Distributed Antenna System PCB Design: PIM, Multiband Isolation and High-Frequency Laminate Selection for In-Building Coverage

2026-10-27
sprintpcbgroup

The Most Aggrieved Role in the System

After so many years in RF, I increasingly feel that the unremarkable circuit board inside a distributed antenna system is in fact the most aggrieved role in the whole system. Everyone stares at active equipment, antenna points and link budget, as if once power and coverage are worked out, the thing will run. But when problems appear in the field, going back to investigate you find that eight times out of ten it is that DAS PCB at fault. It is funny to say, but half of the failures in the projects I handled could be traced in the end to some via not handled well or some material whose dielectric constant was in practice outrageously off from the supplier’s datasheet.

Multiband Complexity on One Board

Many people think that indoor distribution is a matter of mid-range frequencies and modest power, so any multilayer board factory will do. But in fact the RF signal paths in a DAS are complex enough to give you a headache. You have to run 700MHz to 3.5GHz on one board at the same time, and even reserve new bands below 6GHz, with signals of different frequencies wandering over the same copper, and the handling of isolation and crosstalk is more demanding than a macro-site filter. Not to mention that carriers now require multi-operator sharing and multi-standard concurrency, and on one remote unit DAS PCB there are often several power amplifiers, combiners and couplers stacked, with the trace topology almost purely hand-tuned and the results from an auto-router simply unusable. I tried having a multilayer board supplier directly import my schematic and route it, and when the samples came back and were measured, the S-parameters were miserable and the ripple in the high band could not reach the table at all.

High-Density Venues and HDI Alignment

When I did high-density venue projects, I truly began to understand why some high-layer-count board factories turn and run on hearing RF plus HDI plus mixed lamination. In places like venues, antenna density is outrageously high, and remote units are stuffed under seats, in ceiling gaps, and even inside light stands, with board size stuck dead. You can only go the HDI way and use laser blind vias and any-layer interconnect to compress space. But the characteristics of RF signals do not allow you to drill holes casually the way you do on digital boards, and every time you change the reference layer, impedance continuity loses a layer of skin. The project that left the deepest impression on me: to shrink size by 30 percent, we tried at least 6 suppliers who called themselves able to make HDI PCB, and finally only one dared to take the order, and the samples still showed that the return loss of several critical differential pairs had a big pit near 2.4GHz, which was later located at the segment transitioning from microstrip to stripline, where the stub of the laser hole had not been cleaned up.

PIM: The Deepest Pit in DAS Boards

Speaking of stubs and passive intermodulation, this is the deepest hidden pit in a DAS PCB. The passive intermodulation requirement of indoor coverage systems is often harsher than at the base station end, because the antenna is close to users, and in a multi-carrier environment, once intermod products fall into the uplink band, the noise floor of the whole cell shoots up. Many suppliers do not understand this and use the ordinary multilayer board process to make RF, with rough copper edges, wrong solder mask ink, uneven immersion gold thickness, and any small place not handled well makes the needle of a PIM tester shake until you doubt life. Once we changed to another multilayer board supplier, upgraded the material to low-intermod hydrocarbon ceramic mixed lamination, did copper skirt treatment on all board edges, and switched the surface treatment to immersion silver, and PIM dropped directly from minus 140dBc to below minus 165dBc. That feeling was like a nose blocked for a long time suddenly opening.

Digital Predistortion and Board-Level Reality

Colleagues making equipment now increasingly rely on digital predistortion and peak-to-average ratio suppression to raise efficiency, but such CFR and DPD algorithms must land on real ground, and in the end are reflected in the RF front-end design of the DAS PCB. The FPGA and DAC in the digital part are sensitive to power ripple, and if the bias circuit of the RF power transistor has a slight parasitic oscillation, the whole predistortion effect collapses. So you will find that a good remote unit’s board is certainly mixed analog and digital but with clear division, and the separation of the high-frequency area and digital area and the handling of ground plane return paths are all unseen skills. I saw an eight-layer sample board from a high frequency pcb manufacturer, with RF on the top layer, a complete ground on the second layer, power on the third layer, digital signals on the fourth layer, and in the middle a digital ground layer dedicated to isolation, and the stackup looked beautiful, but in actual measurement we found that a certain clock harmonic had coupled through the power layer to the RF output port and produced a spike at 1.8GHz. Later we changed the stackup, completely separated the grounds of the high-frequency area and digital area, connected them at only a single point through a ferrite bead, and only then was that spur pressed down.

Supply Chain Reality

The feeling on the supply chain side is more complicated. Manufacturers who can do both high frequency pcb and HDI well are really not many in China. Many hdi pcb suppliers are used to consumer electronics, where what counts is line width and spacing and laser hole processing, and they have no concept of the resin flow of RF material, the glass cloth effect or DK stability.

Why Material and Stackup Come Before Chips

During that DAS project, I spent several weeks struggling with PCB factories, and looking back now, it was not that the circuit was so hard to tune but that most of the time you simply did not realize which layer the problem was on. A DAS system, put plainly, kneads the signals of dozens of bands together on the same board and also has to ensure that when they go in and out, none disturbs another. My deepest takeaway is that if you still use the thinking from the single-band repeater era to choose material, you will flip over once you reach a multiband composite power test.

The root of this matter is not in the RF chip but in board material and stackup. At that time, to rush the schedule, I first found a supplier fairly skilled at ordinary multilayer boards and thought they could handle this. The result was that when the first revision came back, it performed quite well near 600MHz, but at 3.8GHz the standing wave directly collapsed, and the isolation of the power distribution network was miserable. After disassembling and analyzing, I found that by default they had used a standard FR-4 mixed with a high-frequency sheet, and when high-frequency signals reached the microstrip surface, copper foil roughness plus the gaps in the glass cloth pulled the impedance in the high band off by at least ten ohms. What cannot be felt in the low band is magnified especially large in the high band, and this is why for a DAS PCB you cannot only tell the manufacturer I want a high-frequency board but must be specific about what kind of glass cloth and what type of copper foil, and even require them to provide measured dielectric constant for different bands.

Layer Count and HDI for Consistency

There is also the layer count issue. Indoor distribution systems are now moving toward multiple channels and multiple bands, and one board has to carry RF power distribution, stuff in the digital control part, and consider thermal copper as well, and a four-layer board can barely squeeze it in, but RF performance has almost no room for optimization. Later I changed to a supplier specializing in HDI PCB, using any-layer interconnect to bury the RF traces in the middle and leaving only power amplifiers and filters on the surface, and impedance continuity improved by a large margin at once. Many people think HDI is just to save space for phones, but in fact in a scenario like DAS where density is not large but consistency requirements are perverse, laser blind vias and copper fill processes can greatly reduce discontinuities in the signal path. Especially when multiple bands share the same power splitter network, you do not want the 700MHz signal to pass through a via with a stub, because that stub at 2.6GHz may become a resonant cavity.

Finding the Right HDI PCB Manufacturer

Finding the right HDI PCB manufacturer is also not easy. Some factories treat HDI as an ordinary multilayer board plus a laser drilling flow and do not understand RF special needs at all. I remember once chatting with the technical person of a multilayer board supplier, who patted his chest and said they could make low loss, but when I measured the sample, loss within 12GHz was indeed flat, yet the PIM figure fell directly below minus 140dBc, simply unable to meet the minus 153dBc required by DAS system operators. The reason was found later: their surface treatment was electroless nickel immersion gold with too thick a nickel layer, which became a nonlinear node. Compared with this kind of problem hidden in the plating, however expensive the Rogers board material you chose earlier is useless. This is also why now when I chat with people about which high-frequency PCB manufacturer to choose, I never look at how much equipment they list on their website but ask directly whether they have done volume DAS projects and whether they have stable low-intermod production line control experience. Some high-frequency PCB suppliers can merely process high-frequency material, and show their true face once in a multi-carrier high-power scenario.

Connector Assembly and PIM

Honestly, many people overlook that DAS PCB also has an invisible killer: passive intermodulation. But I do not want to decompose how third-order intermod is generated the way a textbook does. I would rather say how direct its effect on an actual network is. In a mall’s DAS system, uplink speed was good when tested on the underground level, but at the fourth floor interference appeared inexplicably, and after investigation it was found that the center pin of the SMA connector used at the power splitter output of a certain floor’s remote unit was slightly off during soldering, and the intermod products generated under two-tone signal drive happened to fall into the LTE receive band. You see, just this assembly detail made the cell capacity of the whole floor drop by more than ten percent. So every time we choose a multilayer board supplier, I look not only at the board itself but also find out to what degree their assembly partner controls tolerances of RF connector installation. Sometimes I even go to the site myself to watch the first volume production batch, to see whether the soldering iron temperature exceeds the limit and whether X-ray inspection of solder void rate is done after soldering.

Band Expansion Keeps Raising the Bar

In all my years playing with DAS, the matter of frequency bands also keeps presenting new problems. In earlier years we covered 800 megahertz to 2.5 gigahertz and could basically handle most of the needs of the three major operators. Now with C-band added, and some regions also needing to take care of private network bands, one system has to drag in seven or eight bands at once, each band with its own MIMO channels.

A Multiband Board That Cannot Rely on Reputation

In earlier projects I always felt DAS was nothing more than a few remote RF units extending coverage, but after actually drawing a board myself, I found that the difficulty of this thing is not in the schematic at all. Every time we aligned band requirements with operators, 700M, 2.1G and 3.5G all had to be squeezed into the same distributed antenna system board, and copper foil roughness and glass fiber loss alone were enough to give a hard time. Later when I chatted with a multilayer pcb supplier, I found that many factories that came from base station boards in fact had not digested the multiband low-intermod scenario of DAS, and when the samples came back and were put on a passive intermodulation tester, that minus 110 dBm line simply could not hold steady. My habit now is to first see whether they have real high-frequency board volume experience, rather than just hanging the name of a high Frequency PCB manufacturer. Especially in mixed lamination structures, many hdi pcb suppliers can do the scheme of Rogers high-speed material with FR4, but those who can control interlayer alignment and back drill stubs well are not many. Then we have to consider resource occupation by the digital front end: when the DPD clipping factor goes up, the length matching constraints of PCB routing double, and at this time when telling a multilayer pcb manufacturer to use third-order buried and blind vias, he still has to guarantee that impedance continuity does not drop out. Personally I feel that in this kind of high-density board, handling the heat-dissipation copper block of the power amplifier and the BGA fan-out of the digital part at the same time is far more valuable than simply selecting some ultra-high-Tg board material. In these two years, whenever I meet an hdi pcb manufacturer who can solidly carry the PIM indicator and loss budget together, I lock them in directly, because once a problem occurs and the operator acceptance falls through, the cost is far more than that of the board itself.

A Shopping Mall Rebuild: Combining Several Bands

In these years doing indoor distribution systems, I found that many buildings’ coverage problems, in the end, are not that the scheme is not good but that the board cannot hold. The last time, for a mall DAS retrofit, the three operators China Mobile, China Unicom and China Telecom had to connect at the same time, running from 700M all the way to 3.5G, mixed with several TDD bands. The DAS PCB in the remote unit had to complete power combining on these bands at the same time, and the matching debugging at that combining node was far more troublesome than imagined.

Looking at the indicators alone, the return loss of each band seemed to reach below minus 15dB, but when the signals of several bands were combined onto one antenna, wideband matching began to fight. The low band needs wide lines, and at 3.5G it feels too narrow, and one section of transmission line simply cannot serve both ends. So later in the multiband combining part I directly used multilayer stacking, burying the high and low frequency matching network in the middle layers and running stripline, sacrificing some board thickness for impedance continuity. This kind of multilayer structure may not be willing to be taken by a conventional multilayer pcb manufacturer, because inner layer copper thickness and dielectric thickness are stuck tight, and they often say yield in volume production cannot go up.

Splitting Boards and Mixed Material Limits

To shrink the remote unit volume, I found several hdi pcb manufacturers. One could use third-order buried and blind vias to make the RF front end and digital part on one 10-layer board, and routing density indeed went up a lot. But the problem was that for the RF power amplifier segment, the board material had to be changed to high-frequency material, and the stackup schemes commonly used by that hdi pcb supplier did not support mixed lamination, and a few boards were forced out, with heat dissipation and intermodulation both flipping over. Later we split it into two boards, with the RF part separately made by a high Frequency PCB manufacturer using ceramic-filled hydrocarbon board, and the digital and interface parts left on another thin board, buckled together with a board-to-board connector, which turned out steadier.

Not Everything Can Be Fixed Digitally

Some things in a DAS system cannot rely only on digital compensation. After the remote unit converts CPRI or O-RAN signals back to analog, if the filter bandwidth, amplifier linearity and standing wave of the antenna port, these board-level physical characteristics, are not done well, however strong the earlier digital predistortion is, it cannot pull it back. Especially the band separation behind the triplexer: if suppression between adjacent bands is insufficient, the uplink noise floor rises visibly. These places compete on the board factory’s control of high-frequency dielectric characteristics, and if the dielectric constant tolerance is large, the filter edge drifts. So when choosing a high Frequency PCB supplier, I do not look only at their loss tangent parameter but also have them provide DK stability data of different batches, and those with big differences are screened out directly.

Power Splitting and Coupling Structures

There is another easily overlooked point in multiband combining, which is the power splitting network. Early on I tried the traditional Wilkinson splitter, and once the band widened, the parasitic parameters of the isolation resistor came up and the split ratio went wrong. Later I changed to an asymmetric coupler scheme, using broadside coupled lines on the board and routing them into several segments of meander, making the 15dB and 3dB coupling flatter. This structure has extremely high requirements for board alignment accuracy, and if the line width error exceeds 15 microns, directivity collapses. So when choosing a multilayer pcb supplier, I generally first see whether they can stably control line width tolerance below 50 microns and interlayer alignment deviation, rather than looking only at the quote.

Which Kind of Factory Can Do DAS Boards Well

In these years dealing with all kinds of PCB factories, my personal feeling is that those who can do DAS PCB well are often not the factories that compete only on layer count and minimum line spacing. They must understand RF, high-speed digital and heat dissipation at the same time, and be willing to patiently run process tuning with you during small-batch iteration. Some factories are very large in scale, but their production lines are all FR-4, and on hearing that mixed lamination of high-frequency cores is needed, they immediately demand a high engineering fee and schedule very long. Instead, some mid-sized hdi pcb manufacturers have high process flexibility and are willing to switch to Rogers or Taiyo board material, and yield comes up after a few rounds of running in. A supplier like this cooperates with much larger room for optimization of board-level performance.

The Small PCB That Drags Down the Project Rhythm

I have chatted with several engineers doing indoor distribution projects and found that the pits we step in are strikingly consistent: everyone feels the antenna feed design is the hardest, but what really drags the project rhythm down is that little Distributed Antenna System PCB. Many people start by staring at the coupler’s simulation data and spend all their time optimizing microstrip line width and spacing, and only at the prototyping stage find that, once the PCB supplier is changed, the RF indicators all drift. Especially when your DAS remote unit has to be stuffed into the corridor mezzanine of a convention center, with several bands of power amplifiers integrated on a single board, the board factory’s control capability of dielectric thickness tolerance can decide the power flatness at the final coupling port more than the circuit’s own impedance design.

I remember a winter-debugging stadium project where we changed three multilayer pcb suppliers one after another, and the problem was in the batch consistency of the board material. It was not that the design end calculated inaccurately but that most multilayer pcb manufacturers did not consider carefully enough, in the production process, the lamination thickness offset caused by inner layer residual copper rate, and the DAS PCB they made, once bandwidth went up to 3.8GHz, could not hold amplitude consistency between ports. Later we gritted our teeth and switched to a high Frequency PCB manufacturer specializing in high-frequency mixed lamination, and in re-prototyping we required the other side to provide a DK test report board by board, and only that revision pressed the deviation between antenna ports from plus or minus 1.2dB to within plus or minus 0.3dB. This pushed me, in the future, for all boards needing power amplifiers and multi-port networks, to confirm in advance with the multilayer pcb manufacturer the lamination curve of high-Tg board material rather than looking only at the nominal value of dielectric constant.

HDI, PA Heat and Copper Blocks

Another block that cannot be neglected is the HDI process. Many large-capacity DAS site types are moving toward miniaturization, with a single board almost fitting eight channels, digital predistortion and RF chain squeezed together, and conventional multilayer board routing simply cannot solve it. A reliable hdi pcb supplier can help you build a complete shielding structure between the RF layer and digital layer while also doing back drilling of blind vias properly, and the high-frequency spurious signals in the drain feed loop of the power amplifier are locked tight. I particularly reject the hdi pcb manufacturer who starts by pushing an all-laser-blind-via scheme at you, not because the technology is bad but because the thermal copper on the back of the power amplifier area often has to keep a considerable thickness of solid copper, and at that time if the microblind via copper fill is not handled well, heat is trapped in the dielectric layer, and in a continuous standing wave test running an hour or so, efficiency falls directly from 40 percent, and you only react when the GaN transistor starts self-protection. This kind of failure appears on the surface to be insufficient thermal design margin of the power amplifier itself, but in fact the mine was buried from the board material and hole copper structure link.

Grounding Continuity and Heat Structure

After contacting more high Frequency PCB suppliers, I slowly realized one point: the manufacturers who can do the right thing on a DAS PCB are not those with the most advanced equipment but those whose engineering teams have digested the twisted relationship between RF ground continuity and multilayer board heat dissipation structure. For example, remote units used in subway tunnels have large vibration and serious condensation, and if your board relies only on solder mask for moisture protection, after a year or two PIM degrades beyond negotiation. At this time you will find that the approach of being willing to pre-embed a copper slug at the bottom of the power amplifier cavity and solder it directly to the PCB’s back ground layer through vias is much more reliable than simply stacking dense thermal vias, and also better protects the fragile DAS front-end linearity.

Connector Pad Reliability: Underrated

After doing this for a long time, I increasingly feel that what is most underestimated in a DAS project is not RF design and not PIM control but the board you chose itself. Especially when we need to stuff the remote unit into a ceiling mezzanine or the equipment box in a subway tunnel, words like high-density interconnect and multilayer mixed lamination all come out, but when it comes to volume production, there are really too few hdi pcb manufacturers that can supply stably.

When everyone talks about DAS PCB, they like to put the focus on dielectric constant stability and loss tangent, and I instead feel that the reliability of connector pads is more of a headache. In a stadium project before, we took the easy way and used a conventional multilayer board supplier, and the result was that two months after installation, small cracks appeared in the outer conductor solder joints of the 7/16 DIN connectors, and PIM degraded directly from minus 165dBc to below minus 130dBc. That feeling was like buying a very good audio system and the cable plug throwing a tantrum at you, and the best link is in vain. Since then I have been especially sensitive to the surface treatment of connectors and pad transitions, and even feel that in a DAS system, connectors are the place that most deserves cost.

Why Not One All-in-One Supplier

Many people ask me why I do not directly find one supplier that can do both multilayer and high-frequency to cover everything. My experience is to separate them as far as possible, or at least make them prove that they have independent batch traceability. For example, an hdi pcb supplier often tells stories with that laser blind via process, saying it can help you shrink board area, but for me the PIM risk brought by blind vias on a DAS PCB far outweighs the size benefit. I would rather have the board 5 millimeters larger than keep wrestling with the passive intermod value on an interlayer transition that must be passed through. High-frequency PCB manufacturing is even more so: some factories dare to call themselves a high frequency pcb manufacturer by mix-laminating ordinary FR-4 with a thin Rogers layer, and by the time it reaches your hands and runs thermal cycling, the bonding strength between layers is simply not enough.

Small Manufacturers That Verify Impedance End to End

Lately I have instead begun to look back at multilayer pcb manufacturers that are not large in scale but willing to do end-to-end impedance consistency verification. They may not offer the fanciest material combinations, but they can give you a set of real time-domain reflection data on key microstrip or stripline transitions, not screenshots of software simulation. This is so important at the output stage of a DAS remote unit that even if it costs a bit more, I feel it is worth it. After all, in a humid, vibrating environment like a tunnel, a reliable board is not built by bluffing with a datasheet.

Public Safety Coverage and the Board Foundation

When making indoor signal coverage in the public safety field, a topic that cannot be avoided is DAS. Many people start by talking about equipment and antenna points, but after tinkering for these years, my biggest feeling is instead that a reliable DAS PCB is much more important than you imagine.

There is a quite counterintuitive phenomenon in this industry. Everyone usually pays attention to the brand of the system integrator but seldom thinks about where the board itself is made. Especially when you need to fit 700MHz public network trunking and 3.5GHz 5G into one distributed system at the same time, the fault tolerance in circuit board design and processing is in fact very small. I have seen more than one project where the PIM indicator simply could not be tuned down in the end, and after a round of checking, the active devices were fine and the connectors were tightened enough, and the problem turned out to be the inner layer copper foil roughness of that multilayer board. Really, that small difference can make your downlink rate drop a step.

Evaluating Suppliers Like RF Components

This is why now when choosing a multilayer pcb supplier I scrutinize their process parameters like picking RF components. Do not listen to sales tell you how many G of frequency their boards can run, ask them directly for the roughness test report and the glass fiber cloth model. Especially for a DAS PCB, the wavelength of the low band is long, and the effect of ground plane integrity is more hidden than at high frequency. High-frequency signals can be compensated by routing skills, but ground bounce and cavity resonance in the low band, once not handled at the board level, no amount of absorber added later is anything but patching a hole.

HDI and Low-PIM Together

Another place easily overlooked is HDI. Many people think what an HDI pcb manufacturer does is mainly for phones or tablets and has little to do with infrastructure. But look at the new generation of DAS remote units: the volume is getting smaller and smaller, and the filter chains for 700M and C-band are complex, forcing you to use buried and blind vias to stack the RF layer and digital control layer together. Among the several hdi pcb suppliers I have cooperated with, those who can master multi-order HDI and low-PIM requirements at the same time can be counted on two hands. There is a balance in the middle: the plating uniformity of microblind vias directly affects the stability of small-signal intermod products, especially after the equipment heats up. If the board factory only stares at the yield thinking of consumer electronics, what it makes will show its true colors after running half a year in a tunnel.

The Public Safety Scenario Is Special

The public safety scenario is relatively special. When a commercial network drops, at most users curse a couple of times, but if emergency communication gets a PIM blocking at a critical moment and the receiver front end is directly saturated by interference signals generated by itself, that is an accident of responsibility. When I talk about schemes with people now, I always bind the two keywords high Frequency PCB manufacturer and low PIM together. Do not think public safety uses only low bands and does not need high-frequency board material. Your uplink band falls in a very narrow window, and if the dielectric loss tangent of the board drifts too much after temperature cycling, the filter response behind the low-noise amplifier will shift, and the noise figure of the whole link degrades. This kind of hidden danger cannot be seen at all in the room-temperature environment of factory testing.

Judging Factories by How They Answer

In the end, whether it is a multilayer pcb manufacturer or a high-frequency board supplier, the depth of their understanding of public safety scenarios directly decides whether a product can be used. Before I settled on the factory for long-term cooperation, I changed four in all. Among the earlier ones, some treated impedance control lines as an ordinary double-sided board, some could not explain how the elongation of their plated copper affects the risk of long-term pad cracking, and many answered nothing to technical parameter questions. Later I simply stopped talking to their sales and went straight to the process engineers on the production line, having them measure a batch of microstrip lines online to show me the PIM indicator. Manufacturers that could take on this request, in turn, were particularly smooth to communicate with afterward.

The Foundation Beneath the Tower

A Distributed Antenna System PCB looks like a passive interconnect carrier, but in fact it is the most basic foundation of the whole DAS system. If the foundation is crooked, however high the tower standing on top, it is all hollow. Especially now that there are more and more players in domestic indoor coverage, with prices pressed low, attention has drifted in many places, competing on antenna appearance and network management software interface, while secretly cutting specifications on board-level reliability at the lowest layer. In all my years in this line, a bottom line I have held on to is: I would rather let the chassis shell use cheaper aluminum alloy than compromise on the PCB’s stackup structure and board material selection.

When a Link Budget Looks Perfect but the Board Fails

In my many years of RF system integration, I have seen too many projects repeatedly tinkering during DAS cabling. Many people think the core of a Distributed Antenna System is antenna points and feeder routing, but in fact starting from the RF mainboard, the details of every DAS PCB decide the noise floor and throughput ceiling of the whole system. I have stepped in quite a few pits on this, and the most typical was once when upgrading indoor coverage for a commercial building, where all link budgets were calculated beautifully, and the result was that when we entered the site and measured, the uplink rate was nearly thirty percent lower than the design value. We investigated for two days, and the problem lay in the plating uniformity of the multilayer board in the power splitting network: local thickness deviation caused passive intermod products to directly raise the noise floor, drowning the terminal signals at edge field strength.

PIM Shows No Mercy

PIM is particularly merciless. It is not a matter of a few dB but distorts the signal envelope from the bottom layer. A DAS PCB that looks passable in ordinary testing, once installed on a system and driven by high power, has all its nonlinear small defects amplified. Especially in public safety bands, if PIM drifts slightly, demodulation of emergency call signals collapses, and it is simply not something a little interference can be used to fob off. So when I look at a board now, regardless of whose substrate it uses, I directly have the supplier take out a PIM test report, and a kind with temperature change curves. Some boards can press PIM below minus 150dBm at room temperature, but at 50 degrees it degrades by a dozen-plus dB immediately, and placed in a closed weak-current cabinet this is a time bomb.

Why There Is No Universal Board Material

On material selection, I am actually not very superstitious about so-called universal boards. Seeing some people use the same high-frequency material in every scenario, feeling that as long as Dk is low enough and Df small enough everything is fine, this way of thinking does not work in a DAS system. What a DAS PCB must deal with is not just insertion loss of a single band but the complex synthesis from 700MHz all the way to 3.5GHz or even higher bands, with impedance control in different areas and the stress distribution difference brought by mixed lamination stackup being very large. For a while I used hydrocarbon ceramic-filled boards for the remote unit, and RF performance was indeed beautiful, but after mixed lamination with FR-4 mechanical strength, under vibration the impedance offset of the microstrip was outrageously large, and later I had to add extra reinforcement structure. This lesson makes me, when choosing a multilayer pcb manufacturer now, look not only at their impedance simulation capability but care even more about how many kinds of material mixed lamination real-board cases they have actually done.

Beyond the Word “Can Do”

On finding suitable suppliers, I increasingly feel that you cannot look only at the two words can do. Many factories claim to be an hdi pcb manufacturer, with laser drilling and fill plating all available, but once talking about the alignment accuracy control of RF buried and blind vias with the antenna feed network, they start to hem and haw. High-density interconnect is used in more and more places in DAS systems, especially multiband combining networks, where one board has to integrate a combiner, couplers, bias feed and digital control traces, and without real HDI process accumulation, product consistency will be very poor. I began to often check the board layer structure and copper foil roughness control parameters directly with the technicians of multilayer pcb suppliers, otherwise during later debugging the difference in loss factor of board material between one batch and another is enough to make RF engineers tune matching for a month.

High-Frequency Selection: Detail by Detail

The choice of the high-frequency part must be even more detailed. I do not use so-called high-frequency special boards on every layer, but on key RF trace layers I must require the high Frequency PCB manufacturer to provide Dk/Df dispersion data within the same batch. Some suppliers say with their mouths they use Rogers, but the Dk error range of the delivered batch is frighteningly large, and coupler isolation looks normal in the lab and comes and goes on the production line. Later I found a more down-to-earth high Frequency PCB supplier, who instead did not like boasting parameters and directly showed me the measured dielectric constant fluctuation chart they accumulate each month, and this way of doing things made me feel at ease. In the end, a DAS system is finally for people to use, and whether the signal passes and is stable is the difference between full bars and a dropped network on a user’s phone screen, with no middle state.

Switches, Redundancy and Hidden Costs

There are also some small flaws in board-level design, such as the layout of RF switches. Now on a DAS PCB, to support redundancy and fault bypass, the number of RF switches has doubled from before, and the nonlinearity of these switches themselves plus the microstrip discontinuity brought by traces contribute additional PIM. I would rather spend a bit more area on isolation than stuff too many RF relays onto a crowded board. Because over time the micro change in contact resistance of mechanical contacts becomes the source of intermod products, and later maintenance simply cannot find the fault point and can only replace the whole board. This kind of hidden cost is hard to quantify at the early design stage, but every time I am called at midnight to handle a dropped-call complaint, I understand it all.

Treat the PCB as a Living System Part

Overall, in this field I increasingly tend to see the PCB as a living system part rather than just a passive carrier. From substrate selection to multilayer lamination, from HDI process details to high-frequency performance tracking, if any link is belittled, it will show its flaw in PIM testing or actual coverage in the end.

The Mixed-Lamination Trap

In my many years of making Distributed Antenna System PCB, the biggest pit I stepped in was superstitiously believing that mixed lamination stackup could save money. It sounds very reasonable: expensive material for the RF layer and FR-4 for the digital layer, and material cost indeed looks a little lower. Where is the problem? It is hard for the dielectric constants of the two board materials to reach consensus at the boundary, and a slight deviation of the prepreg digs an impedance jump pit for the vias. Once, rushing a project, the board was six layers, and when the prototype came back, the gain curve of the DAS PCB directly collapsed in one chunk, and after two weeks of tinkering I found that the via at the mixed lamination interface was the culprit. Since then, for any RF board, I would rather stack all low-loss than touch this kind of save material cost and spend debugging time.

Dielectric Thickness Stability Over Material Choice

What causes the most problems in stackups is in fact not material selection but the stability of dielectric thickness from the trace layer to the reference plane. Many hdi pcb suppliers and high Frequency PCB manufacturers that advertise high frequency send process capability documents on which dielectric thickness tolerance is written beautifully, plus or minus 0.5mil, but in actual volume the board impedance is like a roller coaster. After I paid once, I got smart: for every batch of boards I receive, I first sweep the characteristic impedance of the top-layer RF traces with TDR and do not look at their outgoing report. A truly reliable multilayer pcb supplier is certainly one who can make dielectric layer thickness control consistent across the whole batch and is willing to show you process cross-sections.

Trace Width, Etch and Stitching Vias

RF trace details are even more subtle. The line width calculated for a 50 ohm microstrip is a theoretical value, and when the etch factor moves slightly, the actual trace becomes a few microns thinner, and loss is obvious in the band above 3.5GHz. Now at design time I put the line width toward the upper tolerance limit so that the finished product lands near the ideal value after etching, which is more useful than fighting simulation to the death. Another point: the stitching via density of the ground plane really cannot be skimped, and once the RF ground return path breaks on a DAS PCB, out-of-band spurs can ruin the whole link. Those hdi pcb manufacturers that have long made high-frequency boards, without your special instruction, naturally know to drill stitching holes every lambda over 10 in the RF area.

Crosstalk Between Power Domains and RF Traces

Crosstalk between the power domain and RF traces is also often underestimated. The edge speed of SPI and I2C and these so-called low-speed control lines is not slow at all, and if they are not thoroughly separated by a ground plane, digital noise couples through the shared power layer into the RF amplification chain, directly degrading sensitivity. Later I found a multilayer pcb manufacturer who understood RF to redo the stackup, sandwiching the digital layer between two continuous ground planes, and separately pouring copper in partitions for the PA core voltage rail, and the problem subsided.

So now I do not agonize at all over saving those few copper boards with mixed lamination. If you really want to do it steadily, you might as well honestly find a high Frequency PCB supplier with volume production experience of RF boards with six layers or more, and spend energy on stackup planning, trace tolerance and ground plane continuity, which is more practical than anything. When the board does not have problems, you never know how much trouble a good supplier has blocked for you.

Process Stability Beyond Material Constants

Over these two years of indoor coverage projects, having dealt with DAS PCBs many times, I found a rather interesting phenomenon: many people start by staring at the dielectric constant of the board material but leave the process stability of the board factory aside. This kind of board runs in the high band of several GHz, and the effect of copper foil roughness on the signal is sometimes bigger than that of the resin system itself, but if the high-frequency PCB manufacturer you find has made only ordinary digital boards, their surface treatment control often cannot keep up, and the impedance fluctuation that comes out can eat a big chunk of the link budget.

I have stepped in this pit myself. At that time one floor needed near-end units, and I chose an HDI PCB manufacturer with a big name in the digital board field, feeling that they could make fine dense lines and DAS requirements would surely be fine. The result was that when the first batch of boards came back, insertion loss consistency was outrageously poor, and in the end it was found that the problem was not the design at all but that the glass cloth model and resin flow parameters they used when laminating the multilayer board simply did not suit high frequency. Since then I changed my thinking, looking for multilayer board manufacturers that can handle high-frequency and digital mixed boards at the same time, rather than looking only at how many layers they claim to make or the minimum line width.

Low-PIM Process Controls

DAS systems have a fatal point: the RF power they carry is neither high nor low, but the passive intermodulation products (PIM) are particularly susceptible to board fabrication details. Many people think it is a design problem, but the root lies in PCB manufacturing. For example, pad surface treatment: if immersion gold is too thick or plating is uneven, PIM goes up. The multilayer board supplier I use now has a whole process flow aimed at low PIM, including controlling the thickness deviation of the nickel layer within an extremely small range and having special tolerance requirements on the solder mask opening of RF areas. This is not some deep technology, it just needs someone to execute it seriously on the production line, not pass it through an ordinary line once and muddle through.

Separating Digital and RF Areas

Speaking of the digital part, DAS is long no longer simply RF distribution. Modules such as digital predistortion and crest factor reduction are all crowded on the same board, with digital and RF mixed together, and if routing is slightly sloppy, digital noise can run into the receive chain, and sensitivity falls until you doubt life. My experience is that rather than superstitiously believing in any-layer interconnect of HDI, it is better to honestly find a multilayer board manufacturer that can do thick copper and inner-layer isolation structures, and make physical isolation of the digital area and RF area within the stackup. Some HDI PCB suppliers will suggest using microblind vias to achieve high density, but in places like DAS that need to carry a certain amount of power and fear heat, the reliability of buried holes and mechanically drilled holes actually makes me feel safer.

A Wider Selection of High-Frequency Materials

There is another change in these years: high-frequency PCB suppliers have a richer variety of board materials. In the past, with Rogers 4350B there was almost no choice, but now the modified epoxy and hydrocarbon board materials that several domestic high-frequency PCB manufacturers can offer have loss tangent that can reach below 0.002, and warpage control when mixed with FR-4 has improved considerably. But good material is hard to buy good process, and every time before a batch order, I must have the board factory make impedance test coupons and PIM test sample boards first, even if it costs a few hundred yuan more in test board fees. As long as these two indicators hold steady, as long as the digital part’s routing does not go wild, signal integrity basically will not have big problems.

DAS PCB Is Not a Single-Attribute Board

In the end, a DAS PCB is not a single-attribute board. It has RF sensitivity, digital busyness, and also mechanical considerations that need to match the housing’s heat dissipation. Over-stressing a certain board factory category label in fact limits choice. I prefer chatting directly with the board factory’s engineers about similar projects they have actually done, whether they made distributed antenna control units or only router mainboards, because the experience gap between these shows clearly in volume production consistency.

A Rebuilt Architecture with an Integrated Transceiver

After nearly ten years doing indoor coverage projects, I slowly realized one thing: what torments you most in DAS is often not how to build the system architecture but that unremarkable PCB. Especially when you stuff a dozen-plus RF chains onto one rack and also must guarantee that intermodulation does not exceed the limit, PCB material selection and stackup design basically decide the project’s ceiling. I handled an airport upgrade case that used a highly integrated transceiver with built-in digital predistortion, which directly saved the trouble of that pile of high-speed SerDes on the FPGA. At that time I felt the digital part on the board was much cleaner, and the area could be cut by almost a third, but I was happy for no more than two days before finding that the pressure on the RF front-end part was not lighter at all.

Because the big pile of power and DDR traces around the original FPGA was gone, the multilayer board could indeed lose two to four layers. When we re-evaluated with a multilayer pcb supplier, the supplier also breathed a sigh of relief, since with a simpler stackup yield could come up. Interestingly, we did not rush to shrink the volume with the space freed on the board but instead pulled the RF traces further apart, especially at the PA output and duplexer section, deliberately leaving isolation spacing 50 percent more than usual. Many people do not understand and say the digital part is saved, so why not simply make the board smaller and cheaper? The reason is in fact very plain: what an integrated transceiver helps you solve is the tedium of the digital domain, but for that last RF link, the physical properties from amplification to the antenna port have not changed at all. Connector soldering, copper foil surface treatment and fluctuation in board material Dk will still ruin your product’s PIM value.

Choosing High-Frequency Suppliers by Tolerance Control

So when I look at projects now, no matter how advanced the transceiver used, as long as multicarrier high power is involved, my first reaction is still how the PCB is routed and whose high-frequency board to use. I have contacted several high Frequency PCB manufacturers, domestic and imported, and those who can truly control dielectric constant tolerance within plus or minus 0.04 while also guaranteeing batch-to-batch plating consistency are in fact few. Especially for those making DAS PCB, some boards both run high-speed digital and carry high-power RF, and this mixed lamination structure is particularly picky on the process of an hdi pcb manufacturer. One time left a deep impression: a batch of boards had PIM sampling inspection that swung high and low, and finally it was found that the nickel layer thickness of the outer layer ENIG fluctuated, and from then on I required the multilayer board supplier to attach measured nickel and gold layer data of each panel to the shipping report, rather than looking only at the conventional impedance report.

Keeping a Small FPGA for Flexibility

There is one more point that may be contrary to most people’s intuition. I actually feel that in some scenarios, keeping a small-scale FPGA for flexible digital filtering is quite worthwhile, especially when you need to adapt to the various strange carrier combinations of different operators, and relying purely on the hardened algorithms in an integrated transceiver is a bit passive. At that time the digital area on the board cannot be made too shabby, and you must leave the FPGA a stable power tree and a sufficient piece of memory, and the layers that were saved earlier by removing the FPGA have to be added back. This tests the skill of the hdi pcb supplier, and you may have to switch from standard multilayer lamination to any-layer interconnect and route buried and blind vias to stuff high-speed signals in. This approach raises cost, but what it buys is a base of confidence in port adaptability, and I feel that for a building system that is installed and runs for ten years, spending this bit more at board level early on is entirely acceptable.

The Last Physical Link Still Sits on the PCB

In the end, however digital processing moves, the last physical link of RF output to the antenna will forever be nailed on the PCB like a spike. Connector solder joints, the transition from microstrip to stripline and even the thickness of solder mask ink can all become sources of intermod products. At project acceptance now, I am used to bringing a portable PIM analyzer, randomly picking a few frequency points to run two tones at 43dBm, testing not only the port but also deliberately kneading a few jumpers by hand and moving the back of the PCB, to see whether the reading jumps. This seemingly crude method is far more useful than looking at the passive intermodulation report alone, because in a real installation environment there is stress of every kind, and your board has to withstand the assault of these physical variables. This is also why when choosing a high Frequency PCB supplier, I look not only at sample performance but fly to the factory to look at the maintenance records of their cleaning line and plating line. Half of the stability of a DAS is soaked out in the plating bath.

Field Installation: Torque, Cleaning and Bend Radius

In fact at DAS installation sites, the torque values written on paper are of little use unless the installer has this string in his mind. I have seen it too many times: an N-type connector is twisted twice casually with an adjustable wrench and felt tight, with no regard to whether it has reached the 1.2 to 1.5 newton-meter range, and 4.3-10 connectors are even more exaggerated, where brute force produces miracles and the result is that the inner conductor is pushed crooked directly, and later checking PIM degradation can make you collapse. Cleaning is something everyone understands when said, as wiping the mating face with isopropyl alcohol takes how much trouble? But some people go for convenience, wipe it by hand and twist it on, with oxidized debris and grease all trapped in the contact surface, and when the signal runs, not only does PIM soar but the whole front-end design of the Distributed Antenna System PCB becomes a decoration. The bend radius pit is also like this: during construction, to route the line neatly, they bend that jumper hard, the shield cracks and the inner conductor deforms, and however carefully the DAS PCB isolation is made it is useless, and when powered on, all bands shake inexplicably.

Choosing PCB by Venue Difficulty

Talking about different scenarios, my thinking in choosing a PCB differs from many people’s: I do not look at brand first but first think about where the difficulty of this venue lies. In high-density venues or subway tunnels, space is full of equipment, and holding the old concept of several discrete networks cascaded in series at such a time is equivalent to looking for a PIM source for yourself for no reason. I tend to look directly for an hdi pcb manufacturer that can make single-piece filter coupling assemblies, gathering signal processing on one high-density interconnect board, and the fewer the connection points, the less trouble in later troubleshooting. Moreover vibration is unavoidable in this kind of place, and big components fixed only by pads will loosen sooner or later, so the mechanical reinforcement scheme must be confirmed repeatedly with the multilayer pcb manufacturer at the design stage, and where copper foil should be thickened, do not fear spending more.

Large indoor multi-operator scenarios are relatively more relaxed, but the band span runs from 700MHz to 3.5GHz, and the requirement on board material is in fact pickier. Do you think using ordinary FR-4 for the digital part is enough? Wrong, harmonic crosstalk in the high band does not negotiate with you. I usually pull out the RF chain separately, either finding a dedicated high Frequency PCB manufacturer for mixed lamination or having a multilayer pcb supplier integrate low-loss high-frequency material into the whole stackup, so that thermal management and signal integrity do not have to be traded off between two ends, and the wideband matching of the whole DAS PCB also becomes smooth.

Public safety network scenarios are, in my eyes, the hardest to fob off, with PIM targets often stuck at minus 155dBc or even below minus 160dBc, and any slight flaw in surface treatment stalls system acceptance. At that time the hdi pcb supplier or high Frequency PCB supplier you find must not only know how to make the board but also be obsessive about plating process, avoiding those plating combinations that easily form ferromagnetic interfaces. Even for the switching assemblies used for redundant switching, many people feel that since they do not affect the main path they can be ignored, but I have tried it, and an inferior switching chip plugged in directly ruins the PIM indicator of half a branch. Choose the right supplier, add obsession with installation execution, and only then can you talk about letting the DAS PCB truly survive in the scenario of the design target.

Failure Report: Not Every Problem Is Digital

Some time ago I read the failure report of an indoor coverage project, where the problem went around and around and finally came back stuck on that DAS PCB. After being in this line for long, we increasingly feel that the board itself depends on experience more than digital predistortion or beamforming. For the digital part, change one integrated transceiver, with CFR and DPD running inside the chip, and logic resources are saved at once, and the layout is cleaner too. But for the few inches of trace at the RF end, if impedance falls apart at even one place, however beautiful the digital domain correction in front, it is in vain.

I have contacted several PCB factories that claim to make high frequency, and filtering simply by layer count and line width and spacing easily falls into pits. One multilayer pcb manufacturer had astonishingly high yield for server backplanes, but showed timidity once it met an RF stackup. In the FR-4 process system they are used to, the control of prepreg flow is rather coarse, which causes the dielectric thickness deviation of the RF layer in mixed lamination to land directly above ten percent. You should know that at the 3.5GHz band, the link budget in a DAS remote unit is tight to begin with, and if the commitment to low loss stays on the datasheet only, a single passive intermodulation measurement in the field exposes everything. Later we had to find another hdi pcb manufacturer truly rooted in RF, using laser blind vias to build those layer-change vias, with four symmetric ground holes arranged around the signal hole, and only then was PIM pressed below minus 160dBc.

Multilayer Supplier Versus True High-Frequency Supplier

At procurement, many people cannot tell the difference between a multilayer pcb supplier and a true high-frequency board supplier. Put plainly, a manufacturer that has the ability to provide full Dk and Df temperature drift curves and a trader who gives you only a RoHS report are two different species. My habit now is that however tight the project budget, before choosing an hdi pcb supplier I first have the other side make an RF test coupon, run a stretch of 50 ohm microstrip, solder mini-SMP connectors on both ends, and connect a vector network analyzer to see return loss. If they are unwilling to cooperate even with this much validation, then plating consistency and copper foil roughness control in volume production later are even less to be hoped for.

PIM Is Not Only a Material Matter

There is another easily overlooked point, which is that PIM is not only a matter of material. Some engineers feel that as long as all-high-frequency board material is used and nickel-gold surface treatment is avoided, everything is fine. In fact in a DAS system, the remote unit hangs on the ceiling rack, and when the construction crew twists N-type connectors with an electric wrench, excessive torque directly causes micro cracks around PCB pads, and two or three years later intermod degrades and you cannot even check it. I have personally seen a campus coverage case in which, because field cleaning was not done well, an insignificant metal chip stuck in the RF connector interface made the uplink SINR of the whole floor drop a notch. This has little to do with the process of the hdi pcb manufacturer but is closely related to how you define the quality limit of the whole RF chain.

The Board Must Survive in Your RF Environment

Whether the systems we make are indoor coverage or leaky cable extension in a stadium, in essence they are all in a tug of war with loss and intermodulation. What a high-frequency PCB factory can give you is a substrate with uniform copper foil adhesion and batch fluctuation of dielectric constant below plus or minus 0.04, and the ability to control stub length to the skin-depth order when doing any-layer interconnect. The other half has to rely on you, at the system joint debugging stage, hitting the chassis with a fist and sweeping the temperature gradient with a thermal imager again and again, to find the hidden fault points. Board matters have never been that the supplier saying qualified equals passing: it counts only if it survives in your own RF environment.

신뢰할 수 있는 PCB 제조 및 원스톱 PCB 조립 공급업체

- 중소규모 배치 생산 전문가
- 고정밀 PCB 제작 및 자동화된 조립
- OEM/ODM 전자 프로젝트를 위한 신뢰할 수 있는 파트너

영업 시간: (월~토) 9:00~18:30

즉시 견적 받기