The Analog Gods Hate Me

[Mr.Roy]

I’ve attached a schematic with which I’ve had excellent assistance from Mooly on diyaudio.
He concluded that I am not breaking any rules and could not see any obvious electrical issues that would prevent it from working.
But it is not working. And I’ve triple-checked my wiring, checked continuity and discontinuity (what should be connected is and what should not be is not).

The aim of the device is mainly to produce three mono outputs from a stereo source:
1. a center channel (mono) output
2. an output consisting of what is unique to either of the stereo channel inputs
and
3. an output consisting of what is common to the stereo channel inputs

In addition, there is a switchable low-pass circuit that can be employed.

Function 1 is simple enough.
Function 2 should contain the sum of one of the stereo channels and the inverted other channel.
The input to this portion is regulated by a linear 2-gang potentiometer acting as a balance allowing the user to pan between the left and right.
For function 3 the outputs of 1 and 2 are combined in a differential circuit (much like done for function 2), essentially producing the opposite of function 2.
The (likely) source unit output is stated as 0.5 volts at 2.2K ohms.

What happens is:
1. The volume seems to be greatly reduced.
2. Neither function 2 nor 3 appears different from the input at all; it’s as if there was no circuitry in place.
3. The low-pass circuit does appear to have some effect.
4. SW2 is a “Break/Make” rotary switch. When I switch between any two positions slowly, thus not making the next connection there is a large squeal. Could someone explain? I didn’t think that an open connection would drive anything nuts.

The schematic and LF353P spec sheet are attached.

 

[rjenkinsgb]

I can see a few of obvious problems:

You are using sequential high value pots at the input, with no buffering, then virtually shorting the two inputs with low value resistors for the centre channel.

Good design keeps the signal level high, to minimise the effect of component noise.

You could try making R4, R5 & R6 1M, but buffering the inputs and having the buffer gain adjustable by pots in series with the feedback resistors would be far better.


Some of your coupling capacitors are rather too low value for full range audio. eg. C5 feeding R14 - that's acting as a high pass. cutting bass. C4 is even worse; if the TDA2050 module has 10K input, the low roll-off will be around 600Hz, off the top of my head? And affected by the line-out load, making it higher still.
C6 would again roll off at somewhere around 100Hz.

As you obviously want low frequencies, having the low-pass filter option, the coupling caps should be far larger.


The differential amp is grossly unbalanced; the reference and feedback resistors (R11 & R12) should be the same value in that circuit. Make R11 500K.

A common way of balancing a diff amp is to use fixed resistors at the input, but one a few percent lower than the other and add a preset with double that difference in series with the lower value.

eg. if the input resistors were 10K, use 10K for one and 9.1K or 9.53K plus a 2K or 1K preset, so the second input can be adjusted slightly higher or lower than 10K to allow for tolerances.


Tip: When feeding capacitor coupled signals to a switch, add eg. 1M resistors to the signal bias voltage (0V for this) - that will minimise switching clicks or thumps.

ps. You would also be better off using TL072 opamps, the have lower offset and distortion. The main advantage of the LF35x is input load, which is not important for this application.

 

[Mr.Roy]

You are using sequential high value pots at the input, with no buffering, then virtually shorting the two inputs with low value resistors for the centre channel.

Good design keeps the signal level high, to minimise the effect of component noise.

You could try making R4, R5 & R6 1M, but buffering the inputs and having the buffer gain adjustable by pots in series with the feedback resistors would be far better.

R1 and R2 probably should be positioned after R3. Their purpose is to undo any difference between the 2‑gang R3 resistances. (They’re there solely for this purpose. As such, they are first set at full CW. Then one is turned CCW until there is no difference. Of course this can’t be done until function 2 works and, of course, with a mono input signal.

• Should I lose the ground on pin 3 and turn R1 and R2 into rheostats?
• One of my complaints is that the volume is extremely low. Would not increasing resistors R4 and R5 make that even worse?
• What would I buffer? How would I buffer?
• I think I understand the wisdom of putting the pots in series with the Rf resistors.
• Should R6 replace R8?
• How would I implement the L/R balance for IC1b using R7? I want to be able to vary how much of each channel goes into the inverting and non‑inverting IC1b inputs.

Some of your coupling capacitors are rather too low value for full range audio. eg. C5 feeding R14 - that's acting as a high pass. cutting bass. C4 is even worse; if the TDA2050 module has 10K input, the low roll-off will be around 600Hz, off the top of my head? And affected by the line-out load, making it higher still.

C6 would again roll off at somewhere around 100Hz.

What would suggest for these values?


As you obviously want low frequencies, having the low-pass filter option, the coupling caps should be far larger.

Again, what would suggest for these values?


The differential amp is grossly unbalanced; the reference and feedback resistors (R11 & R12) should be the same value in that circuit. Make R11 500K.

Understood


A common way of balancing a diff amp is to use fixed resistors at the input, but one a few percent lower than the other and add a preset with double that difference in series with the lower value, eg. if the input resistors were 10K, use 10K for one and 9.1K or 9.53K plus a 2K or 1K preset, so the second input can be adjusted slightly higher or lower than 10K to allow for tolerances.

I am redesigning that area.

Tip: When feeding capacitor coupled signals to a switch, add eg. 1M resistors to the signal bias voltage (0V for this) - that will minimise switching clicks or thumps.

Can you be more specific as pertains to the schematic? I’m a bit lost here.

I have attached the TDA2050 spec sheet. It cites a 500K resistor at pin 1 and a 22K impedance.
Understand that I am NOT using the IC. Rather, it is a packaged speaker-output module with connectors and a volume control.

And thank you for the time you've taken on this.
-Roy

 

[rjenkinsgb]

For buffering, add a pair of unity gain amps between the output of R3 and what that presently feeds.

That means later stages are being fed from a low impedance.

As long as R6 is increased as well as R4 & R5, it won't affect the signal level, as from there it goes to a high impedance input. With the input buffers, you should be OK with 10K or above for R4 & R5, the R6 100K or more.

Just change R11 to 500K, R7 is OK as it is after the buffer.

I'd just use 47uF for all the coupling caps.

Re. the switching noise; you have two "floating" caps feeding CC and D+ at the switch. When switched with audio passing though, those points could have any voltage within the signal range left on them.

Switching back to them while the instantaneous signal level is different will cause a click or thump.

Just add 1M from each to 0V, to eliminate any residual offset voltage.

The audio amp module is shown as having a volume pot included, which is likely 10K. That sets the input impedance of the module. The IC input load is not particularly relevant as that is masked by the pot.

 

[Mr.Roy]

For buffering, add a pair of unity gain amps between the output of R3 and what that presently feeds.

That's part of the plan!

I'd just use 47uF for all the coupling caps.

Are you suggesting that I replace the ceramic 0.1uF caps with hefty 47uF (possibly elec.) ones?

Re. the switching noise; you have two "floating" caps feeding CC and D+ at the switch. When switched with audio passing though, those points could have any voltage within the signal range left on them.
Switching back to them while the instantaneous signal level is different will cause a click or thump.
Just add 1M from each to 0V, to eliminate any residual offset voltage.

The audio amp module is shown as having a volume pot included, which is likely 10K. That sets the input impedance of the module. The IC input load is not particularly relevant as that is masked by the pot.

It's actually a 50K pot and to my surprise is LINEAR!

 

[rjenkinsgb]

Are you suggesting that I replace the ceramic 0.1uF caps with hefty 47uF (possibly elec.) ones?

C1 - 2 - 3 - 4 - 5 - 6 are the coupling caps.
None of those are 0.1uF ??

A word of warning - never use ceramic caps for audio coupling; they are often piezoelectric and can cause strange effects and distortion. Use electrolytic or plastic film types.
(The typical cheap cheap fake-condenser mics from ebay & aliexpress etc. with surface mount PCBs, will pick up tapping and handling noise with the mic capsule disconnected, via the ceramic caps!)

OK on the amp module. Linear is probably just cheaper or easier to obtain.

You still need to allow for a possibly low impedance, as the same signal goes to the line output sockets; you don't want differing loads changing the LF response.

As a very rough, safe, "ballpark" mental roll-off frequency calculation, you can base it on 1uF coupling cap into a 1M Ohm load is near enough flat response down to 1Hz. (That's actually around 6x the -3db point at around 0.16 Hz).

so eg. 1uF feeding 10K would be flat down to around 100Hz.

With a switched or plug-in loads, base the load resistance on the worst case combination.

 

[Ian Rogers]

I don't know if its just me, but why ask for help, then question the help given.. Seems a bit off putting to me.

 

[rjenkinsgb]

but why ask for help, then question the help given

I think the OP is possibly confusing coupling with Decoupling? Just clarification.

 

[Mr.Roy]

Yes, I had it backwards.

"Re. the switching noise; you have two "floating" caps feeding CC and D+ at the switch.
When switched with audio passing though, those points could have any voltage within the signal range left on them.
Switching back to them while the instantaneous signal level is different will cause a click or thump.
Just add 1M from each to 0V, to eliminate any residual offset voltage."
I understand, and I can insert a 1M resistor to ground from the CC and D+ connections. But doesn't C5/R14 present a problem? While these values will change (per recommendations) R14 has to match R13 and IC2 is a differential amp. How do I handle that?

 

[rjenkinsgb]

R14 has to match R13 and IC2 is a differential amp. How do I handle that?

Not quite correct; the way you are using it, the gain is different from each source:
From IC1a, the gain is -1
From IC1b, the gain is +2

It would need a 10K series input resistor between C5 and R14 to work correctly as a unity gain differential amp.

The critical part of a differential amp is that the ratio of positive input resistor : reference resistor (R14) must equal the ratio of negative input resistor (R10) to feedback resistor (R13).

eg. R10 & R13 could both be 100K and it should work the same, though 10K is fine for this.

C5 should not actually be needed, as IC1b is referenced to ground so any DC on the output should be trivial and handled by C6.

 

[Mr.Roy]

Rjen,
There are now buffer amps. And values have been adjusted.
What values would you recommend for R10 and R11 which feed the "center channel" summing amp?
I have inserted a 1M resistor to ground from the CC and D+ connections.
But (for D-) R21 presents a problem: look at R17, R19 and R22. I cannot make R21 1M. Or should I not worry about it?
-Roy

 

[rjenkinsgb]

The input amps have a gain of 2, rather than being unity gain. They also need bias resistors (10K or 100K) from the + inputs to ground and series coupling caps from the input sockets.
With the connection as shown, they will pick up random offset voltages from whatever is connected.

You do not need the coupling caps at the buffer output.

I did suggest the buffers went after the input adjustment pots, to drive the rest of the circuit from a known impedance.
(ie. Between the input pots and the "centre channel" block, C5 & C6 etc).


As R10, 11 & 14 are still fed from pots, they should all be as high as practical to minimise loading on the pots.

D- is already biased to ground; it's only the switch feeds from caps which need the extra resistors.

 

[Mr.Roy]

"The input amps have a gain of 2, rather than being unity gain."

The input amps have a gain of 2, rather than being unity gain.

By "input amps" do you mean the new buffers? If so, I don't mind a gain there unless you believe it'll cause a problem.