Ferrite Beads on Output Pins 2, 3

Help Support GroupDIY:

Piotr

Well-known member
Joined
Jan 23, 2005
Messages
142
Location
Chambery, France
Hi !
I've read through this informative thread, as I'm trying to implement a THAT 1646 output stage in a project of mine.
Would you mind sharing pictures of your output XLR pcbs with EMI/RFI filtering ? If I understand you correctly, the physical layout is of top importance for it to be effective and as I'll be designing mine I would like to grasp the idea a little better.
Concerning the ferrite beads, do you have a leaded part that you use ? I've checked the Mouser catalog, I do get the SMD Wurth option, but would like to check on all the available options, since I don't solders SMDs... Will this project make buy a tweezer iron and accept that I'm wearing glasses now at 46 ???
Thanks for you help ! Have a great day !
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
The properties you're looking for are ~100 ohms impedance at MHz, very low resistance at low frequencies and a few amps current handling. It so happens that these properties are easily obtained with a tiny $0.20 SMT part. Usually I do not use SMT parts in the signal path but I make an exception for EMI filtering parts that are very low impedance at audio frequencies and so this is a good example where SMT is perfectly suitable (and necessary since leaded parts are much less common). Just use whatever component search site to search for ferrite beads that are an SMT package (I would select a part with an SMT package that is a more common slightly larger package, like 1206 (3216), just to minimize the number of different PCB footprints used and to make it a little easier to solder), ~100R impedance and then narrow down the list by selecting the lowest low frequency resistance. Doing this on Mouser brings up BLM31KN471BH1L but I just searched for it, I have never used that part.
 

RSRecords

Well-known member
Joined
Jun 8, 2009
Messages
220
Sorry to hi jack a bit. I've been looking a similar situation on the input.trying to devise a way to keep the RFI caps as close to the input pins as practicable but the recommended scheme for the 1206 would require a third connection to the input PCB. I don't see a way around this because of the R3/C3 connection.

That 1200 RFI.png
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,360
Location
Marcelland
Sorry to hi jack a bit. I've been looking a similar situation on the input.trying to devise a way to keep the RFI caps as close to the input pins as practicable but the recommended scheme for the 1206 would require a third connection to the input PCB. I don't see a way around this because of the R3/C3 connection.
TBH I never felt comfortable with the 120x. IMO it trades ultimate performance of the input stage against the risk of allowing RFI to enter the chassis.
Ideally, the 100r resistors would be in the male plug!
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
Sorry to hi jack a bit. I've been looking a similar situation on the input.trying to devise a way to keep the RFI caps as close to the input pins as practicable but the recommended scheme for the 1206 would require a third connection to the input PCB. I don't see a way around this because of the R3/C3 connection.

For this I made a little PCB that solders directly to the XLR connector like in the attached pic. This keeps everything shunting to ground at the panel.

In this particular implementation, the mic / line input has small chip ferrites but you could swap in any 1206 package. You could use the common cap implementation like in your schem and maybe even squeeze in a small common mode choke. Note that you would probably not want to attach the pin 1 to your cable screen. I would use 0V at the other end.

The one on the right is the line-out version. It also has special TVS diodes. Normal TVS diodes have a lot of capacitance and leakage that can cause distortion. AFAIK, these do not. They're special bidirectional TVS diodes for DSL modems that have 3pF capacitance and 1nA leakage. I can't seem to measure any distortion using these things. I would have used them on the input too but of course you can't use them with phantom power because it would short it.

Note that IIRC the 100R might be important to protecting the 1206 for some reason. But I think if you interpret the datasheet carefully, the chip ferrites and caps make it unnecessary. But you could just put a 1206 resistor in place of the ferrites.

Also, note that the XLRs are the version with the spike in the screw hole (NC3FBF1 for mic / line in). You can see the diagonal metal lead running down and over to the screw hole. So no need to connect anything to the pin 1 wirepad if you use XLRs like that.

1622571460029.png
 

Attachments

  • MicLineEmiCircularPcb.png
    MicLineEmiCircularPcb.png
    86.3 KB · Views: 29
Last edited:

RSRecords

Well-known member
Joined
Jun 8, 2009
Messages
220
For this I made a little PCB that solders directly to the XLR connector like in the attached pic. This keeps everything shunting to ground at the panel.

In this particular implementation, the mic / line input has small chip ferrites but you could swap in any 1206 package. You could use the common cap implementation like in your schem and maybe even squeeze in a small common mode choke. Note that you would probably not want to attach the pin 1 to your cable screen. I would use 0V at the other end.

The one on the right is the line-out version. It also has special TVS diodes. Normal TVS diodes have a lot of capacitance and leakage that can cause distortion. AFAIK, these do not. They're special bidirectional TVS diodes for DSL modems that have 3pF capacitance and 1nA leakage. I can't seem to measure any distortion using these things. I would have used them on the input too but of course you can't use them with phantom power because it would short it.

Note that IIRC the 100R might be important to protecting the 1206 for some reason. But I think if you interpret the datasheet carefully, the chip ferrites and caps make it unnecessary. But you could just put a 1206 resistor in place of the ferrites.

Also, note that the XLRs are the version with the spike in the screw hole (NC3FBF1 for mic / line in). You can see the diagonal metal lead running down and over to the screw hole. So no need to connect anything to the pin 1 wirepad if you use XLRs like that.

View attachment 81504
I guess my concern echoes abbey road, there still needs to be that CM In connection from the xlr PCB. Unless I'm missing something? Maybe in practice is won't matter, the data sheet also offers a simple RF solution with 2 100pf. Maybe that's enough?
 

Attachments

  • that 1200 RFI Simple.png
    that 1200 RFI Simple.png
    41.8 KB · Views: 29

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
I guess my concern echoes abbey road, there still needs to be that CM In connection from the xlr PCB. Unless I'm missing something? Maybe in practice is won't matter, the data sheet also offers a simple RF solution with 2 100pf. Maybe that's enough?

I'm not sure what abbey's concerns are since you can still put EMI filters at the panel regardless of what chip is used. But I'm not sure how vitally important that 4k7 bootstrapping of the EMI filter is anyway. That raises the common mode input impedance but in this case only for RF and I just find it hard to believe that even a moderate imbalance of impedance wrt RF is going to throw off the amp so much that it affects audio frequencies. If you use caps + ferrites at the XLR like I described, that will shunt RF plenty good. Otherwise, to justify using that third wire, you would really need the amp to be right next to the XLR. Maybe if I were making something with PCB mount XLR and the chip was right there I would do it because it's just an extra resistor. But for DIY where you almost certainly have a little shielded cable running between, I would just skip it.

But maybe Whitlock will answer you directly. He's posted a few times here recently as @MisterCMRR.
 
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,360
Location
Marcelland
I'm not sure what abbey's concerns are since you can still put EMI filters at the panel regardless of what chip is used.
I'm not sure what my concerns are either. :)
But I'm not sure how vitally important that 4k7 bootstrapping of the EMI filter is anyway. That raises the common mode input impedance but in this case only for RF
Not really RF, since it works up to about 200kHz.
and I just find it hard to believe that even a moderate imbalance of impedance wrt RF is going to throw off the amp so much that it affects audio frequencies. If you use caps + ferrites at the XLR like I described, that will shunt RF plenty good.
As I wrote earlier, I've never put the 120x series in practice.
My understanding is it improves CM impedance, at the cost of allowing RFI to get into the unit instead of being repelled at the door.
Actually, putting EMI filters at the panel defeats the idea of increasing CM Z, unless no shunt caps to chassis are involved.
Seems to me there are two different approaches, the brute force one that filters RF by shunting it to chassis, and the soft one that nulls it actively.
Combining both seems to be somewhat contradictory.
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
Actually, putting EMI filters at the panel defeats the idea of increasing CM Z, unless no shunt caps to chassis are involved.
Seems to me there are two different approaches, the brute force one that filters RF by shunting it to chassis, and the soft one that nulls it actively.
Combining both seems to be somewhat contradictory.
Wah? The 120x increases CM impedance at all frequencies including audio. Being an audio amplifier I would say the main purpose of the 120x is to increase CMRR at audio frequencies. Yes it increases CM Z at RF as well (actually no, see #39) but if you can shunt that at the panel with a few tiny 20 cent parts then why not just do that? I don't understand your rationale. If the CM impedance is higher because of the 220u Cb bootstrapping cap, then differences in loss between the signal lines due to variation in source impedance will be minimized. What am I misunderstanding here?
 
Last edited:

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,360
Location
Marcelland
Wah? The 120x increases CM impedance at all frequencies including audio.
I don't question that.
It's the addition of an RF filter before that I discuss. Most RF filters involve shunt capacitors, which would decrease CM Z. In addition, tolerance on these shunt capacitors may impait CMRR at high frequency. That's why I suggested the use of just a CM inductor, no caps.
I'm not saying an RF filter shouldn't be used, I'm only warning that the type of RF filter is important.
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
I don't question that.
It's the addition of an RF filter before that I discuss. Most RF filters involve shunt capacitors, which would decrease CM Z. In addition, tolerance on these shunt capacitors may impait CMRR at high frequency. That's why I suggested the use of just a CM inductor, no caps.
I'm not saying an RF filter shouldn't be used, I'm only warning that the type of RF filter is important.

But the RF will have been shunted to the panel through the caps so why do you care about CMRR at high frequency at the amp?

Also, what about differential RF? A CM inductor will not stop that at all. I don't think we should assume the RF noise will always be CM. There might be some flaw in the source gear that's letting RF into the differential signal.

And I'm still struggling to see how two 470p caps are going to be that different. Use 1% NP0 (GCM31A5C3A471FX01D). If you REALLY want to get pedantic about it, use a capacitor array (CA064X102K2RACTU). Even though it's 10%, the difference between multiple elements of the same part should be very low I presume. One $0.23 cent part could beat the CM inductor in performance and price. That particular part is only 200V though.
 

ccaudle

Member
Joined
Mar 18, 2010
Messages
20
Location
Houston
Seems to me there are two different approaches, the brute force one that filters RF by shunting it to chassis, and the soft one that nulls it actively.

You don't necessarily have to cover every frequency range with the same solution, you can use smaller capacitance devices directly at the pins to handle the highest frequencies, then use the Y style approach with CM bootstrap to handle lower frequencies (e.g. AM radio band).
Using ferrites and series resistors is also not mutually exclusive, you are potentially trying to cover 3.5 orders of magnitude (0.5MHz AM radio band up to over 1GHz wireless comm bands) so one cap/ferrite selection is not likely to be fully effective over the entire range.
 

john12ax7

Well-known member
Joined
Oct 14, 2010
Messages
1,848
Location
California, US
I guess my concern echoes abbey road, there still needs to be that CM In connection from the xlr PCB. Unless I'm missing something? Maybe in practice is won't matter, the data sheet also offers a simple RF solution with 2 100pf. Maybe that's enough?

Sometimes you don't have access to the XLRs, like in 500 series, so alternative arrangements can be beneficial.

Also with the previous 120x there are really two things going on. The bootstrapping raises CM impedance, but that's really a lower frequency thing. The RF filter is for higher frequencies. It depends on the needs and environment. The 2 cap solution is often sufficient. As is also a simple differential receiver without bootstrapping.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,360
Location
Marcelland
Also, what about differential RF? A CM inductor will not stop that at all. I don't think we should assume the RF noise will always be CM.
Differential RF is supposedly taken care of by the two 470pF caps. I'm not sure it's efficient, but it seems to be good enough for THAT.

And I'm still struggling to see how two 470p caps are going to be that different. Use 1% NP0 (GCM31A5C3A471FX01D). If you REALLY want to get pedantic about it, use a capacitor array (CA064X102K2RACTU). Even though it's 10%, the difference between multiple elements of the same part should be very low I presume. One $0.23 cent part could beat the CM inductor in performance and price. That particular part is only 200V though.
I don't know. As I wrote earlier, I have never used the 120x series, because I don't understand all its subtleties.
The recommanded RFI protection scheme seems to rely on the inputs not being coupled to ground, but rather to the CM node.
Why does THAT not recommand RF filters?
 

MisterCMRR

Active member
Joined
Jul 1, 2019
Messages
35
Location
Ventura, CA
The two 470 pF caps in series (effectively 235 pF) shunt RF differentially. The bootstrap is to reduce the effective values of those capacitors to much less at the audio frequencies within the passband of the filter formed by the 4.7 kΩ and 100 pF. This keeps the common-mode input impedances much higher than they'd otherwise be (which keeps CMRR from being degraded by the imbalance of driving source's common-mode impedances). When the 4.7 kΩ + 100 pF filter attenuates the bootstrap signal, the 470 pF caps regain their effective values and, in series with the 100 pF, attenuate the RF. It's a way to make the 470 pF capacitances vary with frequency - looking quite small at audio frequencies (and reducing the effect of a standard 5% tolerance), but behaving at their full value to become effective at super-sonic and radio frequencies.
Regarding ferrite beads: Since most beads for audio input line applications are chosen to look like at least 100 Ω at RF frequencies (and hopefully stay that high over a wide frequency range), they're preferred over resistors because ordinary input stages are very sensitive to imbalances resistances. The bead will have audio impedance near zero. But the InGenius input is orders of magnitude more tolerant, so a pair of 100 Ω resistors (of ordinary tolerances) will do just as well as a bead - in fact with an extended frequency range. In output drivers, of course, low impedance in the signal path is very important, so the bead is much preferred.
Most of this is discussed in several AES papers authored by myself and staff at THAT Corp, my licensee for the patent.
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
When the 4.7 kΩ + 100 pF filter attenuates the bootstrap signal, the 470 pF caps regain their effective values and, in series with the 100 pF, attenuate the RF... looking quite small at audio frequencies (and reducing the effect of a standard 5% tolerance), but behaving at their full value to become effective at super-sonic and radio frequencies.

Ah, I see. So it does not increase CM Z at RF as I thought previously. Thanks.

But wouldn't the source impedance have to be really high before differences in the caps would affect the audio range?
 

Latest posts

Top