Integrated DAC + Bridge LM3875 (Balanced)

[Yoshy]

Draft layout

TODOs:
- Remove Ground Plane
- Separate Power Ground from Signal ground
- Find optimal placement for every component

[pacificblue]

Is the gain setting correct on IC4A?

Trust gfiandy about the Zobel network at the output. It is very likely necessary and in the few cases, where the LM3875/3886 can live without it, it doesn't hurt to still have it.

[Yoshy]

Originally posted by pacificblue
Is the gain setting correct on IC4A?

Trust gfiandy about the Zobel network at the output. It is very likely necessary and in the few cases, where the LM3875/3886 can live without it, it doesn't hurt to still have it.

Ahahah I see what I did there... I'll fix IC4A :/ Damn I should trust spice a bit more

As for the Zobel network I'm still trying to figure out a place to put it -- It will be a 3 way active speaker so every board will need different values; how about a "plug-in" board ?

[gfiandy]

Hi,

You will get quite alot of Johnson thermal noise off the 2.2M resistors and whilst some of this will be smoothed out by the servo caps, some of it won't as the opamp output is a driven point it will try to drive the capacitor with the noise signal. (I think, I haven't actually done a noise analysis, if you have a simulator (or are good at math and patient) this would be worth while) To reduce this you could split the 205K resistor into 2 x 100K with a capacitor to gnd between them to filter of the the thermal noise generated. You need to be sure this filter point is at least an order of magnitude higher than the servo itself or you could get sufficient phase shift to compromise stability.

Your schematic shows no decoupling, I hope this is on a seperate sheet as it won't work without it.

Using 360R as the feedback resistor on the unbalancing opamp puts a very high current load on the output. This is likely to cause the opamp some stress and hence an increase in distortion. I would recomend using a resistor more in the 1K (more 820R resistors would be ok) region and changing the input resistors to match. It will also put quite a high load on the output of the opamp driving the -ve input channel. ( I am assuming from the previous post that you have fixed the +ve / -ve input mix up on IC 4A)

Regards,
Andrew

[star882]

Originally posted by Yoshy
Hello

I've been working on my speakers lately and the thought of integrating the amps in the speakers have crossed my mind more then once; not only would that save space that would also cut on visible wires (which gives points for the GAF ).

I currently have a few (16 to be exact) LM3875 that I'd like to use, the crossover will be digital (ADC -> DSP -> DACs). To save space I'd like to have a board with a DAC (PCM1792), 2 LM3875 bridged and de-coupling caps.

I think I've merge the I/V stage with the chip correctly (and this is where I'd like comments), is the AD8620 a good op amp for I/V ? And even more basic - is the circuit correct ? I'd really like not having to convert to single-ended but that's just "because". I'm a bit worried about offset, I haven't put any DC Servo or pots to adjust it. Maybe I'd need output relays too to make sure I don't get a start up pop.

A draft of the PCB layout will be posted in my next post.

Comments are welcomed !

Unnecessary conversions between analog and digital will decrease the quality. At the least, DSP the signal before it is even converted to analog. And since you're dealing with digital, use a digital amplifier chipset. (Yes, pretty much all of them are SMD, but so are pretty much all DSPs, so I assume you have the equipment to work with SMD.)

[Yoshy]

Originally posted by star882

Unnecessary conversions between analog and digital will decrease the quality. At the least, DSP the signal before it is even converted to analog. And since you're dealing with digital, use a digital amplifier chipset. (Yes, pretty much all of them are SMD, but so are pretty much all DSPs, so I assume you have the equipment to work with SMD.)

What?
An analog DSP ?

I don't want to implement a crossover that has delays and an eq in full analog, ADC and DAC are good theses days...

Yes I can solder SMD no problem

[Yoshy]

Originally posted by gfiandy
Hi,

You will get quite alot of Johnson thermal noise off the 2.2M resistors and whilst some of this will be smoothed out by the servo caps, some of it won't as the opamp output is a driven point it will try to drive the capacitor with the noise signal. (I think, I haven't actually done a noise analysis, if you have a simulator (or are good at math and patient) this would be worth while) To reduce this you could split the 205K resistor into 2 x 100K with a capacitor to gnd between them to filter of the the thermal noise generated. You need to be sure this filter point is at least an order of magnitude higher than the servo itself or you could get sufficient phase shift to compromise stability.

Your schematic shows no decoupling, I hope this is on a seperate sheet as it won't work without it.

Using 360R as the feedback resistor on the unbalancing opamp puts a very high current load on the output. This is likely to cause the opamp some stress and hence an increase in distortion. I would recomend using a resistor more in the 1K (more 820R resistors would be ok) region and changing the input resistors to match. It will also put quite a high load on the output of the opamp driving the -ve input channel. ( I am assuming from the previous post that you have fixed the +ve / -ve input mix up on IC 4A)

Regards,
Andrew

Thanks Andrew for helping me out !

I've changed a few things in this version; as for the 2.2MEG resistors, I'm prepared to test it and if the cap doesn't help I'll just remove the DC servo. Or maybe even reduce R14 and R15 to 250k ?

Anyway the schematic is pretty messy - caps got added and removed from other sheets; Eagle doesn't seems to have a "re-annotate" function which would be more then useful right now...

C15 is the cap I added from the split of the 205k res, I was thinking about making it (along C1) 1uF

Also removed are the 360ohms resistors, replaced by 820; this will also make the cost of the amp go down which is great. Forgot to recalculate C11 and C4.

Still trying to lay it down on a PCB... Oh and yes decoupling is on another sheet, 1uF caps on every opamp per supply, X7R type

Many thanks for the comments !

[star882]

Originally posted by Yoshy
What?
An analog DSP ?

I don't want to implement a crossover that has delays and an eq in full analog, ADC and DAC are good theses days...

Yes I can solder SMD no problem

Take the digital signal before it is converted to analog and do DSP on it, then run it to a digital amplifier chipset. Don't convert the signal to analog only to convert it back to digital.

[Yoshy]

Originally posted by star882

Take the digital signal before it is converted to analog and do DSP on it, then run it to a digital amplifier chipset. Don't convert the signal to analog only to convert it back to digital.

I don't have a choice... It's analog out of the preamp then to amps directly and no way I'm replacing my preamp.

[star882]

Originally posted by Yoshy


I don't have a choice... It's analog out of the preamp then to amps directly and no way I'm replacing my preamp.

Converting to analog and back to digital will cause a significant loss in quality. Save the preamp for later and use a direct digital connection.