The Motorola DSP56362 in the Korg CX-3

by Robert Weatherford, Parastream Technologies

Last updated: October 3, 2023

We put a lot of time into creating, gathering, and organizing the information in this page. If you found any of this to be useful to you, especially if it helped you to make or save money, please  to help us continue making content like this available.

Introduction

Both generations of the Korg CX-3 were well designed products for the time period. Through no fault of Korg’s, each generation was plagued with chronic chip failures. The first generation production units suffered from field failures from all 3 Siemens tone generator chips. Some of the failures were spectacular, causing the 10Ω power supply resistors to act like fuses. Later runs of the Siemens chips (SM304A, SM305A3, SM305B) seem to be stable decades later. The second generation production units had field failures of the DSP56362 which would cause the organ to not produce any sound at all.

This article discusses the second generation problems with the DSP56362. I hope it gives the reader a better understanding of the issue and how to deal with it.

I tried in vain in 2023 to get the inside scoop from NXP, who inherited the DSP56000 series from Freescale who inherited it from Motorola. I was told by NXP that the team handling that line was disbanded over 10 years prior and that there was no meaningful archive of the project. I have gathered what is available on-line and in archives such as archive.com, but there is still a lot of missing information. I have had to resort to my experience and best guesses to fill in the gaps. It is also highly unlikely that NXP or anyone else will pick up the product line in the future.

What the DSP56362 Does

From a high-level, the Korg CX-3 has general purpose microcontrollers, a tone generation DSP, an effects DSP, and various analog and digital inputs and outputs. A pair of DSP56362 implements the tone generation DSP. One of them acts like the tone generator of a Hammond, while the other acts like the keyboard switching matrix and mixing transformers. It also implements the percussion envelope.

The two DSP56362 may not fail at exactly the same time, and they sometimes fail slowly. This is why some organs begin to sound or behave strangely for a while and then go silent. If you are attempting to perform a repair yourself, it is been my experience that both chips should be replaced. If one has failed, the other is not far behind.

Failure Modes

I have a story from a customer who purchased a CX-3 in 2000 and kept it unopened until 2023. They unpacked it, set it up, and discovered that it didn’t make any sound and had the infamous DSP failure diagnostic message. I doubt very seriously that it could have left the Korg factory this way, so the DSP chips must have failed just sitting in a box unpowered for over 2 decades. This leads me to believe that the failures are likely process-oriented and not necessarily a design flaw.

I ran into an example of a chip design failure in the 2010s on a contract with The Weather Channel. The audio DSP they were using was made by Texas Instruments and would fail after a predetermined number of power-on hours. It turned out to be a problem with the clock driver transistors, where every clock cycle would damage the transistors a little. TI's workaround was to gate the clock so it is on only when the DSP was being used. While this might be a good solution for a consumer product, it doesn’t work for a system that is on 24/7. I don’ think that is what we’re dealing with in the DSP56362.

Deciphering the Chip Markings

There are 4 chip markings from top to bottom:

1. The company logo. The first-run parts had the familiar Motorola “bat wing” logo, while later parts had the Freescale logo.
2. The part number.
3. The mask revision.
4. The date code. The last 4 digits seem to be the year and week of manufacture. The preceding characters probably indicate the manufacturing plant location.

Dissecting the part number using XCB56362PV100 as an example:

• The first part is either XC or DSP. The XC prefix was used originally, and was switched to DSP somewhere during the 1J21D mask revision run.
• The B indicates a generic programmable part. Other letter codes indicate ROMs for specific applications programmed by the factory.
• 56362 indicates the actual part number in the DSP 56000 series.
• PV indicates a tin/lead plating on the leads, AG indicates matte tin (RoHS).
• 100 indicates a 100MHz part, later parts were 120MHz.

There appears to be several documented and observed mask versions of the DSP56362, including the later ones from Freescale and NXP:

• 0H76G – Motorola chip errata available.
• 0J21D – Motorola chip errata available.
• 1J21D – Freescale chip errata available. This is the one I have seen in non-functional organs.
• 2J21D – Freescale chip errata available.
• 3J21D – The NXP version DSPB56362AG120. There is no information on this mask revision anywhere. These chips had the Freescale logo.

Buying Guide

When NXP decided to EOL (End Of Life) the DSPB56362, the factory stock was quickly gobbled up, largely by what I call the “EOL market.” It took several months, well into 2023, before the availability of the parts began to emerge. After conversations with a couple of these new suppliers, we came to the conclusion that there must be continuing production with these chips. Here’s why. The price of the late date code RoHS chips (like the ones I used to buy for organ repairs) skyrocketed from the $20 to $30 range to the hundreds. However the PV (leaded) parts are still relatively cheap. These parts cannot be used in current production because they are not RoHS compliant.

So here’s the bottom line. If you can afford to spend several hundred dollars for repair chips, by all means go with the AG parts. Otherwise, it gets tricky. Let’s start with the part numbers I do not recommend. Based on the failure log, I would avoid the 1J21D mask regardless of the date code. As the failure log grows, I may update this. I would certainly stay away from the date codes in the failure log.

I would recommend the latest mask revision and date code you can find at a reasonable cost. While 100MHz parts will work fine in the CX-3, I use the 120MHz parts as they are usually indicative of a later date code and perhaps better processing. PV vs AG is not an issue, as the entire CX-3 run used the PV parts. PV is easier for rework anyway.

From 2020 through 2022, I was using the 3J21D mask parts with 2019-ish date codes. These parts were commercially available from suppliers such as Digi-Key and Mouser. Moving forward, I have secured a lot of 2J21D mask parts from 2006. This is the best that can be done at a suitable price point. We are betting on the 2J21D mask and the 120MHz speed rating.

Failure Log

Many of the organ repairs have been logged here, recording the bad chip’s part number, mask revision, and date code. As repairs are done, this table will be updated.