Zarcon Dee Grissom's Idea Page
EV9
Updated 14DEC2015
M.C.C.HPA17r16 iconNot RoHS compliant logo. 3Home > IdeasHPA17r16 Intro.
HPA17r16 Operating Environment.
HPA17r16 RFI filters.
The Main Amplifier Board.
HPA17r16 Vmon & Sources.
HPA17r16 Tests & Afterthoughts.
M.C.C. HPA17r16 RFI filters.
Why havent others done this? The cost of adding them to a product.
Main Menu
Home

Ideas

Meet the computers

About Me

Links

Idea#017 Page 3 of 6 -> On this page; RFI shild overview below - Reducing noise sources from within - Inoput RFI filter - Output RFI filter - Output v2 - and DC Power filter.

The LNA mentality, Remove every noise source possible, and filter out everything we don't want, so we can concentrate on accurate reproduction of only the signal of interest.

"Good filters" don't need to be huge, just effective at what they do. A good filter setup should reduce any RFI emitted from the external wiring (outside world) or internal wiring inside the "dirty box", from contaminating the "inside world" of the shielded enclosure beyond the filter. This can be done with divided shielded zones with a "waveguide below cut-off" between them. A good shield will have sufficient "Skin depth" to block/divert/absorb most of the frequencies it is expected to keep out of the enclosure. A good shielded enclosure will also include the ability to not resonate, unfortunately every metal box has at least one harmonic frequency. The solution is to line the inside of the box with sufficient RF absorbing foam, to absorb a sufficient amount RF bouncing around in the metal cavity.

HPA17r16 RFI Shields and RFI Sources
A good RFI filter is only as good as it's surroundings. An RFI filter is useless if there is an easy path for RFI to bypass the RFI filter and contaminate the sensitive stuff beyond the filter.
Bad Filter Use
One way to prevent the RF from contaminating neighboring wires is to use shielded wires, however that is not always possible, and ineffective if the RF can directly get to the sensitive stuff. So clearly an RF barrier or shield should be between the filters and the sensitive stuff. However there are going to be holes in that shield for wires to enter and exit, That is a path that RF can squeeze past your RFI protection and rain havoc. That is where the waveguide below cut-off comes in, the principle is quite simple.
waveguide below cutoff vs none
There is a a minimum RF frequency that the wires going to the RFI filter can efficiently emit RF into the dirty box. Between the dirty box and the inside world, the wires enter threw a wave guide that is tuned as a low pass filter below the lower limit of the incoming wires emissions. Any high frequencies that get emitted by the incoming wires are reduced by the waveguide before they can get inside the protected zone, and the wires are to short to emit any substantial emissions that would easily pass threw the wave guide.

stuffed holeThe waveguide below cut-off is not the only approach, there is another way involving a grounded screen of sorts. The trick involves using the smallest opening possible, and stuffing it full of ground wires, or shielded wires with the shield grounded. The abundance of grounded wire going threw the small opening acts as a screen of sorts, not allowing an opening large enough to let RF enter the opening easily. This dose not guarantee that the ambient RF can not induce some current on the wires going into the enclosure, it will only prevent most of the ambient RF from passing right through the opening.
There are other methods and fittings for passing wires through a shield, and some are more effective at keeping RFI out then others. The best we can do is limit the majority of noise, and at that, we eventual reach a point where we are overwhelmed by thermal noise and noise from our protected circuits. Even these seemingly "Always there all the time" noise sources can be reduced.

Reducing noise sources from within.
Thermal noise is just that, it is noise produced by the random movement of molecules in a heated object. When the molecules move, the electrons they contain move, and the moving electrons produce some electro-magnetic waves. The hotter an object is, the more the molecules move, and the more intense the electro-magnetic waves are from the moving electrons on those molecules. The ultimate solution seams simple enough, just cool the object down till there is no more heat or movement. That however would require cooling the object down to -459.67F (-273.15C). That's pretty damn cold, and difficult to achieve.
Thermal noise mathThermal noise vs Resistance

What we can do, is make sure that every single source of heat in our chassis is well connected thermally to the classes, so the heat can be removed from the inside of the chassis easily. That keeps the background thermal noise inside our chassis as close to the outside thermal noise level as possible, and reduces excess thermal noise sources. Using solid ground-planes and mounting every PCB to the chassis with copper washers instead of brass stand-offs greatly improves the thermal path out of the chassis, and ultimately reduces the PCB's thermal noise. Just because a part's data sheet states that it can operate at up to 150 degrees Celsius, dose not imply that it will not produce a ton of thermal noise operating at that temperature. All it takes is one hot resistor or chip to produce allot of thermal noise inside the shielded enclosure, so we don't want anything in the chassis operating above room temperature. Even the lowly inductor or capacitor in the RFI filter produces heat that must be dealt with to keep thermal noise at a minimum.
BUF634 bandwidth pin resistor thermal noise

There are other considerations with heat as well. If we are trying to keep thermal noise at a minimum, the last thing we want is a hot PCB trace or wire. Most guides for trace width and wire gage are based on safe Joule heating vs thermal dissipation rate, not coolest operating. The best we can do is balance between thicker conductor vs diminishing thermal noise reduction ratio. Eventually we reach a point where the extra mass at room temperature is emitting more thermal noise then what was reduced from electrical heating. There are practical physical size limitations as well. Remember, even if were dealing with ultra high frequencies stuck on the conductors skin, the non conducting core of the conductor will help move heat away from the skin of the conductor, and cool the wire more efficiently over time.

Thermal noise is not the only noise source in our chassis, and some devices naturally produce emissions as part of how they work. Obviously switching regulators, rectifiers, oscillators, digital circuits and bar inductors produce allot of noise, and should be kept away from the sensitive circuity if not in a separate  shielded enclosure. Resistors and Op-Amps also produce noise, and they are kind of necessary for a headphone amplifier. The noise from these devices can directly effect devices right next to them, and emit noise that gets reflected off the RFI shield back down on components on the other side of the PCB.
Noise from within a shielded enclosure
I specify EMI in the diagram rather then RFI, as RFI sometimes implies high frequency radio emissions, rather then interference magnetic and/or electric in nature of ANY frequency including DC.
The reflected noise can easily be dealt with by lining the inside of the RF enclosure with RF-absorbing-foam. Unfortunately RF absorbing foam only reduces RF, it dose not eliminate it. However we are already talking about extremely weak RF at this point, so it dose not take much to deal with what is produced by the op-amps compensating for there outputs not being dead-on or resistor thermal noise.
Simplistic RF absorber concept
Given that the thermal dissipation is through the PCB mounts and not internal convection, we can fill the shielded box from the lid right down to on top of the components. Possibly with some foam providing RF protection between noisy components and there neighbors. RF absorbing foam is somewhat conductive in nature, so care will need to be taken to keep it off of power and signal traces, as well as component leads. RF absorbing foam will also effect the impedance of signal wires and traces if it is to close to them, so we can not just fill every single void in the box with the stuff.

Bob - Why did they put this stupid piece of packing foam in there? It's useless!
Steve - Don't throw away that "Useless piece of packing foam", it's not packing foam.
C-RAM LFPictured left is C-RAM LF, it is RF absorbing foam that will not shed carbon powder all over the PCB, and is available in weather resistant varieties. Ideal for use in this particular headphone amplifier.
There is one downside to using RF absorbing foam, it absorbs RF by converting it to heat, and that heat results in thermal noise. So we still need to do everything else we can to keep the RF at a minimum in the shielded enclosure.

There are other noise sources that do effect amplifier performance, if not permanently degrade them. I mention that this amplifier is going to be mounted right next to a monitors degauss winding. That can be dealt with by making all the signal paths immune to external magnetic sources, like a twisted pair Ethernet cable. I could also implement a magnetic field diverter around the outside of the entire amplifier. I have not had a chance to look into how well Mu-Metal conducts heat, so it is a questionable approach for me at this time.

There are other noise sources that approach the level of being ridicules, like cosmic rays. I could tell the RoHS crowd to go pound sand and line the entire amplifier in lead (Pb), however lead (Pb) is a lousy heat conductor. Another outstanding cosmic ray shield is High Density Polyethylene (HDPE) or watter, however so is the are we breath. So long as there is enough volume of air around the amplifier for us to keep breathing here on planet earth, the amplifier will not get exposed to high enough cosmic ray levels to effect performance. So we don't need to take the Dremel to that Polyethylene cutting board in the kitchen to make it fit in the amplifier chassis, or stuff the thing in many layers of zip-lock bags. I think it is a good time to stop here and move on before the rat hole gets any worse.

Now that we have covered a majority (Not All) of the possible types of interference that this amplifier will have to deal with, we can move on to the RFI filters for the wires going in and out of the amplifier.


HPA17r16i, Intput 47kohm 133kHz LPF
HPA17r16i chart: vswr, trans, imp, E-delay
800x415
input 47kohm RFI filter schem.
no larger size.
I wanted to set the RCA input "T" Low-Pass-filters at just under 44.1kHz however there wasn't a shielded inductor of addiquite microhenries to do so at 47k ohm impedance. I settled for 133kHz LPF, as the 07M type 56mH inductors were available, shielded, and physically small. The other reason for not going lower, is the inductors phase shift effect is mostly kept out of the audio band with the filter set to a higher frequency.
HPA17r16i_PCBsim
630x480
787x600
*not to any standard printing dpi
Glamor shot
Glamor Shot
640x480
1024x768
Placement shot
Audio Input RFI filter
640x480
1024x768
folded line capacitorFolded line low-pass filterThe other features on the bottom of the board are not ancient magic glyphs, nor some mixture of Maxwell's equations and wizards magic. There "Distributed Element filter" folded line capacitors, and "folded line low pass filters". I placed these filters on the board to take over at frequencies above the 56mH inductors optimum operating frequency. I didn't create the folded line low pass filters, I adjusted the dimensions of a 1.5GHz low pass filter on CST's website to operate at the required frequency for this filter.
MuMetal shield
The inductors are shielded to reduce there magnetic emissions, however they are still bar-inductors. If an external changing magnetic field is properly aliened with the inductors, they may act like tape-heads. To prevent this from happening I wanted to place a MuMetal box around each channels inductors. MuMetal sheets are not cheep (around Two-Hundred USD and up for "Engineering sample Kits"), and this may never happen. It all depends how sensitive they will be next to my monitors. The ferrite encasing around the inductors may be addiquite for the job.

Revision 1A
filter/soundcard interaction
Sound card interaction with the reflection.
The Fix, Revision 1a.
Revision 1a

I adjusted this filter to absorb the reflected signals to tame the SWR induced increase in the sound cards upper frequency range. The 10k-ohm resistor across the first inductor is to reduce it's reflectivity a bit, and the 22k-ohm resistor and 147 picofarad (100pF+47pF) capacitor before the first inductor is to absorb what is reflected by the first inductor. The 470k-ohm resistor across the fifty picofarad capacitor after the first inductor, is to tame some phase angle effects in the 7MHz range. I will eventually make a new board to accommodate the additional components. However, the current filter PCB seams to respond fine with the placement of the additional components.
x-talk above 1.5kHzThere is a small amount of crosstalk starting at 5kHz and up, that is probably from magnetic coupling of the inductors. The crosstalk disappears when this board is bypassed, and is insignificant compared to the crosstalk of the source sound card. About 0.0dB@5kHz to 4dB@43kHz over the baseline of the soundcard.

Revision 1B or version 2.0.
47kohm input RFI filter Revision 1B schem.47kohm input RFI filter Revision 1B Response.HPA17r16i Revision-1B is in the works, more parts need to be ordered at this point. The new 15mH inductors, and a replacement for the 50pF capacitors have already arrived. I need more resistors at this point
I discovered when I bypassed the input filter to work on the output filter, the IMD+N dropped considerably. The more the inductors were allowed to resonate, the worse the THD+N and IMD+N became. Even tho Revision-1A fixed most of the filters disappointments, the IMD+N can be improved. So I have started to look into other ways to keep the inductors from acting up. One promising method seams to be locally dropping the impedance across the inductors above a set frequency with a resistor and capacitor across each inductor. 
Prototype Input RFI filter v2.0 messPrototype Input RFI filter v2.0 new caps
I dropped the first inductor from 56mH down to just 15mH, as this seams to have drastically dealt with the IMD+N issue. I then used a resistor to drop it's localized impedance down to "32k-ohms plus Xc", to get it's cut-off frequency below it's resonant frequency. I also used a similar setup to tame the second inductors resonance, with considerable fine tuning resistor adjustments. The abundance of fine tuning resistors have taken there tole tho. It's time to replace all them mismatched things with just one resistor, along with the doubled up ceramic capacitors.



HPA17r16o, Output 40ohm 41.2kHz errr 700kHz LPF
Old filter
Old 40ohm LPF Schem
Revision-1a Output RFI Filter
New 40ohm LPF Schem

The output filter was easier to get a lower frequency cut off, Due to the much lower impedance. However it needed to work with a wide range of impedances from 40ohms to 100 ohms. The other requirement is not excessively loading down or reflecting signals back at the amplifiers output buffers within there operating frequency range (DC to 210MHz).
It took considerable time adjusting values of parts to get the filter close. I was not entirely happy with the VSWR spike at about 15kHz, however I thought this was irrelevant at audio frequencies. In lieu of the flat response on the other charts, I was not inclined to change the filter until I saw how the VSWR effected the output at 15kHz. No one I asked new if the VSWR spike would adversely effect an audio amplifier, now we know.

The 1.35 to 1 VSWR at 15khz, caused a +7dB increase at 15kHz in the amplifiers feedback loop. The other +1.5dB spike just before this filters cut-off was caused by the old revision of the 133kHz input filter.

Old filter charts
VSWR vs Headphone Impedance
VSWR vs Output Impedance
Input Impedance vs Headphone Impedance
Input Impedance vs Output Impedance
Transmission dB vs Headphone Impedance
Transmission, dB vs Output Impedance
Envelope Delay vs Headphone Impedance
Envelope Delay vs Output Impedance
Old Effect
The old LPF cut-off started roughly at 41.2kHz at 40 ohms, and climbs slightly with higher impedances headphones. The temporary replacement filter is completely different with a slope approximately from 30kHz down to a -6db shelf from 300kHz out to about 700kHz, followed by a -24db/octave slope that steepens significantly at about 350Mz.
PCB sim-cap
HPA17r16o_PCBsim
571x480
753x633 eg 300dpi
Old filter
Headphone jack RFI filter
640x480
1024x768
New filter (revision-1a)
New Filter
640x480
1024x758
Placement shot
HPA17 output RFI filter
640x480
1024x768
I got the idea for the folded Low-pass filter stage on the input of this filter, from the filters in the 2-meeter amplifier on the previous page. The one-eighty degree bend acts as an inductor of sorts to keep ultra high frequencies from the headphones from getting past the disc capacitor. The disc capacitor shunts most of the ultra high frequencies. The other capacitors are be hind a ten ohm resistor to keep them from shorting out the upper frequency range of the amplifiers output buffers, and to absorbs any signal the larger inductors reflect back at the amplifiers output. The rest of the filter is a regular malty stage low pass filter. I decided to place the last stage capacitor directly on the headphone jack, to deal with most of the RFI before it gets very far into the amplifier.

Revision-1B.
I have yet had time to build or test any possible upgrades to revision 1-A. I am not entirely happy with the filter cut-off being way above 40kHz, I will address this after the input filter is done. In the mean time, revision 1-A dose work.

HPA17r16o Revision-1B is a minor adjustment to Revision-1A, According to the simulation software. A RN65D 162 ohm resistor is placed across each of the 5.6 uH inductors, and the 0.01uF Polypropylene Metallized Film Capacitors on the headphone jack are replaced with 12pF 1kV COG/NPO ceramic disc capacitors. It has not been field tested yet.

Revision 2.0 (Dec 2015)
I've decided to completely ditch the former output RFI filter in favor of a topoligy similar to the input filter for many reasons.

HPA17r16o v2
I'm still fine tuning the particular component values, however the simulations and testing so far is very promising.
HPA17r16oV2_Output at 39ohmsHPA17r16oV2_Output at 133ohms
The Preliminary is looking much better then all the other filter types I've tried so far. So it is time to make a new PCB.



DC Power RFI filter
As for the power filter, I find that with most low power stuff here (less them 1000mA), a single 1uf metallized polypropylene film capacitor, and a diode is more then addiquite to dampen the 4kHz PWM noise at the device. The diode prevents the capacitor from trying to stabilize the entire power bus, so it can provide more stable power to the device.
Simple DC filter
For this amplifier, I wanted something a little more robust.
DC-in RFI filter schem
A lot of low impedance capacitors, with some inductors in the mix. Then a quick check to make sure the hole thing will not oscillate. The inductors, also reduce the initial load on the power buss, when power is applied.
PCB sim-cap
HPA17r16pf1_PCBsim
858x453 eg 300dpi
Glamor shot
DC-in RFI filter
640x480
1024x768
Placement shot
DC power RFI filter
640x480
1024x768


DC Power Riple filter
DC-in Riple filter schem
This board is basically some more 1uF caps after the second toroid on the previous board, to further reduce HF ripple in the power. The arc-suppressor is to limit the spike of the power switch arc when it is switched on or off.
PCB sim-cap
HPA17r16pf2_PCBsim
602x227 eg 150dpi
1204x453 eg 300dpi
Glamor shot
DC-in Riple filter
640x480
1024x768
Placement shot
DC power riple filter
640x480
1024x768


Credit for the Distributed element filters go to;

Wikipedia "Distributed element filter" 7 May 2010  [External Link]

CST - Computer Simulation Technology, "Microstrip Bandstop and Lowpass Filters" Article ID: 248, 6. Jan 2006  [External Link]

H. Peddibhotla and R.K. Settaluri, “Compact Folded-line Bandstop and Lowpass Filters,” Micro. Optical Tech. Letters, vol. 42, issue 1,  pp.44-46,May 2004.

H. Peddibhotla and R.K. Settaluri, “Miniaturized High Performance Lowpass and Bandstop Filters for Wireless Applications,” Proc. IMAPS Conf. on Ceramic Interconnect Tech., Apr. 2003.

J. Brian Thomas, “Cross Coupling in Coaxial Cavity Filters: A Tutorial Overview”, Apr 2003

And the staff at Deep Space Network.

<PG2 (Operating Environment)< - ^Ideas^ - >PG4 (The Main Amplifier Board)>

Home > IdeasHPA17r16, Intro.
HPA17r16 Operating Environment.
HPA17r16 RFI filters.
The Main Amplifier Board.
HPA17r16 Vmon & Sources.
HPA17r16 Tests & Afterthoughts.

Valid XHTML 1.0
My Email link. copy and past. zarcondeegrissom@yahoo.com  
or is the status "coffee is good" on yahoo messenger