By "garbled audio", do you mean the RX audio? I found this information that might be helpful:
Problem: RX audio distortion. {This problem seems to be more prevalent in early production radios. There may have been a recent factory component change to reduce receive distortion.} There is one designed-in source of distortion and several other possible sources of distortion. The designed-in source of distortion exists because not enough forward bias current is applied to the switch diodes that select the SSB [D23] , AM [D24] , and FM [D25] audio detectors. Here's why: A mixer is a nonlinear device. Nonlinearity and distortion go hand in hand. Diodes make good mixers when their forward current is in the range of 0.05mA to about 0.6mA. At currents above 1.5mA, diodes are reasonably linear and they make good switches. The switch diodes in the TS-440S have enabling currents of from 0.2mA for FM to 0.28mA for SSB, so the switch diodes are operating in the region of maximum nonlinearity, which results in distortion. The fix is simple: Increase the forward bias DC-current through the diodes to roughly 2mA. This is accomplished by decreasing the resistance of each DC-bias resistor to about 1k Ohm. When these resistors are decreased in value, the 5000 Ohm impedance of the low-level audio circuits decreases to about 500 Ohms and the values of the coupling capacitors must be increased accordingly to prevent a rolloff of the low frequencies. Similarly, the resistors that are in series with the audio signals in these circuits must be decreased in value to offset signal attenuation. The capacitance of the filter capacitors between switch diode bias resistors (for SSB: C52, 4.7µF) must be increased to compensate for the decreased value of the bias resistors.
Fix for SSB audio detector and D23 switch: In the left, rear corner of the IF Unit, change: R71, R73, R74, and R85 to 1k Ohm; C53 to 47µF, 10V [XL=68 Ohms at 50Hz]; C52 to 22µF, 16V.
I'm guessing that similar changes can be made in the AM and FM detectors. This project can become tricky since some of the needed changes can also affect the transmit÷receive transition performance of the radio. Thus, it may be necessary to compromise by lowering the switch diode bias current to about 1mA. In general, this can be done by using 2k Ohm, instead of 1k Ohm, bias resistors.
If, after the changes are incorporated, you can still hear RX distortion on SSB, the problem may lie at the
[right-adjacent} product detector. Possible FIX: Install the missing injection-oscillator terminating-resistor at the product detector. This resistor is 62 Ohm, 1/8W or 1/4W. It is soldered under the IF Unit PC board, near L5. The terminating-resistor is soldered to the junction of R69 and R70 and the ground foil at the edge of the PC board. If the distortion persists, you may have an unmatched set of Germanium product-detector diodes: D19, D20, D21, and D22. These diodes can be replaced with Schottky diodes. The product detector balance should then be checked and adjusted as follows: Connect an RF detector to the emitter of Q9/R77. R77 sticks up from the board at one end so that the test connection can be easily made. The RF detector can be an oscilloscope, detector/probe for a DMM, or a 455KHz receiver coupled through a 10pF capacitor. With no signal input to the ANT jack, turn the RF Gain on the 440 to zero. Set mode to USB, IF Shift to detent. Adjust TC1 [also installed backwards, like TC2] for minimum RF. End of test. Reset the RF Gain control to normal.