DCX2496 audio mod

[Moby]

Here is the preliminary version based on suggestions before. Free for criticism  ;D

For Comm mode choke I decided to go with http://www.we-online.com/katalog/de/we/katalog/index.php?language=en&key=Inductors%2FCommon_Mode_Chokes%2FTHT%2FWE-VB
To be honest I'm not sure is it appropriate, also I'm not sure which value to go with  :-\

 

[Harpo]

Here is the preliminary version based on suggestions before. Free for criticism  ;D

Mounting holes? 
Maybe you increase C14 and C15 from 82pF to 270pF. Phase lead compensation @720kHz looks a little on the high side and a -6dB rolloff @220kHz won't damage your audio. YMMV.

 

[Moby]

Mounting holes?  Wink

Thanks for suggestion. I will probably add one hole but I planned that XLR will hold it. I think It's light enough 
Give just a minute to do a sim about C14,C15. I think that that I decided value 82p because of phase response.

 

[Moby]

Here we go...
82p

270p

Not a big deal, maybe U right 

 

[Harpo]

why did you move your clamping diodes @ ADC input before the -10dB U-Pad (the 2x 332 series / 332R shunt with the cap across)?
This way the max. 5Vpp will only be max. 1.67Vpp @ the ADCs input.
Maybe missing something.

 

[Moby]

why did you move your clamping diodes @ ADC input before the -10dB U-Pad (the 2x 332 series / 332R shunt with the cap across)?
This way the max. 5Vpp will only be max. 1.67Vpp @ the ADCs input.
Maybe missing something.

Good question... 2x332's /332+cap are on the DSP/ADC board. Maybe to remove them and to put on this board? Not sure? What you think about it? Also, what do you think about phase response with 270p?

 

[jdbakker]

Here is the preliminary version based on suggestions before. Free for criticism  ;D

What Harpo said.

Besides:

- move the two series resistors (R10A/B on the Beh schematic) to your board, lower them to 68...100Ohm
- remove the shunt resistor (Beh R6); a pad in that location will cost you more in distortion than it gains in noise performance
- if you lower R10A/B, raise Beh C31. Agree with Dan: replace it with a C0G or other high-quality cap
- connect R5/R6 the way you did in your earlier schematic
- most op-amps, including the OPA2604, have much more distortion in non-inverting mode than in inverting mode. Change them to inverting, possibly with a slight loss (~6dB) to allow for larger input swings
- move C1/C2 before the OPA2604s. The small DC offset of the 2604s will most likely not matter much. Better yet, replace the OPA2604 with a proper line receiver like the THAT12xx series.
- the protection diodes D1-D4 must be on the same board as the converter, with minimal trace length to the converter's power and ground lines. Your IC3 is likely a standard linear regulator; these can't sink current, so there's nothing stopping the OPA1632 from pulling IC3's output up through the protection diodes.
- set your total system gain (or more likely loss) to adapt expected line levels to your converter's input. With the values you have it looks like the converter clips when the input is ~20dBu, does that offer you enough headroom?
- the OPA1632 runs pretty hot at +/-15V; consider running it at a lower voltage.

About your PCB:
- the decoupling capacitors are much too far from the OPA1632. Use SMD ceramics; try to place them as close to the chip as possible.
- have a contiguous ground plane on one side of the board
- I would suggest also using a good (=C0G) SMD cap for C14/C15

JDB.

 

[radiance]

Moby, how would this pcb fit inside the existing Behringer case? Is there enough room inside the DCX? Or are you going to cut away the input section from the original  pcb?

I don't think this pcb will fit inside my DEQ2496 without removing some of the original pcb...

Or are you planning to do an new output section as well in which case we could leave the original input/output pcb intirely 

 

[Harpo]

Also, what do you think about phase response with 270p?

For usual you want the phase response in the audible range (20Hz - 20kHz) not exceed +/-5°. YMMV.
Rolloff frequency for this lpf is 1/(2*pi()*R*C) with R in ohm and C in farad.
Phase response in degree is ARCTAN(testfrequency/lpf)*180/pi()   where testfrequency of interest for usual is 20kHz and lpf is the rolloff calculation from the line above.
The loss in dB @ testfrequency is log10(1/(squareroot(1+(testfrequency/lpf)^2)))*20.

Just running the varying numbers of interest for you (82pF, 220pF, 240pF, 270pF)

R in ohm	C in pF	lpf in Hz	testfreq.in Hz	dB loss	phase err.°
2,700	82	718,857	20,000	0.0034	1.59
2,700	220	267,938	20,000	0.0241	4.27
2,700	240	245,609	20,000	0.0287	4.66
2,700	270	218,320	20,000	0.0363	5.23

So a 240pF might be a better fit on paper than the 270pF. I probably won't notice the difference.
(At least now I know how these table icons operate )

 

[Moby]

Thanks guys you are really great  All comments are welcome 

move the two series resistors (R10A/B on the Beh schematic) to your board, lower them to 68...100Ohm
- remove the shunt resistor (Beh R6); a pad in that location will cost you more in distortion than it gains in noise performance
- if you lower R10A/B, raise Beh C31. Agree with Dan: replace it with a C0G or other high-quality cap
- connect R5/R6 the way you did in your earlier schematic

Done  ;D

most op-amps, including the OPA2604, have much more distortion in non-inverting mode than in inverting mode. Change them to inverting, possibly with a slight loss (~6dB) to allow for larger input swings
- move C1/C2 before the OPA2604s. The small DC offset of the 2604s will most likely not matter much. Better yet, replace the OPA2604 with a proper line receiver like the THAT12xx series.

Cool, that's a great idea. I think that 10k inverting mode is OK. Am I right? More than that will be bad noise vice, less too hard to drive from passive attenuator 
Regarding THAT chips I would like to avoid that, too hard to find in my area  At least I can swap 2604's for something better one day 

- the protection diodes D1-D4 must be on the same board as the converter, with minimal trace length to the converter's power and ground lines. Your IC3 is likely a standard linear regulator; these can't sink current, so there's nothing stopping the OPA1632 from pulling IC3's output up through the protection diodes.

Unfortunately there is no way to mount diodes on the ADC board  :-\ Behrngrr did two separate boards for audio and DSP/ADC and only thing I can do is to put complete new audio board uder existing DSP board (new 1HE box). Leads will be around 5cm's. Yes I know that it's not perfect but there is no better way. Maybe if I reorganize complete unit in new case (one day) audio input board will be closer, say 1-2cm to ADC... Hmmm, yes, I decided to go with small package 7805 but no problem to change it to TO220.I hope that will work better.

- set your total system gain (or more likely loss) to adapt expected line levels to your converter's input. With the values you have it looks like the converter clips when the input is ~20dBu, does that offer you enough headroom?

This unit (crossover )  i s connected after monitor volume control. No way that I will touch 20dbu. Maybe accidentally but that's the reason I decided to put diodes.

About your PCB:
- the decoupling capacitors are much too far from the OPA1632. Use SMD ceramics; try to place them as close to the chip as possible.
- have a contiguous ground plane on one side of the board
- I would suggest also using a good (=C0G) SMD cap for C14/C15

Cool about this too  .
I still don't know about common mode choke. Is that critical part ? Or let me ask this way... Which one to choose  ???

Moby, how would this pcb fit inside the existing Behringer case? Is there enough room inside the DCX? Or are you going to cut away the input section from the original  pcb?

I don't think this pcb will fit inside my DEQ2496 without removing some of the original pcb...

As I mentioned before I will mount this in separate box or put complete unit in new box Yes, output section is my second thing to do. I will continue to bother you about  8)
Harpo, thanks for calculations 

 

[Moby]

jdbakker, if I raise the C31 and remove R6 I have ugly transient square response.. If I lower the C31 to 820p square (10k) looks much nicer. Any logical explanation?

 

[dmills]

Further on the PCB:

Pin one on the XLR should be returned via a short connection to case ground (PE), not to the circuit reference (Pin one problem).

A common mode choke on its own does you little good, you also need to shunt the output side to PE (not circuit reference ground) with a few hundred pF (This is something else that those THAT parts make easier to do right).
Are you providing any input transient protection? Even a hundred ohms in series with each input will help (And will kill the Q of the CM coke and cap).

C2 does need to be CLOSE to the ADC chip, the currents here are RF not audio!

I still hate that input stage, there are better two opamp designs available  http://www.dself.dsl.pipex.com/ampins/balanced/balfig13.gif for example. This way you do not rely on a handful of matched parts for your CMRR. A THAT 1203 is still by far the better way to do this however.

Why the dc bias trimmer? Most ADC parts have a HPF in the digital domain that can be enabled that will remove trivial DC bias. Good design eliminates trimmers as far as possible.

HTH.

Regards, Dan.

 

[jdbakker]

Regarding THAT chips I would like to avoid that, too hard to find in my area 

Fair enough, but the OPA1632 isn't exactly easy to get either. BTW, THAT chips are available from our friend Wayne, he ships worldwide. And Mouser has free/cheap shipping to Europe for orders over ~60EUR these days, I believe.

Maybe if I reorganize complete unit in new case (one day) audio input board will be closer, say 1-2cm to ADC... Hmmm, yes, I decided to go with small package 7805 but no problem to change it to TO220.I hope that will work better.

That makes no difference at all. Look at the internal schematic of the 7805 (or the 317, or almost any linear series regulator). You'll see that there's just a pass transistor between input and output. All the regulator can do is allow more or less current to flow from input to output if the output voltage is too low. It has no way to make current flow to ground if the output is pulled too high, which is exactly what happens when D1/D2 start conducting.

Another reason to connect D1/D2 to the ADC's supply is that maximum input voltage is defined relative to supply level.

I still don't know about common mode choke. Is that critical part ?

Only if your input wires are long, your equipment is in a high-RFI environment (think broadcast transmitters) and/or the preceding equipment outputs a lot of HF common-mode junk. I believe Samuel Groner had a discussion on CM chokes for mic preamp inputs somewhere; can't find it right now.

jdbakker, if I raise the C31 and remove R6 I have ugly transient square response.. If I lower the C31 to 820p square (10k) looks much nicer. Any logical explanation?

What's your definition of ugly?
Is this in the sim or in a real circuit?
Can you show the circuit that's giving the 'ugly' behaviour? Between all the versions and modifications that have been tossed around I've lost track.

JDB.

 

[Moby]

Thanks again guys  I will include all suggestions and draw a new schemo. Regarding 1632, I have few samples around. Ordering from rest of the word from SErbia? You must be kidding. Well It's possible but last time I tried it turned in to nightmare because of money transfer. But that's another story. I will try with Wayne and prepare two different boards. Opamp http://www.dself.dsl.pipex.com/ampins/balanced/balfig13.gif  and THAT....  ;D

 

[Moby]

Guys, what you think about diffamp similar to this, driving 1632. No way I can order THAT's  :-\

I also did simulation for "C31". Ugly on pic maybe the real world change something . Not photogenic one 

 

[Igor]

I absolutely agree with jdbakker. I'd use caps before opa2604's, in unity gain (inverting or not) connection.
Don't see any reason to use THAT receiver with OPA1632, 1632 is already balanced.
Regarding CMRR, just use good MF resistors.
Even 1% DALE from same box will give better CMRR than THAT receivers-it prooved on m/s matrix
(app where perfect balance is more important than ot line receiver placed after short <5m cable).
My $0.02

 

[Moby]

I absolutely agree with jdbakker. I'd use caps before opa2604's, in unity gain (inverting or not) connection.
Don't see any reason to use THAT receiver with OPA1632, 1632 is already balanced.
Regarding CMRR, just use good MF resistors.
Even 1% DALE from same box will give better CMRR than THAT receivers-it prooved on m/s matrix
(app where perfect balance is more important than ot line receiver placed after short <5m cable).
My $0.02

Cool than. I believe that 2 THAT's configured in diff mode can sound good but I hope an believe that 2604's will do the job too. At least it's worth of trying and it's gonna be better than stock configuration  Just not sure about "(inverting or not)". Help me to decide  ;D
I already decided to exclude diodes from this board, I will mount them together with cap on some piggy board few mm from ADC and steal the +5V as suggested from ADC. Also, I'm not sure do I have to mount series resistors from 1632 to the same (piggy)? The rest of details like COG caps, close position of PSU ceramics are clear 

 

[dmills]

Keep the rest on the analogue board, the high speed stuff is the loop formed by that cap and the ADC chip. You want to keep the feedback connections around the driver short and dont forget to twist the leads between the analogue and digital boards.

I guess my experience is probably biased by live sound and broadcast day jobs where 100M balanced lines are common, sometimes with ten volts or so of common mode RF. That kind of thing makes you real sensitive to anything that helps with noise rejection, and I found that a suitably protected 1203 plus starquad cables was about as good as anything out there (A good transformer with interwinding screen is better, but this is cheaper).

With short cables almost anything will work, but near a radar site, with long lines you separate the stuff that is good from the stuff that is prosumer (And some kit that should know better fails miserably, a certain expensive mic manufacturer springs to mind - really stupid pin one problems).

Regards, Dan.

 

[jdbakker]

I also did simulation for "C31". Ugly on pic maybe the real world change something . Not photogenic one 

Please show the entire circuit. If I had to guess what's happening I'd say that for the larger C31 values the bypass caps between the output and the input of the OPA1632 are too small, and you get ringing comparable to an opamp with too much capacitive loading.

[...] 2 THAT's configured in diff mode [...]

While you could do that, I would just go with a single THAT line receiver driving one of the input resistors of the OPA1632 with the other one tied to ground. Noise performance will likely be better that way around.

Just not sure about "(inverting or not)". Help me to decide  ;D

Most op-amps have highest distortion in non-inverting low gain configurations; read Samuel Groner's Opamp Measurement Series, especially the parts about common-mode distortion, if you want a second opinion on that. Admittedly inverting topologies have somewhat higher source gain, but with 5532-class opamps or better the discussed resistor values will dominate noise performance anyway.

JDB.

 

[Moby]

Here's the circuit i used for simulating. I have to note that it's not the final circuit but problem with ringing remains whatever I do...