Very true (albeit probably underestimating the current generation of PCB buyers quite a bit)!
Then again, this was in reply to twest820, not Tom. 😉
I was merely interested in picking Tom's brain on the matter.
But since he didn't make the comparison, I rest my case.
Cheers,
Sebastian.
Maybe I am maybe I'm not. Dunno as I have nothing to sell. And I am sure if Tom could have got better performance from SMT he would have gone that way.
I'd expect that to take up twice or more board area. Depending on the core material chosen for the common mode choke, they could also introduce distortion.
I don't understand your fascination with common-mode chokes. Would you care to elaborate?
To the balance? Do you mean the few mΩ of difference between the two chokes? I suspect that'd be swamped by differences in the cable resistance.
When I measured the THAT Driver, I used a real world XLR cable from the THAT Driver output to the APx525 input. The cable was about 6' (180 cm) long.
Tom
Potentially, although the two ferrites aren't exactly small either.
I have no recommendation for a particular non-SMD part, though.
True, but applicable to chokes in general.
I have no stake in common mode chokes. 🙂
Also I'm not naïve so as to lecture you on chokes. 😉
But I also have no stake in X2Y capacitors, for example, yet I would favour them over two discrete caps for RFI rejection wherever applicable (i.e. in SMD circuits, but you get my point).
They're another example of an objective improvement, so why not use them?
The answer is surprisingly simple: THAT couldn't go that far when determining the data sheet recommendations back in the nineties.
They had to consider their market at the time (and let's not get started on the incredibly richer RFI spectrum of today).
It certainly isn't be a deal breaker, yes.
For that length THAT doesn't actually recommend to use an output filter at all... 😀
Cheers,
Sebastian.
Better matching is available at lower cost with individual C0Gs. I trust you're not asserting three terminal 0805 and 1206 components belong in a through hole offering, so not seeing the relevance here.
The 16x6 datasheet is dated 2011 (consider fact checking before posting, eh?). I'm also not following the assertion X2Ys offer meaningful improvement to a 48V protection circuit.
Last edited:
twest820,
I notice you repeatedly reacting to my replies not adressed to you, picking single lines out of context and then mocking me with them.
Something on your mind? 😕
Value matching of three terminal caps works exactly the same way (just across the respective device pins).
The residual systematic asymmetry (if any) is by far outweighed by the benefits of the lower inductance.
The lower cost argument is likely true, depending on where you source your capacitors.
We're talking about a high value PCB and BOM project here and I didn't claim to make a case for cost reduction, though...
That's exactly what I remark, exactly one line above the one you so conveniently quote, yes: SMD...
That particular document, sure.
My copy claims copyright 2007 though – and THAT used the circuitry in question before that point.
Now, consider project R&D times and you're back a couple more years...
Wrong function, again. This time, the components in question are part of the RFI filter, not the 48V surge protection...
PS: Are you perhaps confusing X2Y MLCC for RFI suppression (link) with something else?
I notice you repeatedly reacting to my replies not adressed to you, picking single lines out of context and then mocking me with them.
Something on your mind? 😕
Value matching of three terminal caps works exactly the same way (just across the respective device pins).
The residual systematic asymmetry (if any) is by far outweighed by the benefits of the lower inductance.
The lower cost argument is likely true, depending on where you source your capacitors.
We're talking about a high value PCB and BOM project here and I didn't claim to make a case for cost reduction, though...
That's exactly what I remark, exactly one line above the one you so conveniently quote, yes: SMD...
That particular document, sure.
My copy claims copyright 2007 though – and THAT used the circuitry in question before that point.
Now, consider project R&D times and you're back a couple more years...
Wrong function, again. This time, the components in question are part of the RFI filter, not the 48V surge protection...
PS: Are you perhaps confusing X2Y MLCC for RFI suppression (link) with something else?
That may be. However, my board is designed to be able to handle just about anything the builder throws at it. I wouldn't want to paint myself into a corner by offering a board that imposed limits on the cable length, for example. If the builder doesn't like the output filter or deems it to be unnecessary in their application, they're free to replace the inductors with wire links and not populate the caps.
It seems your comments can be summarized as: "The THAT Driver circuit could have been designed differently, using different components, with different tradeoffs made". Is that a fair summary?
If yes, then you're most certainly right. I could have designed the circuit in any number of ways. I chose the THAT1646 because it provides stellar performance. Unlike many SE -> DIFF circuits out there, the THAT1646 provides good performance even with one of the pins in the differential pair shorted to ground (as it would be if you plugged it into a single-ended load). The performance of the THAT Driver can be viewed on my website (THAT Driver: Differential line driver / preamp). I think the measurements speak for themselves. It's a darn good circuit.
Tom
PS: Based on my experience in the semiconductor industry, a couple of years of R&D time for a project like the THAT1646 would be outrageous. If THAT were that slow, they wouldn't be around.
Last edited:
Nope. The tradeoff of mildly increased stopband depth from lower inductance in X2Y packages and matching effects is not actually one sided. In typical layouts, particularly those realistic at DIY price points and skills, total inductance is dominated by nH on the board. So pH optimizations within the package typically offer only a few dB improvement. This can come at the expense of more dB in matching considerations.
That would be 2005. Which, last I checked, would not be in the 1990s as indicated in post 1904. Please do let us know once you've converged on a date.
And what circuit unrelated to the THAT Driver shall the next post discuss?
Again, I disagree, since my experience says otherwise.
Particularly, the topic is Tom's PCB design skill and price point, not anyone else's.
Again, please refrain from twisting other peoples words to make yourself look superior in a discussion.
It's considered trolling. That's something entirely different than an off topic sidetrack!
The data of post 1906 indicates approximately 4dB improvement, consistent with post 1908's remark of "a few dB improvement". Have you additional citations?
PCBs designed to DIY pricing and assembly preferences necessarily share a common implementation space. It's been a pleasure mentoring Tom from time to time over the years on how to get the most from such layouts; as measurements here show he's a capable student.
Yes, but I respect Tom's settling nudge not to sidetrack the topic any further, since the design decisions turn out to be irrevocable.
Discussing minute changes to RFI immunity design without mention of 'real life' trials is just wanking.
In the 80's, I designed microphones and the biggest customers were the broadcast organisations.
All my mikes had EVIL 10n ceramics on the XLR pins. There were no NPO ceramics of this value in dem days but you should always look at the complete system.
Much later, when EU certification became mandatory, I was amused to find that their dreaded RFI/EMI specs were trivial to meet.
Also much later, I found that Dip Ing Wuttke of Schoeps used exactly the same EVIL method as I did though our circuits were completely different 😱
_____________________________________
Tom's amp is a 4ppm THD item but it's also Line+ level. Today, 10n NPO/COG ceramics are also readily available.
RFI/EMI immunity is a LOT more than choosing your bits. The layout is AT LEAST as important .. along with earthing, grounding bla bla.
If a circuit + layout + box + bits has been extensively tested for RFI immunity, ANY changes should be looked at with suspicion until they have been extensively tested.
If you don't sell stuff to the BBC and other organisations that object to being alerted when you receive an SMS, you may safely ignore this bullsh*t. 🙂
______________________________________
PS I've never used CM chokes for real. They just weren't available in dem days.
Anyone know if X2Y cap networks are available NPO/COG? Links?
In the 80's, I designed microphones and the biggest customers were the broadcast organisations.
All my mikes had EVIL 10n ceramics on the XLR pins. There were no NPO ceramics of this value in dem days but you should always look at the complete system.
Much later, when EU certification became mandatory, I was amused to find that their dreaded RFI/EMI specs were trivial to meet.
Also much later, I found that Dip Ing Wuttke of Schoeps used exactly the same EVIL method as I did though our circuits were completely different 😱
_____________________________________
Tom's amp is a 4ppm THD item but it's also Line+ level. Today, 10n NPO/COG ceramics are also readily available.
RFI/EMI immunity is a LOT more than choosing your bits. The layout is AT LEAST as important .. along with earthing, grounding bla bla.
If a circuit + layout + box + bits has been extensively tested for RFI immunity, ANY changes should be looked at with suspicion until they have been extensively tested.
If you don't sell stuff to the BBC and other organisations that object to being alerted when you receive an SMS, you may safely ignore this bullsh*t. 🙂
______________________________________
PS I've never used CM chokes for real. They just weren't available in dem days.
Anyone know if X2Y cap networks are available NPO/COG? Links?
Last edited:
No design decision is irrevocable. However, I choose to base my design decisions on facts, measurements, and data rather than "so-and-so on the Internet said..." Thus far, all I've seen from you has been "you could do it differently" and "what if". I haven't seen any data from you that indicates that a change is necessary. The THAT Driver circuit performs remarkably well. I see no reason to change it. It's not broken. It does not need fixing. In fact, it works very well.
If you believe you can improve on the circuit, you're more than welcome to buy a board and put your ideas into practice. I'd be very curious to see your measurements, level of improvement, and whether the improvements can be reproduced.
Tom
Digikey has them up to 1 nF:
http://www.digikey.com/product-sear...t=0&page=1&quantity=0&ptm=0&fid=0&pageSize=25
If the link doesn't work, just type in X2Y in the search bar on www.digikey.com, click search, and select ceramic capacitors from the list.
Tom
the 1000pF and 470pF are non stock.
The largest value NP0/C0G available seems to be 220pF.
Umm, 1000pF? Guess if you need a 470pF it's a Mouser order though.
If one's doing pro audio type assemblies with board mount connectors RF filtering at the pins becomes more interesting. Even the lower connection in a part like the NSJ12HC still sees an inductive discontinuity on the order of 10 nH from plug wiring, jack contacts and leads, PCB routing, and surface mount filter components. X2Y caps are perhaps an order of magnitude more helpful in this situation. But there's really not much to be done for the actual filter efficacy limitations until such time---ha!---the industry adopts more compact connectors designed with EMI/RFC in mind. Filter optimization doesn't hurt but, yeah, more valuable to minimize SCIN and other noise injection mechanisms.
As an aside, I should probably note tolerance sims for balancing show THAT's inlet filter for the 1200 is reasonably effective at mitigating mismatch between 2Y caps. There's the inductance of the earth cap in this arrangement but the extra nH and some is not that extravagant.
Mmm hmm. Jim Brown's published a good amount of data and analysis on this, with bonus a good bit is not behind paywall. The X2Y parts are kind of novel and hence fun to look at. In a typical DIYer setup with panel mount connectors, in box cabling, terminal blocks, and earth return through some substantial board and chassis loop with possibly ground lifts and whatnot dropping a couple hundred pH in the package is pretty well irrelevant due to the large loop inductance back to the connector. To borrow Bill's phrase, maybe 4dB gnat fart improvement but likely a couple orders of magnitude below 4 dBV. Quite a bit earlier in this thread---a hundred pages, maybe more---Neutrik's EMC connectors came up. Since DIY typically adheres to the home audio convention of using XLRs for balanced line level these are probably the most effective option for most builds.
If one's doing pro audio type assemblies with board mount connectors RF filtering at the pins becomes more interesting. Even the lower connection in a part like the NSJ12HC still sees an inductive discontinuity on the order of 10 nH from plug wiring, jack contacts and leads, PCB routing, and surface mount filter components. X2Y caps are perhaps an order of magnitude more helpful in this situation. But there's really not much to be done for the actual filter efficacy limitations until such time---ha!---the industry adopts more compact connectors designed with EMI/RFC in mind. Filter optimization doesn't hurt but, yeah, more valuable to minimize SCIN and other noise injection mechanisms.
As an aside, I should probably note tolerance sims for balancing show THAT's inlet filter for the 1200 is reasonably effective at mitigating mismatch between 2Y caps. There's the inductance of the earth cap in this arrangement but the extra nH and some is not that extravagant.
I am seriously considering 3 pin DIN (well 5-pin sockets as that gives me more flexibility. Whilst lacking the 'home improvements' style manliness and pro look, the WNTL factor is getting higher by the day for me.
WNTL? For DIY I've been tempted by any number of alternate connectors at various times but pragmatism has always won out----pro audio best practice is proven sufficient to avoid nearly all problems and the few that do come up are rarely relevant to home audio except in the abstract. Non-problems have a way of needing non-solving, though there's a certain enjoyment in considering what would be needed to solve them anyway.
Besides, for high RF immunity it's difficult to improve on digital interconnect and an integrated DAC and power amp. The miniDSP and Modulus combination Tom put together over in the build thread is a good example. Downside is the central amp results in longer speaker cable runs than powered and hence more opportunity for noise. One can have a fine discussion as whether a given amount of noise injection through unshielded, untwisted pair on an amp output at the end of the gain chain is better or worse than a lower amount of injection through shielded and possibly twisted pair before amplifier gain. The basic answer, however, is practical experience doesn't indicate difficulty with either approach. A more nuanced answer is decent interconnect cables offer more dB of rejection than most power amplifiers offer dB of gain. So, if all the pin 1 stuff and whatnot is got right, plate amps/powered speakers/monoblock style setups offer higher overall system immunity in typical installations.
Perhaps the most interesting twist DIY offers in this is one has control over the amp connectors and hence opportunity to move the design points around by using shielded twisted pair speaker wire or such. It's probably gilding the lily (as a non-problem needing non-solution) but I don't know anyone who's actually done it. Verifying it's in fact a non-problem isn't a bad thing.
Besides, for high RF immunity it's difficult to improve on digital interconnect and an integrated DAC and power amp. The miniDSP and Modulus combination Tom put together over in the build thread is a good example. Downside is the central amp results in longer speaker cable runs than powered and hence more opportunity for noise. One can have a fine discussion as whether a given amount of noise injection through unshielded, untwisted pair on an amp output at the end of the gain chain is better or worse than a lower amount of injection through shielded and possibly twisted pair before amplifier gain. The basic answer, however, is practical experience doesn't indicate difficulty with either approach. A more nuanced answer is decent interconnect cables offer more dB of rejection than most power amplifiers offer dB of gain. So, if all the pin 1 stuff and whatnot is got right, plate amps/powered speakers/monoblock style setups offer higher overall system immunity in typical installations.
Perhaps the most interesting twist DIY offers in this is one has control over the amp connectors and hence opportunity to move the design points around by using shielded twisted pair speaker wire or such. It's probably gilding the lily (as a non-problem needing non-solution) but I don't know anyone who's actually done it. Verifying it's in fact a non-problem isn't a bad thing.