Some LME49600 implementations

[Avro_Arrow]

I actually use this supply for a wide variety of projects and experiments...that's why the heat sinks are oversized.
You could indeed use smaller heat sinks.
 
It would be very easy to add a ground plane.
 
I will look at rearranging the filter caps.
 
I already have a similar power supply I built last year but
I lost the files in a Hard Drive crash. I decided to design
this one to replace it so I am wide open to suggestions.
 
Quote:

Just some quick remarks here for now.
 
The heat sinks can be made the same size as on the amplifier boards.
 
Large capacitance is needed more on the input side than on the output side of the regulator. Additionally such large capacitors might benefit from some film or ceramic decoupling capacitors or even snubbers.
 
Seeing that you have the board space available, consider increasing the size of C1 and C2. Up to at least 470uF is a good idea to improve ripple rejection.
 
The very large V+ and V- copper areas form an effective antenna loop, together with the wire between PSUs and amp. A top layer ground plane would solve this issue almost completely (just like I described in my post above).
 
Cheers,
Sebastian.
 
PS: By joining the negative side of the positive supply to the positive side of the negative supply you effectively form a (local) ground star. That's fine, but you could leverage on this by having the screw terminal offer multiple ground point connections. The board space is there.

 

 

[Avro_Arrow]

Yes, I have done double sided boards before, exactly as you described. I even used cut off resistor wire...
The heat sink can easily be tied to any potential you wish with a jumper to the mounting pin.
 
One of the goals was to try and keep it as simple as possible for people with only beginner level home etching skill.
 
If I were to do a board for a board house, it would be much tighter and complex.
 
Quote:

 
With regard to this I find the concept of guard rings around pins and traces very interesting. Here's an introduction to some ground related issues (from TI).
 
Do you think you can handle a two-sided PCB manufacturing process, with the upper layer being ground plane only? All it takes for this to be effective are a reasonable number of vias (i.e. hand soldered on both sides using resistor wire cuttoffs).
 
 
There's no reason not to ground the heat sink (granted the buffers are mounted in an insulated way), especially if EMI is taken into consideration. OTOH it's not absolutely necessary either.
 
Cheers,
Sebastian.
 
 

 

 

[Avro_Arrow]

I should add that Wakibaki is doing the board house quality board.
I'm just doing an alternate (simpler) version.
I don't want to seem like I am hijacking his thread...
 
Quote:

I actually use this supply for a wide variety of projects and experiments...that's why the heat sinks are oversized.
You could indeed use smaller heat sinks.
 
It would be very easy to add a ground plane.
 
I will look at rearranging the filter caps.
 
I already have a similar power supply I built last year but
I lost the files in a Hard Drive crash. I decided to design
this one to replace it so I am wide open to suggestions.
 


 

 


Quote:

 
With regard to this I find the concept of guard rings around pins and traces very interesting. Here's an introduction to some ground related issues (from TI).
 
Do you think you can handle a two-sided PCB manufacturing process, with the upper layer being ground plane only? All it takes for this to be effective are a reasonable number of vias (i.e. hand soldered on both sides using resistor wire cuttoffs).
 
 
There's no reason not to ground the heat sink (granted the buffers are mounted in an insulated way), especially if EMI is taken into consideration. OTOH it's not absolutely necessary either.
 
Cheers,
Sebastian.
 
 

 

 

[wakibaki]

Quote:

I should add that Wakibaki is doing the board house quality board.
I'm just doing an alternate (simpler) version.
I don't want to seem like I am hijacking his thread...

No problem at all Avro_Arrow, I'm happy to see anybody posting layouts they're working on, in fact I was thinking of suggesting sek@ post one of his own since he obviously has software and skills, it's much easier to see the advantages of a particular strategy with an example in front of you, rather than add in people's suggestions, design by remote control, as it were.
 
I think the EMC considerations are, to a degree, secondary here, neither board has a clock. Crosstalk between tracks is a concern though. Certainly you want to keep RFI out of the circuit, but the reference design, while doublesided, has no extensive ground plane, and we are concerned to keep stray capacitance in the area of the active device pins at a minimum as oscillation is a real concern with such high bandwidth devices.
 
I've made some changes, the regs have gone to TO220 instead of SOT223-3. This gives me a bit more than 3 sq. in. of board area to cool the 49600's, about the same as on the reference board. I was aiming at using some 16V wallwarts I've got here, if the regs are set to 18V out they will be dissipating about 0.4W for a 250mA peak sinewave output, which I think means they will cope without heatsinking, music power should be 10dB down or more re. sinewave.
 

 
w
 

 

[Avro_Arrow]

I'm trying to get my head around what Kevin said...
In my schematic, the gain resistor is R1, the feedback resistor is R3.
Is the servo resistor R5?
 
Quote:

The servo resistor should be at least a factor of 10 greater than the gain resistor
to ground. Then you can adjust the feedback resistor for the overall gain
you require.

 

 

[kevin gilmore]

in your schematic...
 
R1 is the resistor to ground, in the schematic 1k.
R5 is the servo resistor, currently 10k, but can go higher depending
on the offset of the audio opamp.
R3, currently 1k is the gain setting resistor, so calculate the
R1,R5 pair, and set R3 for the voltage gain required.
i'm using 4.7k for R3.

 

[Avro_Arrow]

Thanks Kevin
 
Quote:

in your schematic...
 
R1 is the resistor to ground, in the schematic 1k.
R5 is the servo resistor, currently 10k, but can go higher depending
on the offset of the audio opamp.
R3, currently 1k is the gain setting resistor, so calculate the
R1,R5 pair, and set R3 for the voltage gain required.
i'm using 4.7k for R3.

 

 

[sek@]

Hi wakibaki,
 
Quote:

I was thinking of suggesting sek@ post one of his own since he obviously has software and skills

 
I'd be glad to, but unfortunately most of it is work related (robotics, microcontrollers). That stuff can't be released here.
I wish I'd have the time to get to my DIY projects as often as I seem to come across like. And please don't make me look more experienced than I am.
 
My current headphone related project is a balanced amplifier with integrated analog volume control. 
 
Here's part of the schematic. It's prototyped (one iteration) and almost set, give or take a couple of component values:

 
And here's part of the board. It's tow-sided, currently unrouted, but mostly figured out. Power supply and some of the bypassing are omitted in this detail:

 
Other than that, the things I mentioned are specifically those that caused me problems in the past, both with DIY projects and with actual products. That does of course not imply that builders of the projects in this thread will, though. YMMV.
 
 

I think the EMC considerations are, to a degree, secondary here, neither board has a clock.

 
I tend to think that - after everything has been wired correctly and a good parts choice has been made - EMI is the only concern left! 
 
At the same time, it's a phenomenon of perpetually increasing relevance. The number of wireless devices (and clocks, as you narrow it) around us (and on us) rises considerably and this rise will most likely even become steeper in the future. 
 
Thus, a design that worked well in the past may have GSM or ISM related RFI problems today. A design that works well today...
 
Cheers,
Sebastian.

 

[sek@]

One more thing!
 
Quote:

The heat sink can easily be tied to any potential you wish with a jumper to the mounting pin.

 
Thinking about this approach, this could help to solve one of my remaining issues: heat dissipation.
 
If I were to mount the voltage regulators upright (on heat sinks) and then were to tightly solder the heat sink to the negative supply copper, the sinks could help the dissipation via the copper plane.
 
I wonder how strong the effect would be...
 
But enough with the threadjacking, back to wakibaki's design!

 

[Avro_Arrow]

There are solderable board level heat sinks for SMD devices such as this Wakefield Heatsink.
 

 

[sek@]

Quote:

There are solderable board level heat sinks for SMD devices
 

Sure, I'm using those with THE WIRE. But they take a lot of horizontal space, i.e. they don't allow for close spacing of the buffers (because the fins would overlap).
 

 

[kevin gilmore]

here is what the default servo filter does
http://gilmore.chem.northwestern.edu/defaultfilter.jpg
amplitude in green, phase in red

 

[wakibaki]

Is this a Proteus plot Kevin? Certainly looks like one. Where are the probe and stimulus placed?
 
Changing the resistor changes the range of servo authority as I understand it, i.e. the default resistor permits the servo to influence the DC level from rail-to-rail, being the same value as the feedback resistor (the one from output to inverting input), and because it is in parallel with the resistor to ground (being the same value as that one too) and because the servo opamp output is a virtual ground it increases the circuit gain to 3.
 
1 + Rf/(Rg||Rs) = 1 + 1000/(1000||1000) = 1 + 1000/500 = 1 + 2 = 3.
 
Increasing it to 10k would decrease the servo authority to (+/-)1/10 * rail voltage, and reduce the circuit gain to 2.1
 
1 + Rf (Rg||Rs)= 1 + 1000/(1000||10,0000) = 1 + 1000/909.09 = 1 + 1,1 = 2.1.
 
What other negative effects arise from using the default values? I can't visualise what these are purely on the basis of the gain/phase plot.
 
w

 

[kevin gilmore]

more clear

http://gilmore.chem.northwestern.edu/defaultfilter2.jpg

 

[wakibaki]

 
Both circuits have gain virtually flat @ just over 1Hz, phase ~0 degrees by 10 Hz, I don't think the differences are anything to get excited about. It'll be a rare CD player having output (and a rare CD having content) @10Hz. There's no real inconsistency with the datasheet's claim that the servo cutoff is over 2 decades below 20Hz, which should be entirely adequate to avoid any audible perturbations.
 
I was surprised when I first saw the values for the servo in the datasheet. I doubt that there's any necessity to give the servo that much authority, but the main concern, as I see it, is that a fault might result in the servo pulling the circuit to a large DC offset. Setting the authority to 1/10th. however, merely restricts the offset in a fault condition to 1.5V (for, say, 15V rails), which is probably still catastrophic for most headphones, and is why the circuit I have drawn includes output offset protection, which might not have been considered necessary in an AC coupled circuit, depending on the exact implementation.
 
I will, however, use a value of 10k to set the servo authority to 1/10 in the light of your input (I see no real necessity for any greater) and modify the gain setting resistors accordingly if experience of the first prototype shows it to be desirable. Thanks for taking the time and putting in the effort to do the simulations and post them. Please get back to us if I have misunderstood the significance of the plots or if you have any further observations
 
w