What is the optimal way to implement a CS3318 based multichannel volume control with balanced signals? Is it...
- use individual channels for (-) and (+) maintaining balanced signal through the chip, but halving number of available channels and doing who knows what weird stuff if paired channels somehow get out of sync either through programming error, glitch, whatever
- use a balanced input receiver and balanced output line driver as I/O buffers, running unbalanced through the chip
- other?
- use individual channels for (-) and (+) maintaining balanced signal through the chip, but halving number of available channels and doing who knows what weird stuff if paired channels somehow get out of sync either through programming error, glitch, whatever
- use a balanced input receiver and balanced output line driver as I/O buffers, running unbalanced through the chip
- other?
Optimum depends on the system, requirements, and preferences in selecting tradeoffs; for a proper answer you'll need to do a THD+N workup or similar on your system.
A starting point which may be helpful is to consider THAT 1200 + CS3318 + THAT 1646 versus differential passthrough on the CS3318. The latter requires roughly 80dB CMRR on receive to render the common mode error residual from the CS3318's 0.1dB channel matching comparable to the linearity of the 1200 and 1646. 80+dB CMRR is difficult to guarantee but the matching error residual is only significant if input and output signal levels are high enough to afford sufficient SnR to expose it.
A starting point which may be helpful is to consider THAT 1200 + CS3318 + THAT 1646 versus differential passthrough on the CS3318. The latter requires roughly 80dB CMRR on receive to render the common mode error residual from the CS3318's 0.1dB channel matching comparable to the linearity of the 1200 and 1646. 80+dB CMRR is difficult to guarantee but the matching error residual is only significant if input and output signal levels are high enough to afford sufficient SnR to expose it.
OK, that is only partially Greek. At least I have some reading material to chew on. 🙂 In my possible use, this would come between multichannel balanced out DAC with signal at FS and multichannel balanced in amps. Am trying to decide whether keeping the signal balanced all the way through (let alone just through the CS3318) is worth the effort. I could just use unbalanced out from the DACs and maintain that throughout.
There's enough variation in gain structures it's difficult to make general statements. But often it's difficult to do meaningfully better than digital volume control at the DAC and differential direct to the power amp. Lots of FUD on the volume topic so it's best to do the math for your specific case; trouble most folks get into is they don't.
Balanced is highly recommended for interconnects, as trying to go for unbalanced tends to be inherently flawed and begging for trouble these days. Within devices though, it's not strictly necessary unless you have a high RFI environment such as a Class D power amp. In any case, I would prefer to shed common-mode components as quickly as possible after the input.
Read this. Since I don't think there's any true balanced PGA out there (that would have actual CMRR, not just two unbalanced channels tied together), you'd treat the PGA as an unbalanced "problem component".
Read this. Since I don't think there's any true balanced PGA out there (that would have actual CMRR, not just two unbalanced channels tied together), you'd treat the PGA as an unbalanced "problem component".
I wanted DAC with Balanced Analogue Out. This is the solution I came up with: http://www.diyaudio.com/forums/digital-line-level/283319-dac-preamp.html
I'm novice compared to the Electronic Engineer experts around here.
I think I just got lucky cause it sounds awesome. I'm not surprised that the Preamp part works well as it was well designed by someone that really knows what they are doing. But the DAC part was an unknown Chinese thing, and it's much better than I expected. For me the luck is that when I cobbled all these pieces together there's no hum, buzz, etc,.. it sounds really good.
I'm novice compared to the Electronic Engineer experts around here.
I think I just got lucky cause it sounds awesome. I'm not surprised that the Preamp part works well as it was well designed by someone that really knows what they are doing. But the DAC part was an unknown Chinese thing, and it's much better than I expected. For me the luck is that when I cobbled all these pieces together there's no hum, buzz, etc,.. it sounds really good.
Interesting, as doing volume control just before DAC is certainly an option and from both cost and system complexity standpoint would be nice. I'm sure you can guess I was convinced I should avoid that because of truncation of digital signal at lower volume levels. Another option and the simplest of all is to do volume control much further upstream, before ADC and all the multitude of digital surround processing, crossovers, EQ, etc. But then obviously I had assumed that digitally filtering signals derived from substantially less than full scale analog signals, where noise is a higher proportion of the signal and the full resolution of the ADC is not utilized, was to be avoided.
So I've made plenty of assumptions, but am happy to more methodically evaluate and compare the tradeoffs. Do you have any suggestions on where to get started in getting a handle on the concepts and calculations required to do so?
Analog, THAT, and TI have all published extensively on analog amplifier behaviour; focus on noise and CMRR. There's an abundance of introductory DSP texts on the digital side; I don't have a specific recommendation. DIY Audio's gain structure article may also be helpful in understanding why most of the trouble tends to originate in power amp implementation conventions and differences consumer and pro audio gain structures. Broadly speaking, if analog levels are set properly 24+ bit digital volume is fine. It's when one tries to do digital level setting and digital volume that trouble comes, particularly in 16 bit. Ultimately it's all mitigation of not doing volume in the ideal location, which is power amp VAS, but to the extent levels can be configured to move volume downstream the better the overall system will perform.
It's a common DIYer mistake assume output accuracy is limited by digital quantization in the DAC; no worries. Datasheets or eval boards typically include the linearity versus level information needed to be specific about this; usually linearity doesn't begin to degrade until around -100 dBFS and the fall off is relatively slow beyond that. DACs proper are also pretty quiet in the analog domain. Usually the limit is more noise of the output and input buffers. For example, the -106 dBu noise floor of a THAT 1200 is comparable to a DAC+output buffer measuring 119 dB-A DnR, give or take a few dB depending on supply voltages. It's not easy to maintain this across an interconnect. Assuming a typical balanced receiver offering around 35dB CMRR, in this example the ground offset between the DAC output buffer and power amp line receiver required to create an error at the same level as the noise floor is 18uV.
It's a common DIYer mistake assume output accuracy is limited by digital quantization in the DAC; no worries. Datasheets or eval boards typically include the linearity versus level information needed to be specific about this; usually linearity doesn't begin to degrade until around -100 dBFS and the fall off is relatively slow beyond that. DACs proper are also pretty quiet in the analog domain. Usually the limit is more noise of the output and input buffers. For example, the -106 dBu noise floor of a THAT 1200 is comparable to a DAC+output buffer measuring 119 dB-A DnR, give or take a few dB depending on supply voltages. It's not easy to maintain this across an interconnect. Assuming a typical balanced receiver offering around 35dB CMRR, in this example the ground offset between the DAC output buffer and power amp line receiver required to create an error at the same level as the noise floor is 18uV.
Thanks, the additional insight is helpful. I've been busy reading datasheets, articles and presentations, forum discussions, etc. on gain structure and tradeoffs in digital volume implementations. It seems that at least the specific device I'm considering, the miniDSP 4x10Hd using a 24bit DAC/codec with optimally 114dB SNR, may not be horrible for digital volume. Maybe even be only as bad as analog options? Lol. That would be nice. More to learn.
I think you're right about the digital volume in the miniDSP 4x10. In my build I linked to above, I originally planned a future upgrade to be this: Prices. But this 6 channel version is very expensive. I would have placed in after the DAC, in a single ended configuration on the input of the preamp.
I've found the digital volume in miniDSP nanoDIGI works well and has remote control too.
To work out specifics of the digital versus analog tradeoff you'll need the required speaker drive level and relevant power amp gain in addition to the output swing selected on the 4x10.
Interfacing sources with pro audio levels (such as the 4x10) to common power amps tends to be a bit awkward. For about 85% of folks it's desirable the power amp have unity gain rather than the typical 20 to 35dB. The most suitable commercially available amp I'm aware of is a Modulus-86 with the THAT 1206 option. But that's still 14dB.
If it's helpful, the Scarlett 18i20 I use is based on the same -100dB THD/114dB-A DnR DAC found in the 4x10 (Cirrus acquired it from Crystal back in the 1990s) and Focusrite's implementation of 0dBFS is about 10dBu on most of the line outs. I'm in the <2V drive group nearly all of the time and my current power amps are unity gain. Digital level matching costs about 12dB in this configuration but I've never particularly felt it was worth the effort to have the amps attenuate. Mathematically it'd clearly be better to do so but it's close enough to optimal other projects have consistently proved more interesting.
Interfacing sources with pro audio levels (such as the 4x10) to common power amps tends to be a bit awkward. For about 85% of folks it's desirable the power amp have unity gain rather than the typical 20 to 35dB. The most suitable commercially available amp I'm aware of is a Modulus-86 with the THAT 1206 option. But that's still 14dB.
If it's helpful, the Scarlett 18i20 I use is based on the same -100dB THD/114dB-A DnR DAC found in the 4x10 (Cirrus acquired it from Crystal back in the 1990s) and Focusrite's implementation of 0dBFS is about 10dBu on most of the line outs. I'm in the <2V drive group nearly all of the time and my current power amps are unity gain. Digital level matching costs about 12dB in this configuration but I've never particularly felt it was worth the effort to have the amps attenuate. Mathematically it'd clearly be better to do so but it's close enough to optimal other projects have consistently proved more interesting.
Last edited:
Is it a cardinal sin if, after evaluating the specific gain structure for a system all digital through volume control into dac, then straight to amp, and finding that several or more dB of attenuation would be required for 0dBFS to better match maximum desired amp output, to implement a simple fixed attenuation, perhaps even passive, right before the amplifier?
I have several pieces of the puzzle likely roughed in, such as most likely drivers I'll wind up with, but not all pieces. Biggest one is suitable amplifier. I'm probably going to be using a boatload of channels and I'm still looking at best value way of getting there. Pro amps are a definite possibility so that might be useful especially if gain is adjustable. Another strong possibility is a DIY solution using assembled off the shelf amp modules, chip amp or class d, to make several of my own multichannel amps. I guess that leaves a lot of room for screwing things up!
I have several pieces of the puzzle likely roughed in, such as most likely drivers I'll wind up with, but not all pieces. Biggest one is suitable amplifier. I'm probably going to be using a boatload of channels and I'm still looking at best value way of getting there. Pro amps are a definite possibility so that might be useful especially if gain is adjustable. Another strong possibility is a DIY solution using assembled off the shelf amp modules, chip amp or class d, to make several of my own multichannel amps. I guess that leaves a lot of room for screwing things up!
Only if you dork up the fixed divider implementation. 😉 If a truly clean input is desired be careful of amplifier CMRR against introduced common mode, source impedance issues, and source loading; like any precision circuit it gets involved fast. (There are better divider implementations but I'm not aware of any modable commercial amp employing them at reasonable cost per channel.)
billshurv's concept of dBGF---dB gnat fart---may be a useful guide here. I generally consider errors 100+dB down to be dBGF territory; approaching such levels is often interesting engineering. Additional reduction is usually about specmanship or simple enjoyment of optimization rather than any meaningful subjective improvement (it's often fair to call it measurebation).
billshurv's concept of dBGF---dB gnat fart---may be a useful guide here. I generally consider errors 100+dB down to be dBGF territory; approaching such levels is often interesting engineering. Additional reduction is usually about specmanship or simple enjoyment of optimization rather than any meaningful subjective improvement (it's often fair to call it measurebation).
opc's The Wire LPUHP or MPUHP modules are configurable as anything from unity gain -> ~18dB or so ...
Is there current availability? Didn't dig through the hundreds of pages in the various threads but the most recent PCB offering I found with a bit of searching was 2012.
If by MPUHP you mean the LME49830 lateral FET version (I get no hits on MPUHP proper) be aware the LME49830 is end of life. The LME49600 the LPUHP relies on is also end of life and the most comparable replacement I know requires a board spin.
If by MPUHP you mean the LME49830 lateral FET version (I get no hits on MPUHP proper) be aware the LME49830 is end of life. The LME49600 the LPUHP relies on is also end of life and the most comparable replacement I know requires a board spin.
I second this. You get very good performance this way. You could skip the THAT1646 and use impedance-balanced (quasi-balanced) output if you like.
This thought process has me moving more and more towards 16+ ohm speakers for my next diy build*, where the voltage gain is at least largely masked by a decrease in voltage sensitivity on the speaker-side (conceptually you can look at it as another -6 dB). Bonus is that it also mitigate crossover distortion issues. And, yes, integrating volume control on the digital side, as all my music comes through the DAC.
(which will likely be a multi-channel 7293 composite with a global 20dB gain; 26 min recommended gain on the 7293+OPA1652, -6 from the THAT1206)
I'm long since tired of having my volume control barely above minimum.
Wondering how to reconcile the above sentiments (and I agree, I typically do not listen at insanely loud levels, and even with the modest composite sensitivity I will likely end up with in a speaker I would be at a few watts much of the time) with the goal of reproducing reference level peaks in home theater use and the power required to get there.
I believe boards are available from opc via his thread in vendor's section using google spreadsheet and PM to order. MPUHP is a bridged version of the LPUHP 😎. See measurements here - http://www.diyaudio.com/forums/vend...-here-bal-bal-se-se-lpuhp-69.html#post4282803 - (Announcement in Post #689 and measurements in #692 of that thread) But, yes it will need a board revision between now and whenever LME49610 are no longer available for purchase.
Last edited:
Like many home audio issues there's a solution in pro audio; most gear offers ~20dB level pads. Is only hard problem if you make it one. 😉 An electrically ideal implementation of such a pad is to switch the input and output between, say, 0dB and +20dB amplifier modules---a classic fat box power amp build has the chassis height for this, making it a matter of build time and expense. Less involved implementations switch a resistive lpad (level pad) in or out. Lots of forms of that.
Current iterations of the build mentioned here are flexible in accommodating different input levels and selecting output level. I've been debating to also support an lpad switchable via a pushbutton mountable wherever one wanted on an amplifier chassis. Would be curious to hear how interested folks are in this feature and what preferred pad amounts might be. It'd formally be a turn amp off, push button, turn amp back on kind of thing but probably wouldn't thump egregiously if one happened to forget (or be feeling lazy). Customizing the pad amount would not be difficult.
Thanks for the link; good choice if one doesn't mind throwing money at problems. Personally I'm inclined to do a bit more design, less building, and spend less. Hence the creation of the topology Derf describes. For example, a complete build of such an amp module costs rather less than just the 496x0s on the MPUHP did at their reduced end of life price. With proper compensation and good fill in of support circuitry and layout of that approach one arrives at the Modulus-86. Tom did a lot of build work to validate a lot of maths and sims I'd done and got a not bad little business out of it too (after some sims of his own). But, apropos of Derf's remarks, I suppose I should note the first Mod type layout had the composite part operating at unity and therefore offered from -6 to 0dB gain depending on 120x selection.
Last edited:
- Status
- Not open for further replies.