This article explores a scenario where audio quality issues (static, robotic sound) in an Amazon Connect contact center are misattributed to the Connect service, while the actual root cause lies in the Salesforce CTI Adapter integration layer
Question
How do I troubleshoot audio quality issues (static, robotic sound) in Amazon Connect when using the Salesforce CTI Adapter, especially when the issue doesn't appear in call recordings?
Short Answer
If agents experience static or robotic audio only when using the CCP embedded in Salesforce (but not when using the default CCP directly), the issue is likely in the Salesforce integration layer rather than Amazon Connect itself. Collect browser console logs and HAR files in addition to standard CCP logs to identify the root cause.
Detailed Resolution
Step 1: Isolate the Problem Layer
- Have the affected agent open the default CCP directly (not through Salesforce)
- Have the agent take several calls using the default CCP
- If the audio quality issue disappears when using the default CCP, the problem is in the Salesforce integration layer
Key indicator: If call recordings sound normal but the agent hears static/robotic audio in real time, the issue is on the agent's audio output path — not the Connect media pipeline. Call recordings capture audio at the Connect service level, before it reaches the agent's browser.
Step 2: Collect the Right Logs
For Salesforce-embedded CCP issues, standard CCP logs and WebRTC dumps may appear clean because the Connect layer is functioning correctly. You need integration-layer data:
| Log Type | How to Capture | What It Shows |
|---|
| CCP Logs | CCP Settings gear → Download Logs | Connect-side events, session state |
| WebRTC Internals | chrome://webrtc-internals/ → Create Dump | ICE candidates, codec negotiation, packet stats |
| Browser Console Logs | Ctrl+Shift+J (Windows) or Option+⌘+J (Mac) → Console | JavaScript errors, CTI adapter failures, iframe communication |
| HAR File | F12 → Network → Export HAR | HTTP requests/responses, failed API calls, timing |
Important: Have agents open DevTools before taking calls and enable "Preserve log" in the Console to retain logs across page navigations.
Step 3: Check Common Salesforce Integration Issues
- CTI Adapter version: Ensure you're running the latest version. Older versions may have known audio handling bugs.
- Duplicate CCP instances: Verify that only one CCP iframe is loading. Multiple instances can compete for the audio device, causing intermittent static.
- Cross-origin configuration: The Salesforce domain must be properly allowlisted in your Connect instance's approved origins.
- Chrome extensions: Ad blockers, security extensions, or corporate browser policies can interfere with WebRTC media streams. Test with extensions disabled.
- Audio device contention: Salesforce Lightning components or other browser tabs may compete for microphone/speaker access.
Step 4: Review Network Configuration
Even if the default CCP works, the embedded CCP may route traffic differently:
- Zscaler/proxy: Ensure Connect domains (
*.transport.connect.<region>.amazonaws.com) bypass SSL inspection. WebRTC media, encrypted via DTLS-SRTP, can break if intercepted.
- DNS resolution: Verify the embedded CCP resolves Connect TURN NLB endpoints correctly from within the Salesforce iframe context.
Step 5: Evaluate CloudWatch Metrics
Check the ToInstancePacketLossRate metric in CloudWatch:
- Namespace: AWS/Connect
- Dimensions: Instance ID, Participant=Agent, Stream Type=Voice, Type of Connection=WebRTC
- Statistic: Maximum
Values above 0.2 indicate connectivity instability. However, if this metric shows spikes but CCP logs are clean, the packet loss may be occurring at the browser/iframe level — reinforcing the integration-layer hypothesis.
Next Steps: If the Issue Persists
If the above steps do not resolve the issue, open a support case with the following information:
- CCP logs from the affected calls
- HAR file captured during the issue
- Browser console logs with timestamps
- WebRTC internals dump
- Confirmation of whether the issue reproduces on the default CCP
- Agent's network configuration (proxy/VPN details)
- CTI Adapter version and Salesforce environment (Classic vs. Lightning)
This information enables the support team to pinpoint whether the issue lies in the Connect media path, the CTI adapter, or the network layer between them.
Resources