Zarcon Dee Grissom's Idea Page
EV9
Updated 21DEC2011
M.C.C.HPA17r16 iconNot RoHS compliant logo 4.Home > IdeasHPA17r16 Intro.
HPA17r16 Operating Environment.
HPA17r16 RFI filters.
The Main Amplifier Board.
HPA17r16 Vmon & Sources.
HPA17r16 Tests & Afterthoughts.
The Main Amplifier Board.
RFI can have adverse effects on circuits incapable of handling them. There are ways, to prevent this.
Main Menu
Home

Ideas

Meet the computers

About Me

Links

Idea#017 Page 4 of 6 -> On this page; introduction below - Feedback Loop Capacitor - PCB trace and feture layout - Input & Output decouplingMB power system - V-reg & V-ref - MB Ground Plane - and Thermals

I settled on the op-amp and buffer topology for many reasons. Most of all, I didn't want to build a single-chip solution, or re-invent an amplifier from the ground up. I wanted to build a solid amplifier, and focus on making the circuit RFI immune. Single chip solutions have very limited options for improving there circuit, and most lack dedicated power pins for the finals and voltage gain stage. Also I wanted to use good headphones at my desk, and most single-chip Headphone amplifiers max out at five volts supply. Inadequate for driving most high-end headphones.

I reference the PIMETA v1.0 from tangentsoft, because it was the only op-amp and BUF634 design I came across with any kind of detailed description of the circuit design. The PIMETA was designed for a completely different role then this amplifier, so please don't think that I think less of the PIMETA. It was the only other source for insight other then the BUF634 datasheet, and weighed heavenly in the selection of component values for this amplifier.


The Main Amplifier board.
Introducing, The Almost-Cryogenically-Cooled, LNA for Audio.
Original shot
MB inside RFI box
640x480
1024x768
Adjusted feedback loop
Adjusted feedback-loop
640x480
1024x468
The M.B.
The M.B.
640x480
1024x768
PCB sim
HPA17r16mb sim
150 dpi [602x482]
300dpi [1203x963]
Now that I have gone out of my way to ensure this amplifier board is only dealing with accurately amplifying audio, and nothing else. Now to look inside the copper RFI shielding box inside the outer layers of filters and shielding, and enter the inner sanctuary of the amplifier.
BUF634 fig5BUF634 Fig4
The fundamental amplifier is based on two separate schematics in the BUF634 Datasheet [PDF 796,111 bytes]. Figure 5 shows an audio amplifier, and Figure 4 describes using different op-amps with the buffer. The big difference being the picofarad capacitor tyeing the "Interstage" to the Feedback loop.
HPA17r16mb Audio amplifier stages 10pfSome traces and 10pF capacitor
Not that much of a deal with audio frequencies, however the capacitor helps stabilize the circuit at higher frequencies. This has something to do with the Envelope Delay of the buffer within the feedback loop of the op-amp. The data sheet calls for a 220pF capacitor, ultimately being to large for the impedance across the Interstage and Feedback loop of this amplifier. Tangentsoft used a 10pF capacitor on the ground channel of the PIMETA. I have yet to decide if a larger capacitor at the sacrifice of slew-rate is at all required or necessary to build an RFI hardened circuit.

Feedback loop Capacitor vs Frequency ResponseFurther testing reveals that the 10pF capacitor reduces the audio gain starting at about 16kHz through about -0.00025db at 20kHz. Some audiophiles may complain about that, and in agreement with Tangentsoft a slightly smaller capacitor like 8.2pF through 4.7pF would remedy that. I am not concerned about such a minuscule high end reduction, given that most sound cards and radios are infinitely worse in there high end frequency characteristics. The smaller capacitor reduces the RF tolerance of the circuit, so if I ever reduce the capacitor, a 6.8pF capacitor is the best for my requirements.

sig tracesThe input pins on the OPA2134 have a nominal impedance of ten tera-ohms (10E13 ohms), the terminators are 470 kilohms (47E4 ohms). Guess which one would fare better at the ends of the trace, errr microscopic straight-wire antenna. With the lower impedance devices nailing down the ends of the trace, the ultra high impedance input pin is better protected from transient effects in between the other lower impedance devices. The same goes with the eight Meg-ohm input pin of the buffer on the Interstage trace.

Guard ring
Guard ring schem
Know thy circuit, before attaching guard rings. Guard Rings can limit stray voltage effects, provided the circuit there connected to is capable of handling the load. Ideally in a non-inverting amplifier, the guard ring would be connected to the negative input pin of the op-amp (The feedback loop). In this amplifier the feedback loop trace would get dragged all over the place by the guard ring, causing all kinds of adverse effects. In this amplifier, the ultimate place to connect the guard ring to is directly to that channel's Vref. Unfortunately there is a ground plane and/or a power trace in the way. So the only stable place to connect the guard ring to without killing the ground plane, is the output pin of the op-amp. It is not perfect, however it is far better then the alternative.

Grounded Coplanar Waveguide
The non-reflective Grounded Coplanar Waveguide, can prevent waveform reflections from adding up at the device input. If implemented correctly with a slight gradual increase in trace capacitance and a slight gradual decrease in impedance at all the device pins. These spots in the length of the trace can absorb some RFI energy before it gets to the device input and prevent the RF from reflecting off the device pin back down the length of the trace.

HPA17r16mb Audio amplifier stages I/O decoupling
I didn't have a choice with the input decoupling capacitors, given the source of power. I went with the ridiculously large farad values to get the majority of the capacitors frequency dependent phase delay out of the audio band. I decided to go with the decoupled output rather then a floating ground for many reasons. The doubled up output capacitors are to reduce there total ESL and ESR effects on the output signal.


HPA17r16mb Power
The on board capacitors are more for decoupling and power reserve, then power supply filtering. With the large electrolytic capacitors on the board, the chips are not limited by what they can get through the inductance of the power wires going to the board. I wanted enough reserve power on the board to keep the voltage steady through whatever draw the buffers may pool. And again, the multiple electrolytic capacitors are to limit there total ESL and ESR effects.

The forward diode between the P-amp Power and the Op-Amp Power (OAP+), is to limit any voltage fluctuation at the op-amps caused by the draw of the buffers (or bus fluctuations). The reverse diode is to prevent the Op-Amp from driving a voltage significantly higher then the buffers V+ into the buffers input pin.

I chose the diodes over a toroid for a few reasons. The OAP+ voltage drift is limited, not removed regardless of what is used to isolate the power rails. A toroid will only limit high frequencies, a diode is frequency independent provided the primary OAP+ capacitors are large enough. A toroid will not prevent the buffers from draining the OAP+ capacitors under heavy low frequency loads, a forward diode will limit this draining effect regardless of frequency, limited by the reverse diode and OAP+ caps. I would have liked to place two reverse diodes here in series, to increase the potential isolation, there wasn't room on the prototype. I wanted to go with two 25 volt 560 micro farad ESRL capacitors on the OAP+ rail, they were not available when I ordered the parts, and there was not room for two one-thousand micro farad cans here on the prototype.

HPA17r16mb V-RefRemember, in this amplifiers world, everything is in reference to ground, from the input, to the output. This in combination with a lack of guarantee that the power voltage will remain the same under all circumstances, a voltage divider is the last thing I want in this world. I don't want a mid point, I need a fixed voltage reference (Vref) relative to the negative rail (Ground). This followed by the need to gradually rise the Vref during power up, to prevent a surge of power flowing out the input and outputs. I also didn't want the feedback loops and terminators of each channel, effecting the other channel's Vref. Hence the independent Vref buffers for each channel.

The independent Vref buffers allowed me to use separate chips for each channel, this has many advantages. The biggest stems from each chips op-amp functions. While one op-amp is pooling the output signal high, the other op-amp is responding to the change on the Vref by pooling that down, and vise versa for the negative side of a signal. In essence a semi balanced load on the power for each chip, albeit attenuated by the massive capacitors on the Vref buffers output. There is also a complete isolation of the audio signal, with the exception of the shared power rails and ground plane between the op-amps. Speaking of ground plane...

Some forget about the other part of there circuit, the ground plane.
Sim ground plane
HPA17r16mb sim ground plane
Ground plane Jumpers
HPA17r16mb ground plane Jumpers
Ground plane Jumpers
HPA17r16mb ground plane jumpers
The ground plane is not just a means to return current to the power source, it also shields the sensitive traces on the underside of the PCB among other things in this amplifier.
This prototype has some nasty voids in the ground plane. The worst runs between the op-amps ground pins and the buffers, occupied by the OAP+ power trace. This is compounded by a input signal trace running at right angles to the power trace on the other side of the board. I mended this by adding bare copper jumpers across these voids on the side with some ground plane remaining, to reinforce the ground plane at these points. Some call this cheating, or simply pore design. I call it a creative way to get around redesigning the entire board sacrificing everything else, just to save the ground plane.

The other jumpers leading from the chips to three of the six mounting points, are for thermal dissipation. This board is not cooled by internal convection, this board is cooled by direct conduction to the chassis by the PCB mounts. The four corners, and a mount at each buffer, are connected with 0.5in OD copper washers to a 0.025inch thick copper sheet that is mounted flush to the chassis. The mounting bolts go through the whole thing. The chassis and copper sheet provides thermal dissipation, and significant mechanical support for the boards.
Buf634 mount
This arrangement gives me a better then 0.538 C/W thermal connection for each buffer (From Buffer Tab to Chassis). Provided I could have found the chassis I wanted, you can't get much better then this without Cryogenic cooling. If I had simply mounted each buffer to a heat sink inside the chassis, there would still be the challenge of getting that heat out of the chassis, Assuming there is an atmosphere in the chassis to cool the heat sinks.

Referring back to the BUF634 mount cross-section view diagram above, I was stunned by how thin the copper layer is on a 1oz/square-inch PCB when I made that diagram, just one pixel thick to the approximate scale of everything else. It best illustrates why heat can be moved threw a PCB from one side to the other far better then latterly along one side. The amount of copper moving heat along one surface is so thin, that it has a limited thermal dispersion range from a heat source. A grid of 20mil plated through holes 0.05 inches on center, has far more conductive copper threw the PCB from top to bottom, then the same aria of PCB conducting heat laterally along it's surface with no plated through holes. Another thing is something as small as a 22AWG wire soldered across the surface of a PCB, drastically increases the amount of copper to move heat across the PCB. During the prototype design stage, I was not able to find a simply way to insert copper wire segments or a "Thermal slug" in the plated through holes under the buffers, Perhaps in the future I will find a way to do this and preserve the flat mounting surfaces for the parts.

Full MB schem.
HPA17r16mb Schem full
Larger Schem T.B.D.


Sources referenced, during the design of the main board.

I would be omissive if I left out Tangentsoft, and the collection of referenced articles at Tangentsoft's site. [External Link]

Another site being Microwaves101. [External Link]

Eur Ing Keith Armstrong CEng MIEE MIEEE, Design Techniques for EMC & Signal Integrity Part 1 through 6, Partner, Cherry Clough Consultants, Associate of EMC-UK.

And, The staff at NEPP [External Link], and JPL [External Link].


Other Sources referenced or "kept in the back of my mind", during the design of the main board.

Louis Fan Fei, "Developing Designs For CMOS Power Amplifiers", Microwaves&RF, ED Online ID #17507, November 2007. [External Link]

B. Aja, E. Artal, MA L. De La Fuente, M. Detratti, J.P. Pascual, "Equalize Gain In Millimeter-Wave Amplifiers", Microwaves&RF, April 2005. [External Link]

Ian Piper, Al Ward, "Design An E-pHEMT 4.9-to-6.0-GHz LNA", Microwaves&RF, ED Online ID #10135, April 2005. [External Link]

Saman Asgaran and M. Jamal Deen, 
Design of the Input Matching Network of RF CMOS LNAs for Low-Power Operation, IEEE Transactions On Circuits And Systems, Vol. 54, No. 3, March 2007, pp. 544-554.

List T.B.C

(Excellent impedance matching article, with lots of formulas. Came out after this board left for the fab)
Baimei Liu,  Chunhua Wang, "Low-Power LNA Drops Noise At 2.4 GHz", Microwaves&RF, ED Online ID #22405, February 2010. [External Link]


<PG3 (RFI filters)< - ^Ideas^ - >PG5 (Vmon & Sources)>

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