← Back to the plan

Soft starters, hard start kits, and SCE billed demand at 1660 9th St

Research memo · compiled 2026-09-06 · sources cited inline

Bottom line: the $4,500 / 7-unit install will not lower SCE billed demand. Not "a little", the effect is roughly 0.05 kW, and SCE bills demand to the nearest whole kW. Victor is right and Farzad is right. Ari's and Greg's theory that compressor start-up inrush drives the 19.5 kW peak is wrong, for two independent reasons proven below.

Confidence: high on the physics and the arithmetic. Moderate on the specific site numbers, because nobody has yet pulled SCE 15-minute interval data or measured a single compressor at this building. Every kW figure here rests on stated assumptions, listed where used.

1. The category error: a hard start kit is not a soft starter

Victor proposed "Supco Hard Starts or Compressor Saver 5-2-1." Both of those are the same device class, and it is the opposite class from a soft starter.

Hard start kit = a start capacitor plus a relay (potential relay or PTC thermistor) that dumps extra phase-shifted current into the compressor's start winding, then drops out. It adds torque. Supco advertises the SPP5 at "300% torque increase," the SPP6 at 500%, the SPP7S at 600% (Supco SPP series datasheet).

It does not reduce peak inrush. Bryan Orr of HVAC School put an oscilloscope on this and published the traces:

"Hard start kits don't decrease starting amps at the moment of start; they can't. What they can do is reduce the time it takes to get the motor started... Hard start kits do increase the current on the start winding.", HVAC School, Start Capacitor & Inrush, Facts & Myths Part 4

"Hard starts reduce the time-averaged starting current because they get the compressor to start up more quickly (therefore, the starting current is higher for a shorter time). However, hard starts do NOT reduce the spike of current upon startup.", HVAC School podcast, Electrical Myths

Micro-Air, who make an actual soft starter, say it plainly:

"Hard starts are simply a start capacitor in series with a PTCR. Hard starts actually increase the start-up amperage, but slightly reduce the start-up duration.", Micro-Air EasyStart performance paper

Mike Sokol's scope traces make the same point from the generator side: a hard-start capacitor "still pulls the maximum LRA... but for a longer time of 1.5 seconds compared to 0.150 seconds of the stock starting capacitor" (RV Electricity).

Why technicians believe otherwise: clamp meters in inrush mode sample over ~100 ms. A compressor with a hard start is already near full speed at 100 ms, so the meter reads lower. The meter is measuring how fast it started, not the peak.

5-2-1 Compressor Saver (CSRU1 / CSRU2 / CSRU3) is the same thing, the manufacturer's own product sheet calls it "Potential Relay & Start Capacitor" (CPS product sheet). Its marketing claims "reduce damaging in-rush current by up to 50%," but the mechanism it describes is time: it "significantly reduces the time the compressor draws high amperage at start-up." That is a duration claim wearing a magnitude claim's clothes. No independent measurement supports a 50% peak reduction, and the scope traces above contradict it.

Nothing Victor quoted is a soft starter. If the goal were genuinely inrush reduction, the parts list is wrong on top of the strategy being wrong.

2. Comparison table

Note on the SureStart three-phase line: the manufacturer's own catalog states "Start Current Reduction: 70% (single phase); 40% (three phase)" (SureStart catalog). Three-phase motors start far more gracefully to begin with , which is part of why Victor said he can't add a hard start to the 3Ø units, and why he doesn't need to.

3. The math (this is the whole memo)

3.1 What SCE actually bills

SCE's tariff language, verbatim:

"The Maximum Demand for each TOU period shall be the measured maximum average kilowatt (kW) input indicated or recorded by instruments, such as SCE metering, during any 15-minute metered interval... Where the demand is intermittent or subject to violent fluctuations, a 5-minute interval may be used.", SCE Schedule TOU-GS-2-RTP, Special Conditions §4

"The Billing Demand shall be the kW of Maximum Demand determined to the nearest kW." (§5, same document)

Three things follow immediately, and they decide the case:

1. It is an average over 15 minutes, not a peak instant. A half-second event is 1/1800th of the interval.
2. It is kW (real power), not amps and not kVA. Locked-rotor current is mostly reactive and SCE does not bill it.
3. It rounds to the nearest kW. Anything under 0.5 kW is invisible by definition.

The 20 kW threshold that governs Greg's PCOC / rate-change goal is measured the same way: TOU-GS-1 eligibility turns on whether monthly maximum demand, that same 15-minute average, has exceeded 20 kW (TOU-GS-1-RTP, Applicability).

3.2 Real power during inrush is not 5× running power

This is the step most people skip. LRA is an apparent current at a terrible power factor. During locked rotor the motor is essentially an inductor: measured power factor runs 10-20% through most of the start, spiking only in the final moments as the rotor reaches speed (Energy350, Digging Deeper into Motor Inrush Current).

Assume a representative 4-ton single-phase condensing unit at this site:

Running real power: P_run = 240 × 20 × 0.85 = 4,080 W ≈ 4.1 kW

Real power during the inrush: P_inrush = 240 × 110 × 0.25 = 6,600 W ≈ 6.6 kW

The current is 5.5× running. The real power SCE bills is 1.6× running. The scary number on the nameplate is mostly reactive power that never appears on the bill.

3.3 What one start adds to the 15-minute average

Take the most pessimistic framing: the unit was off before the start, so the entire 6.6 kW is new load, and stretch the start to a full 1.0 second.

Energy in the inrush: E = 6.6 kW × (1 s ÷ 3600 s/h) = 0.00183 kWh

Contribution to the 15-minute average: ΔkW = 0.00183 kWh ÷ 0.25 h = 0.0073 kW = 7.3 watts

One compressor start adds about 7 watts to billed demand.

Using the more realistic 0.5 s start and counting only the excess over the running level the unit would draw anyway: (6.6 − 4.1) kW × (0.5 ÷ 3600) h ÷ 0.25 h = 0.0014 kW = 1.4 watts

3.4 All seven units starting inside one interval

Worst imaginable case, all 7 compressors start within the same 15-minute window, all 1.0 s, all treated as pure addition:

7 × 0.0073 kW = 0.051 kW ≈ 0.05 kW

The combined inrush of every air conditioner in the building contributes about one twentieth of one kilowatt to billed demand. SCE rounds to the nearest kW. It contributes zero.

Even if SCE invoked the 5-minute interval clause for "violent fluctuations": 7 × 0.00183 kWh ÷ 0.0833 h = 0.154 kW

Still zero after rounding.

3.5 Therefore: what soft starters save

A best-in-class soft starter cuts peak start current 65-75%. Applied to a quantity that is 0.05 kW, the saving is:

0.05 kW × 0.70 = 0.035 kW = 35 watts

Installed cost $4,500. Billed-demand saving: 0.035 kW. That is $0.00 per month at any demand rate SCE charges, because the number rounds to zero before it reaches the bill.

And soft starters do nothing at all to running current. Every manufacturer says so; the SureStart is explicitly "out of the circuit after startup, consumes less than 2 W average in running mode." That is the point of a bypass. The compressor draws its full RLA the entire time it runs, with or without the device.

3.6 What actually produces 19.5 kW: concurrency

Same 4-ton unit, all-in per system (compressor + condenser fan ~0.25 kW + indoor blower ~0.5 kW):

4.1 + 0.25 + 0.5 ≈ 4.85 kW per system running

For smaller 3-ton units (RLA 15 A): 240 × 15 × 0.85 = 3.06 kW + fans ≈ 3.8 kW per system.

Four to five simultaneous compressors is the entire 20 kW budget, before a single light, console, mini fridge, LED wall, or laptop. That is the mechanism behind every spike this building has recorded.

Duty cycle counts fractionally, which is the good news. A unit running 8 of the 15 minutes contributes 8/15 of its kW to that interval's average. Demand is not "how many are on", it is "how much compressor-minutes landed in this window." That is exactly what a demand controller manipulates, and exactly what a soft starter cannot touch.

3.7 Two site-specific flags the AC theory does not explain

The Tesla charger. Greg's own note records 11.3 kW of Tesla charging observed on site. That single load is 58% of the 19.5 kW ceiling and 2.3× the largest AC unit here. It is also a sustained load, so it lands on the 15-minute average at full value, unlike an inrush, which does not. If Tesla charging overlapped the January and February peak intervals, it explains them by itself with room to spare. Confidence: moderate, pending interval data. This is the single highest-value thing to check, and it costs nothing to check.

January. The 19.5 kW peak was 2026-01-22, and the 17.6 kW peak was 2026-02-28. Seven air conditioners do not run concurrently in Santa Monica in January. If those units are heat pumps in heating mode the picture changes, and if they have electric resistance strip heat it changes a great deal, strip heat is typically 5 kW per stage of pure resistive load, which would dominate everything. Either way, "compressor start-up inrush in winter" is the least likely explanation on the list. Confidence: moderate. Determining whether these are cooling-only or heat pumps, and whether any have strip heat, is a 20-minute question for Victor and it materially changes the plan.

The site also has a 3-phase sauna (resistive heaters are commonly 8-15 kW), an LED video wall, and the mini fridges SCE itself named as a common culprit. None of these are addressed by anything Victor quoted.

4. Independent corroboration that this is a known, named myth

This is not a contrarian reading. It is the textbook position across motor engineering.

"Soft-start equipment on big electrical motors cuts utility demand charges... Demand charges from utilities, however, are not affected. The electric meter measures the average kW consumed over each 15 or 30-minute period. In contrast, soft starters affect motors' power draw over the course of just a few seconds.", Machine Design, 5 Common Motor Myths

"Soft starters typically ramp up the voltage applied to a motor over a few seconds at start-up... but it doesn't affect utility demand charges. That's because the electric meter averages the kilowatts consumed over each 15-30-minute period.", Maintworld, Common Misconceptions About Motors

"In virtually all cases demand for an industrial plant is based on a 15- or 30-minute average. Thus, brief high peaks, such as those that are present during the starting of large motors, are averaged because the starting is of very short duration with respect to the demand-averaging interval.", Edward Cowern, P.E., Baldor Motor Basics: Factors that Determine Industrial Electric Bills

"The engineering consensus is clear: soft starters do not reduce energy bills in standard applications... expecting measurable savings on utility demand charges is a common misconception." , Industrial Monitor Direct engineering analysis

Energy350's field case is the closest analogue to a controlled test: an 800 hp motor with 1,500 A inrush, logged. Real power spiked to 800 kW for under one second. The 15-minute demand window average came out at 262 kW, indistinguishable from normal operation. Their conclusion: "peak demand occurs during peak operation, not at motor start-ups."

A soft-starter dealer on the Mike Holt forums, arguing against his own commercial interest:

"Most utilities use a sliding 15 to 30 minute demand window, so any starting surge experienced is not seen, only the actual running load is. In all but those few cases, soft starters have no tangible direct effect on peak demand charges."

He goes on to explain the two documented cases he investigated where a customer did see demand savings after installing soft starters: in both, the real cause was that operators started shutting equipment off during breaks once they stopped fearing restarts. The soft starter changed behavior, not physics. That is a genuine and relevant mechanism, but it is a $0 policy change wearing a $4,500 price tag.

5. Warranty and install notes (for completeness, since these are the usual objections)

• Copeland scroll compressors accept soft starters, with a hard constraint: "Normally, the ramp-up time should not exceed 1 second" for standard scrolls, and 3 seconds on 20-40 ton models (Copeland application guidelines; AE4-1388). A soft starter configured with a long ramp to maximize inrush reduction can violate this. If anyone does install these, the ramp setting is the thing to check.
• Copeland notes that scrolls "start unloaded, even if system pressures are not balanced... starting components are normally not required." Start assistance on a healthy scroll is solving a problem the compressor does not have.
• The 5-2-1 marketing claim "approved by all major compressor manufacturers" refers to the 3-wire potential-relay method, which Copeland does document in AE10-1272. That is true but narrow: it means the wiring method is sanctioned, not that the device reduces inrush.
• Copeland's own single-phase soft starter, SecureStart, is positioned explicitly for "applications where in-rush current is regulated", i.e. utility interconnection rules and generator limits. Not demand charges. Its listed value drivers are warranty savings from low-voltage and reverse-rotation protection.
• Hard start kits carry a real downside: a stuck potential relay leaves the start capacitor in circuit and cooks the start winding. HVAC School: "Adding in a hard start adds more current at start and the possibility of potential relay sticking that can definitely cause start winding stress."
• Three-phase units cannot take a hard start kit at all. Victor is correct on this. They also need it least.

6. Recommendation

Do not spend the $4,500 on soft starters for the demand goal.

It buys 0.035 kW of billed-demand reduction against a 20 kW ceiling. Restated as a decision rule: if a proposed fix acts for less than about a second, it cannot move a number that is averaged over 900 seconds. Nothing about this site changes that arithmetic.

There are three legitimate reasons to buy a soft starter, and Greg should know them so the decision is made on the true grounds:

1. Light flicker. If the LED video wall or studio gear visibly dips when a compressor kicks on, a real soft starter (EasyStart Flex or SureStart, not a Supco kit) fixes that. This is a production-quality issue in a recording studio and might be worth money on its own merits.
2. Generator or battery backup. If Kora ships, or any inverter/generator is added, LRA is the binding constraint on that hardware and soft starters become genuinely necessary. Worth revisiting then, sized to the battery.
3. Compressor longevity. Marginal, real, unquantified.

None of those is the demand charge. If Greg wants them, buy them with eyes open, and buy the correct device class, a true soft starter at ~$300-390/unit, not a $20 hard start kit.

Do these instead, in this order

1. Pull the SCE 15-minute interval data. Cost: $0. Do this before anything else. Download Green Button interval data from SCE My Account for the 2026-01-22 and 2026-02-28 peaks, plus a full recent month. Find the exact 15-minute intervals that hit 19.5 kW and 17.6 kW, then reconstruct what was running. Every other decision here is guesswork until this exists. Right now the entire $4,500 proposal rests on a theory nobody has tested against a single data point.

2. Lock down Tesla charging. Cost: $0. 11.3 kW is over half the ceiling in one load. Set the vehicle or charger amperage limit, and schedule charging outside business hours entirely. If the goal is confidence the building will never exceed 20 kW, uncontrolled Level 2 charging is incompatible with that goal on its own, regardless of what the AC does. This is the highest-leverage free action available.

3. Answer the equipment question. Cost: one Victor visit ($380). For each of the 7 units: nameplate RLA, LRA, voltage, phase, tonnage, cooling-only vs heat pump, and presence of electric strip heat. Also the sauna's kW rating. This is what Victor's 2-hour minimum should be spent on, a load inventory, not installing hard start kits. With that table, the concurrency budget above becomes exact instead of assumed.

4. Install the real-time monitor Greg already wants. Cost: ~$1,000-2,500 installed. Circuit-level submetering (eGauge, Verdigris, or similar) on the main plus each AC circuit, the sauna, the EV charger, and the video wall. This turns "we think it's the ACs" into a live answer, and it is the prerequisite for any automated demand limiting. It also gives Greg the confidence he says he needs before filing the PCOC, you cannot declare a permanent change in operating conditions you cannot see.

5. Automate the AC concurrency limit. Cost: ~$2,000-4,000 installed. The staff policy of staggering AC turn-ons by 15 minutes (in force since 2026-03-17) is exactly the right idea and exactly the wrong implementation, it depends on humans, and one slip in one 15-minute interval sets the demand charge for the entire month. Two ways to make it mechanical:

• Energy Sentry 9388B commercial demand controller , reads the building's actual kW through a CT, projects whether the running 15-minute average will breach a set limit, and sheds the lowest-priority compressor before it does. Configurable 15-minute averaging period to match SCE's exactly, 8-16 relay outputs, fixed-priority or rotating shed strategies so no zone stays off too long, and per-load minimum on/off times up to 20 minutes to protect compressors from short-cycling. Set the limit at 18 kW and the building physically cannot produce a 20 kW interval. This is the direct instrument for the stated goal.
• Smart thermostats plus Encycle Swarm Logic, cloud coordination that prevents units from running concurrently. Documented results on comparable sites: 13% average peak HVAC demand reduction at a 27,000 sq ft California office with 13 RTUs, 16% peak demand reduction at Southern California retail sites with 4-6 units each, 18% at a healthcare site with 7 rooftop units. That last case is the closest match to TRC's 7-unit configuration.

Either approach is an electrician plus a controls integrator, not an AC contractor. Note that a demand controller and a soft starter are not competing purchases: one addresses the thing SCE bills, the other addresses the thing it does not.

6. Interlock the sauna and treat the video wall as a scheduled load. Resistive sauna heat is sustained load and lands on the demand average at full value. It should not be permitted to run concurrently with peak AC.

7. Keep Kora in play, but not as the plan. A battery genuinely does shave 15-minute demand, because it supplies energy over the interval rather than milliseconds of it. But delivery has slipped four times (June → end of July → July-August → August-September) and it is still pre-certification. Treat it as upside on top of items 1-6, never as the thing that makes the PCOC safe to file.

On filing the PCOC

Greg's stated goal is confidence the building will "never again exceed ~20 kW." Items 1-3 tell you whether that is even achievable with the current equipment. If seven systems really do total 26-34 kW of connected running load, then no device, soft starter or otherwise, makes 20 kW safe, only enforced concurrency limiting does, and it needs to be automatic before anyone signs a declaration. Filing on the strength of a soft-starter install would be filing on a mechanism that provably does nothing.

Appendix: assumptions used

Every number above marked "unverified" is resolved by recommendation 3, at a cost of one Victor visit. Demand rates and the PCOC mechanics are being researched separately by the tariff agent and are deliberately not duplicated here.