Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Opamps stage design questions
New Posts  All Forums:Forum Nav:

Opamps stage design questions

post #1 of 23
Thread Starter 
Altrough I'm reasonably skilled moder of mainbords and GFX cards and all electronics stuff, I did not play into the audio field much.

So I think I better ask these who know about my ideas how to modify a existing design.

This one:



Complete scheme:


Input swichboard:


As soon as the signal hit the switching part of the amplifier, it is shorted to ground with C99 cap, a 0.1nF one. I think this is completely unnecessary blurring of the sound, as capacitor in general act to prevent voltage changes, so it has to "blur" a little the amplitude to prevent fast and rapid changes of it.
Do I get it right?

Later the signal go thru a 0.9 resistor divider, witch is probably used to put some small load (11k to ground) on the audio source. Is this value optimal for the X-Fi equiped with LM4562 opamps...?

What about keeping just ONE of these dividers and adjust it to the 0.68 as the result of two dividers (0.9 and 0.76) after themselves are in the end.
The aim is less distortion in resistors - or even using a audio grade resistors such as Vishay Audio Resistors:
diyAudio Forums - Vishay Audio Resistors

I also fear that the combination of R and C components can create a slight RC filter that in the end make the "blur" effect of the C99 stronger a little. Right?

After it pass thru the switch, here come another ground-shorting cap, a C1. Now with 0.33nF capacity.
Why?

Then come another resistor divider, this time 0.76 and directly after him a first decoupling cap, a C3 - 10uF 25V.
As far as I understand audio, the blocking caps is necessary for the filtering of the DC offset. What if my X-Fi has very low DC offset? Is not no cap better for audio that ANY cap, even quality audio one?

I think the C3 is entirely unnecessary one. I think only one decoupling cap in the whole spekers (or none) is best solution - and placed directly before the output amplifier.
Right?

And it get worse. Just after the opamp, there is another decoupling cap! A C9 - again 10uF 25V for all except CENTER and SW channels. First thing I did not like is that the capacity on the output is same as on input - should not be bigger? Maybe is the level of signal not that high still, but... it just did not feel right.
Second thing I did not like at all is the fact that we already removed the DC offset before the opamp, so, why now? Sure, a badly balanced of sucking opamp could produce some DC voltage at the output, but... why not balance it better or remove it and use quality one instead that does not need second decoupling?

I think with the LM4562 or perhaps better AD8599 I can remove these.
Right?

And right after the potentiometer we have another decoupling cap - a C13! In fact, he is in serial circuit with the C9, witch bring the ending capacity down to half... not to mention that with the huge resistance between then the impact on the signal can be high.
I hope I'm wrong on this one, but... IIRC the most clean voltage filering is a RC way. Only with the problem that it's output voltage differ with different current - so current has to be always the same and stable...

I think the designer of this speakers just put together the recommended way of the used circuits and then these double-triple decoupling caps are the result.

Suggestions?
post #2 of 23
Thread Starter 
The problem

Plaing with the amp, like tracing the signal, is very hard to do, almost extremly. It is even hard to diassemble, and can't really run in such state... that just added to the trouble...

Nevermind, how to stop the opamps from oscilating is the question. I did not have any other opamps to do replacement ATM and for like weeks, till I can recieve the new/different opamps

That is not good. Not to mention the aim was to improve the quality, so, use new opamps is the goal. Possibly AD8599 later, however it looks like the samples will took like a month (!) to get there, and I need 2 shipments - only 2 pcs per one... (likely 4 shipments, if I use AD8599 on X-Fi too...)

So the question is, how to change the circuit to stop the oscilation from happening?



Shorting to ground? Not need. Unplugging the X-Fi is enought and there is silence then
Suddently. Obviously the opamps are oscilating, why then hell I did not figure that out myself...

X-Fi output DC offset measuring:
--------------------------------------
L: -190mV DC, 10mV AC
R: -174mV DC, 10.2mV AC
RL: -210.2mV DC, 10.1mV AC
RR: -204.7mV DC, 10mV AC
CENTER: -209.5mV DC, 10mV AC
SW: -231.9mV DC, 10.2mV AC

Good? Bad? Terrible? Usable?

What does worry me is the constant output signal noise... I cant hear it, but around 5.2 to 5.6kHz signal is there... Even the X-Fi is not playing anything ATM. I used the X-Fi control panned to check on the signals, when mesuring AC voltage, then letting the card say "left, right, center..." is clearly visible on the scope

PS. come to think, you are right. No capacitor removal (and certainly not AFTER the opamps) could trigger such noisy oscilations as I experiencing. Well, what about the C117 then? It is this a part of the "to be removed" stuff from the amp too?
post #3 of 23
Thread Starter 
Hoooray! I figured what I did wrong!

It is very embarrassing, but these 10uF SMD ceramics I added to the opamps legs, well... I added them to the legs 1 and 5, not 8 and 4 ...

No wonder the amplifier did not play well...

So, when I figured this silly mistake made, I fixed it and - hoooray - it now play MUCH better. There is only one catch - it is still very noisy. There is, as soon as the amplifier is powered on (w/o connecting to my X-Fi), still noise in all channels :thumbdn:

Much less that before, but there it is. So I put all my changes back. I give back all the ceramic blocking caps, even these for the opamps feedback (witch should not be there IMHO), and no change.
So as last desperate attempt I even removed my voltage filtering caps, but quess what - no change at all. It is not these caps...

Just instead of the original C9, C10, C35, C36, C61, C62, C52, C64, C3, C4, C29, C30, C55 and C56 10uF 25V CapXon crap caps I used Elna RFS 22uF 25V audio caps. (true, C52 is 1uF and C64 is 0.1uF by original design, but all these caps are used only for the DC offset filtering, so they can be bigger... at least C64 does not limit the bass line now..)

And of course, instead of JRC4558 opamps are there now a LM4562 ones...

So, can anyone tell me, how to get rid of this little (3x hoooray!) oscilation ... ? What about the C117? I never seen that capacitor in any design... and the C7 might also trigger the oscilation, right?

I read some there:
Working with Cranky Op-Amps
...but I did not yet come to any practical solution. Help?

A good example of opamp circuit:


See? No C117 from my amplifier...
post #4 of 23
In many cases somw tweaks are bad tweaks, like improper supply decoupling. In some designs you cannot decouple from a rail to the ground, in others you cannot decouple between the rails as you did. First of all be sure to remove all such "improvements" and then try op-amp rolling in the original circuit. C117 is a weird idea, indeed but might be helpful for the NJM4558D specifically. You can try to desolder it when placing another op-amp. AD8599 is pretty safe because it's slow, just alright for audio as intended by Analog Devices. LM4562 shouldn't be cranky, either but it's sound leaves much to desire. Actually, the given design shouldn't be problematic for fast op-amps if only the traces obey the high frequency rules. As it was designed for the 4558's, you can't be quite sure about it.
post #5 of 23
Thread Starter 
majkel -
Quote:
In many cases some tweaks are bad tweaks, like improper supply decoupling.
Looks like that. But in my case, I just did the decoupling with ceramics wrongly, decoupling the output and input to ground

Quote:
In some designs you cannot decouple from a rail to the ground, in others you cannot decouple between the rails as you did.
I decouple rails to ground. I think this is the best I can do for the opamps...?

Quote:
First of all be sure to remove all such "improvements" and then try op-amp rolling in the original circuit.
Done! It play now, but the noise is annoying...

Quote:
C117 is a weird idea, indeed but might be helpful for the NJM4558D specifically. You can try to desolder it when placing another op-amp.
Tried today. No help on the noise, but it might be beneficial to not have it for the quality... Damn, I wish I have a 4pcs of 4558D opamps...

Quote:
AD8599 is pretty safe because it's slow, just alright for audio as intended by Analog Devices.
Yep, yep, but I still have only two samples from AD... But I might be soldering them on the sockets and trying them... because the noise, is, well... terrible. Should not I rather put these AD8599 ones into the X-Fi that works as sound source?

Quote:
LM4562 shouldn't be cranky, either but it's sound leaves much to desire.
Shouldn't, but it seems it is. Not even shorting the input to ground quiet these oscilations...

Quote:
Actually, the given design shouldn't be problematic for fast op-amps if only the traces obey the high frequency rules. As it was designed for the 4558's, you can't be quite sure about it.
I can take a shoot of the traces, at least I can show the added ceramics, when I add them back there, but I dubt they obey any rules at all


Well, the circuit in question looks like this now:


Opamps nicely in sockets, all modifications reversed back to original. Still noisy. Not like 1:1 to the signal, as it was thanks to my silly mistake, but like 0.1:1 to the signal. Useless.

So todays work - add a voltage filtering caps missing in the design near the opamps:


This 470uF 16V Samxon GC suxxka is supporting now the +15.8V for the IC9 - a subwoofer-dedicated opamp.

No change. Noise from subwoofer still comming.

Later I added two pieces of Panny FM 1000uF 16V caps to filter these +12V and -12V for the IC6, 7 and 8 - eg. for the rest of the opamps.
Even I think I did good job, no change on the noisy behaviour.

So I started again (sigh) removing weird looking caps, started with C117, continuing on the C118, C119, C120, C121 and C122.

And quess what. No change at all. Still noisy, oscilating.
I beginning to think that these LM4562 was either damaged (but all of them? and also - audio is playing right now as I type and well, except the noise, of course) or too cranky to be used in this design.

What to try now?

Put back the 10uF 16V Murata X7R 1206 caps near each the opamps legs? Or use the much smaller values (0.1uF or even 0.01uF) as suggested there?
Working with Cranky Op-Amps
Quote:
you may need to go down to 0.01 µF for faster op-amps
I still not understand why use so small values, because as I look into the Murata specs, these ceramics go way up to GHz, regardless of capacity and when come to voltage filtering, then the more capacity the better...?!

If not help, then remove the C7 and siblings? (that mean C8, C33, C34, C59, C60 and C63)

And what if that does not change a thing?
post #6 of 23
Thread Starter 
Votage filtering for IC6, IC7 and IC8 - all channels input opamps in short, powered from +/-12V, now this is filtered:

by 2x Panny FM 1000uF 16V ( P12366-ND ) caps.

Voltage filtering on the opamps pins

Done by 10uF 16V Murata X7R 1206 caps, ( 490-3911-1-ND ) 7x of them.

No change in behaviour. Still the damn noise is there. Side note - upon powering the amplifier on, the noise seems to none, then sharply the amp become very noisy and then the noise get lowered to what it is for the rest of the time. That happen in like 0.3 sec.

Is this significant?

Another question - the opamp gain is determined by the R9 / R7 size ratio, right? In the Chu Moy's example this is 3.91 ...
In the Genius case it is 1.
Does that mean the opamps gain is ZERO?
Should not that pose a problem for the majority of opamps?
What is the minimal gain to keep the amp stable?

As for trying the other opamps, well, does Fairchild NE5533 count as generic opamp?
Fairchild Semiconductor - Site Search - Operational Amplifiers

I just recieved 4pcs of the DIP8 ones today and - no change at all.
In a desperate attempt to cure the problem, I even desolder the shorts I made on the input switcher, because the RR channel was sometimes nonpresent - eg. the contact is not good anymore there...
No change again.

To make things real simple, I included the switch part in the schematic and keep just the L channel in it:



Comments? Suggestions?
post #7 of 23
In a non-inverting opamp, gain is 1 + R9/R7 = 2.

You really should have the decoupling caps as close to the opamps power pins as possible. You could try soldering a small ceramic cap (eg 1 uF) on the bottomside of your board from pin 4 to pin 8. When you added the sockets, you increased the distance to the caps.

You could try to limit the bandwidth with a cap in parallell with the feedback resistor = C7/C8 in your schematic, maybe 10 - 20 pF. As I understand, you've removed those caps.

You could try adding a snubber resistor on the output of LM4562, maybe something like 100 ohms instead of C9. I can't really see the need of C9 when using opamps with low DC-offset like LM4562. There's still C13/C14 in front of TDA7269.

I've found LM4562 to be very prone to pick up interfering signals, more than any other opamp I've tried. Some kind of shielding have been necessary.
post #8 of 23
Thread Starter 
Thanks for you suggestions, since I'm out of ideas, I definitively try these.

10 - 20pF sounds resonable to limit the hi-frequency oscilation. There was a C7 cap for this case and it has a 100pF... a bit too much, IMHO. However removing or putting it back has little change on the noise (I think that W/O the C7 the noise level is a bit higher, tough) ...

As for small ceramic caps, I already did that:

As you can see and as I explained bellow. I only used 10uF caps, and not 1uF. I think the bigger the better, right?

Quote:
snubber resistor on the output of LM4562, maybe something like 100 ohms instead of C9
This sounds interesting. The Chu Moy's example has one (10 ohms only, tough) too. Worth the try.

Quote:
I've found LM4562 to be very prone to pick up interfering signals, more than any other opamp I've tried. Some kind of shielding have been necessary.
No shielding there The wooden box did not have any... And come to think, the pump (Eheim 1250) is just like feet away from the sub...!!!

Okay, okay, calm dowm. The sub generated the oscilating noise even when used/tested on my balcony, like 10m from the Eheim... So this is not the case, but I have been thinking about some kind of protection inside of the wooden box...
Check down there of what it looks there:
Jaký software budeme potøebovat


Could it be, that the LM4562 / NE5532 opamps simply did not like so low values on the gain controlling resistors?

I mean - in the Chu Moy's example they are 470k / 120k ... in my case they are 10k / 10k ...

That is like 12x less at least.

Could that be a reason to the oscilations, producing this nasty noise?!
post #9 of 23
Quote:
Originally Posted by trodas View Post

Could it be, that the LM4562 / NE5532 opamps simply did not like so low values on the gain controlling resistors?

I mean - in the Chu Moy's example they are 470k / 120k ... in my case they are 10k / 10k ...

That is like 12x less at least.

Could that be a reason to the oscilations, producing this nasty noise?!
LM4562 is unity gain stable. It says so in the data sheet, and I've used it that way. I've never used as high value resistors as 10k. I don't think they're the problem.

When using those very high resistor values (470/120k ), you definately will get some noise.
post #10 of 23
Thread Starter 
Okey, dokey. I will not touch these resistors - yet. First I need to get the amp stop producing the cursed noise


...and...


Big success!

I managed to quiet the R, L, RR and RL channels! At least for like 30% of the volume, witch is what I use daily. So, this is very good for me - if it was not that the center speaker brum is not gone ... (and subwoofer also product some noise, but at very low level, since the noise is like 5 - 6kHz and the subwoofer is optimized for much lower frequency anyway)

What I did over the original schematic?

Removed C117, C118, C119, C120, C121 and C122.
Replaced the C7, C8, C33, C34, C59, C60 and C63 with 12pF caps (original 100pF).

That almost instantly kill most of the noise from the R, L, RR and RL channels. Adding R100 resistors instead of C9, C10, C35, C36, C61, C62 and C64 seems to helped a bit too.

No hearable pot noise as others suggested that this is why these caps are there in the first place.


Current situation.
-----------------
The noisy center is VERY ANNOYING. However hear this - the the volume pot is at like 30% of volume, the R, L, RR and RL channels seems dead silent. But as soon, as I increase the volume, then the center channel noise quiet and all other channels become noisy...!!! :wired:

This is nuts.

It, however, I think clearly demonstrate one thing. That the oscilation is happening because of the output resistance.

I'm I right?

Also caps size on the opamp output seems to play a role, too. Notice that I complain about the noise in the center channel most. (because at 30% volume are the other channels dead silent, so... even later I wand them to be silent at ANY volume, right now I want get rid of the noise at all costs - except for unpludging the center speaker, that it is ) Then notice the C5 cap capacity. 0.22uF for center output from opamp?! Are you kidding me! I want there 22uF 25V Elna RFS cap and I think it will stop the noise - at least in the 30% volume settings.
Also notice the C116 cap - a 470pF one. It is NOT present in the recommended TDA 7360 schematic. I vote for removal...

Input resistances
TDA 7269A - 20k (R, L, RR and RL)
TDA 7360 - 50k (CENTER)
TDA 7296 100k - (SUB)

So, do I get it right that these input resistances are maybe too high, and that cause the oscilations, because high input resistance mean high voltage and that cause high feedback and that, possibly, cause these oscilations?

I replaced almost all the audio decoupling caps with the Elna RFS 22uF 25V ones. All the input ones, and all four for the R, L and RR, RL channels. Can that be significant too?
post #11 of 23
Thread Starter 
Yesterday FedEx delivered the 4pc of RC4580 opamps and 4pc TLE206 opamps, listed my TI as compatible with NJM4558. Both produced very similar oscilation noise, so, no help. Probably not THAT compatible...

So, I got the idea that I can make the opamps to run with like 10k load. That mean put a resistor 12k parallel to the 50k pot = 10 or 9.6k load resistance for opamp.

I tried it for center channel first and the noise it did really limit a lot the oscilations for center channel, but only at given volume, so I tried it for all channells and it make matters a LOT worse after power-on, but in just a short while 10 - 15 sec the noise deacrease to level witch is very nice.
Still noisy, but much more enjoayble. now.

I also tried bump the capacity of C7 & siblings from the 12pF to 330pF, but no change. Maybe it is even slighly worser now...

Maybe the grounding is not perfect and the voltage supply suxx too? Dunno.
post #12 of 23
Thread Starter 
WoW!

NJM4558 operating current - 3.5mA typical ; 5.7mA maximum.

What worry me is, that I can't find in the LM4562 spec what current it draw
trodas: LM4562 - Dual High Performance, High Fidelity Audio Operational Amplifier

But there I find: http://www.head-fi.org/forums/f6/sta...9/#post3536731
That the LM4562 draw 5-6mA/amp - but it is double amp (10 - 12mA) and that is true, when "amplifier current draw when using 3.7V lithium batteries"...!!!

While using +/- 12V voltages, then we talking probably a hell of more current!

I see the guy claiming that LM4562 draw notably more that others - (whole amp= 500m ) while with TL082 it is just whole amp current draw is 320mA.

So instead of the stuuupid R11 and R12 resistors and nonexisting D4 and D5 Zener diodes (they are NOT present in reality!) I slap a 7812 and 7912 regulators there and we see what happnen then.

With the TO-220 package, they should do about 1000mA, so heatsinks is probably not yet required
post #13 of 23
Thread Starter 
Work in progress update
I used L7812CV and L7912CV from STMicroelectronics in TO-220 to make better and nicer the +/-12V voltage supply for opamps. Luckily, I check the datasheets.... the 7912 is NOT like the 7812 - eg. pins view from top is INPUT, GROUND, OUTPUT - the damn 7912 is GROUND, INPUT, OUTPUT...!!!

And I'm NOT use ANY resistor now Recommended caps near these regulators are 0.33uF on input and 0.1uF on the output minimal. I used 1uF SMD ceramic on the input, and 0.1uF on the outputs of both ones. Soldered almost directly on the L7812 and L7912 legs.
Replaced C49 from the cursed 10 000uF CapXon to Samxon KM - also 10 000uF, but 25V only, as it is only 16V on the cap top.
Replaced C112 from bad cap Su'scon 1000uF 10V to Panny FM of same capacitance and voltage.

But damn! Brum is not gone... even the voltages going to the opamps are now very nice, a -11.96V and 11.95V - VERY close, very nice... but the amp play now MUCH louder, so the resistors are obviously overhelmed with the replacement opamps current, so, this was a good change...

Sadly no other changes. The amp still produce a lot of brum at start, and it slowly get away... as things heat up or what the hell...

Time to revert most modifications in the opamps stage back to what things way. I keep the voltage filtering and new regulation as well, as the quality caps, but put back the C7 & spol. caps of 12pF values as that sound best and most quiet, IIRC.
I consider the C117 & spol., but I would like it to keep off.
I remove all the 12k resistors added on all the opamps outputs to ground.
I put back C9 & spol.

But if that does not fix things up, well... I'm out of ideas
post #14 of 23
Thread Starter 
Work in progress update
I removed all the 12k resistors added on all the opamps outputs to ground.
I put back C9 & spol. deblocking caps.
Caps C41 & C42 are now nice Nichicons VR 4700uF 35V suxxkas.
Caps C47 & C48 are now Rubycon XYF 1000uF 16V, taken from one old PSU for DVD player, looks genuine and sure better that Su'scons
Replaced R89 and D8 (near C26, top right part of schematics, close to the front L/R repros) with a L7805 regulator, 1uF ceramic on input (16V cap and the voltage there as well) and 100nF on the output. PCB under R89 was very much darker, so I consider that as safe replacement. Also the output was like 5.3V and not 5V as it should be.

Resulting changes.
Removing the 12k resistors increased the noise level to the before-known situation. Not good, not cool. Caps have no change on the subject, only the C64 seems to put the bass line under control (it was weird, overblown and just blurred before...).
I regret removing the resistors and I planing on puting them back to battle the noise.

Still weird.
It still act kinda weird. When I power the amp up, it lack the before-mod strong SUB kick, but very shortly a strong (oscilating?) noise come from all speakers, and it slowly fade (as things heat-up?) to more acceptable levels. Weird at least. Still lack caps to do complete recap and starting to fear the caps was not the issue, as from the main power caps only C37 & C39 remain as original bad caps...

Futher testing.
I'm somewhat confused that when I power the amp on my balcony, where I solder and work on it, there is no noise from the subwoofer. It might not be even when I power it on with all the speakers, because they simply produce too much noise, that I can't be sure about that. I'm sure about all the 5 other channels, tough.
So first I power the suxxka w/o my PC/watercooling pump nearby and so on. No change.
Then I got idea. I unplug the PCB with the opamps from the rest of the amp. That way, only the output stages are "in game" and they should be very quiet, no noise, as it was before? Right?
Well, wrong.
SUB and CENTER seems dead-quiet, but these L, R and RL, RR channels are full of - wait a minute - MUCH stronger noise that WITH the opamps (and resistors on their inputs to ground!) ...!

What I think of that?
I strongly beginning to suspect that when I at first connected the ceramic 10uF caps to the wrong opamps pins (1 and 7 to ground for all the there opamps), it has consequences. The oscilating noise was unbearable, true... So that lead me to question the TDA7269A amps.
CENTER and SUB use different amps, but the L, R and RL, RR channels use the two TDA7269A ones. What if they are somewhat damaged, so they produce the noise all-by-itself? It is normal that amp produce so strong noise when not connected to any source? I doubt that, and at least CENTER and SUB are fine then.
Their voltage filtering caps (C41 & C42) are quality new Nichicons from Digikey now, so... can't be a issue there.

This would ALSO explain the noise in SUB and CENTER channel too. These channels are interconnected by R109, R111, R156 and R157, so the noise CAN get there by this way.

It would ALSO explain why the resistors helped to battle the noise. Of course opamps does NOT need so low load, but if the noise come the other way, then these quiet it down for obvious reasons. 12k is reasonably low, so it helped...

I think I should remove the R109, R111, R156 and R157.
If that kill the noise from CENTER and the very little noise from the SUB, then these TDA7269A are damaged and replacing them fix the problem. Higher quality parts drop-in replacement (okay, I willing to add few components as well, but the basic pinout has to be the same) suggestion welcome.



oshifis -
Quote:
Is it something similar you want to achieve with your current project?
Heh. I probably already did. It is "loudest noise comming from amplifier" or "worsest dB ratio"... Actually this is no anymore about the difference between signal levels of noise and signal - but more likely now about between the signal and noise...


AndrewT -
Quote:
I think those current consumption graphs are for dual polarity supplies even though the Voltage scale shows V (indicating V+).
Note that the typical 10mA matches the 15V current value.
You could be right. But check on the maximal ouput current as well, 26mA. That is on the top of the 10mA, probably.
Each way around, there was SIGNIFICANT, at least 2x higher (if not more) volume after I added these 7812 and 7912 regulators AND kicked the R11/R12 off the amp.
So that was a good idea and good change in the end. However the resistors should be keept there... the noise is not significantly higher
post #15 of 23
Thread Starter 
This is a little attempt to show how the amp behave on power on and power off/on by remove control. Turn the volume up to get an idea what I hear now all the time...
YouTube - subwoofer problems
New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › Opamps stage design questions