diff --git a/readme.md b/readme.md
index bcb8080..5c3c724 100644
--- a/readme.md
+++ b/readme.md
@@ -1,11 +1,10 @@
 # `adis1647x`
 
-no_std driver for the Analog Devices [adis1647](https://www.mouser.com/new/analog-devices/adi-adis1647-imus/) family of IMUs. Currently only the adis16475 is implemented.
-
+no_std driver for the Analog Devices [adis1647](https://www.mouser.com/new/analog-devices/adi-adis1647-imus/) family of IMUs. Currently only the adis16475 is supported
 
 ### Usage
 
-You need an embedded_hal_async spi driver
+You need an embedded_hal_async spi driver and delay implementation
 
 ```rust
 
@@ -14,8 +13,9 @@ use adis1647x::delta::*;
 
 async fn main() -> Result<(), SomeSpiError> {
     let bus = some_spi();
+    let delay = some_delay();
     
-    let imu = ADIS1647X::new(bus).await?;
+    let imu = ADIS1647X::new(bus, delay).await?;
     
     let mut angle: DeltaAngle = Default::default();
     
@@ -28,4 +28,5 @@ async fn main() -> Result<(), SomeSpiError> {
         
         println!("yaw: {} degrees", angle.angles().y);
     }
+}
 ```
diff --git a/src/lib.rs b/src/lib.rs
index c3bd3cb..dcaa958 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -1,6 +1,6 @@
 #![no_std]
 use delta::{DeltaAngle, DeltaVelocity};
-use embedded_hal_async::spi::SpiDevice;
+use embedded_hal_async::{delay::DelayNs, spi::SpiDevice};
 use nalgebra::Vector3;
 use registers::*;
 
@@ -45,38 +45,44 @@ impl Model {
     }
 }
 
-pub struct ADIS1647X<T> where T: SpiDevice {
+/// microseconds between requests
+const STALL_TIME: u32 = 16;
+
+pub struct ADIS1647X<T, D> where T: SpiDevice, D: DelayNs {
     bus: T,
-    model: Model
+    model: Model,
+    delay: D,
 }
 
-impl<T: SpiDevice> ADIS1647X<T> {
-    pub async fn new(bus: T) -> Result<Self, T::Error> {
+impl<T: SpiDevice, D: DelayNs> ADIS1647X<T, D> {
+    pub async fn new(bus: T, delay: D) -> Result<Self, T::Error> {
         //let bus = BlockingAsync::new(bus);
 
-        let mut query_imu = Self { bus, model: Model::ADIS16475_1 };
+        let mut query_imu = Self { bus, model: Model::ADIS16475_1, delay };
         let range = query_imu.read_u16(RANG_MDL).await?;
 
         let model = Model::detect(range);
         
-        Ok(Self::with_model(query_imu.take_bus(), model))
-    }
-
-    fn take_bus(self) -> T {
-        self.bus
+        Ok(Self::with_model(query_imu.bus, model, query_imu.delay))
     }
 
     /// use a known model instead of detecting it
-    pub fn with_model(bus: T, model: Model) -> Self {
+    pub fn with_model(bus: T, model: Model, delay: D) -> Self {
         //let bus = BlockingAsync::new(bus);
-        Self { bus, model }
+        Self { bus, model, delay }
+    }
+
+    async fn stall(&mut self) {
+        self.delay.delay_us(STALL_TIME).await;
     }
 
     pub async fn read_u16(&mut self, addr: u8) -> Result<u16, T::Error> {
         let request = [addr & 0x7F, 0];
         self.bus.write(&request).await?;
+        self.stall().await;
         let mut dat = [0;2];
         self.bus.read(&mut dat).await?;
+        self.stall().await;
 
         Ok(u16::from_be_bytes(dat))
     }
@@ -84,6 +90,7 @@ impl<T: SpiDevice> ADIS1647X<T> {
     pub async fn write_u8(&mut self, addr: u8, data: u8) -> Result<(), T::Error> {
         let request = [addr | 0x80, data];
         self.bus.write(&request).await?;
+        self.stall().await;
 
         Ok(())
     }
@@ -108,6 +115,7 @@ impl<T: SpiDevice> ADIS1647X<T> {
 
         // first address
         self.bus.write(&read(addr[0])).await?;
+        self.stall().await;
 
         let mut responses = [0; N];
 
@@ -119,6 +127,7 @@ impl<T: SpiDevice> ADIS1647X<T> {
         for i in 1..(addr.len()*2) {
             let addr = addr[i/2] + (i as u8 % 2) * 2;
             self.bus.transfer(&mut buf, &read(addr)).await?;
+            self.stall().await;
 
             partial[partial_idx] = i16::from_be_bytes(buf);
             partial_idx += 1;
@@ -132,6 +141,7 @@ impl<T: SpiDevice> ADIS1647X<T> {
 
         // last address
         self.bus.read(&mut buf).await.unwrap();
+        self.stall().await;
         responses[addr.len() - 1] = partial[0] as i32 + (i16::from_be_bytes(buf) as i32) << 16;
 
         Ok(responses)