Ferrite Beads on Output Pins 2, 3

Help Support GroupDIY:

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
What would be a good part for ferrite beads on pins 2 and 3 of a balanced output?

Would a small axial inductor (like Bourns 78F101J) be ok even though it has ~8 ohms DCR? Or do I really need something like a ferrite core with a large gauge wire?

Also, regarding Fig 8 of the THAT1646 datasheet, why are the beads before the caps and not after if they're supposed to filter RF coming in?



 

john12ax7

Well-known member
Joined
Oct 14, 2010
Messages
1,848
Location
California, US
You need the caps to be as close and as short a path to the chassis as possible to shunt RF.

The ferrite helps too, but is also provides stability. Essentially you want a low Q inductor here, which is what the ferrite bead essentially is. Jensen makes a load isolator JT-OLI-3 you could use. Other options are an inductor and resistor in parallel. Or just 50 ohm resistors.
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
Ok, so from looking at the JT-OLI-3 datasheet and other posts on this site I understand what these ferrite beads are doing.

If the load has some capacitance (like a long cable, a heavy transformer, a long bus connected to an op amp virtual ground summing input, ...) then this "load isolator" is near zero impedance (< 0.5 ohms) at all audio frequencies but starting at ~1MHz or so the inductance increases impedance to the point where the parallel resistor is the limit (39 ohms in the case of JT-OLI-3).

But the JT-OLI-3 part seems unnecessarily large (and probably expensive although I could not even find it for sale anywhere). So after some searching I have found some possible parts:

Code:
Capacitive Load Isolators

Part
Z
DCR
I
size
cost

Laird Technologies 28L0138-10R-10
45 @ 25 MHz
0.01
5A
4.5mm x 3.5mm Axial
0.10 USD

Wurth 74279215
80 @ 100 MHz
0.03
3A
1206 SMD
0.24 USD

Wurth 742792121
300 @ 100 MHz
0.06
3A
1206 SMD
0.24 USD

Taiyo Yuden BKP2125HS331-T
330 @ 100 MHz
0.075
1.5A
0805 SMD
0.11 USD

Wurth 742792040
600 @ 100 MHz
0.15
2A
0805 SMD
0.24 USD

Wurth 782853561
560 @ 100 MHz
0.1
1.5A
0805 SMD
0.21 USD

Wurth 7427511
272 @ 25 MHz
0.011
5A
8mm x 5mm SMD (possibly tricky soldering)
1.41 USD

Wurth 7427521 common mode
34 @ 25 MHz common mode OR 136 @ 25 MHz using both legs in series "2 turn"
0.003
5A
9mm x 6mm SMD
2.70 USD

Some of these parts are tiny and they seem to fit the bill. Does anyone see a problem with using these? A 0805 package for 6.2 cents for 10+ is hard to beat.

If one really wants to get fancy the common mode part might be interesting since it's one part for both legs of the transformer or pins 2, 3 of the output. Even though the Z is only 34, would it not be higher for common mode noise?
 

gridcurrent

Well-known member
Joined
Aug 28, 2009
Messages
590
squarewave said:
What would be a good part for ferrite beads on pins 2 and 3 of a balanced output?
In his 990 paper, Deane describes the output inductor as 40 turns of #30 wire,
wound around a 39 ohm, Allen-Bradley 1 watt resistor. 
 

Monte McGuire

Well-known member
Joined
Jul 26, 2013
Messages
344
I've used the Wurth 74279215 ferrite bead as a line output isolator / filter, and it seems to work well. It's rated at 3 amps current, and adds only 30mΩ DC resistance, with an impedance around 80-100Ω around 100MHz and up. Further, I cannot measure any distortion spurs using this, down to the -150dBC level with an APx-555 and lots of FFT averaging.

The key seems to be selecting a part with a really high DC current rating. If you can stay far away from the saturation point, distortion will be extremely low, even with not quite so amazing ferrites. Different materials will provide different impedance curves, but to avoid distortion, make sure to have a healthy current margin to completely avoid saturation.
 

kambo

Well-known member
Joined
Apr 24, 2009
Messages
1,979
Location
CA
JT-OLI-3  : $4.65
http://www.jensen-transformers.com/transformers/line-output/
scroll down to "Inductors and Load isolators"
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,445
Location
New Jersey, USA
Monte McGuire said:
I've used the Wurth 74279215 ferrite bead as a line output isolator / filter, and it seems to work well. It's rated at 3 amps current, and adds only 30mΩ DC resistance, with an impedance around 80-100Ω around 100MHz and up. Further, I cannot measure any distortion spurs using this, down to the -150dBC level with an APx-555 and lots of FFT averaging.

The key seems to be selecting a part with a really high DC current rating. If you can stay far away from the saturation point, distortion will be extremely low, even with not quite so amazing ferrites. Different materials will provide different impedance curves, but to avoid distortion, make sure to have a healthy current margin to completely avoid saturation.
Excellent. That is exactly the response I was hoping for. Despite all of the right parameters, I was having trouble accepting that such a small part was equivalent to other parts like 10x6 beads.

Strangely I was not able to find an axial leaded package. It seems they are only available as SMD chips.

For posterity I have added Wurth 74279215 to my list above along with current limits. But since you've already tested that part I will just use that. Thanks. The 1206 package is still small but it should be easier to solder than the 0805.

It's still not crystal clear why the caps are after the inductor in the THAT1646 datasheet. At audio frequencies the outputs are still basically driving the 100p directly so I don't see a difference. It seems like it's just a missed opportunity to make an LC filter. Regardless, I layed it out as described in the datasheet:

5UGohOu.png
 

Monte McGuire

Well-known member
Joined
Jul 26, 2013
Messages
344
Glad to help!

Cap after the inductor is an attempt to piss off the op amp a little less, especially when it's run at low gain and the unity gain frequency is high. The impedance of the bead will isolate the capacitive load a little better and keep the phase margin of the driving amplifier a little nicer. Sure, it's not such a big difference at lower frequencies, but if you use high gain bandwidth amplifiers, it's nice to maintain as much pf a phase margin as possible. That output shunt cap essentially gets added to the cable capacitance, so it's nice to have all of the inductance right after the amp and then all of the shunt capacitance after the inductor.

Soldering SMD is not so  bad, especially big chips like 1206. A quality SMD tweezer iron is a great help - you'll find that it can be faster than a through-hole system and ultra-reliable. I'm using a Weller WXMT tweezer for small stuff, and while it might be a little overpriced, it really is pleasant to use. It's nice and small, and the cartridge tip design keeps  it perfectly aligned without much hassle (no goofy hex screw mounting screws to fiddle with). A single fine point iron can work too, but it's a bit clumsier.

Overall, I'm a big fan of SMD, especially for stuff like this. It prevents one from having to poke a hole into a perfectly good ground plane, and the small component size can be a great help in making a layout that's ideal even at super high interference frequencies.
 

john12ax7

Well-known member
Joined
Oct 14, 2010
Messages
1,848
Location
California, US
Do these parts have any issues at lower frequencies? At 1 MHz the impedance is low  so a misbehaved opamp layout could still have some issues. In a modern high impedance environment do we really gain anything over a simple high quality 50 ohm resistor?
 

Samuel Groner

Well-known member
Joined
Aug 19, 2004
Messages
2,935
Location
Zürich, Switzerland
Iron/ferrite core inductors will generate distortion if significant current is flowing through them. I don't have a good estimate for the order of magnitude at hand, but as in this position quite a bit of current flows air-core might be beneficial.

Samuel
 

ricardo

Well-known member
Joined
Apr 28, 2011
Messages
1,429
Monte McGuire said:
Overall, I'm a big fan of SMD, especially for stuff like this. It prevents one from having to poke a hole into a perfectly good ground plane, and the small component size can be a great help in making a layout that's ideal even at super high interference frequencies.
The caps have to be AT THE PINS.  They must be SMALL ceramics joining p2 & p3 to the case with the SHORTEST POSSIBLE LEADS.

There are pros & cons of using inductors, ferrite bead or resistors  between the pins and your PCB but this PHYSICAL POSITION of the caps is VITAL.

If you have the caps etc on a PCB joined to the pins by 1" of ribbon cable, they are useless.

Their most important role is to get rid of EMI/RFI.

There's a couple of papers explaining the Neutrik EMC-XLR series connectors
http://www.neutrik.com/fl/en/audio/204_1603252336/EMC-XLR_Series_productlist.aspx

They show the ideal arrangement for p1 but it applies to p2 & p3  too.
 

Monte McGuire

Well-known member
Joined
Jul 26, 2013
Messages
344
Samuel Groner said:
Iron/ferrite core inductors will generate distortion if significant current is flowing through them. I don't have a good estimate for the order of magnitude at hand, but as in this position quite a bit of current flows air-core might be beneficial.

I agree, and want to reiterate the concept of "significant current". The key here seems to be using an inductor that saturates at current levels that are 3-5 orders of magnitude higher than the expected signal currents. When this is done, I have empirically found that a ferrite bead rated for 3A has absolutely no effect on a line level signal of around a volt loaded into something as low as 600Ω, down to the residual level of an APx-555 with heavy FFT averaging, basically to the -150dBC level.

Inductor distortion seems to rise rapidly with signal level, but similarly, it seems to fall rapidly with decreasing signal level. So, if you stay far away from saturation, it's really difficult to detect any effect whatsoever with current test gear. One would think that "iron hysteresis" would cause problems, but again, this doesn't seem to result in "bad numbers".

I think there's an aversion to using inductive components for high resolution audio because of these suspicions, and I'm not sure that they're well founded, either by measurements or listening tests. Sure, inductors with high permeability cores can be misapplied, but just the same, if some simple concepts are applied, it seems to me that they can be used without causing detectable distortion.

I'm not sure what the margin of signal current to inductor saturation should be, but given that "easy to apply" parts offer current overload points of 3A and we can use them for line level signals of around 1mA and get great results, it might not require so much thought. I can't measure problems, and any of the simple models won't predict problems, so maybe it just works well?
 

Monte McGuire

Well-known member
Joined
Jul 26, 2013
Messages
344
john12ax7 said:
Do these parts have any issues at lower frequencies? At 1 MHz the impedance is low  so a misbehaved opamp layout could still have some issues. In a modern high impedance environment do we really gain anything over a simple high quality 50 ohm resistor?

The situation of a headphone amplifier is a good example of how a high quality resistor can be worse than an inductor. Many headphones used for professional audio have impedances around 50-100Ω over the audio band, so simply placing a quality 50Ω in series with the driving amplifier will cause predictable response problems. In this situation, using an inductive component to add isolation from the amplifier to the load and its attached cable can prevent amplifier instability without causing response problems in the audio band.

These inductors need to be chosen so that they behave well at low frequencies and at the expected signal currents, but I contend that this is possible to do with available components, and can result in better overall results than a simple resistive isolator.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,359
Location
Marcelland
squarewave said:
What would be a good part for ferrite beads on pins 2 and 3 of a balanced output?

Would a small axial inductor (like Bourns 78F101J) be ok even though it has ~8 ohms DCR? Or do I really need something like a ferrite core with a large gauge wire?
The ferrite beads there are the most basic "one-turn", similar to these
http://www.digikey.com/product-detail/en/laird-signal-integrity-products/28L0138-10R-10/240-2438-1-ND/806798
They are specified to about 75-200 ohms @100MHz, increasing up to several GHz, which is what you want if you want to prevent RFI from cell phones, WiFi and remote controls. With only one turn, saturation of the core is not an issue at the negligible current (compared to line current) of an audio line transmission.
As Ricardo mentions, the capacitors to ground are much more critical than the inductors.
Now, if you had high-speed digital circuitry in your product, my answer would be somewhat different.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,359
Location
Marcelland
Martin Griffith said:
Will a common mode inductor be any good, rather than two inductors?
Yes, for common-mode noise, as the name implies. Since most EMI/RFI in balanced connections is common-mode, indeed it is good. However, it may not be so good at rejecting whatever diff-mode noise may appear. I know some particularly punctilious designers who use both common-mode AND diff-mode chokes; belts and braces...
 

ricardo

Well-known member
Joined
Apr 28, 2011
Messages
1,429
If you use 'good' inductors to go from the XLR pins to the PCB, they will form a resonant circuit with the Ceramic caps (10-22n on a microphone) which should be DIRECTLY on the pins.  If you are meticulous about not having supersonic funnies, you need to use a Zobel or other method to damp any peaks.

I did a lot of stuff like this in da previous Millenium including using transformers to achieve nice supersonic filtering.

The advantage of beads is that their inductance and ESR is 'non-linear'.  Just slipping a couple on the XLR leads gives inductance & ESR nicely placed & damped for RFI/EMI but with little or no effect at audio.

 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,359
Location
Marcelland
ricardo said:
If you use 'good' inductors to go from the XLR pins to the PCB, they will form a resonant circuit with the Ceramic caps (10-22n on a microphone) which should be DIRECTLY on the pins.  If you are meticulous about not having supersonic funnies, you need to use a Zobel or other method to damp any peaks.
Indeed! Filtering out EMI/RFI is about filters. For proper design, one must know what is to be filtered out and the topology and impedances (in and out) of the circuit. Cookbook recipes may work, but they are known to contain many errors.
 

arnyk

Active member
Joined
Sep 13, 2016
Messages
42
ricardo said:
If you use 'good' inductors to go from the XLR pins to the PCB, they will form a resonant circuit with the Ceramic caps (10-22n on a microphone) which should be DIRECTLY on the pins.  If you are meticulous about not having supersonic funnies, you need to use a Zobel or other method to damp any peaks.

But Zobels are resonant circuits, too. They may be damped, but they are resonant circuits as well.

Because their leads like all wires have inductance, ceramic capacitors resonate, too.

Noticing that mic cables usually have substantial length and inductance, their contributions to all this shouldn't be ignored.

If mic circuits were truly impedance matched circuits, and all mic outputs, all mic preamp inputs, and all mic cables,  direct boxes, mixers and distribution boxes actually had the same characteristic impedance then  we would be in a different world.  But they aren't, and we can't ignore the substantial inductance of mic cables.

Because of their series resistance and damping, zobels don't attenuate the highest EMI frequencies  (now up near or above 1 GHz courtesy of cell phones and WiFi, )  as well as a simple appropriate sized parallel, low ESR cap with minimal length leads and good grounding.

 

ricardo

Well-known member
Joined
Apr 28, 2011
Messages
1,429
arnyk said:
Because of their series resistance and damping, zobels don't attenuate the highest EMI frequencies  (now up near or above 1 GHz courtesy of cell phones and WiFi, )
In this context, the purpose of the Zobel is not to attenuate EMI frequencies.  This is done by the Ceramic caps on the XLR pins and the inductors/beads which connect the XLR to the rest of the circuit.  They form a filter which , if 'good' inductors are used, will resonate with a high Q peak.

The Zobel here, is to damp this filter peak.

The type & length of cable etc also affects this peak so the Zobel is chosen to give 'good' performance at the limits of what may be connected to the device.  You may need other methods too .. to ensure all sensible cases are covered.
 
Top