero at one end of the sweep to nearly 100k at the other measured from both ends to the centre, the split in the middle is more like 12/77 than 50/50.
The resistors R216 and R217 measure correctly.
R214 and R215 also measure correctly; the caps C209-C213 all have the correct labels but my probe (a Uni-T UT116C) won't show a reading for them in situ.
My probe did measure C205 and it shows a correct reading of roughly 10uF. I also checked the other electrolytics around the tone amp for good measure, and they are all within spec.
I took some more scope readings, and I noticed that the noise had something of a pattern to it:
and when the treble controls are turned all the way down, it stabilises into this (note the different time scale):
All of which shows that this is a complex feedback loop that I don't yet understand the details of, and leaves me wishing there were a tone bypass switch that would isolate this stuff to make diagnosing it a bit easier. 🙂
I also noticed just how sensitive this circuit is - when adjusting the EQ knobs, the capacitance of my body sent the output up to 200mV.
I measured the voltage rails, and they both have around 2mV of noise measuring back to the main ground connection where everything meets.
I took the cover off and got to the tone amp board, and aside from the big film cap between the bottom two pots having a bit of a bumpy shape, everything looks normal and original:
The pots work more-or-less as expected - they're not linear variable resistors, they're stepped, and whilst the resistances go from zero at one end of the sweep to nearly 100k at the other measured from both ends to the centre, the split in the middle is more like 12/77 than 50/50.
The resistors R216 and R217 measure correctly.
R214 and R215 also measure correctly; the caps C209-C213 all have the correct labels but my probe (a Uni-T UT116C) won't show a reading for them in situ.
My probe did measure C205 and it shows a correct reading of roughly 10uF. I also checked the other electrolytics around the tone amp for good measure, and they are all within spec.
I took some more scope readings, and I noticed that the noise had something of a pattern to it:
and when the treble controls are turned all the way down, it stabilises into this (note the different time scale):
All of which shows that this is a complex feedback loop that I don't yet understand the details of, and leaves me wishing there were a tone bypass switch that would isolate this stuff to make diagnosing it a bit easier. 🙂
I also noticed just how sensitive this circuit is - when adjusting the EQ knobs, the capacitance of my body sent the output up to 200mV.
My mistake on the pots... the 12/77 split will be because they are a 'C' taper which is a logarithmic pot. So I think we should say they are OK.
This is strange then. The pattern to the noise does not look like normal 'hiss' which just doesn't look like that. This image from the web shows exactly how hiss looks on a scope... but it needs a wideband scope to show it like that.
Random question. What speakers are you using. Are they extremely high sensitivity (db/watt) like a horn speaker that would highlight noise?
We can play around with the feedback network as a test if you have a selection of resistors available. All we do is use two resistors to set the gain and couple them to C205. It may prove something.
Also do have a means to play an MP3 music file into the preamp? I 'm going to say you should look at the output on a test signal and make sure it looks OK and with a flat response.
This is strange then. The pattern to the noise does not look like normal 'hiss' which just doesn't look like that. This image from the web shows exactly how hiss looks on a scope... but it needs a wideband scope to show it like that.
Random question. What speakers are you using. Are they extremely high sensitivity (db/watt) like a horn speaker that would highlight noise?
We can play around with the feedback network as a test if you have a selection of resistors available. All we do is use two resistors to set the gain and couple them to C205. It may prove something.
Also do have a means to play an MP3 music file into the preamp? I 'm going to say you should look at the output on a test signal and make sure it looks OK and with a flat response.