Physiotherapy Device PCB Safety Design: High-Voltage Isolation, Creepage Control and Supplier Selection for Stimulators

In a Physiotherapy Device PCB, safety is decided by the physical board as much as by the schematic. Copper-edge burrs, solder mask pinholes, dielectric thickness drift, resin voids beside isolation slots and unbalanced stackups all become leakage paths, charge residue or pulse distortion. Choose multilayer, HDI and high-frequency PCB suppliers by their batch cross-section reports, process change control and medical high-voltage experience, keep hardware protection independent of firmware, and match board technology to the real pulse spectrum instead of chasing the most advanced label.

Why the PCB Factory Matters More Than the Simulation

In my years of working on physiotherapy equipment, I have increasingly felt that whether the PCB factory is chosen correctly affects the final result more than how beautifully the circuit simulation runs. A Physiotherapy Device PCB in particular does not merely connect components. Much of the interference and occasional loss of control at the treatment output, when traced back, is not a schematic problem at all; the PCB had already put the ground plane and the inner-layer routing off at the manufacturing stage. Many people plunge into topology and parameters from the start, which I think is putting the cart before the horse.

When I first made a handheld nerve stimulator, the board had boost, control, sampling and display drive crowded together in a small area, and on a double-sided board the ground loops ended up completely fragmented. The prototypes came back with inexplicable glitches always superimposed on the stimulation waveform. Later we switched to a factory that makes multilayer boards, redesigned with a six-layer structure and ran the analog ground and power ground on separate layers, and only then did the output become clean. When looking for a multilayer pcb manufacturer, you cannot look only at whether it can make multilayer boards; you also have to look at how well it controls inner-layer alignment and dielectric thickness tolerance. Between different multilayer pcb supplier options, boards made from the same files differ very noticeably under mixed high-voltage and small-signal conditions.

Later, on an even smaller portable model, the board could not even fit conventional vias and we had to go to HDI. At first I casually picked a factory that was taking HDI orders; its laser drilling positions were badly offset, and a few boards also developed barrel-wall cracks after reflow, making the stimulation output intermittent. After switching to a dedicated hdi pcb manufacturer, the problems dropped sharply. When I choose an hdi pcb supplier now, I first ask about their blind via diameter and stackup structure and then look at their batch reports; otherwise comparing only price is meaningless.

One more thing is easy to overlook: some physical therapy methods use high-frequency oscillation or RF energy, for example some therapy electrodes need carrier modulation. Ordinary FR4 behaves unstably here; once the dielectric constant deviation is large, the output frequency drifts, directly affecting the controllability of the treatment dose. So for this kind of board I tend to go to a high Frequency PCB manufacturer familiar with high-frequency materials, and have the high Frequency PCB supplier provide measured dielectric constant and loss tangent data rather than just a nominal value. In a high-frequency environment, tiny parameter deviations of the PCB become uncertainty at the output, and on a treatment device that uncertainty is risk.

So I no longer treat the PCB supplier as a pure contract manufacturing step. Physical isolation, energy control and output stability all finally land on the manufacturing consistency of that board. However conservative the circuit design, if the PCB factory cuts corners on hole copper, inner layers or laminate, the treatment device may still deliver energy it should not under a single fault. The cost saved at this step may end up as trouble on the clinical side.

Isolation, Moisture and Batch-to-Batch Consistency

For the PCB in physical therapy equipment, especially the kind with a high-voltage pulse output, many people put all their attention on isolation devices from the start, thinking that adding a few more optocouplers or isolated op-amps will solve the problem. But after doing several similar projects myself, I feel that what really separates results is the laminate and the processing, not those few components on the schematic. The problems exposed after mass production are almost all concentrated in the physical characteristics of the PCB itself.

In the early days, to save cost, I found a very ordinary multilayer pcb manufacturer. The first batch of prototypes was completely normal, but in the third month of mass production, insufficient creepage distance between the high-voltage side and the low-voltage side began to appear one after another. When we cross-sectioned the boards, we found that core thickness differed by several tens of microns between batches and the resin fill near the isolation slots was uneven. At that point you are not struggling with circuit design but with machining tolerance. So later I switched to an hdi pcb manufacturer that specializes in medical and high-voltage boards, and I also check whether it has a stable-supply multilayer pcb supplier channel. The benefit of the HDI process is that laser drilling can leave the edges of isolation slots cleaner, with far fewer glass-fiber burrs than conventional mechanical drilling, which is critical in high-voltage small gaps. When choosing an hdi pcb supplier, I care more about whether it can provide a cross-section report for each batch than about verbal promises that it can do anything.

There is another easily overlooked point: although the output frequency of this kind of equipment is not especially high, the pulse rising edge is very steep, and some people will directly suggest high Frequency PCB manufacturer materials, thinking that low loss means signal integrity. But my experience is that for pulse stimulation of only tens to hundreds of kilohertz, using a high-frequency laminate may make ringing more obvious, because the dielectric constant is too low and the trace impedance is on the high side, and the slightest mismatch at the driver end overshoots. I am not saying the materials of a high Frequency PCB supplier are bad; they are simply not always a plus in this constant-current pulse scenario, and it depends on the specific output stage topology. In one project I tried replacing ordinary FR4 with a low-loss high-frequency material, and the waveform leading edge gained nearly 15% overshoot, which was only suppressed by finally adding a damping stage. So my habit is to first confirm whether the output waveform really needs that much bandwidth and only then decide whether to ask a high Frequency PCB manufacturer for a special laminate.

After mass production, environmental factors affect isolation far more than in the laboratory. I once received after-sales feedback that when the device was used in the rainy season in southern China, patients felt a slight pricking sensation. When it was taken back and opened, it turned out the isolation device had not failed; the PCB solder mask had tiny pinholes at the edge of the isolation slot, and after moisture seeped in, a weak but real path formed between the high-voltage side and the low-voltage ground. This kind of problem simply cannot be measured at the prototype stage because laboratory humidity is well controlled. So now in the mass-production documents I explicitly require added solder mask thickness on both sides of the isolation slot, and a sampled 85/85 test before shipment. Many companies understand isolation purely as optocouplers and transformers, but I think the physical isolation structure on the PCB is the bottom line of mass-production consistency.

When all is said and done, the difficulty of a Physiotherapy Device PCB is not in the schematic but in how you fit high voltage, narrow pulses and long-term safety into one board while keeping it stable across different suppliers and different batches. That requirement sounds simple, but every step of doing it is detail. Especially when you depend on both an hdi pcb manufacturer and a multilayer pcb supplier, the communication cost is higher than the technology itself.

physiotherapy device pcb products

High-Voltage Spacing, CTI and Independent Safety Circuits

In my years of working on physiotherapy equipment, my biggest takeaway about the Physiotherapy Device PCB is that the high-voltage part does not pass just because the circuit has been calculated correctly; the physical state of the board can overturn your design at any time. The numbers 360 V and 100 mA are just two parameters on a schematic, but once they land on the PCB, copper edge burrs, solder mask pinholes and the plating thickness on via walls all become leakage paths. When I first made prototypes, I left 2 mm spacing in the high-voltage area of the board, and when the boards came back and were hit with a withstand-voltage tester, flashover began before the nominal value was reached. Cutting it open later, I found a dozen-odd microns of residual copper from etching at the copper edge. Since then I no longer trust theoretical spacing; I trust only physical testing.

On the supplier side, I have stepped into quite a few pits. An ordinary multilayer pcb manufacturer may give you FR-4, but the CTI performance of FR-4 from different manufacturers differs greatly in a humid environment. I had a batch of boards placed in the southern damp season, and the high-voltage-to-ground leakage current rose nearly threefold; after they were taken down and dried, they were fine again. Later I switched to a multilayer pcb supplier that had specifically made medical high-voltage boards and had them run 100% withstand-voltage testing on every board, and only then did things stabilize. The HDI process I actually do not recommend for this kind of product, especially do not go to an hdi pcb manufacturer in the name of miniaturization and press the high-voltage area into fine lines as well. I worked with an hdi pcb supplier whose laser blind vias worked well in the low-voltage digital part, but in the high-voltage area it was still safer to leave ordinary vias and wide spacing separately. There is room in physiotherapy equipment, and no need to make the board as dense as a mobile phone.

The high-frequency part is more troublesome. Some physiotherapy devices superimpose a medium-frequency carrier and require a stable laminate dielectric constant, and then many people go straight to a high Frequency PCB manufacturer. But a high-frequency board factory does not necessarily understand medical safety regulations, especially when high voltage and high frequency exist at the same time, where the partial discharge resistance of solder mask and laminate is critical. I tried a factory very famous for RF boards, and when the high-voltage pulse was applied to the prototypes, very fine dendritic traces appeared on the board surface, which were in fact burned by partial discharge. Later I settled on a factory with both high-voltage and high-frequency experience; although it is not strictly a high Frequency PCB supplier, it can mix materials according to my requirements and uses a tracking-resistant solder mask in the key areas.

On the independent safety circuit, I personally insist on physical disconnection first. Between the 360 V energy storage capacitor and the output electrode plates, I place two solid-state switches in series, not controlled by the microcontroller, but cut off directly by hardware comparators. Although a 100 mA treatment current is not large, the instantaneous peak current of the capacitor discharge can reach several tens of amperes, so the layout must leave a short, wide return path, otherwise ground bounce will reset the digital part. One version of mine shared a section of ground between the safety shutdown driver and the ordinary signal ground, and as a result the microcontroller restarted occasionally whenever a pulse started; it took two weeks to find.

Finally, I want to talk about batch consistency. Prototypes passing tests does not mean every board in mass production will. Each time a PCB factory changes ink or lamination parameters, your withstand voltage may change. The agreement I now sign with a multilayer pcb supplier states explicitly that any critical process change must be notified in advance, and that a certain proportion of each batch be pulled for high-voltage aging. This money cannot be saved, because if something goes wrong it is on the patient, not on your test bench.

Constant-Current Sources Depend on Manufacturing Consistency

In my years working on physiotherapy equipment, my biggest feeling is that many people put their effort in the wrong place. When constant-current sources come up, everyone stares at closed-loop accuracy, DAC bit count and dynamic response, as if getting the circuit simulation to work made the product stable. When it comes to volume production and clinical use, the trouble brought by patient impedance fluctuation is far less than the problems of the PCB itself. I saw a project where the constant-current source was designed with great care and the feedback loop kept enough phase margin, but when the first batch of boards came back, the spacing between the high-voltage traces and the low-voltage control area was nearly one third smaller than on the drawing, and periodic leakage finally appeared on the patient end; it took more than two months to pin down that lamination registration was off. This made me fully understand that for a board like the Physiotherapy Device PCB, electrical design is only the first half; the second half depends entirely on whether the manufacturing side can hold your safety boundary.

When I later took part in physiotherapy products, my first reaction was not to open the schematic but to ask where the board would be made and which hdi pcb manufacturer would be used. With high-voltage pulses, a precision analog front end and digital control stacked on the same multilayer board, a slight layer misalignment can create extra coupling between the constant-current sampling trace and the high-voltage node. This coupling cannot be seen at all on a low-impedance patient, but in a high-impedance case such as dry skin or poor contact, intermittent oscillation can appear on the feedback loop and the current waveform is completely distorted. I would rather find an hdi pcb supplier with medical board experience from the start, even if the cost per board is twenty percent higher, which is cheaper than reworking after problems blow up on the clinical side.

High-frequency factors are also often simply ignored. The fundamental frequency of physiotherapy stimulation is indeed not high, but the harmonic content brought by the boost stage in front of the constant-current source and by the pulse edges is not small at all. An ordinary multilayer pcb manufacturer building a conventional FR4 stackup can see the dielectric constant deviate by ten-odd percent near tens of megahertz, and pulse overshoot and impedance mismatch follow. A high Frequency PCB manufacturer with an RF background will give suggestions down to details such as copper foil roughness and glass-weave style, and these places often decide whether the edge is crisp or sluggish. Conversely, a high Frequency PCB supplier that has only made base-station antennas may not fully understand medical safety spacing and creepage distance, so screening suppliers takes a good deal of effort.

When all is said and done, the difficulty of making a constant-current source deliver current into a patient steadily has never been in the control algorithm. When patient impedance varies from a few hundred ohms to a few thousand ohms, the power stage must withstand the transient, the high-voltage node must keep enough distance from low-voltage traces, and after the board heats up the material expansion must not shrink the creepage distance. All of this is physical-layer work that circuit simulation simply cannot discover. So I tell the team that, rather than agonizing over DAC bit count, it is better to free up the energy to find a multilayer pcb supplier with acceptable process consistency. Only when the board is made stably do the earlier fine designs truly count.

Capacitor Discharge, Power Stage Layout and Residual Energy

I have handled several home physiotherapy devices, and from the first prototype to mass production, what gave me the most headache was neither constant-current accuracy nor waveform distortion, but whether the high-voltage loop is actually clean after power-off. Many people place the safety boundary on the output current, thinking that as long as the current is clamped everything is fine, but I have always felt this approach solves only half of the problem. The other half is in the capacitors.

When I was making a physiotherapy device, the board had a group of large electrolytic capacitors storing energy for the high-voltage bus, and after each power loss, if they were not actively discharged, that voltage could hang around for a long time. At first I put a high-power resistor across the capacitor, thinking slower was fine, but when I later probed with an oscilloscope, touching the test point by hand more than ten seconds after shutdown still gave a tingle. That made me change my design completely: afterward, every high-voltage energy storage node got a dedicated discharge MOS and bleed loop, and once power-loss detection triggers, it opens a low-impedance path directly, bringing the bus voltage down in a very short time rather than relying on a resistor to consume it slowly.

The power stage of a Physiotherapy Device PCB really cannot copy the approach of ordinary consumer products. Enough creepage distance must be kept around high-voltage traces, and if the power ground and signal ground are mixed, the feedback loop will jitter inexplicably. When choosing a PCB factory, I look especially at whether the other party has high-voltage board experience. Many a multilayer pcb manufacturer can now make thick copper and sufficiently large inner-layer spacing, but the real difference lies in the consistency of solder mask and lamination. I worked with a multilayer pcb supplier whose four-layer board always had the middle dielectric thickness off, so the isolation withstand voltage failed to meet the requirement, and things stabilized only after switching to another one.

In the last two years I have gradually leaned toward using the HDI process for the main control and sampling part of physiotherapy devices, because the analog front end has high routing density and also needs impedance matching. But I still place the high-voltage power part in a separate area, not crowded together with the high-density interconnect area. When looking for an hdi pcb manufacturer and hdi pcb supplier, I first ask whether they can do local thick copper and hybrid lamination; some factories become vague as soon as they hear it, and then there is basically no need to talk further.

Another easily overlooked point is high-frequency noise. Although the main power frequency in a physiotherapy device is not high, there are very steep edges at the instant of switching, and if the sampling trace is too close to the high-voltage switching node, the collected signal will carry glitches. Later I moved the sampling line to an inner layer and wrapped it with ground planes, and at the same time had a high frequency PCB manufacturer help adjust the laminate dielectric constant, and the problem came down. When looking for a high frequency PCB supplier now, I also first confirm the loss tangent of the laminate and do not go by price alone.

When all is said and done, the bottom line of this kind of board is not how fancy the functions are but that the high voltage is cut off cleanly when it should be. There is stored energy in the capacitor, and the discharge path must be independent of firmware, otherwise when the MCU hangs the whole safety logic is gone. I only truly engraved this lesson in my mind after taking a fall.

Isolation Slots, Residual Copper and Stray Coupling

In my years working on physiotherapy equipment, I have gradually stopped seeing the isolation problem of the Physiotherapy Device PCB as a matter of which isolation device to choose; it is more about how the PCB factory understands your loops and processing boundaries. Drawing an optocoupler on the schematic does not mean the patient side and the low-voltage control side are really disconnected, because if a little copper remains under the isolation barrier or the solder mask window is off by a few tenths of a millimeter, the actual withstand voltage drops very sharply. I once had the same file proofed by two factories; one measured with no problem at all in isolation withstand voltage, while the other leaked as soon as high voltage was applied in a humid environment, and when cut open, the lamination fill had voids at the edge of the isolation slot, which means the loop had not been cleanly disconnected at all. So now when I follow a project, I confirm in advance with the multilayer pcb manufacturer how they handle the creepage distance and layer alignment of medical boards, rather than only pressing for delivery time. When choosing an hdi pcb manufacturer I also ask specifically about barrel-wall roughness after laser drilling and the copper fill quality, because if vias in the high-density routing area are near the isolation barrier, burrs on the barrel surface can eat up the safety distance.

Charge, Isolation and Pulse Safety at the Physical Level

During my time working on physiotherapy devices, my biggest feeling was that what is truly hard to deal with on the Physiotherapy Device PCB is never the functional circuit but how isolation and pulse safety are delivered at the physical level. Many people designing a scheme look only at the current limit, thinking that pinning the output current inside the specified range makes it safe, and when the prototype is measured it trips on charge. Charge and current are not the same thing: with the same current amplitude, widening the pulse width slightly makes the accumulated charge jump up, and this constitutes a risk independent of current, directly related to whether the heart can bear it. So in hardware I press the discharge time constant of the energy storage capacitor very hard, limiting charge accumulation at the source rather than adjusting parameters after testing exposes a problem.

On isolation, the pits I stepped into were basically all tied to the PCB factory’s process capability. Drawing an isolation slot and marking a creepage distance on the schematic looks quite standardized, but it turns into something else when the board is actually made. I changed to another multilayer pcb manufacturer, and with the same design files the withstand-voltage results differed by several times; the investigation found it was due to inner-layer core thickness tolerance and uneven resin flow during lamination, causing inconsistent dielectric thickness near the isolation barrier. This made me understand that the process control of a multilayer board supplier is far more important than design rules. Later I simply looked for a factory that works as both an hdi pcb manufacturer and a multilayer pcb supplier, whose laser drilling and layer alignment precision leave the isolation barrier edges cleaner and less likely to leave copper foil burrs, and in a humid environment those burrs are potential conductive bridges. In fine dense routing areas, an hdi pcb supplier does indeed handle isolation better than an ordinary factory; it costs more but saves the trouble of rework.

High-frequency signal routing is also common in physiotherapy equipment, for example modulation waveforms or internal communication. Running high-frequency signals on ordinary laminate gives outrageous reflection and loss, and the waveform reaching the output is completely deformed. Later I settled on a dedicated high Frequency PCB manufacturer, who are indeed professional in dielectric constant stability and copper foil surface roughness control, with much better impedance consistency than a general-purpose board factory. A high Frequency PCB supplier will also usually produce more detailed impedance test reports, so I do not have to adjust line widths over and over to make it fit.

I generally place Y capacitors far from the isolation slot. They really do suppress EMC, but if placed badly they create a weak point on the isolation barrier. My approach is to put the Y capacitor pads as close to the low-voltage side as possible, keep the area near the isolation barrier clean, require the PCB factory to give full solder mask coverage at the isolation barrier edge, and then apply conformal coating. These details add up to something far more reliable than simply increasing creepage distance, and they also lower the test risk related to charge and limits by a notch.

In my years working on physiotherapy equipment, my biggest feeling is that everyone puts attention in the wrong place. When the output stage comes up, everyone asks how high the voltage can be pulled and whether the pulse width is stable, but what really keeps me up in the lab until midnight is often the invisible capacitive coupling and charge residue of the board itself. A Physiotherapy Device PCB is not an ordinary control board; what happens between copper foils and between layers after transients of several hundred volts hit simply cannot be seen from schematic simulation alone.

I remember a project where the positive and negative phase pulses looked quite symmetrical on the oscilloscope, but with a current probe on the ground return, the negative phase side always had an extra tail. At first I thought the program was not written well, and adjusting the duty cycle for a long time did nothing. Later, when the board was cut open and the cross-section examined, I found that the reference copper on the middle layer had been cut into small pieces; as the rising edge of the high-voltage pulse passed, the adjacent layer directly induced a small patch of charge, which slowly discharged after the pulse ended, showing up on the output as a tail. The measured parasitic capacitance was only a few picofarads, which sounds small, but under pulses of several hundred volts at microsecond scale, the charge stored in a few picofarads is enough to make the positive and negative phases unclean. Later I switched to a multilayer pcb supplier that understood high-voltage spacing better, readjusted the stackup, and sandwiched the high-voltage traces between complete reference layers, and that tail was suppressed. After this I understood that the output accuracy of physiotherapy equipment is determined to a large extent by the PCB stackup and copper integrity, and cannot be solved by raising the sampling rate.

Some people like to go straight to an hdi pcb manufacturer, thinking the finer the lines and the smaller the board the more advanced. But on physiotherapy pulses I actually avoid it. The interlayer dielectric of HDI is too thin, and the small distance between trace and reference layer is enough to make distributed capacitance larger than you want to accept. With larger capacitance, pulse edges become blunt, and under high-voltage transients ringing may also appear: at best the waveform is unclean, at worst the local electric field concentrates and the insulation degrades over time. I would rather make the board larger and give the key loops enough interlayer distance. As for materials, I do not blindly pursue high-frequency laminates; although some high Frequency PCB manufacturer will recommend low-loss dielectrics, the spectrum of physiotherapy pulses is not as high as that of communication equipment, and what really affects the waveform is often copper foil roughness and the lamination structure. Choosing materials that are enough is fine, and saving the cost for creepage distance and insulation treatment is more practical.

Of capacitors and charge in physiotherapy equipment, what I care about most is what happens after power-off. Shutting the device down does not mean it is safe; if nobody manages the several large capacitors on the bus, the residual voltage can give you a shock. My habit is to reserve a place for a bleed resistor near the output end; even if the resistance is chosen larger, that is more stable than relying entirely on control logic to discharge. Because once the main controller fails to reset, the charge stays there and becomes a hazard at the next power-on or maintenance. As for the approach of separating the positive-phase and negative-phase capacitors, I have seen some designs, and it indeed improves net DC symmetry, but I think separating the capacitors alone is not enough: a small difference in loop inductance, or two power traces of different length, will skew the current path of the reverse phase pulse, and the charge finally reaching the load will still differ by a little. True symmetry comes from layout and cannot be guaranteed by matching two capacitor values.

When all is said and done, seventy percent of the stability of a physiotherapy equipment output stage is on the PCB. Some people spend a lot of time tuning algorithms, trying to compensate hardware faults with software, which I see as putting the cart before the horse. A reliable multilayer pcb manufacturer must understand the electric field distribution under high-voltage pulses and not just route like an ordinary digital board. An hdi pcb supplier of course has its value, but in this scenario it is not necessarily suitable. High-frequency laminates are the same: first work out how fast your pulse edges really are, and then decide whether to spend several times the money for those few nanoseconds of improvement. When I talk with project teams now, the first thing I do is not look at output parameters but pull out the board stackup and routing and go over them; many problems show signs at this step.

In the years working on the Physiotherapy Device PCB, my biggest change has been that I no longer put the emphasis on control algorithms but move my energy to the copper and pads that current actually flows through. The H-bridge discharge loop drawn on the schematic is often far from the path on the actual board, especially when the pulse edge is steep, where a few millimeters of trace difference makes the charge released by the two capacitors different. Some people like to select the positive and negative bus capacitors in the BOM as the same part number, thinking that makes them symmetrical, but I do not see it that way. 

physiotherapy device pcb manufacturing equipment

Hardware Protection Independent of Software

In my years of commissioning physical therapy equipment, what has made me feel least at ease is staking all the safety logic on software. It is not that I distrust the MCU, but in a situation like electrical stimulation output, hardware backup is always one step faster than software judgment. On a Physiotherapy Device PCB, if only firmware shuts off the high voltage, once the program runs away or a status bit is disturbed, the patient side may directly face current that should not appear. I have met a situation where the output was briefly out of control at the instant of reset, and only after I independently added a hardware window protection built from comparators did I feel more settled. Software can do monitoring and regulation but cannot be the only safety gate.

On the H-bridge, many people stare at the switching transistors themselves, but I think the parasitic parameters of the power loop and gate loop in the layout deserve more time. If the trailing edge of a biphasic pulse is unclean, residual charge will keep flowing through the patient during the pulse gap; you may not see anything when measuring average current, but over a long time the net DC drifts. To handle the trailing edge I tried letting the two low-side transistors briefly overlap in conduction during turn-off, but later found that if the overlap time was not controlled well, there was a risk of shoot-through in half a bridge. I now prefer to put a small independent bleed transistor at the output, so that the bleed path is more controllable and the bridge-leg devices do not also have to deal with body-diode recovery at turn-off. If the delays of the two isolated driver channels are inconsistent, the commutation timing goes off, and both waveform symmetry and charge balance are affected; this can only be handled by looking at channel-to-channel skew when choosing the isolation device and by keeping the drive loop short on the PCB.

Patient impedance is a rather interesting parameter: it is not a fixed number marked in the manual but changes with electrode fit, skin moisture and even muscle state. If during constant-current loop commissioning you use only a pure resistor as the load, trouble easily occurs in actual use. At low impedance the loop gain is high and the pulse front overshoots or even rings; at high impedance it may be limp and fail to rise to the set value. I generally put a small capacitor in parallel at the output to simulate cable and skin capacitance, then put a resistor in series and sweep several points, to see whether the constant-current loop shows any tendency toward marginal oscillation over the entire range. If enough phase margin is not left here, changing the person or the electrode pads on site may change the waveform.

When making this kind of board I will not casually look for an ordinary double-sided board factory. Power and signal are mixed together, routing density is not low, and impedance must be controlled, so when I look for a multilayer pcb manufacturer or multilayer pcb supplier, I generally ask directly for the lamination structure and impedance test reports. Once a batch of boards had lamination not well controlled and the ringing on high-voltage pulse edges clearly grew, and it calmed down only after switching to a supplier with more experience in impedance control. If the board also has a high-frequency sampling or wireless communication part, then you have to see whether the high Frequency PCB manufacturer or high Frequency PCB supplier can make dielectric constant consistency good. When the density gets a bit higher, the laser drilling and micro-via process of an hdi pcb manufacturer or hdi pcb supplier will also affect yield. Many people think that as long as a supplier is cheap it is fine, but in electrical stimulation equipment these details are all reflected in patient sensation and treatment consistency.

My habit now when making this kind of product is that hardware protection must be independent of software, and three things, namely the H-bridge bleed path, the delay matching of the isolated drivers and full-load stability of the constant-current loop, have to be forced to be verified at the PCB stage. It cannot pass in one go every time, but at least when something goes wrong you know which direction to check.

Redundancy, Stackup and Impedance Drift

In my years working on physiotherapy equipment, my biggest feeling is that everyone puts the word safety redundancy on too high a pedestal. If the main controller fails, who protects the patient? Many people’s first reaction is to add an independent watchdog and hardware comparators, but if the board’s own stackup and impedance control are a mess, no number of redundant channels is more than psychological comfort. A Physiotherapy Device PCB is usually stuffed into a very small enclosure, and once the pulse output comes on, ground bounce can stir the voltage sampling signal beyond recognition. If this is not straightened out first, however beautifully the constant-current loop is tuned, it will still jitter in actual operation.

I have paid quite some tuition on the supply chain. Early on, to be cheap, I went to a small factory for multilayer boards, and although the part number was the same in every batch, the dielectric constant was like opening a blind box. With the same impedance design requirement, the result drifted from seventy-odd ohms to ninety-odd ohms, and the constant-current output was directly disordered. Later I switched to a manufacturer dedicated to HDI PCBs; the unit price went up, but impedance consistency finally stabilized. So when I pick an hdi pcb manufacturer now, I first ask whether they have impedance test reports and lamination parameter control capability, rather than looking only at the quotation. The laser drilling accuracy and layer alignment of an HDI PCB supplier directly bear on the high-voltage isolation spacing of physiotherapy equipment, and if creepage distance is not enough, however much the main controller monitors afterward, it cannot pass safety regulations. The same goes for the multilayer pcb manufacturer side: the dielectric thickness tolerance of the laminate must be asked about clearly, otherwise the impedance you worked hard to calculate is just a number on paper.

I do not much like verifying the loop only at nominal impedance. Whether the patient’s skin is sweaty or not, how tightly the electrodes are attached and what range the impedance of aged electrode pads can reach are far too uncontrollable. My habit is to let the device run through the entire load range, sweeping from a few tens of ohms to a few thousand ohms, to see whether the constant-current loop suddenly oscillates at some point. The premise of this test is that the board’s own impedance is stable: if the dielectric loss of the laminate supplied by the high Frequency PCB manufacturer is on the high side, the pulse edges will slow down and the constant-current response will also seem sluggish. The copper foil roughness of a high-frequency PCB supplier also cannot be ignored: if roughness is too large, signal attenuation is obvious, and problems an ordinary digital board does not expose all emerge in an analog feedback chain.

What gives me the most headache in the feedback chain is the bandwidth of the isolation amplifier. The operating frequency of physiotherapy equipment is not high, but the pulse edges are steep, and if bandwidth is insufficient, a fast-changing current signal is integrated into a slow ramp. If the ADC sampling rate of the main controller cannot keep up, an extra phase delay is introduced and the dynamic response becomes sluggish. In this situation, do not rush to change the algorithm; first check whether the dielectric thickness and copper foil roughness given by the multilayer pcb supplier have changed. These parameters directly affect the fidelity of the signal rising edge, and the source of many so-called unstable algorithm problems is actually at the PCB factory.

For the detection of open circuit, short circuit and poor contact, I do not like to pile on complex state machines. Synchronously sampling voltage and current in hardware is enough, but the sampling timing is particularly delicate. There is a polarization voltage between electrode and skin, and if you sample just after the pulse ends, the reading is outrageously high, and if you wait several tens of microseconds it returns to normal. If the detection window is designed too narrow, false alarms can annoy you to death. I would rather keep the voltage sampling trace away from the power loop at the layout stage, so that pulse current does not induce glitches on the sampling line. Otherwise the voltage signal sampled back carries high-frequency noise, and no judgment logic at the back end is reliable.

Is an independent safety circuit necessary? Yes, but it should not be deified. An external window comparator directly monitors the voltage across the current sampling resistor and cuts off the enable when a threshold is exceeded, which is far more reliable than a software watchdog. But there is a premise that is easily overlooked: the reference voltage of the comparator must use a separate low-noise supply and cannot be shared with the main controller. Once the main controller runs away and the GPIOs jump randomly, if the reference level of the protection circuit sways with it, it is worse than having none. True safety redundancy first cuts noise and interference clean through the layout and the power architecture, and only then does hardware protection mean something.

physiotherapy device pcb inspection equipment

Current Paths, Supplier Hype and Selective Technology

A couple of years ago I helped a rehabilitation team tune a physiotherapy device, and after changing PCB factories three times I gradually understood that the real difficulty of the Physiotherapy Device PCB is not the number of stackup layers but how the current paths are arranged. Many people start by asking whether eight or ten layers can be made, but the output stage current of physiotherapy equipment is not small, and trace width, copper thickness and via positions matter much more than adding two more inner planes. I suffered on my first version: I routed the feedback sampling line beside the power inductor, and when the output was loaded, the current waveform read by the analog front end was all glitches and the microprocessor directly reported a fault. Later I moved the sampling point close to the output connector and poured a separate ground, and the problem calmed down.

When choosing suppliers, I advise not being led by the talking points of an hdi pcb manufacturer. The advantages of HDI are small line width and spacing and the ability to fit more routing, but the board area of many physiotherapy devices is not tight; what is truly lacking is reliability and repairability. I later split the main control board and the power board: for the main control board I found a multilayer pcb manufacturer to get it done with four layers, the power board went separately through a thick-copper process, and at mass production I switched to another multilayer pcb supplier that specializes in thick copper boards because the earlier one could not fit us into its capacity. An hdi pcb supplier also came to quote and said it could make laser blind vias; I asked what this board needed blind vias for, and the other party could not answer. Of course, if you are making a small portable neuromuscular stimulator, where component density is indeed high, it is also reasonable to find an hdi pcb supplier for a first- or second-order board, but applying a consumer electronics mindset to physiotherapy equipment will waste quite a lot of money.

For a small board responsible for pulse modulation, because it had to run at a rather high frequency, I first used an ordinary laminate and the waveform edges were somewhat blurry; only after switching to a hybrid lamination scheme from a high Frequency PCB manufacturer was signal integrity pulled back. But the lead time of a high Frequency PCB supplier is long and the cost of a hybrid board nearly doubles, so I used high-frequency laminate only in the few centimeters of the RF front end and conventional FR-4 elsewhere. This experience holds equally for the pulse control circuit of physiotherapy equipment: spend where you should, and not a penny more where you should not.

The role of an SCR in a protection circuit is often simplified to trigger on overcurrent, but in practice it is not that clean. One of our prototype versions did trigger the SCR in a short-circuit test and the bus capacitor energy was discharged, but after the test the device could not be powered up again immediately, because the SCR remained in conduction and had to wait for the current to fall below the holding current. This means that an active reset is needed after a protection action, or a current-limiting resistor in the output loop to help it turn off. In addition, the trigger loop had better not share a power rail with the main controller, otherwise when the main controller supply is pulled down, the SCR trigger pulse may not be crisp enough. 

Medical Isolation, Charge Paths and Supplier Experience

In my years working on physiotherapy equipment, I have become less and less willing to look only at surface parameters when it comes to isolation on the Physiotherapy Device PCB. Many people regard isolation as something solved by choosing an optocoupler or DC-DC with a high enough withstand voltage, but when the actual board comes back and is measured, the charge that should leak still leaks and the places that should spark still spark. The problem often lies at the PCB factory. The understanding of medical boards of many a multilayer pcb supplier is still at the level of consumer electronics: thin prepreg between layers, casual solder mask openings, and creepage distance not controlled to IEC 60601 requirements at all. Then however good the isolation device you use, the bare board itself has already broken the safety boundary first.

Later I switched to an hdi pcb manufacturer that specializes in high-reliability boards, and things stabilized. The benefit of the HDI process is not only that line width and spacing can be small, but that the position of laser blind vias and the stackup can more actively control the electric field distribution. When high-frequency pulses hit in physiotherapy equipment, the copper foil on both sides of the isolation barrier forms parasitic capacitance, and charge crosses along this path. Many designs look only at whether the charge at the output end exceeds the limit and forget that those inconspicuous copper patches, vias and even residue under the silkscreen on the board are all participating in charge transfer. Once high-speed edges come up, there are many more paths than imagined. Some high Frequency PCB manufacturer plants are busy pushing loss and impedance down but leave excessively large reference-layer windows near the isolation barrier, which instead makes stray capacitance harder to predict.

So when I choose a supplier now, I first ask whether the other party has made medical power supplies or physiotherapy boards, especially a high Frequency PCB supplier, because such a factory’s understanding of dielectric stability and creepage control is completely different from that of an ordinary PCB factory. A higher layer count in a multilayer board is not better; symmetrical stackup, uniform dielectric thickness and machining accuracy of isolation slots are what matter. Some hdi pcb supplier plants can make isolation slots very small but do not control slot wall roughness, and in a humid environment insulation performance falls straight down; this kind of board cannot be detected at delivery and gets into trouble after a few months of use.

The same goes for charge: limiting pulse width in software can only be the last line of defense. True safety redundancy should be built at the physical level, so that charge which should not exist has no possibility of forming a path, rather than waiting until it appears and then limiting it. On a safety-critical path, every millimeter of creepage distance and the stability of every dielectric layer matter more than that bit of BOM cost. Going back to check the PCB factory process after a problem occurs costs far more than choosing a reliable hdi pcb manufacturer or multilayer pcb manufacturer in the first place.

Multilayer Experience, HDI Records and Wireless Modules

I have worked on several physiotherapy devices with current stimulation, and the boards were changed from two layers all the way to six, with more pits along the way than imagined. Many people start by talking about safety redundancy and independent hardware protection, but in actual projects I care more about whether the PCB factory has the ability to keep impedance and withstand voltage stable. Especially for a board like the Physiotherapy Device PCB that carries both high-voltage pulses and precision sampling, a slight drift in stackup and routing changes the output waveform completely, and all later debugging is in vain.

In the early days we casually found a factory that made ordinary double-layer boards. The prototypes had no problem, but small batches had all kinds of strange problems and could not even pass the basic withstand-voltage test. Only after switching to a supplier specializing in multilayer boards did the situation improve. When we talked with a multilayer pcb manufacturer, the other party directly pointed out that the stackup order of our original scheme was unreasonable and the high-voltage ground and signal ground were not truly separated; after the change, many inexplicable restarts and interferences disappeared. But I also compared two multilayer pcb supplier options whose prices could differ by a factor of two, and the more expensive one was indeed stable in copper thickness uniformity and solder mask withstand voltage, while the cheaper one was prone to ion migration near high-voltage nodes, causing leakage over time.

Later, as board density rose, with digital parts and the analog front end crowded together, an ordinary through-hole board was hard to route, so we tried HDI. When proofing with an hdi pcb manufacturer, we found that even among HDI, the laser drilling quality and copper fill processes differ a lot between factories. One hdi pcb supplier gave sample boards that looked beautiful, but on power-up testing the inner-layer micro-vias had hidden cracking and sparked directly under high voltage. This made me understand that for this kind of equipment you cannot look only at the quotation and delivery time; the supplier’s process records are more important than anything, especially since medical products have small volumes and you have no chance to spread the quality-control risk through large batches.

On hardware shutdown, my approach is actually different from many people’s. Some like to make safety protection entirely an independent loop detached from software, but I hold some reservations. In one of our designs, overcurrent protection and the discharge circuit were all built in hardware, and as a result the board grew a size larger, the failure points increased instead, and repair was laborious. Later we changed to a hardware comparator as the last shutdown, with software for state monitoring and logging, and in testing reliability did not drop while production and maintenance became much simpler. For capacitor discharge in the shutdown state, we simply connected a normally-closed bleed resistor at the output and did not add an extra SCR trigger, and so far there has been no problem. I think as long as the power rating and failure mode of that resistor are chosen correctly, there is no need to turn the system into a layered protection network, which would instead introduce new failure points.

If a physiotherapy device has a wireless module or high-frequency modulation, the laminate and surface finish also have to be chosen. We had a prototype with Bluetooth and 2.4G that used ordinary FR4, and signal attenuation was severe, with communication distance less than half of the datasheet. After consulting a high Frequency PCB manufacturer, switching to low-loss material and readjusting the stackup, the problem was solved. The high Frequency PCB supplier also advised us to control copper foil roughness. These details are easily overlooked in medical products, because most hardware engineers care only about the schematic and take whatever the PCB factory says.

Overall, I think that for this kind of board, rather than piling a lot of protection circuits on the schematic, it is better to get the PCB process and supplier right first. Multilayer, HDI and high-frequency boards are not better the more high-end they are; the key is that they match your design and that the supplier can produce them stably and repeatably. Hardware protection must of course exist, but do not make the system too complicated in pursuit of absolute independence, as that introduces new failure modes, and in the end even you cannot say where the problem lies.

Frequently Asked Questions

Why is the PCB factory so important for a Physiotherapy Device PCB?

Many output interferences, leakage events and occasional loss of control trace back to manufacturing, not the schematic. Copper-edge residue, solder mask pinholes, dielectric thickness drift, resin voids beside isolation slots and layer misalignment all reduce real creepage and withstand voltage, so the same design files can perform very differently between factories and between batches.

How should isolation and creepage be handled on the board?

Treat isolation as a physical boundary in the layout: keep isolation slots clean of residual copper, require extra solder mask thickness at the slot edges, place Y capacitors toward the low-voltage side, sandwich high-voltage traces between complete reference layers, and verify with sampled 85/85 humidity tests and cross-section reports for every batch. Do not rely on isolation devices alone.

Should a physiotherapy device use HDI or high-frequency laminates?

Only where they fit the real need. HDI suits the dense analog front end and portable designs, but its thin dielectric raises distributed capacitance, so high-voltage power areas often stay with ordinary vias and wide spacing. High-frequency laminates help for carrier modulation or wireless parts, but for pulses of tens to hundreds of kilohertz they can increase ringing, so check the actual pulse bandwidth first.

What hardware protection and discharge design is recommended?

Keep hardware protection independent of firmware: use hardware comparators and series solid-state switches for shutdown, give the comparator reference its own low-noise supply, add dedicated bleed paths for high-voltage storage capacitors, keep short wide return paths to avoid ground bounce, and limit charge accumulation at the source through the discharge time constant rather than the current limit alone.

More Posts

Leave Us Message
Drag & Drop Files, Choose Files to Upload You can upload up to 5 files.
Please upload your Gerber files or BOM, and we will provide a quote promptly.

Your trusted PCB manufacturing and one-stop PCB assembly supplier

• Expert in Small-to-Medium Batch Production
• High-Precision PCB Fabrication & Automated Assembly
• Reliable Partner for OEM/ODM Electronic Projects

Business Hours: (Mon-Sat) From 9:00 To 18:30