I used a groundplane based PCB for the power amplifier just once. But it was a chip amp and the PCB was quite small. It worked/works well. For a big power amplifier, I would not use the groundplane concept, at least not for the whole PCB. But maybe yes for the input stage part of the power amplifier PCB.
I guess I'm glad this topic finally gained attention, 7 years later. 😉
Say we have a 1mm thick PCB. Traces in parallel on the top can be closer together than parallel traces on either side of the board. So say we have two parallel traces on the top that are 1mm spaced center to center, but the edges of the traces are 5mil apart. At what frequency will the current in the traces move over to the closest parts of the traces, rather than moving through the centers?
Say we have a 1mm thick PCB. Traces in parallel on the top can be closer together than parallel traces on either side of the board. So say we have two parallel traces on the top that are 1mm spaced center to center, but the edges of the traces are 5mil apart. At what frequency will the current in the traces move over to the closest parts of the traces, rather than moving through the centers?
I always use groundplanes with success. Nowadays with so low prices for fabricating PCB's is this possible, former times, when everything were etched and drilled at home this was impossible. I got good results on power boards (power amplifier) or low signal boards (MM_MC_preamp)
There are more details seen from Eagle files Top and Component sides with Ground Planes. I prefer this kind PCB’s for very low noise in term my ultrasonic schemes after 35 years researching low noise, ultra fast, linear, straight up to MHz high end CF power amplifiers, two step passive RIAA equalisation (0,1dB flat) filters and 3-way active filters for HighEnd (18˝ CerwinVega bass, 8˝ Audax middle, 1.5˝ Audax T34 TW) marble OB speakers.
with all those slots it hardly qualifies as a plane.
How does the current in the plane follow the trace below where the trace crosses over a slot?
How does the current in the plane follow the trace below where the trace crosses over a slot?
In Eagle you freely decide hatch or solid ground plane. Depend how tight you want, in my case I ever decide 40mil isolate and hatch profile.
I had read a paper where the person had carried out measurements on the effect of
1. slots (due other signal traces) in ground plane,
2. just through hole component pads breaking the continuity of plane and
3. uniform ground plane.
cant seem to find it again or dont know what particular google keywords I had used.
The result was, Slots had worst effect on return current flow, whereas through hole pads and uniform ground plane were the best.
I think he measured the voltage drop on ground points between which the current was flowing. Although I am not sure, as I have a just a vague memory of the test method used.
Another paper visualized return path, where at low frequency, current took the shortest path and as frequency increased, return current took the flow path.
my 2$$
1. slots (due other signal traces) in ground plane,
2. just through hole component pads breaking the continuity of plane and
3. uniform ground plane.
cant seem to find it again or dont know what particular google keywords I had used.
The result was, Slots had worst effect on return current flow, whereas through hole pads and uniform ground plane were the best.
I think he measured the voltage drop on ground points between which the current was flowing. Although I am not sure, as I have a just a vague memory of the test method used.
Another paper visualized return path, where at low frequency, current took the shortest path and as frequency increased, return current took the flow path.
my 2$$
Last edited:
it would be best to keep in mind that isolation should be decided based on voltage.
for example, mains would require minimum of 2.54mm or 100mil of clearance between live and neutral as per some standard.
HOWEVER, there are many standards and confusing requirements (for me) however. So I avoid making any further comments on isolation.
I do not see the advantage of hatched versus plain vanilla solid ground plane.
I have had very good success with power amp designs (and not only power amp designs) that have a solid ground plane under the routing layer. Altho many believe that return path at audio frequencies is not an issue, they seem to miss the point that it is highly desirable with good audio amplifiers to have a power bandwidth much much higher than just audible frequencies (20kHz). At such -slightly- elevated frequencies one can easily observe the return path currents in a ground plane using a handy current tracer. In doing so, it be comes evident that not having a solid ground layer under the forward path or one that is criss crossed by some traces is doing capital damage to the signal return paths and so will force the return current's unnecessary detours, in the worst case throughout the entire board. These detours effectively create many little RF slot antennae which will radiate the return signals EMI e.g. the Love quite nicely. These slot antennae will also happily pickup all sorts of stray fields, the amplifiers very own stray fields. Needless to say that most such designs using modern components tend to be unstable and lean to happy oscillation as many found out the hard way. Such amplifier with such terrible layout practice will pollute its very own signals and be a total mess. Sure the tendency of oscillation can be suppressed by adding pF capacitors everywhere (which incidentally help to provide AC (RF) return path for signals that are in desperate need of one) but what is the addition of such capacitors going to do to the power bandwidth of the amplifier ?
And what is a spoiled power bandwidth gonna do to THD ?
As Nigel #35 suggested, the layout should be done with care such that certain ground signals, generally those carrying high currents, will not connect straight into the ground plane but connect by a deliberate trace to the common ground reference point.
To sum it up the concept of a plane grounding in combination with star grounding gives very good results when applied in a sensible matter.
I think most hobby designers don't consider much the return path when layouting a board, they don't seem to understand that at DC one set of layout rules applies while on AC (or RF if you will) another set of design rules is applicable. They should instead familiarize them-self with the concept of return path and loop area and high speed layout practice guidelines than flat out dismiss or ridicule the application of a ground plane in audio amplifier designs.
Just because in the distant past with inferior components designers got away with single layer boards and no ground plane concept whatsoever does not mean that this is true today with modern superior components.
Ground plane or general plane layers is the 101 of modern design practice and imperative to adhering to EMI standards.
There is plenty of information on this subject found on the net. And if all fails a training course for EMI compatible design practice should open some eye's, I suppose.
And what is a spoiled power bandwidth gonna do to THD ?
As Nigel #35 suggested, the layout should be done with care such that certain ground signals, generally those carrying high currents, will not connect straight into the ground plane but connect by a deliberate trace to the common ground reference point.
To sum it up the concept of a plane grounding in combination with star grounding gives very good results when applied in a sensible matter.
I think most hobby designers don't consider much the return path when layouting a board, they don't seem to understand that at DC one set of layout rules applies while on AC (or RF if you will) another set of design rules is applicable. They should instead familiarize them-self with the concept of return path and loop area and high speed layout practice guidelines than flat out dismiss or ridicule the application of a ground plane in audio amplifier designs.
Just because in the distant past with inferior components designers got away with single layer boards and no ground plane concept whatsoever does not mean that this is true today with modern superior components.
Ground plane or general plane layers is the 101 of modern design practice and imperative to adhering to EMI standards.
There is plenty of information on this subject found on the net. And if all fails a training course for EMI compatible design practice should open some eye's, I suppose.
I prefer to use ground plane because I was a digital person in my engineer career. I'm sure there is no ground plane issue in a digital domain. PCBs for a digital circuit has no choice but multi-layer. It means they have ground plane. No ground plane, no performance in digital domain.
I recently posted my experience about a power amplifier having ground plane to another thread. I couldn't have positive feedback because an analog person usually dislikes multi-layer. They say no need for the audio band to use a ground plane; star ground is the best solution.
I thought so before. I built my first version with two-layer PCB without ground plane and the second one was with four-layer PCB with a ground plane. The latter is superior to the former because the latter has stable ground potential. You can connect ground wire of a scope wherever you want. Star ground or two-layer PCB isn't so. What does this mean for a circuit?
An analog circuit requires the stable ground potential. SIM is also. But in real PCB, this is a very difficult condition to achieve. Star ground is one solution to do this but means compromising way. An analog circuit can manage to work under unstable ground potential with a virtuoso performance. But now multi-layer design is an easy and cheap solution for everyone to achieve stable ground potential even if it needs some experience with PCB design, IMHO. No need to deny ground plane even in the audio band.
I recently posted my experience about a power amplifier having ground plane to another thread. I couldn't have positive feedback because an analog person usually dislikes multi-layer. They say no need for the audio band to use a ground plane; star ground is the best solution.
I thought so before. I built my first version with two-layer PCB without ground plane and the second one was with four-layer PCB with a ground plane. The latter is superior to the former because the latter has stable ground potential. You can connect ground wire of a scope wherever you want. Star ground or two-layer PCB isn't so. What does this mean for a circuit?
An analog circuit requires the stable ground potential. SIM is also. But in real PCB, this is a very difficult condition to achieve. Star ground is one solution to do this but means compromising way. An analog circuit can manage to work under unstable ground potential with a virtuoso performance. But now multi-layer design is an easy and cheap solution for everyone to achieve stable ground potential even if it needs some experience with PCB design, IMHO. No need to deny ground plane even in the audio band.
Thanks. When you plan a PCB there are so many rules and influences, not only ground planes at the end by finishing. GP are help to cancel parasite capacitance and horizontal induction from wires, but unfortunately it has spherical shape. Like you mentioned, there is a good measure to place the active layers between Com and Sol planes with only holes in GP. At first you have to planning IN side and OUT side, OUT side need many current, input have gentle small signal source with separate signal ground, normally 10R up from allover ground. First line, input, in my case everything is discreet with servo, need 1mA, second 3mA, third 7mA and drivers about 27mA at full voltage +-50V or even more. Output have to be open to full bandwidth, normally limited with capacitance of gates with choke resistor or transit speed of BJT types everything strictly DC with no capacitors. With about 30mA on drivers is enough current for both models. Then you do not forget on electronic brake for max current on, in my case compound style outputs, with negative compensated FETs, also extremely complicated VBE thermal stabilizer, where you are unable to divorce small signal and high current. With growing currents from 1-27mA into PCB you got first rule, transit frequency and fastness are higher and higher, you can not get any oscillations. I always aware in current feedback schemes the capacitor in feedback, while leading to oscillate everything. It is also critical to decide which pair or resistor are used to feedback. In my case with 3mA in noninverting side modules are ideal 5K6 via 180E, that mean about 32 gain factor, quite enough for 2x3 channels driving over active network. After all years of study Mr. Linkwitz exposing, this are the only way for good 3-D and micro details on site. One pair output transistors are also enough for brilliant reproduction, you need more fastness in amplifier vs. high amperage (W). If you send some amperes to speaker, you got immediately backward current, in this term you need very very fast amplifier with low output resistance to manage this phenomena in real time. At the end, this kind of amplifiers you can not sell or implement with normal user with lack of knowledge to start this kind of system.
Spacings, if in doubt use the IPC-2221 Table 6.4 figures.
IPC-2221B PCB Trace Spacing / Clearance by Voltage
I agree, don,t cross hatch a ground plane.
By the way a ground plane (or planes if you do splits) are a contiguous layer of copper, there are no holes or slots... If you have holes or slots it is termed a copper pour.
Love ground planes, been singing their praises for 30+ years, shed loads of info...
But providing a return current can be more problematic and often is, BUT it is the most critical part of any signal so requires consideration.
One scheme I like for power amps is:
Proper component placement, isolation of small signal and large signal blocks. Sensible signal flow across the PCB, in at one end out at'other no looping back and forth.
4 layer board, they are cheep these days, so its a no brainer.
Main power supply on separate board, isolate that B***** whether its linear or SMPS, then filter the DC coming in.
Layout the small signal stuff with a contiguous ground plane on the layer below the components, use SMD where possible (limits cuts in the ground plane and minimises loop areas), attempt to get ALL signals on the top layer, minimise layer changes.
Put power on layer 3, LDO isolated preferably keep that low level analogue clean, and smother the bottom layer with another ground pour, with plenty of stitching capacitors especially round the edges. Extend the low level analogue planes about 10mm past the circuitry block.
Power should be at the opposite end of the PCB with a clear demarcation between the power section and the low level stuff.
Power delivery from the PSU and power outputs, best routed as broadside coupled buss bars, on the 4 layers you would have thick tracks + - + -, this gives a low impedance, higher capacitance feed (the capacitance helps filter any high frequency noise) that you can control and isolate from all the low level stuff. Doing this right means you don't have to lift signal ground etc. to get a low noise design.
That's a simple description, depending on the complexity boards can take 1 to 4 weeks (and in one very complex 14 layer flexi rigid 8 months so far!!!) so don't expect miracles over night and when you are starting out, play with your placement and if your CAD system has rubber banding of the connections look at the patterns of connections with different placements, as you develop with practise you will be able to see how these route. Simple connectivity with minimum criss-crossing of signals is critical, if you can highlight nets colour your power nets, placements is a mixture between optimum signal layout and optimum power delivery AND the best return path you can create... forgetting that last point adds noise, ground loops and problems.
Love ground planes, been singing their praises for 30+ years, shed loads of info...
But providing a return current can be more problematic and often is, BUT it is the most critical part of any signal so requires consideration.
One scheme I like for power amps is:
Proper component placement, isolation of small signal and large signal blocks. Sensible signal flow across the PCB, in at one end out at'other no looping back and forth.
4 layer board, they are cheep these days, so its a no brainer.
Main power supply on separate board, isolate that B***** whether its linear or SMPS, then filter the DC coming in.
Layout the small signal stuff with a contiguous ground plane on the layer below the components, use SMD where possible (limits cuts in the ground plane and minimises loop areas), attempt to get ALL signals on the top layer, minimise layer changes.
Put power on layer 3, LDO isolated preferably keep that low level analogue clean, and smother the bottom layer with another ground pour, with plenty of stitching capacitors especially round the edges. Extend the low level analogue planes about 10mm past the circuitry block.
Power should be at the opposite end of the PCB with a clear demarcation between the power section and the low level stuff.
Power delivery from the PSU and power outputs, best routed as broadside coupled buss bars, on the 4 layers you would have thick tracks + - + -, this gives a low impedance, higher capacitance feed (the capacitance helps filter any high frequency noise) that you can control and isolate from all the low level stuff. Doing this right means you don't have to lift signal ground etc. to get a low noise design.
That's a simple description, depending on the complexity boards can take 1 to 4 weeks (and in one very complex 14 layer flexi rigid 8 months so far!!!) so don't expect miracles over night and when you are starting out, play with your placement and if your CAD system has rubber banding of the connections look at the patterns of connections with different placements, as you develop with practise you will be able to see how these route. Simple connectivity with minimum criss-crossing of signals is critical, if you can highlight nets colour your power nets, placements is a mixture between optimum signal layout and optimum power delivery AND the best return path you can create... forgetting that last point adds noise, ground loops and problems.
I should have also mentioned, that if for some reason the ground plane must be opened, as some modern components can require to have no ground plane underneath (such as certain OpAmps for example), the layout designer must make sure that traces do not cross over the open part of the ground plane.
Generally, pretty much most audio amplifier circuits are simple enough and can easily be placed on dual layer boards while having a single ground plane. Of course a single routing layer is a routing challenge when using SMD components. If you hate to use jumper wire bridges the next stop is then a 4 layer board.
Using through hole components an experienced designer can easily manage to route an amplifier board on a single routing layer. Generally I do not cut, open or interrupt my ground plane but if I absolutely have to I make exception after very careful consideration of the alternate options and signal return paths.
Deliberately separating an AC signal froward path from its return path is the 101 of microwave slot antenna theory.
I do not believe in audio amplifiers it is desirable to transmit or receive any RF signals, therefor integrated slot antennae designs should be avoided. 😀
Generally, pretty much most audio amplifier circuits are simple enough and can easily be placed on dual layer boards while having a single ground plane. Of course a single routing layer is a routing challenge when using SMD components. If you hate to use jumper wire bridges the next stop is then a 4 layer board.
Using through hole components an experienced designer can easily manage to route an amplifier board on a single routing layer. Generally I do not cut, open or interrupt my ground plane but if I absolutely have to I make exception after very careful consideration of the alternate options and signal return paths.
Deliberately separating an AC signal froward path from its return path is the 101 of microwave slot antenna theory.
I do not believe in audio amplifiers it is desirable to transmit or receive any RF signals, therefor integrated slot antennae designs should be avoided. 😀
Some very high speed op-amps require the ground clearing under input pins to minimise parasitic capacitance, unlikely that that speed will be reached for domestic audio designs.
What do you mean Deliberately separating an AC signal forward path from its return path, you never separate a signal from its return path.
My idea has mainly been to route an excellent single layer board (which causes me troubles with no end), and then use a ground pour for the bottom layer. As I understand, if your ground plane is connected to only one node, so no currents can be forced through it, then it can only short out fields it intercepts, and cannot retransmit any of that energy. To me that means that a ground plane with only 1 connection to the rest of the circuit cannot in any way be worse than the same circuit without the ground plane. But the requirement of only 1 connection is very strict.
Of course, I could just be ignorant.
Of course, I could just be ignorant.
Oh yes you do, if you want to build a decent slot antenna, you do that on purpose and you would tune to 1/2 wavelength for best results.🙂
Reference
http://wireless.ictp.it/school_2007/lectures/Struzak/5Anten_theor_basics.pdf
-The basic makings of a slot antenna-
Page 7
"Any slot / opening in the screen of a device / cable carrying RF current"
Page 8
"Any discontinuity in transmission medium"
Page 14
Slot antenna: a slot is cut from a large (relative to the slot length) metal plate. The center conductor of the feeding coaxial cable is connected to one side of the slot, and the outside conductor of the cable - to the other side of the slot.
Page 15 Traditional Slot
Reference the image Page 15 Traditional Slot and the text Page 14 describing the connection.
Do you see how the signal forward path has been purposely and deliberately separated from the return path ?
Do you see that the AC current would have to flow around the slot from the center conductor to the shield conductor as described ?
Same would be true if you choose not to twist (physically separate) the conductors of a forward and returning signal or for example if you choose to separate these two conductors from each-other physically, here I believe Loop antenna theory would be applicable as you would purposely increase the loop area.
This is the reason why we would want to twist wires (e.g. not separated), to minimize as much as possible the loop area and have at best a very very poor loop antenna. Forward and returning AC signal conductors in physical close proximity e.g. twisted such that would cancel each-others magnetic fields then also reduces unintended radiation.
Also we want to avoid having any slots / cuts / openings in a ground plane to avoid incidental and unwanted slot antenna action if a signal is routed over such slot.