Few questions about a filter/crossover

[sgrossklass]

Ha! My now broken X-Fi Xtreme Music had the JRC4556:

View attachment 789100

I guess it was one of the earlier ones without a heatsink on the X-Fi chip?

Those had a tendency to die. My own little theory is that since it's a BGA chip, it may be suffering from cracked solder balls due to thermal stress. Have you ever tried reflowing a BGA before, heatgun and Al foil style?

Yeah, the 4556A is a classic, and Creative used them for many years.

I would love to, but pots are F expensive here, and as I said limited functionality is fine. But given there will be only 2 dual gang pot (forget about finding a 4 gang one) probably will be variable...

OK, if that's a problem, you might consider using socketed resistors. People make single lead sockets from IC sockets.

And before I forget, I think a gain/sensitivity switch will be good because of all the different sources with different power, some too weak, some too strong.

Hmm, this would be easiest at the U3/flat amp level (switch in another resistor in the ground leg in parallel), but that obviously will affect noise level, too. Well, I guess we better make the circuit not too noisy at highest gain then.

If our power amp gives us 26 dB, required preamp gain should be anywhere between 0 and 20 dB max. I'd say include a ~6 dB (+2) normal and ~16 dB (+6.5) high gain, and you should generally be set.
Rf = 4k7
Rg = 4k7 (|| 1k for high gain)
That would give us about 6 µV of output noise in high gain @ +16.5 dB and 2.4 µV in low gain @ +6 dB with a 5532. I have yet to check (simulate) what the crossover will add, but so far things look promising. If we get to 8-12 µV, that would still make for "HiFi-worthy" output noise levels of ~160-240 µV.

Nah, the idea is either a simple opamp or not even that, and as I will need the baxandall tone, after this will be enough. It's just for normal headphones, not high impedance ones. And remember that the NE5532 can drive big loads...supposedly...
I made this thing years ago and drives normal headphones strongly:

View attachment 789114

I still have a little BTech NE5532-based headphone amp floating around somewhere, too. For 100, 150, 300 ohm headphones of decent sensitivity, these are quite adequate. And think of how many cMoys have been built over the years - the OPA2134 is quite comparable in drive capability.

That said, if you have the power available, a little more oomph in your headphone amp is definitely preferred. Often an opamp will be complemented with a discrete (or IC) buffer in the feedback loop.

My fears. That and any other kind of noise and distortions due to PS (or bad desing).
And something I forgot to add to the black diagram. But yeah the original idea was to use the one PS filtering after it and use a single supply for the opamps to simplify (as I don't have to create a negative voltage from the single supply notebook charger). I know it works fine and the C200 has that. As it says in the other thread, the C200 has the switching PS +15v with it's filtering, goes to the board feeding the 7379 directly and the with the opamps, there's an SMD inducor or similar before a 7809 that feeds them in single ps mode.

Sounds like a decent enough scheme. Since you have 19 V available, you could easily step it up to a 78(M)12... maybe even a 78(M)15 if you feel lucky. By the time the SMPS has dropped to ~17 V it's likely to be seriously unhappy anyway.

You will have to plan out the grounding very carefully, but that's a given in any case.

Speaking of case, if you don't have a solid metal enclosure you'll have to worry about shielding as well. Having at least 2-layer circuit boards with a dedicated ground plane will go a long way already, but I don't know whether you have that sort of luxury available.

But on the other hand all the shcematics have to be changed, a lot...So if there's a good proper way to turn a single PS into a dual ps (the 7379 will be fed from 16v to 18v).

The Art of Electronics in its current edition has a nice overview of rail splitting techniques... from R/RC (like your headphone amp), R/RC buffered by a suitable opamp (not afraid of driving capacitive loads with no resistor in series), dedicated rail splitter IC (e.g. TLE2426) and I forgot what else it was. But "rail splitting" is what you want to look for.

Another commonly-used option would be using a buck regulator as an inverter to generate a negative supply. The only caveat of this otherwise unproblematic circuit (well, aside from the usual switch-mode fun factor - filtering, common-mode chokes etc.) is that the reg has to be able to take the full input-output differential, i.e if you are generating -12 V from +19 V it needs to sustain 31 V.

Going single supply isn't that hard though.
1. You need a clean +Vs/2 reference.
2. Inverting opamp circuits: (+in) gets referenced to +Vs/2 instead of "real" ground.
3. Noninverting opamp circuits:
a) (+in) input bias resistor referenced to +Vs/2 instead of "real" ground.
b) Feedback goes to ground via a suitable large coupling capacitor. (You could use +Vs/2 but it often can't take much of a load and/or isn't as low in impedance as you'd like.)
4. Since the circuitry is now operating around +Vs/2, ground-referenced inputs and outputs need coupling capacitors to reflect that.
5. Opamp supplies: V+ = +Vs, V- = GND.

Why? Aren't these kinds of "balanced" just as normal inputs, you know, with 2 poles, signal and ground?

Yes, but when switching unbalanced signals you are often going 1-pole for simplicity, leaving the ground always connected.

No idea.

Ask the multimeter. The ohms range should tell you whether there is a direct connection or resistor. If neither, the capacitance range will show you more.

Yes, and I expect nothing better! But it's the only PS I can afford for the moment so I have to make it work. Through heavy filtering I guess...a thing that needs coils...things hard to find here! So I have to use simple stuff.
Of course I could obtain a normal transformer, but they are absurdly expensive and the quality...

Hmm. Having a collection of misc. electronic junk to salvage parts from may prove handy here.

Your circuit confuses me too much! I'm really not sure what I'm seeing! Haha!
I mean it's very similar to the Edifier C200 near the input, but apparently balanced. Remember I don't need that.

What exactly is confusing you? Note the use of "American" R and L symbols, these just happen to be the default in LTspice. There are prettier ways of drawing a common-mode choke, but anyway.

You'll be quite glad you have a (decently) balanced input the moment you implement the 5V regulator to supply your BT receiver. Guess what the most common problem among DIY mobile boombox projects is? Ground loops because people are trying to supply everything from a single battery.

Yes, this is also suitable for a full-fledged balanced input. I am lazy and figured I'd do it properly once, then I can reuse it later.

Hey Sgrossklass, I made 4 schematics of most of the circuit.
I tried to make all the changes you proposed in earlier comments, but also did the baxandall in parallel version that you don't like, just to see if it's valid anyway.
Most sections were actually made separately for different reasons at different times, so I modified what I could register. The opamps models for example are mostly 4558 because that's what was at the original, you tell me which to change (I want to simply use 5532 and 4558).

U3 is definitely a candidate for 5532.

Anywhere a 4558 is driving an output in the crossover, I would add another 4558 unity-gain buffer.

5532 is not an ideal candidate for a balanced receiver since it has some symmetry issues with input capacitance that reduce high-frequency CMRR. That said, I wouldn't necessarily use a 4558 either.

Using scheme "C" now for discussion:

With R1, R4 at 820 ohms, your volume control range is now limited to roughly 26 dB. This is not likely to be adequate, I'd want at least 40 dB - try 150 ohms or so.

I'd like you to combine the balanced input and the common-mode choke, as I discussed earlier. I also think C32/33 at 1n are going to restrict U1 GBW too much.
I suggest splitting R56-59. First 1k-2k2, then your 1n between both legs, then 10k (or 9k1 if you have).
C36 and C109 etc. should be the same value.
C34 and C35 should have a counterpart at L/R IN.

I thought the random TDA7379 pin labels were hilarious, that one still needs some editing, I guess.

The 7805 for USB supply is going to be very inefficient coming from +19V. If all you need is about 200 mA, it's not too bad, but beyond that it gets toasty quite fast. I would suggest about 2x 10 ohm / 5 W in series with the input to keep regulator dissipation at an acceptable level (still about 2 W at 500 mA, so some sort of heatsink required).

 

[Fergutor]

Ops! Forgot the phase switch.


This is from the JBL Array 1500.

Will this work or do I need the other type, like the BU2 I posted earlier next to this one?

 

[Fergutor]

I guess it was one of the earlier ones without a heatsink on the X-Fi chip?

No, but, after it died, put it in a bag and let it around, the heatsink fell off. But I don't think it suffered any shock...

Those had a tendency to die. My own little theory is that since it's a BGA chip, it may be suffering from cracked solder balls due to thermal stress. Have you ever tried reflowing a BGA before, heatgun and Al foil style?

I had a 8800GTS 640MB that I bought just as it came out. This cards had that problem of solder cracks you mention. So I tried the oven method but it didn't work.
I will look for something to try with the X-Fi but I don't think it's a solder crack because the fail it's no sound from surround outputs (can't remember about LFE), terribly distorted sound on the left channel when using stereo speakers and hissing at the right channel (same output!) when using headphones (22ohms if I remember correctly).

OK, if that's a problem, you might consider using socketed resistors. People make single lead sockets from IC sockets.

That's an idea (in fact I already had separated IC sockets for transistors...didn't worked, leads too thick, haha). I will see.

Hmm, this would be easiest at the U3/flat amp level (switch in another resistor in the ground leg in parallel), but that obviously will affect noise level, too. Well, I guess we better make the circuit not too noisy at highest gain then.

If our power amp gives us 26 dB, required preamp gain should be anywhere between 0 and 20 dB max. I'd say include a ~6 dB (+2) normal and ~16 dB (+6.5) high gain, and you should generally be set.
Rf = 4k7
Rg = 4k7 (|| 1k for high gain)
That would give us about 6 µV of output noise in high gain @ +16.5 dB and 2.4 µV in low gain @ +6 dB with a 5532. I have yet to check (simulate) what the crossover will add, but so far things look promising. If we get to 8-12 µV, that would still make for "HiFi-worthy" output noise levels of ~160-240 µV.

Got it, but are you sure rf=rg?

That said, if you have the power available, a little more oomph in your headphone amp is definitely preferred. Often an opamp will be complemented with a discrete (or IC) buffer in the feedback loop.

Too many opamps! I think I will start bringing out the TL074's...
(not for that of course hehe)

Having at least 2-layer circuit boards with a dedicated ground plane will go a long way already, but I don't know whether you have that sort of luxury available.

I have yet to see prices of boards. Here, you know what I'm going to say: are too expensive. The logic always seems to be: "we don't have enough clients: let's charge more!"

Usually I make my own PCBs (1 sided only) but when they are too big like this one, the iron method is a headache.

The Art of Electronics in its current edition has a nice overview of rail splitting techniques... from R/RC (like your headphone amp), R/RC buffered by a suitable opamp (not afraid of driving capacitive loads with no resistor in series), dedicated rail splitter IC (e.g. TLE2426) and I forgot what else it was. But "rail splitting" is what you want to look for.

Ok I will take a look at that.
That TLE2426, what's the difference between that and the usual voltage divider?

Going single supply isn't that hard though.
1. You need a clean +Vs/2 reference.
2. Inverting opamp circuits: (+in) gets referenced to +Vs/2 instead of "real" ground.
3. Noninverting opamp circuits:
a) (+in) input bias resistor referenced to +Vs/2 instead of "real" ground.
b) Feedback goes to ground via a suitable large coupling capacitor. (You could use +Vs/2 but it often can't take much of a load and/or isn't as low in impedance as you'd like.)
4. Since the circuitry is now operating around +Vs/2, ground-referenced inputs and outputs need coupling capacitors to reflect that.
5. Opamp supplies: V+ = +Vs, V- = GND.

While mostly I understand, I am still not sure how to apply to all the different cases in the schematic.

Ask the multimeter. The ohms range should tell you whether there is a direct connection or resistor. If neither, the capacitance range will show you more.

No resistance, no connection and my multimeter doesn't measure capacitance.

U3 is definitely a candidate for 5532.

Ok, done.

Anywhere a 4558 is driving an output in the crossover, I would add another 4558 unity-gain buffer.

I will TL074 the F out of this circuit!

If I use 5532 would I need another for that same reason?
What if instead of the 4558s I use the 5532 in all the cases? remember the sections of the schematic are from other projects and I dind't changed the opamps.

5532 is not an ideal candidate for a balanced receiver since it has some symmetry issues with input capacitance that reduce high-frequency CMRR. That said, I wouldn't necessarily use a 4558 either.

What do you suggest? LM833? I have one and I think I can find more of those.

With R1, R4 at 820 ohms, your volume control range is now limited to roughly 26 dB. This is not likely to be adequate, I'd want at least 40 dB - try 150 ohms or so.

I was trying to follow your advice: it was 820r in the first attached shcematic, you asked me why "now" it was 75r in the second, so I "reverted" here.
But done, 150r now.
What pot value do you suggest? Log or linear?

I'd like you to combine the balanced input and the common-mode choke, as I discussed earlier.

Ok, I have two problems with this.
1-Still not sure how do do it, still not sure if the one I drew, copy from the car amp is proper...I still don't undertsand yours (should I ignore the "Bal 2 SE section completely? Why is so different form the car amp?) to implement in my shcematics.
2-Notice the HIGH LEVEL inputs, they are different in each type of input. Again copied from respective types of inputs (no idea if they are right, why they are different, etc...). I guess with a combination I should use the one in the "C" and "D" schematics...?

I also think C32/33 at 1n are going to restrict U1 GBW too much.
I suggest splitting R56-59. First 1k-2k2, then your 1n between both legs, then 10k (or 9k1 if you have).

If the problem is the value, what value do you suggest?
Or is it also the topology?
Ok, now looking at it, I think it was my mistake when I copy it, as the ones in the car amps range from 10 to 100 pF. Guess I simply forgot to change the value as "1nF" is the default.

C36 and C109 etc. should be the same value.

Same thing here as C36-C37 are 22p in all schematics.
C109 and C111...I guess I used 100pF because that's the value in the C200
Ok, so what values should I use in C36, C39 C109 and C111?

C34 and C35 should have a counterpart at L/R IN.

They are like this in the original car amp schematics. So I should add 1 from signal to input ground?

I thought the random TDA7379 pin labels were hilarious, that one still needs some editing, I guess.

yeah..In the other thread I edited them correctly. Actually I edited this ones too initially, but I did it so many times that I forgot to do it in these final ones...
The problem is the damn Proteus that doesn't have things for audio or packages to edit, like the multiwatt15.

The 7805 for USB supply is going to be very inefficient coming from +19V. If all you need is about 200 mA, it's not too bad, but beyond that it gets toasty quite fast. I would suggest about 2x 10 ohm / 5 W in series with the input to keep regulator dissipation at an acceptable level (still about 2 W at 500 mA, so some sort of heatsink required).

Ok, I do the changes or just remove the thing completely and use an USB charger (well, not me, the ones who want to use bluetooth, I don't care about it).
Or what if not the 7805?