DCX2496 audio mod

[Andy Peters]

I'm asking myself the same thing, but must be because of filter at the end of first stage. It looks like some kind of pre-emphasis filter or something....I wonder is it possible to avoid this filter and also to add some more current capable amp to drive 332+332ohnm terminated input of AD5393  :?:

The 332-ohm resistors aren't "terminations" that require any significant current source. Look again at the schematic -- those resistors are in series with the driving op-amp, not a load to ground!

These resistors help isolate the driving op-amp from the capacitive load presented by the ADC and the input load cap (the cap across the ADC inputs).  Reducing them could be bad if the op-amp can't handle the load. Increasing them beyond some value will increase setting time.

So the resistors aren't the problem -- the capacitance is.

-a

 

[Moby]

Driving the inputs beyond the power-supply rails is never a good idea.

Yes I know that, but what are circumstances in this case? Burned ADC or just clipping? Again, I doubt that I will ever feed the input with more than +/-2V pp but ... I see that passive output stage is used often in some of upgrades but I never saw direct ADC driving.
Regarding "termination, yes, I see series resistance , but what about R5,R7 across AIN+/-?

 

[Igor]

Hi Moby!
What is DCX2496?
From a glance at posted schemo/pcb layout:
+/-  op-amps supply traces seems need to be fatter. Use 0.8 mm trace at least.
opa1632 is the best substitute to good 4:1 balancing input transformer.
If circuit on 2 op-amps used at the input, to get proper CM and good balancing,
0.1% MF resistors are needed.
For some dog-ear'ed persons, SMD resistors are
adding color in precision circuits like this.
Input filter's corner minds A LOT for transient responce.
Silver mica/polyprop caps preferred by dog-ear'ed testers in these circuits as well.
Proper designed input trunny is the best winner.
Just my $0.02....

 

[Moby]

What is DCX2496?

It's a DSP crossover, cheap Behngrr , 3 analog inputs. 6 analog outputs ... http://www.zzounds.com/item--BEHDCX2496
I'm using that to cross my main studio monitors. Stock unit sounds really poor, but hearth of the system (DSP and algorithm) are cool.Analog stage sucks, but many people are happy with upgrading with opa2134 and changing cheap electrolytic with Elna's or similar audio caps. As I mentioned before I'm not desperate with the overall sound but I would like to go step further.
So Igor U vote for TX passive approach? I tried that but with OEP's and there were a lot of coloration so I gave up. I didn't tried Lndahls , but I'm afraid that will be just a bit more straight.  ???
1632 was my first option but I heard from people who tried that they had overheating problem. 
Hmmm, so you think that input buffer stage is a must? No way to feed it directly to ADC?

 

[Igor]

AAAAA! If the purpose is studio monitor's crossover....
No trunny which doesn't adding color.
But good trunny sounds nice!
OEP's are crap trunny's. Good for low-mid saturation and adding some phazzzshhhhft on highzzz.
I don't think it is best to use trunny's in monitoring soundpath.
For input ADC, when recording...yes!!!!
Try Jenssen's A/D converter input trafo's, JT-6110K-B for xample.
Despite Jenssen winding them on E-I lams (???!!!), they absolutely symmetric, clean,
have single easy-to-dump ( ) overshoot, and surprising by sound.
If 31267 had twice less turns on secondary's!!!!!
(hm, if grandma had the balls, she were grandfa...anyway

IMHO the best for your situation will be OPA1632.
I never had probs with overheatting of them running on +/-15 V, not more.
Connection is exactly like appnote says. VERY clean.

Direct feed to ADC is no good.
Most AD chips needs at input: low output imp., short wires, kinda filter and CM offset.
Trunny or 1632 doing the best job IMHO.

 

[Moby]

Thanks Igor. Yes I know about OEP's .. price tells enough . On the other side Jensen is great but it will probably add some "nice" coloration for a lot of money  ;D Cool, 1632 is my way to go 

 

[Moby]

Ok than  I would really like to redesign this analog input stage of the AK5393 ADC because stock one looks and preforms really ugly.
After lot of reading, lurking and messing about in other threads I found that I need some help to design this.
What I learned so far?
1. ADC input buffer need to be capable to drive some current (around 40ma in this situation)
2.It have to be clean as possible from coloration because I need this for studio monitoring.
3.ADC need to be protected from over-voltage by clamping diodes (Shottky)
4.Diodes have to be chosen (low capacitance) say, BAT82
5.OPA1632 is best candidate for ADC buffer but it has to be low input impedance because of low noise performance.
6.OPA1632 needs additional input to achieve reasonable high input impedance , say 10K (I still don't know how it should look)

Of course, PCB has to be designed carefully , probably the SMD is best way to go.

Did I forgot something?

 

[Igor]

Hi Moby, all right, at #6: just use schem. from datasheet of opa1632, page 6.
Here, 1K and 270R feedback resistors used. It is possible to add just unity gain buffers at the input,
opa2604 will do the job perfect.

 

[Moby]

Thanks Igy  ;D I will draw somethig going in that direction 

 

[radiance]

Still have not started on the first "opa2134" mod for the DEQ, so I'm very interested....

So Moby, you've done the "opa2134" mod and you just did not like the outcome of it?

Did you try the LM4562 on the input as well??

 

[mhelin]

Example of OPA1632 application here by one of the WinISD guys, he is a perfectionist PCB designer (designs all stuff like it were a RF circuit, and never separates digital and audio grounds), I've got one 4 layer PCB he designed for a 8 ch remote volume controller (PGA2320) which I'll use with DCX:

http://koti.mbnet.fi/jahonen/Audio/DIY/DSP56371TestBoard/DSP56371TestBoard_Schematics.pdf (the xover)
http://koti.mbnet.fi/jahonen/Audio/DIY/8%20Channel%20Volume/ (8 ch vol controller)

That design (if you browse the web page given in the link) would kill DCX easily. I think he didn't use the DSP for much as the design includes the Altera Cyclone II FPGA which was used to implement the FIR xover filters.  I guess he may still have couple of PCB's left for the project (or at least gerbers) in case you might want to build that one instead of modding the DCX.

 

[jdbakker]

http://koti.mbnet.fi/jahonen/Audio/DIY/DSP56371TestBoard/DSP56371TestBoard_Schematics.pdf (the xover)

That design (if you browse the web page given in the link) would kill DCX easily.

No argument there.

It doesn't look entirely optimal, though. All those OPA2134s in unity-gain configuration will introduce rather more CM-induced distortion than necessary, especially given that the DAC-out buffers could easily have been inverting. It's a pity that the converter clocks are routed through the DSP and the FPGA, as that will introduce quite a bit of jitter, especially considering that both the DSP and the FPGA are in a QFP-package which has higher power/ground lead inductance. It would have been nice if the switching power controller and the microcontroller were run synchronously with the sample clock, too.

Don't get me wrong, it looks like an interesting design, and it will surely outperform a DCX2496. All I'm saying is that with the same parts (plus a few logic gates to build a clean clock distribution network) the performance could have been better.

JDB.
[a big fan of routing boards pretending everything's RF]

 

[Moby]

Still have not started on the first "opa2134" mod for the DEQ, so I'm very interested....

So Moby, you've done the "opa2134" mod and you just did not like the outcome of it?

Did you try the LM4562 on the input as well??

OPA2134's are performing better than stock but they are getting hot and simply they are not capable to drive ADC in original topology. I lost my eyes desoldering and soldering that small bastards so I'm sure that i will never  do the same thing again. At last, it doesn't sound good enough  I was in the hurry to finish my main speakers and I wanted to go fast as possible so that was the reason of that first upgrade. To be honest I was bit suspicious ... It doesn't sound horrible, just not good enough. I didn't tried LM's it's possible that they will preform better but as I mentioned before designing from scratch is more elegant way 
Mihelin , thanks for sharing the schemos. It looks interesting . I wonder about opa134 input buffer in unity gain. How good is that? About output stage... hmmm I'm not blow'd away 

 

[Igor]

(designs all stuff like it were a RF circuit, and never separates digital and audio grounds)

LOL.... It was good!

 

[Moby]

OK, something like this? I'm not sure do I need c5/c6 and not sure about c3/c4 value...

 

[dmills]

Operating U1 and U4 as simple non inverting buffers means that the matching of R1,R2, R8,R9 C3,C4 & C5,C6 are critical to the common mode rejection, and with 100u and 1K the 3db point is 1.6Hz, but at 50Hz you will still have enough phase shift to ruin CMRR if the caps are at opposite ends of the tolerance range.

The inputs to U1 and U2 need some protection (both transient and RFI) and I would add a common mode choke.
I don't see the need for Z1.
What is R6 supposed to do? The low frequency feedback is taken from this point anyway so it will not effect the gain at low frequency.
I would make R5 and R7 much larger to raise the corner frequency of the reference filter.

As long as you can tolerate the bias currents flowing in the feedback network, you can probably do away with C5 and C6 provided that any resulting DC offset can be filtered out in the software you are using, remember that this bias current will also increase the heating in the chip.

C2 really needs to be COG or similar dielectric and needs to be close to the ADC pins, value should be as per the data sheet for the ADC in use.

You will want to decouple the power supply pins well, some of those parts are fast.
One advantage of the 1632 is that you will not tend to get large ground currents flowing as most of the current will be rail to as C2 is driven differentially.

As drawn the circuit is inverting with respect to the polarity shown on the signal source.

Regards, Dan.

 

[Moby]

Thanks a lot for comments and helping. I tried to improve the thing with your suggestions, so let's see 

Operating U1 and U4 as simple non inverting buffers means that the matching of R1,R2, R8,R9 C3,C4 & C5,C6 are critical to the common mode rejection, and with 100u and 1K the 3db point is 1.6Hz, but at 50Hz you will still have enough phase shift to ruin CMRR if the caps are at opposite ends of the tolerance range.

Yes, I know about 100u is a bit low but to achieve a good phase it should be at least 1000uf  Now, I wonder how it gonna sound? Unfortunately , no way for film cap in that value. On the other hand BB suggest 1k or less at input of 1632 because of low noise performance. Personally I will be much happier with 10K so the input buffer can drive it and the cap will be much smaller. Not sure about  :-\  here is the simulation of both situation, total noise... What you think about it?

As long as you can tolerate the bias currents flowing in the feedback network, you can probably do away with C5 and C6 provided that any resulting DC offset can be filtered out in the software you are using, remember that this bias current will also increase the heating in the chip.

I'm not afraid of the input DC offset, I can keep that controlled or to put cap in front of buffer but I'm not sure is it problem for buffer op--amps to have that bias on the output. Heat? Hmm, yes, that can be a problem. How about sound vice?

What is R6 supposed to do? The low frequency feedback is taken from this point anyway so it will not effect the gain at low frequency.
I would make R5 and R7 much larger to raise the corner frequency of the reference filter.

R3,4,6 and C2 are on the other (DSP board) mounted close to the ADC chip. I really don't know do I need R6 and I don't know why is there  ??? If I remove it hf response is changing so I suppose it's lowering the input impedance of the ADC. Am I wrong? Yes. R5,6 have to be larger, I will change them for, say, 22k 22k or something 

You will want to decouple the power supply pins well, some of those parts are fast.
One advantage of the 1632 is that you will not tend to get large ground currents flowing as most of the current will be rail to as C2 is driven differentially.

Cool, I wanted to make this preliminary circuit simpler , yes I will add 10U+100n on the PCB design 
Thanks again, and please add some more comment if you have some time  Also, can you help me about choosing right one common mode choke? Something like Epcos B82796C0/S0? Good?

 

[Igor]

I used 1632's with 5k resistors on input as well... You have to re-calculate other R/C values in this case.
BTW schem shows R3, R4 as 332...strange.

 

[Moby]

Thanks Igor  . Than, if I go with 5k on input of 1632 do I need buffer in front ? 5k is OK to drive with your Passive relay controller?

BTW schem shows R3, R4 as 332...strange.

Look at the 1. page of this Topics. There is the Bhrngrr schemo of the ADC. You will see the 332 resistors...

 

[Moby]

About common mode chokes again... I see a lot of widely available but have no clue what I'm looking for  :-\ For example, would any of these work ?
http://www.we-online.com/katalog/de/we/katalog/index.php?language=en&key=Inductors/Common_Mode_Chokes