Welcome to Solid State Guitar Amp Forum | DIY Guitar Amplifiers. Please login or sign up.

March 28, 2024, 05:39:50 AM

Login with username, password and session length

Recent Posts

 

TDA2003 - Reducing Voltage Gain

Started by blackcorvo, March 06, 2013, 12:14:51 AM

Previous topic - Next topic

J M Fahey

#15
Sorry but that Handmades.br layout has errors.

You can't trust it.

In the original schematic, clipping diodes D1/D2 have *always* one end connected to ground, the other one gets connected or not to R6 through Sw1a to provide clipping at will.

In the layout no end of D1/D2 is connected to ground and anyway R6 gets connected straight to ground (the "wrong" diode ends) or "Rx", which he *had* to add or on "Boost" the amp would actually "mute".

So the designer instead of correcting an error,  added an extra one.

That flawed design *may* work, in a fashion, but in a crooked way.

Not surprising you don't like it.

PS: *maybe* somebody noticed and corrected that later, but I'm not reading 35 pages of posts to find it ;)

blackcorvo

#16
But they ARE connected to ground. There's a jumper right next to Rx (which has been added to keep the volume of the Drive channel about the same as the Clean channel).

On another note, I'm playing around with the amp still. I have an LM386 preamp right now, with the tone control I posted. I get squeals with the tone at about 75% on the treble side and high-gain settings, so I'm guessing it is actually a gain problem that's causing it. I might change the 100k volume-pot for a 50k pot + 47k series resistor, to keep the max volume at non-squeal levels.

- - - - - -

[UPDATE]

After playing around a little, I ended up with the following circuit:
http://img836.imageshack.us/img836/7029/l1.png

I based the tone control on the Marshall 18w. I scaled it down based on this: http://www.aikenamps.com/ToneControlScaling.html , then I tweaked R5 to taste (it was supposed to be 27k, but I felt 12k gave it a more noticeable bass cut). I also had to change C3 to be connected AFTER C4, otherwise it wouldn't work. In practice, it doesn't change much the way it works.

It sounds nice even when I drive the TDA2050 into distortion. I'll keep it as a single-supply design, because that way I can use my 12v/7Ah battery with it in case I wanna have a jam somewhere without mains outlets nearby, and I can use common power supplies instead of something with an specific connector.

I still get feedback at full volume, but this time it's the good kind of feedback, from being too close to the speaker. Samples to come very soon, and maybe even a layout!

[UPDATE 2]

Here's a (rather crappy) sound sample of the amp, working off a 12v/7Ah sealed battery, still on protoboard so it has some background noise:

https://soundcloud.com/corvolino03/18w-ss

Roly

Quote from: blackcorvoI'm guessing it is actually a gain problem that's causing it.

Instability comes from a combination of forward gain, a feedback path, and the feedback being in phase (or positive), or phase shifted so it ends up being positive around the loop.

The last point relates to the tone controls which not only alter the amplitude of the signal passing through, they also alter the phase, so any signal feedback that goes around the tone controls is very likely to result in some setting of the controls that causes oscillation.

In this case I've had a look at the circuit and PCB layout.  It's one thing to connect all the points together on a PCB layout, but it's another to take care of mutual ground and power distribution paths.

In this case there is a path in common between the main chip input (being the ground connection of the master volume), and the output circuit of the main chip.  In particular the Zobel network, R9 and C8, are required for RF and VHF stability and therefore must go as directly between the output pin and chip ground as possible.

Here the chip ground is not clearly defined and the Zoble is grounded by a long run back to the power supply filter caps, then back along the edge ground to the chip area.  This grounding arrangement is an invitation to problems and I'm surprised other haven't encountered instability.  This board layout is about as far as you can get from the concepts of single point earthing, and of not having power and grounds shared in common between stages.

Appended is a suggested mod that would improve the situation by removing ground return currents from the ground shared with the main volume pot (and other parts of the preamp), but really, this needs to be layed out again with an understanding of why preamp and main amp grounds have to be both segregated and well defined.  The preamp ground point is between decoupling caps C12 and C13, yet the ground path from there back around via the main filter caps and back to the preamp shares much of its path with output stage currents and pretty well defeats the whole object of decoupling.

I'm not at all surprised it's unstable.
If you say theory and practice don't agree you haven't applied enough theory.

blackcorvo

Quote from: Roly on July 04, 2013, 03:37:55 PM
I'm not at all surprised it's unstable.

I see. When I draw my layouts, I personally try my best to have all grounds in a single piece of trace instead of using jumpers like in this layout. I imagined that could be a huge contributor to this issue.

Anyways, could you please give me some piece of mind on my own design I posted on the 2 updates added to my latest post? What are your thoughts on it?

phatt

If You want to reduce voltage swing then loose IC2a.

Connect the tone input at C8 output and see if it works to your liking. Drop VR3 as well,, not needed.

Try  R12 at 10k  and make R9 A pot for gain trim control also hang a 100pF cap across pin 6 and 7 for some HF stability.
have fun,, Phil.

blackcorvo

Quote from: phatt on July 05, 2013, 02:24:06 AM
If You want to reduce voltage swing then loose IC2a.

Connect the tone input at C8 output and see if it works to your liking. Drop VR3 as well,, not needed.

Try  R12 at 10k  and make R9 A pot for gain trim control also hang a 100pF cap across pin 6 and 7 for some HF stability.
have fun,, Phil.

That's not exactly what I meant... I like how it sounds now!
I don't really need a clean amp since I don't have any pedals... but if I wanted to, I could simply add a couple of switching jacks between IC2a and C4 and use them as an FX Loop...

What I meant is to know if you think I should change something to make it sound better, or if you spot any errors. You mentioned a cap in parallel with R9, and I noticed I forgot to add that for R7 in my schematic!
Do you think a 100pF in parallel with each of those resistors are gonna do the job?

Roly

Well the jumper is not going to help, but the basic idea behind single point earthing is to avoid different stages sharing common ground impedances.  If you have a look inside your typical Asian stereo amp in the power supply area you will notice that separate tracks arrive at the main power supply filter cap +ve and -ve, even when they may run in parallel across much of the board.

If you could see the currents you would observe that the high currents from the output stage were flowing down one of the tracks, and the low currents for the preamps down another (and the tuner, etc, etc, down others).  The whole point is that they don't share the resistance of any ground trace because it is this shared resistance in common that couples the high level and low level stages together in exactly the way we try to prevent with decoupling.

So a good power supply distribution scheme will radiate separate supply and ground return paths for each of the sections of the amp, each only coming together at a single point, not a "daisy chain" of one to the next.

If you look at this layout around the headphone socket you can see that there is a temptation to use what is effectively the preamp ground (master vol ground) as the "ground" for some of the main amp and output wiring, and this temptation has been given in to, resulting in output stage currents flowing in the preamp ground circuit (and incidentally right past the EQ).  This mixing of ground (or supply) paths is simply begging for instability, leaving you at the mercy of the width and thickness that determine the resistance of the PCB tracks.


R9 = 100k
R7 = 220k

Cx = 100pF

The corner frequency, fc, occurs when the reactance, Xc, of the capacitor is equal to the resistance.

At what frequencies is Xc = 100k and 220k?

Xc = 1/ (2 Pi f C)

transpose

f = 1/ (2 Pi Xc C)

Xc = 100 x 103 = 105 ohms
C = 100 x 10-12 = 10-10 farads

1/(2 * Pi * 10^5 * 10^-10) = 15915.4943091 or 16kHz

For the 220k it's going to be 2.2 times higher or 16*2.2 = 35.2kHz

Considering that typical guitar speakers cut off at around 5kHz both of these figures are a bit on the high side, so I'd be inclined to make the cap in parallel with R9 about three times larger, say 270pF or 330pF, and the one in parallel with R7 some six times higher, say 560pF or 680pF.

I would also add something like a 0.1uF right across the power supply pins 4 and 8 of the TL082.

HTH

If you say theory and practice don't agree you haven't applied enough theory.

blackcorvo

#22
Quote from: Roly on July 05, 2013, 07:14:38 AM
Considering that typical guitar speakers cut off at around 5kHz both of these figures are a bit on the high side, so I'd be inclined to make the cap in parallel with R9 about three times larger, say 270pF or 330pF, and the one in parallel with R7 some six times higher, say 560pF or 680pF.

I'm not using a *real* guitar speaker, I'm using a generic 8 ohms, 8" Philips full-range that has been re-coned to be somewhat like a guitar speaker. It has a pretty deep cone, and sounds decent enough for me. Not sure if that could influence anything on the choice of these capacitors...

Quote from: Roly on July 05, 2013, 07:14:38 AM
I would also add something like a 0.1uF right across the power supply pins 4 and 8 of the TL082.

I just did that on the layout I made. Doesn't look like anything professional, but I think it's functional. I'll add it as an attachment.
Please don't mind it not looking very nice, it's just that I haven't done a large layout like this in a long time.

By the way, the jumper between pin 7 of the TL082 and the 22nF is in case I wanna add an FX loop, which I can do in an external board.

Roly

Well start off with the free air resonance of the speaker you are using, multiply by the rule of thumb x80 to find the top end roll off, then recalculate as above for a frequency that is "a bit more" than this new cutoff.  Apart from being supersonic, a 32kHz cutoff isn't a lot of help with stability - it needs to be just comfortably above the highest frequency of interest.

Very often the op-amp power supply bypass is kluged onto the copper side of the board under the chip, and that is in fact the best place for it, right at the op-amp pins.  As audio op-amps get better so their gain/bandwidth product goes up.  When I converted the preamp in my Twin-50 from transistors to op-amps I used the LM833 which has a unity gain of 3 MEGAhertz, so it was not entirely surprising that on first power up it promptly started oscillating around 450kHz!  I eventually ended up crowbaring the bandwidth down to about 35kHz, which was still more than ample, but it's a caution that modern components are getting good enough to test even the best layouts.


Looking at your layout a couple of points arise.

The first is that you still have a ground path in common with the main and preamp between the main amp and the power supply filter cap (the mod is diagrammatic, the track should be reasonably fat).

The second is that the Zobel network (1r + 220nF) is a bit far away from the chip when we consider that it has to be effective into the HF/VHF region, which means the shortest possible path between the chip output terminal and ground terminal.

I've appended your layout with a modified ground and indicating where the Zobel could be moved.
If you say theory and practice don't agree you haven't applied enough theory.

phatt

@*blackcorvo*,
If I can butt in for a mo/
What Program do you use to get those rather neat looking layouts? :)
Reminds me of old board designs from the 60's.

And thanks *Roly* Good to be reminded as I to am guilty of forgetting all about ground path.
I think i spend so much time learning how to use the darn software that the old brain just reaches overload.
I guess if you do circuit design everyday it becomes easier but as a hobby you only make a few here and there so by the time you have to make a new circuit you forget a lot. xP
Carry on chaps,, Phil.

blackcorvo

Quote from: Roly on July 06, 2013, 08:30:46 AM
Well start off with the free air resonance of the speaker you are using...

I have no idea how to find that. I know nothing about speaker enclosure design, I just put it in a box and if it makes sound, I'm happy with it.

Quote from: Roly on July 06, 2013, 08:30:46 AM
Looking at your layout a couple of points arise.
(...)
I've appended your layout with a modified ground and indicating where the Zobel could be moved.

I'll put the modified layout based on your suggestions (and with a couple other small tweaks) as an attachment.

Quote from: phatt on July 06, 2013, 09:20:02 AM
@*blackcorvo*,
If I can butt in for a mo/
What Program do you use to get those rather neat looking layouts? :)
Reminds me of old board designs from the 60's.

Carry on chaps,, Phil.

It's called DIY Layout Creator. It's a free software and can be found here:
https://code.google.com/p/diy-layout-creator/

Roly

Method A - look at the specs, "fo".

Method B - take it out of the enclosure and sweep it from, say, 20Hz to 200Hz, and somewhere in there it will go crazy as it resonates - that's the free air resonance.

Method C - since what we actually want to know is the high cutoff, measure it directly by sweeping the speaker in the enclosure, say 1kHz - 10kHz, and note the frequency it drops the bucket, that will be the high cutoff (or your ears, whichever comes first).


The ground for the preamp looks good, but that Zobel ... oh dear, you don't seem to be getting the point here.  The Zobel must go as directly as possible between the output pin and ground pin of the TA2050, not a grand tourissimo via Zagreb.  Where does the 0.022uF now "ground"?

So when I say "short and direct" I mean SHORT and DIRECTCapiche?
If you say theory and practice don't agree you haven't applied enough theory.

blackcorvo

Quote from: Roly on July 06, 2013, 01:32:42 PM
Method A - look at the specs, "fo".

Method B - take it out of the enclosure and sweep it from, say, 20Hz to 200Hz, and somewhere in there it will go crazy as it resonates - that's the free air resonance.

Method C - since what we actually want to know is the high cutoff, measure it directly by sweeping the speaker in the enclosure, say 1kHz - 10kHz, and note the frequency it drops the bucket, that will be the high cutoff (or your ears, whichever comes first).

I have no idea how to do that, and all I found about this speaker were these pictures of a pair of it being sold in MercadoLivre (our version of Ebay, but not as reliable), which are in bad shape but have the original cone in them:

http://img843.imageshack.us/img843/69/l0yb.jpg
http://img833.imageshack.us/img833/2289/mxml.jpg
http://img191.imageshack.us/img191/6752/tlm7.jpg

And here's mine on it's temporary "home":

http://img835.imageshack.us/img835/8003/nvep.png
http://img194.imageshack.us/img194/7173/8p7u.png
http://img834.imageshack.us/img834/164/aljr.png

Quote from: Roly on July 06, 2013, 01:32:42 PM
The ground for the preamp looks good, but that Zobel ... oh dear, you don't seem to be getting the point here.  The Zobel must go as directly as possible between the output pin and ground pin of the TA2050, not a grand tourissimo via Zagreb.  Where does the 0.022uF now "ground"?

So when I say "short and direct" I mean SHORT and DIRECTCapiche?

Wow, I didn't mean to piss you off man. I'm sorry  :-[

Does this look better to you?

phatt

Ha, No Roly is just being his normal *Expressive* self so don't panic,,, Yet.  :lmao:

Re the Direct thing.

The Circuit Common Zero point is the Negative pin of that 4,700 Filter cap. Roly means to get that 220nF Cap to go direct to that track (Or make a dedicated separate track) so it has the shortest possible return path to circuit common.
Sadly it might require a bit of fiddling with the tracks yet again. :'(

Fear not it's worth it I've spent days and weeks trying to rework tracks to make things go right you do have to walk away a lot and come back fresh.

Hey thanks for the link I'll have good look at it. :tu:
Yes I do remember the first vero board version which was not what I needed but many first time builders find it useful.
But this looks far more useful. :)

Phil.

Roly

Quote from: blackcorvoWow, I didn't mean to piss you off man. I'm sorry  :-[

Quote from: phattHa, No Roly is just being his normal *Expressive* self so don't panic,,, Yet.  :lmao:

True dat; I'm a tech and musician, not a politician or social worker.

...and...  ()ing YO!    :dbtu:That's that way ya do it, money for nothing, and your chicks for free...



In the case of the Zobel I do actually mean the ground pin of the IC, not ground at the main filter cap (for once) because it's effect is needed local to the IC to place an increased load on the IC output at HF and VHF frequencies, and because of these short wavelengths it's an exception case to the normal rule of tying everything back to the main filter cap terminals, which in the case of VHF are too much stray inductance away to help.  {But then 8/10 audio engineers don't know what the Zobel is there for in the first place - thank you Doug Self; e.g. in how many amps have you seen the Zobel placed on the speaker side of the output inductor, the wrong side?}


There are a pile of freeware signal generators, but to test your loudspeaker an audio sweep generator is more useful than just a spot signal generator.  I've downloaded this one and a) it works, and b) didn't seem to virus or crash my system.  Your computer sound card may well have enough grunt to drive the speaker directly for this test.
If you say theory and practice don't agree you haven't applied enough theory.