DCX2496 audio mod

[dmills]

D4/5 are there to catch pathological input cases (someone sticks P48 up the thing for example), but I suppose it is arguable that given the 6db gain of the input stage we could drop them to 12V or so and run then in reverse breakdown the whole time which might improve their speed.
The intention is that the caps catch short surges and spikes and the zenner clamp deals with more major things.

As to the two opamp instrumentation amp, you could do a three opamp one, but the key thing is that to get good noise performance you need to avoid large series input resistors because of the thermal noise (R7 & 8 are ok because they are in parallel with the source impedance which should be much lower then they are), a simple minded single opamp difference amp will be the dominant noise source if you are trying to get the differential input impedance higher then a few K.

Given the stated source characteristics (Highish Z, only a few metres of cable), I felt that the two opamp instrumentation amp would be sufficient, but it is a matter of taste.
And yea, be careful with spice, it is a useful tool but the results always need sanity checking.

Regards, Dan.

 

[jdbakker]

D4/5 are there to catch pathological input cases (someone sticks P48 up the thing for example), but I suppose it is arguable that given the 6db gain of the input stage we could drop them to 12V or so and run then in reverse breakdown the whole time which might improve their speed.
The intention is that the caps catch short surges and spikes and the zenner clamp deals with more major things.

Yeah, I figured as much. My worry is/was that with those naked '+'-inputs you'll get the op-amp's protection diodes conducting before the zeners.

JDB.
[this is not a problem with the THAT chips, as their input CM range extends beyond the power rails]

 

[Moby]

Guys, at last I decided how to manage whole thing.. I decided to split input buffer on 3 pcb's because 1632 have to be close to DSP board connector. So, that PCB  will be plugged directly into socket. Because of small space around and also because my curiosity to try few versions of input buffers that will be the 2. PCB mounted on top of 1632 PCB (piggy board) , with enough space for air flow but short to the 1632 PCB. 3. PCB is holding clamping diodes and ADC capacitor and it will be mounted close to ADC , somehow, still not have idea, but it can be glued at last  It will aks to mak one more "floor" for 1U box but that's not a problem.  ;D
Regarding instrumentation amp I will make 3 versions. One with Dan's schemo , one with unity amps and 3. with THAT 1203, so one day when I find the source for that (THAT) I can just replace "piggy". I suppose that input resistors of OPA1632 can be  as they are 1k5... Can they?
Dan do you want to share version with THAT receiver? You mentioned some passive filter network included  8)

 

[Moby]

Here's the rough version of 1632 PCB based on the Dan's circuit. It holds all connectors so I can add later output buffers . Comments welcome  Of course this is not real dimmension 

 

[dmills]

Assuming those OPA134s are the 2.5V bias generators then you only need one, but I would keep R18/C16 (my numbers) separate for each channel (Speaking of which, where is my C16?).

There are some spurious vias.
You also seem to have a rather curious mix of through hole and SMT involved.

Regards, Dan.

 

[Moby]

Assuming those OPA134s are the 2.5V bias generators then you only need one, but I would keep R18/C16 (my numbers) separate for each channel (Speaking of which, where is my C16?).

Cool, I missed C16, good to save a space with one OPA134 

There are some spurious vias.

They are suspicious because I don't have vias metalization (self etching) so I added few more near parts which I can't solder under.

You also seem to have a rather curious mix of through hole and SMT involved.

Huh , just don't push m to the SMT resistors. I will loose my eyes 

 

[jdbakker]

What Dan said.

Additionally:

- Have a solid ground plane on one side of the board. 'Solid' here means no traces, no SMT components, just copper (plus a few isolated pads for through-hole parts, naturally. Have no vias other than ground connections; the design is simple enough for that to be doable.

- Consider putting small resistors (~10R) in the supply lines of the OPA1632, between +/-15V rails and the '1632 and its decoupling cap. This helps HF decoupling and it makes layout easier as you can use the resistors as jumpers away from the crowded feedback network of the '1632.

- C5/C6 on your board are still too far from the OPA1632. Try to keep the sum of both trace lengths (from OPA to cap, from cap to OPA) to 10mm. This is possible without resorting to heroic measures.

- Identify signal loops, and reduce their area.

...plus the stuff I mentioned earlier.

JDB.

 

[Moby]

the design is simple enough for that to be doable

I understand what you are talking about but with connectors onboard I  think that I can't do that . I don't say that it' impossible, but my design skills are limited    Anyway, I appreciate your efforts to help  Edit.. I really don't know how to put SMT components on bottom layer... huh...

 

[Moby]

Cool. I figured out , thanks Google  8)

 

[Moby]

Any better ???

 

[Moby]

Better?

 

[Moby]

No comments about PCB? That means OK or you tired?  Before I start with 2. PCB design I must ask can I substitute Shottky's dual SMD to through hole shottky's? BAT82 or similar?

 

[dmills]

Busy, day job....
Yea, you can swap the diodes, I used them because I needed to fit 8 channels on a 100 * 160 eurocard in the design I copied that section from.

At some point I will upload the whole drawing.....

Regards, Dan.

 

[Moby]

Cool Dan. Does this drawing include DAC audio section? I'm asking because it need to be modded too  ;D

 

[dmills]

Nope, that was on a separate set of cards (This thing had 16 inputs feeding 64 outputs, with a load of DSP in the middle (I love the AD DSP Parts)).

There are in any case plenty of good DAC output designs out there, it is in most ways a far simpler thing to design then the input stage.
As ever, board layout makes a huge difference and you will want to get the (if appropriate) current to voltage stage and first filter very, very close to the DAC chip.

Do not get fooled by the presence of Agnd and Dgnd pins on mixed signal converters by the way, for most parts you want them connected to the same ground plane (The issue is bond wire inductance), and remember that converters are high frequency (RF) devices with clock edge harmonics well into the VHF region.

Regards, Dan.

 

[Moby]

There are in any case plenty of good DAC output designs out there, it is in most ways a far simpler thing to design then the input stage.
As ever, board layout makes a huge difference and you will want to get the (if appropriate) current to voltage stage and first filter very, very close to the DAC chip.

DAC's are AK1393. From datasheet I can see that they recommend non inverting diff output with appropriate r/c filter (how about L/C?). I'm not sure is the chip capable for I to V stage since max I out rating is 3.5ma  ??? What you think about it? Also, I will need some mute circuit to skip  on/of bump . Here's the schemo of the stock version. How bad is that transistor mute? (sound vice)

 

[dmills]

That chip is voltage output so you do not need a I/V stage.

I never liked that cross coupled output stage, it asks a lot of the resistor matching and it is far from clear to me that it actually helps anything (And stability can become marginal into some loads).
Pin one of the XLR should also be connected directly and locally to chassis ground, not to the analogue ground connection.

I don't normally bother with a power on mute, amps should only be powered on after the processing anyway, but a relay shorting the output would be the better way if you wanted to go there.

Given this is an output on XLR I would like to see some protection against P48 accidents!

Regards, Dan.

 

[Moby]

That chip is voltage output so you do not need a I/V stage.

Yes, it's probably done under the chip, so than I need buffer stage with ratio 1:4. It needs to be balanced of course  Do you have some recommendation about topology?

I never liked that cross coupled output stage, it asks a lot of the resistor matching and it is far from clear to me that it actually helps anything (And stability can become marginal into some loads).
Pin one of the XLR should also be connected directly and locally to chassis ground, not to the analogue ground connection.

Cool, I posted the stock schemo just to show why it need to be modded ;-)

I don't normally bother with a power on mute, amps should only be powered on after the processing anyway, but a relay shorting the output would be the better way if you wanted to go there.

Ys, but I don't want to think about possibilities when somebody turn of the unit in the middle of the session or something like this. So I must have mute option. Than I can go with transistor mute but driving relay? Am I right? Small spike from relay on-off is not an issue  ;D

Given this is an output on XLR I would like to see some protection against P48 accidents!

What the Fck is P48 accident  ???

 

[dmills]

Given that the DAC is voltage output, the circuit around IC1A looks fine to me. I might choose a classier opamp, but I don't see anything much wrong with that stage.
AC Coupling to the following stage might be a good plan however as the drawing will pass any DC offset from the DAC to the output.


A P48 accident is when someone plugs the output of this thing into a mixing desk and accidentally enables the 48V phantom power on that channel....
Particularly with the thing powered down, pushing that back into the opamp inputs is not going to do anything any favours.

Regards, Dan.

 

[Moby]

Guys, is it possible to use opa1632 for D/A output stage to? Also, what for is enable (pin7)? Is it possible to use it for on/off mute?