DecisionHelmdesign decision intelligence

Quieter drives.
Lighter fixes.
Fewer late changes.

DecisionHelm cascades your vehicle targets down to the component specs that control them. Every trade gets a price in mass and dollars, before tooling locks it in.

See what you can start now Talk to Hasan →
housing mode gear mesh order motor order whine peak frequency → speed
Fig. 0 · Order mapDiagram, not data
01Ranked fixesEach with its expected dB and its mass cost.
02Targets you can defendEach limit traced to the vehicle target it serves.
03Earlier decisionsTrades settled in design, ahead of the dyno and the customer.

Founder's experienceLucid Motors · Apple · Turntide Technologies · Boeing 747-8 program · University of Cincinnati, sponsored research for BMW, P&G and GE Aviation · IIT Madras

01Why DecisionHelm

Small teams ship electric drives without a deep NVH bench. The whine shows up late, on the dyno or in a customer's ear, and every fix then costs mass, money and weeks. I built DecisionHelm to move that decision earlier. It links each vehicle target to the part that controls it and prices every trade in kilograms and dollars. Your engineers see which target is at risk, by how much, and what closing the gap costs.

Hasan Pasha, founder

02Who it's for

Teams without a deep NVH or target-setting bench.

Your product moves, spins or switches, and people will hear it. Noise sits with one engineer, or with nobody yet.

EV and e-axle

Electric drives

Gear and motor tones with no engine to mask them.

eVTOL

Electric aircraft

Motor and propulsor noise that shapes certification and community acceptance.

Robotics

Actuators

Gearbox whine at ear height, in homes and warehouses.

Energy storage

Packs and power

Pumps, fans and inverters with a tight mass budget.

03Engage now

Four ways to start this month.

AFixed scope
2 to 3 weeks

Drive-unit NVH diagnostic

A focused look at why your drive whines, buzzes or radiates. You pick two or more workstreams.

You getRanked fixes with expected dB and mass cost, plus a one-page target cascade.

  • Gear whine across the full torque range
  • Housing radiation and bearing-seat stiffness
  • Mounts under drive and regen torque
  • Dyno and test plan review
BScoped to
your program

Target-cascade work

Take a product-level noise or vehicle target down to the component limits that control it.

You getA cascade your team can defend in a design review.

  • Vehicle target to subsystem to part
  • Bearing force, transmission error and mount stiffness limits
  • Each limit tagged measured or claimed
  • Every trade priced in mass and dollars
CBody, frame
or panel

Static stiffness from a tap test

Torsional and bending stiffness from an impact test with a hammer and accelerometers. No static load rig.

You getA stiffness number you can defend, from a test you can run in a day.

  • Torsional and bending stiffness, each with its interval
  • First global modes from the same test
  • A comparison against your stiffness target
  • The mass cost of closing any gap
DConsumer products
and vehicles

Sound quality for consumer positioning

Start from your brand positioning. Find what good sounds like to your buyer, then engineer toward it. The click, the thud or the hum that tells a customer your product is worth its price.

You getA product that sounds like its price, with sound targets in engineering units.

  • Jury listening studies on your product and its rivals
  • Psychoacoustic metrics: loudness, sharpness, roughness, fluctuation strength, tonality
  • A composite metric that predicts how buyers rate the sound
  • Sound targets your engineers can design to
  • A sound brief that ties your positioning to those targets

All four are powered by DecisionHelm. You own every output. Scope and price follow a short first call. The stiffness method builds on my PhD research and on papers published at IMAC in 2014 and 2015.

04Field notes: gear whine

Why your drive whines, and where to fix it.

Gear whine is a steady tone at the gear mesh frequency and its harmonics. In an electric drive it sits between one and four kilohertz, where the ear is most sensitive. No engine noise covers it.

Source

The mesh

Transmission error, mesh stiffness variation and axial force shuttling drive it. Friction and trapped air and oil add to it.

Path

Shafts and bearings

They turn mesh error into bearing force. That force is what the housing feels.

Receiver

Housing, mounts, body

Panels radiate the tone. Mounts carry it into the cabin.

Synthetic map of motor speed against frequency. The gear mesh order and motor order rise with speed. Where the mesh order crosses a housing mode near 2.2 kHz, the tone peaks.
Fig. 1Find the crossing. A whine peak is a mesh order meeting a structural mode, so the fix can live at the gear or in the housing. Synthetic data, for illustration.
Synthetic chart of transmission error against motor torque. Without relief, error rises with torque. With tip relief, the curve dips to a quiet point near 30 percent torque, where a duty-cycle histogram below shows the most time is spent.
Fig. 2Place the quiet point. Tip relief puts minimum transmission error at the torque your customers spend the most time at. Synthetic data, for illustration.

Six lessons from the test cell

  1. Watch force as well as error. Mesh force is mesh stiffness times transmission error. A stiff mesh with less error can still be the loudest.
  2. Tune the flank for the torque your customers use. Tip relief and crowning put the quiet point at one chosen torque. Pick it from real duty-cycle data.
  3. Check the gear blank before blaming the flank. A thin web can have a mode inside the mesh band. No flank change fixes that. The web shape does.
  4. Pair the source fix with a path fix. Micro-geometry gains shift with torque. An enclosure's insertion loss holds at every torque. In decibels the two add.
  5. Judge the tone by how it stands out. Listeners hear a tone well above its neighbouring noise. Set targets in tone-to-noise ratio and prominence ratio.
  6. Derive the error limit from the cascade. A rule-of-thumb micron value is a starting point. The real limit comes down from the cabin target, through the housing and bearings, to the mesh.

Grounded in public literature: Welbourn's gear noise survey (1979), the Ohio State Gear Dynamics and Gear Noise Research Laboratory, the University of Cincinnati vehicle NVH course, and the ECMA-74 tonality metrics.

Have a whine you can't kill? Send me a waterfall plot →

05Field notes: static stiffness

Static stiffness, measured without a static rig.

Torsional stiffness is the torque it takes to twist a body or frame through one degree. The usual test clamps the structure in a rig and loads it. A free-free tap test reaches the same number from vibration data, with far less setup.

Step 1

Hang it free

Support the structure on soft springs, clear of the ground. No fixtures and no load rig.

Step 2

Tap and measure

Excite it with an impact hammer. Read the response with accelerometers at the load points.

Step 3

Filter and read

Combine the responses to isolate twist or bending. The low-frequency level gives the stiffness.

Hasan Pasha at a laptop test station in a vehicle test lab, cables and data acquisition hardware on the bench.
PhotoAt the test station during a body test for my PhD research.
Hasan Pasha with Dr. David L. Brown in a vehicle test lab, a vehicle body behind them.
PhotoWith Dr. David L. Brown in the test lab. He served on my PhD committee.
Synthetic log-log plot of compliance against frequency. The raw response is dominated by rigid-body motion at low frequency. The filtered twist response flattens to the static compliance, one over stiffness, below the first torsion mode.
Fig. 3The static answer sits below the first mode. The raw response is swamped by rigid-body motion. Filtered for twist, it flattens to the static compliance, and stiffness is its inverse. Synthetic data, for illustration.

Five lessons from the lab

  1. Geometry does the filtering. Responses at symmetric points, combined, cancel rigid-body motion and leave the twist or bend you want.
  2. Averaging steadies the number. Many response points and repeat taps average out noise and local effects.
  3. A known added mass helps. Testing with and without a known mass gives a second view of the same structure.
  4. One test, two answers. The first global torsion and bending modes come from the same data, so stiffness and modal targets get checked together.
  5. Price the gap in kilograms. Closing a stiffness gap adds mass. Know the cost before choosing where to add it.

Grounded in public work: my PhD dissertation, Estimation of Static Stiffnesses from Free Boundary Dynamic (FRF) Measurements (University of Cincinnati, 2014), and my IMAC papers from 2014 and 2015.

Need a stiffness number before the rig is free? Let's talk →

06Field notes: sound quality

Your product's sound is a price tag.

Buyers judge quality in a fraction of a second: the click of a cap, the thud of a door, the hum of a drive. Marketers tend to treat sound as a jingle. Engineers treat it as a decibel limit. The value sits between the two, and few teams work there.

Hasan Pasha standing inside an anechoic chamber lined with sound-absorbing wedges, microphones on booms around a test stand. The test hardware is blurred.
PhotoInside an anechoic chamber, where every reflection is engineered away and what's left is the signal itself. Test hardware blurred.
Engineering value

What it does

Stiffness, sealing, precision and power, built into the product.

Perceived value

What the buyer hears

A crisp click, a damped thud, a quiet hum, in the first second of use.

Positioning

What it says

Premium, safe, sporty or reliable. The sound has to match the price and the brand.

target zone Rival A Rival B Your product dullcrisp sound character (sharpness) cheappremium jury rating
Fig. 4Position the sound the way you position the brand. Map your product and its rivals on what buyers hear and how they rate it. Then set engineering targets that move you into the zone your price demands. Diagram, not data.

Six lessons from engineering and the business school

  1. Marketing translates engineering into perception. A product can be built well and still sound cheap. That gap loses value in translation.
  2. Context flips the meaning. The same gear tone reads as power at a hard launch and as cheap in city traffic. Every sound target needs its use case.
  3. Segment by ear. Music to one buyer is noise to another. Jury studies can group buyers by sound preference, the way marketers group them by needs.
  4. Brands can own a sound. Payment brands engineer checkout chimes to signal a safe transaction. A door thud or a cap click can carry a brand the same way.
  5. Sound can earn price. Sensory marketing research links sensory cues to perceived quality and willingness to pay.
  6. Make it a spec. A composite metric turns jury ratings into targets in engineering units, so design, marketing and suppliers aim at one number.

Grounded in public literature: Zwicker and Fastl, Psychoacoustics: Facts and Models; Lyon, Product Sound Quality; Aradhna Krishna's work on sensory marketing; and my Wharton coursework in consumer behavior and brand management.

Does your product sound like its price? Let's find out →

07The cascade

Four vehicle targets. One currency.

Every program trades between the same four targets. DecisionHelm prices each trade in mass and dollars, so the chief engineer and the part owner read the same number.

Gear whine is a quietness problem first. That is why the drivetrain reports to Refinement.

08How the numbers stay honest

Rigor you can audit.

R1

Measured and claimed look different

Supplier and simulation numbers stay claimed until a test establishes them.

R2

Every number shows its source

Each value carries its unit, its origin and its interval.

R3

Decisions at the right level

A trade goes to the person who owns both sides of it, with the cost attached.

09Founder story

The unwanted signal.

Hasan Pasha, founder. How a rattling laptop led to engineering what customers feel.

In 2007 my laptop rattled. A curved plastic lip over the speakers buzzed on certain songs. Most people would never notice. I had spent two years learning to read resonance as data, and I could not unhear it. One question stayed with me: had anyone who built it asked a customer what good should sound like?

For years I treated sound as a quantity to drive down. Then a film score gave me goosebumps. A documentary showed a fungus-treated violin beating a Stradivarius in a blind test. Good sound turned out to be physics, and physics can be decoded.

In sponsored research at Cincinnati for P&G, working with psychologists and a consumer panel, I grouped psychoacoustic metrics into three dimensions: loudness, product feedback and modulation. The composite metric predicted how jury listeners rated the products.

Electric drives made the lesson urgent. With no engine to mask it, the cabin becomes an acoustic magnifying glass. A faint whine at city speeds reads as cheap.

So I built DecisionHelm to engineer what customers feel, early, and to price every fix in mass and dollars. To the customer, perception is the product.

The full story: The Unwanted Signal, on LinkedIn →

Track record

2024 to nowDecisionHelm and CLVHelmFounder.
2021 to 2024AppleLead product design engineer, NVH and systems architecture. Set target-cascading methods for electromechanical systems.
2020 to 2021Turntide TechnologiesPrincipal engineer, NVH and multiphysics. Led the fix for a field motor noise issue.
2016 to 2020Lucid MotorsPowertrain and vehicle NVH specialist. Built the powertrain and drivetrain NVH function from the ground up.
2015 to 2016Achates PowerSenior engineer, dynamics and NVH for advanced powertrains.
2012 to 2015University of CincinnatiPostdoctoral research fellow on three industry-sponsored programs.
  • BMW, Structural Dynamics Research Lab: static stiffness from dynamic tests on car bodies. The methods were adopted by BMW North America.
  • P&G, UC Simulation Center: objective sound quality metrics for consumer products, with engineers, psychologists and a consumer panel.
  • GE Aviation, Aeromechanics Lab: a method to evaluate damping coatings on engine fan blades.
2010 to 2014UC Structural Dynamics Research LabPhD in structural dynamics under Dr. Randall J. Allemang and Dr. David L. Brown, with several papers co-authored with both.
2006 to 2008IIT MadrasM.Tech in mechanical engineering, acoustics. Built a solver for sound fields in partially open cavities.
2001 to 2010Boeing, through Tata Consultancy ServicesConsultant on the Boeing 747-8 program: design automation and structural optimization for fuselage components. Earlier, software delivery for enterprise clients.

Experience is listed as background. DecisionHelm's methods come from public literature and my own published research. No employer's work, data or code is in the product.

10Contact

Tell me what your drive does under load.

hasan@decisionhelm.com

A short first call is enough to see which of the four fits your program. Or find me on LinkedIn.