Behringer DCX2496 digital X-over - Page 291 - diyAudio
Go Back   Home > Forums > Source & Line > Digital Line Level

Digital Line Level DACs, Digital Crossovers, Equalizers, etc.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 8th January 2011, 09:56 PM   #2901
diyAudio Member
 
Join Date: Jan 2007
Quote:
Originally Posted by sek View Post
Yes, I'm open for other ideas. Unfortunately I don't know the DACmagic from the insides. Could you point me towards schematics or an analysis.
Schematic attached. If you reuse the structure, you will need two of those, each connected to both inputs, but with opposite polarity. The ground level of the auxiliary op-amps can be changed, for example to match input offsets in the OPA1632. I propose increasing the resistor values to have higher input impedance and less loading on the main opamp.

Quote:
Well, maybe I'm just a little bit reluctant to use the fully differential amp as the first (input) and only (anti-aliasing filter) stage. But as I see from a lot of evaluation boards and commercial products, there seems to be no reason to be afraid of doing so. That's why I made a modification to the schematic (see attachment).
Well, if you don't have any input buffering, the OPA1632 will see the input common-mode. Especially if the source is distorted (some pro gear might expect the next stage to be very immune to common-mode) or the XLR cable picks something up, this might affect you. Also, if your input is single-ended, this might affect sound quality badly. There are enough eval boards and commercial products with bad implementations, so I advise you to go with your own common sense instead of looking at what others do.

--
Greetz,
MatchASM
Attached Images
File Type: png op275stage.png (37.5 KB, 336 views)
  Reply With Quote
Old 9th January 2011, 12:42 AM   #2902
sek is offline sek  Germany
diyAudio Member
 
sek's Avatar
 
Join Date: May 2003
Location: Berlin
Quote:
Originally Posted by MatchASM View Post
Well, if you don't have any input buffering, the OPA1632 will see the input common-mode.
Of course, that's the job (and struggle) for any input stage. The main question is wether the OPA1632 is up to it - and I guess it is. Let me explain:

If I do have input buffering, the buffers are the ones that get to see the input common-mode. It's not obvious to me as to why shifting the problem to the prior stage would make solving it any easier. It's still common-mode, the buffers are still opamps, still having limited specs.

The OPA1632 is specified for a common mode input voltage range of -12.5V to +14V for a +/-15V supply voltage. That's half of what you get with a high impedance buffer, still it's way more than what most sources present. I could live with that, especially considering that even instrumentation amplifiers and differential receivers don't surpass this. As the input is protected (clamped) internally, I don't see any risk from common mode input, just possible distortion from overloading the diff stage. If in doubt, just clamp with additional zeners.

Quote:
Especially if the source is distorted
If the source is distorted, then there's no hope to regain the original signal. For any type of input stage, thus I guess this is not what you meant. In case you refer to unwanted DC bias due to asymmetric distortion I don't see how the resulting common mode voltage could surpass the source's supply voltage. Within the +/-12V range the OPA1632 will damp this down by about 90dB (resulting in ~0.4mV output common-mode), above this range the clamps will just pass on the distortion.

Quote:
(some pro gear might expect the next stage to be very immune to common-mode)
I don't intend to use pro gear that shows such behaviour, nor do I see the necessity to protect against any and all problems imaginable.

The only situation where this type of problem arises is with high source output impedance (like instrument effects gear or DI boxes), a property that isn't to be expected in quality line level interconnections.

Besides this, gear that delivers ill defined common mode behaviour can't be expected to deliver good noise immunity at the same time, thus sound would be degraded anyway.

It's usually the (truly) proven pro gear that is actually free of such symptoms, i.e. any transformer coupled output. Even low budget gear incorporates regular low-impedance drivers (like opamps) and balanced output circuitry (like in Behringer's DCX circuit itself, similar to Rod Elliot's Balanced Line Driver, fig. 2).

Quote:
or the XLR cable picks something up
The balanced cable run picking up common mode voltage is to be expected! That's why balanced runs are preferred, as they (ideally would) make any noise pickup common mode instead of differential.

I still don't see the difficulties (with common mode input voltage) that an instrumentation amplifier would not have at or around unity gain...

Quote:
Also, if your input is single-ended, this might affect sound quality badly.
True, but only because the circuit is DC coupled, as any DC component that is imposed on the input signal would now be amplified and passed through. The most widespread solution against this problem is capacitor coupling at the input, I'd rather prefer a high quality blocking capaciter after the input stage. My quality recommendation would be transformer coupling and proper cabling, although capacitor coupling is a lot easier and cheaper. See Jensen notes AN003 and AN004 as well as Rane Note 151, fig. 5 for further details.

I'd also assume that using the OPA1632 as a balanced input stage is similar to choosing any differential to single ended input receiver like INA134, a widely accepted choice in this type of application.

Quote:
There are enough eval boards and commercial products with bad implementations, so I advise you to go with your own common sense instead of looking at what others do.
Understood.

But I still don't see the culprit with common mode distortion in this application. I'm aware of the reduced impedance level at the inputs, but it's still appropriate for the intended application.

Please also keep in mind that the DAC input is fully balanced with it's own CMMR specs, i.e. it has an internal high pass filter that elminiates any DC offset within it's own common-mode range.

Quote:
Schematic attached.
I perceive it as a differential to single ended converter with servo balanced biasing and capacitor output AC coupling.

Quote:
If you reuse the structure, you will need two of those, each connected to both inputs, but with opposite polarity.
My idea was to reduce complexity, not increase it...

Now, applying two of them to the input of a fully balanced opamp would theoretically increase CMMR with the added benefit of offset control and AC coupling. Interresting idea, ideally the common-mode reduction ratios of the buffers and the balanced amp would add, as the balanced amp only sees the common-mode output of the buffers - which is the input common mode attenuated by their CMMR.

It's practical application looks difficult, though. The frequency compensation and transfer function doesn't look suitable to leveraging the OPA1632 in the next stage, also complicating calculation and making the whole thing more prone to component quality and tolerances (as imbalances in the two buffer circuits' transfer functions will greatly deteriorate CMMR of the following stage, probably beyond the point of making the gained CMMR futile).

Take a look at the attached schematic. It's Behringers original differential input buffer and AC coupling. That's basically your proposal minus the actively driven feedback loop (and plus rubbish components). Cloning this stage and reversing the input feeds towards the pair would theoretically yeld the same CMRR increase, yet it's never used. I can only explain this observation with difficulties that likely come with mirroring it's behaviour.

Lastly, that Vocm capability of the OPA1632 is a feature that renders control of absolute offset voltage unnececcary, as the output will be elevated to the (constant, input independent) ADC bias level anyway. Also, I don't see the need for input or output offset cancellation in this balanced, DC eliminating ADC drive application. Can you please elaborate? I might as well be wrong on this, it's been a weekend full of hard work at work.

Cheers,
Sebastian.
Attached Images
File Type: png original.png (35.3 KB, 333 views)
  Reply With Quote
Old 9th January 2011, 07:48 AM   #2903
qusp is offline qusp  Australia
diyAudio Member
 
qusp's Avatar
 
Join Date: Oct 2009
Location: Brisbane, Australia
I'm glad you finally got around to mentioning the VCOM pin, ; a fantastic feature that solves most of what you are talking about previous to that point as that is the whole reason its there. provided errors are common mode, which given a reasonably well realized diff cct is entirely possible; the result will be very acceptable. I love this chip along with others in its family, the SQ is as good as it is handy IMO. also if you use the package with the pad on the underside the current capability is quite high

Last edited by qusp; 9th January 2011 at 07:51 AM.
  Reply With Quote
Old 9th January 2011, 08:43 AM   #2904
sek is offline sek  Germany
diyAudio Member
 
sek's Avatar
 
Join Date: May 2003
Location: Berlin
Quote:
Originally Posted by qusp View Post
I'm glad you finally got around to mentioning the VCOM pin
Only now did I mention it, but look how I prepared to use it from the beginning by extending the signal from the pin in the schematic.

It's the "VOCM" pin, btw.: output common-mode.
The fun property is that the Vocm input voltage replaces the output DC component of the (differential) input signal, independent of gain and frequency compensation. The block diagram (see attachment) shows that this is actually realized by something like a tapped servo loop...

Quote:
provided errors are common mode, which given a reasonably well realized diff cct is entirely possible; the result will be very acceptable.
That's also my reason for choosing it. It sounds like you know it, then. How would you solve the input stage? What would be reasonably well realized according to you?
Attached Images
File Type: png THS4130-block.png (52.4 KB, 315 views)
  Reply With Quote
Old 9th January 2011, 10:23 AM   #2905
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
is that the genius circuit?
__________________
regards Andrew T.
  Reply With Quote
Old 9th January 2011, 10:39 AM   #2906
sek is offline sek  Germany
diyAudio Member
 
sek's Avatar
 
Join Date: May 2003
Location: Berlin


It's what it says it is, the simplified block diagram of the THS4031.

The similarities between the THS403x series (generic ADC driver) and the OPA1632 (specified for audio) are no secret, AFAIK. At least not since TI got into the DIY news when licensing parts of it from Nelson Pass.

It's not the holy grail of audio nirvana, but it gets the job done pretty fine I suppose...

PS: Andrew, I'm really interrested in your input regarding my buffer stage uncertainty.

Last edited by sek; 9th January 2011 at 10:44 AM.
  Reply With Quote
Old 9th January 2011, 10:58 AM   #2907
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
I am starting to refer to balanced connections as balanced impedance connections.
Many Members do not realise that the advantages of balanced are thrown away if the send and receive impedances are not balanced.

The circuits showing +IN & -IN with different receive impedances are not balanced impedance.

Some believe that common mode impedance does not need to be matched, but I can't get my head around the physics of that stance.
__________________
regards Andrew T.
  Reply With Quote
Old 9th January 2011, 12:12 PM   #2908
qusp is offline qusp  Australia
diyAudio Member
 
qusp's Avatar
 
Join Date: Oct 2009
Location: Brisbane, Australia
correct I stuffed up the acronym ;D I admit I didnt start far back, I just read MatchASM's long post and then yours addressing his and was surprised at how long it took to be mentioned considering it makes much of the discussion irrelevant. i'm no expert at design, but I have a fair bit of experience with this chip and its brethren such as THS4151 (from memory, so many numbers in my little brain) in other peoples designs.

these chips are very similar to the pass susy designs, in fact Ti had to pay Nelson some dollars for using his patent so to speak (i'm not privy to the details) and its this servo that is at the heart of it, so you might have a look for some of these discrete designs for inspiration.

I would not presume to be able to tell you how to design your cct, i'm only just starting to play around with my own designs now and up till now have relied on the brains of others to create them, only exercising judgement on what I feel works best for my purpose and taste, mixing and matching, often modding. I dont know much about the design of this ADC, or ADCs in general and have only used these as dac output stages, despite them actually being designed for ADC. I would look at what AVCC/2 is for your ADC units, do they operate on a common mode level that is half the supply voltage and if so, sample that voltage and drive it into VOCM pin with a quality single instrumentation amp with very low input offset; then let the servo do its work on the output.

as far as the input filters that can be a bit tricky with differential designs due to the tolerance of even high quality caps (resistors arent a problem given enough money, so I would tend to try to do any filtering inside the OPA1632 feedback loop to keep any error to a minimum.

I do understand you correctly here in that you are looking to use this to drive the to drive the INPUT?
  Reply With Quote
Old 9th January 2011, 12:15 PM   #2909
sek is offline sek  Germany
diyAudio Member
 
sek's Avatar
 
Join Date: May 2003
Location: Berlin
Quote:
Originally Posted by AndrewT View Post
I am starting to refer to balanced connections as balanced impedance connections.
I am referring to balanced impedance as an equal impedance as seen from inverting and from non-inverting in- or output.

Quote:
The circuits showing +IN & -IN with different receive impedances are not balanced impedance.
Yep!

BTW, do you see this in any of my schematics attached here?

Quote:
Many Members do not realise that the advantages of balanced are thrown away if the send and receive impedances are not balanced.
Well, some differentiate.

For an interconnection to be fully balanced, the inverting and the non-inverting current paths need to be equal in transmitter output impedance, gain/damping, transmission line frequency response/transfer function and receiver input impedance.

They do not necessarily need to have equal transmitter impedance and receiver impedance, i.e. 50 Ohm output impedance and 50 Ohm receiver input termination, respectively. This would of course be advantageous to high frequency signal transmission (RF) like an AES/EBU digital audio connection, but it's absence does not ruin the gained common-mode reduction due to the symmetry.

I have attached a diagram of input vs. output impedance. It's published by Eberhard Sengpiel on his website, I unfortunately couldn't find an English source for that kind of illustration (and I don't support all opinions of Sengpiel).

It shows voltage (U), current (I) and power (P) against impedance ratio of output impedance (Ra) divided by input impedance (Ri).
The extreme cases are short circuit ("Kurzschluß") at Ra/Ri = 0 and idle ("Leerlauf") at Ra/Ri -> inf.
Inbetween lies the range having maximum power transmission ("Leistungsanpassung").
As can be seen from the logarithmic ordinate, the relationship changes pretty quickly.

Consequentially, fitting a low output impedance transmitter to a high input impedance receiver almost universally makes any audio interconnection a case of "voltage fit" in the attached diagram. It's almost never a power fit and virtually never a short circuit. Those are for transmission lines and power feeds.

As of my understanding, going balanced and going for matched input and output impedances are two entirely independent design decisions, whereas the case Ra = Ri is entirely misplaced in the (analog) audio interconnection world due to established standards and proven best practice.

Cheers,
Sebastian.
Attached Images
File Type: png Ra-Ri.png (194.7 KB, 294 views)
  Reply With Quote
Old 9th January 2011, 12:19 PM   #2910
qusp is offline qusp  Australia
diyAudio Member
 
qusp's Avatar
 
Join Date: Oct 2009
Location: Brisbane, Australia
haha I left the above post half finished while I had something to eat, I see you also mentioned the same pass vs Ti and the THS series ha
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



New To Site? Need Help?

All times are GMT. The time now is 01:19 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2