Building New Voice BoardsBy David Clarke [21030085++] <firstname.lastname@example.org>
Looking at the picture I spotted something that looked interesting. Specifically I saw a different type of key damper, and I surmised that this might be a Chroma Pressure Sensor. I figured that even if the rest of the keyboard was ruined it might be worth getting - if only to gain some insight into the pressure interface.
After having received the Chroma I found the keyboard did have a pressure interface and I was able to jot down a few notes on it (see Chroma Pressure Sensor & Interface).
The keyboard itself ended up being much more complete than the description of 'carcass' would tend to imply.
As the Chroma wasn't terminally ill I couldn't bring myself to just use it as a parts machine. The primary stumbling block to restoring it to full functional status was to replace the missing 7 dual channel voice boards.
Back in January 2002 there was some discussion on the Chroma List regarding the repair of dual channel voice boards. At that time, I made a remark akin to "...if worse comes to worse, a board could be created from scratch..." (see thread Voice Board Repairability).
I decided to put my money where my mouth is and attempt make a set of replacement boards.
The following article provides the details behind the Dual Channel Voice Board Replacement Project.
This project had as its goals:
- To generate a complete parts list, identifying all components on the board (including values of otherwise unlisted components);
- To identify a source for all components on a dual channel voice board, including Web addresses (if appropriate) and cost (to assist with a new build or with repair of existing boards);
- To add a trace to the circuit board so that no wire straps would be required;
- To track the full cost to build a board from scratch;
- To validate the schematic in the service manual;
- To show that it could actually be done (i.e., that a complete replacement board could still be created and built with components available today);
- To generate a circuit board that a Chroma owner could use should a board need to be replaced (due to physical damage, etc.);
- Generate a schematic which 100% represented the component values installed on the board; and,
- To build a set of voice cards that could be used in the 'Chroma Carcass' to make it operational once again.
Details of the effort are contained in the following sections.
Original/Existing Reference Material
When I reviewed the technical material available to reproduce the board, I noticed that it was neither complete nor necessarily consistent.
The service manual provides a 'parts list'; however, it does not cover all parts on the boards. Specifically, it omits most general components, such as resistors and capacitors. From the point of view of a service manual this is understandable; however, to recreate the board, all components need to be accounted for.
There are two set of schematics for the dual channel voice boards, one from the "preliminary service manual" and one from the more common Rhodes service manual. The newer service manual has diagrams showing component placement on the boards. These diagrams also show component values. In some cases, component values differ between the old service manual schematics, the newer service manual schematics, and the newer service manual placement diagrams.
In addition, there were service notices which made further changes to the values intended for the boards (e.g., Field Change Notice FCN2-001).
Even if the material had been fully complete and consistent, the presentation presents issues in and of itself. First, from a usability point of view, it is difficult to locate a specific part (i.e., where on the schematic would you find resistor "R81"). Further, from the point of view of generating a new board, the paper representations of the circuit traces found in the service manual are of little or no use.
Because of these conditions, it was decided that the best way to proceed was to actually start at the beginning and redo the original circuit layout.
The New Build
It was decided to recreate the schematic with all part reference designators and values. The schematic would then be used to create a 'net-list' to identify the electrical connectivity between the individual parts. The net-list would then be used to generate the new Printed Circuit Board/Printed Wiring Board (PCB/PWB).
While there are many schematic capture programs available, I decided to use the one that I was most familiar with (and which was already available to me.) This was OrCAD's Schematic Design Tools (SDT).
While quite old (1991), I knew this program would be able to support the schematic entry as well as the eventual generation of both a Bill of Materials (BOM) and a net-list.
The generated schematic has nine main sheets, as follows:
- The top level;
- One sheet each for VCO A and B;
- One sheet each for VCF A and B;
- One sheet each for VCA A and B;
- One sheet for the Analog Sample and Hold circuitry; and,
- One sheet for the data latches.
The following voce_sch.zip file contains the OrCAD schematic and library files, respectively. In addition, the schematics are available in PDF format.
The searchable PDF will hopefully allow owners to locate parts faster (and better) than on the original paper schematics.
The net-list is a representation of the connectivity between individual parts or "nets." Without identifying the exact parts, this list determines what gets connected to what.
OrCAD SDT can generate net-lists in several different industry standard formats. The one available here is specifically in a format that is recognized by the OrCAD PCB II circuit board layout program.
- dualvoce.net (plain text file)
OrCAD PCB was used to layout the net-list.
Since the board was being totally recreated, components could have been placed in different locations (in an effort to 'improve' upon the layout); however, to ensure that the new boards behaved like the old ones (as well as to help with future serviceability and ensure that existing documentation was still relevant), it was decided to keep the same general physical board orientation and signal trace positioning (without becoming obsessive) as the production Rhodes dual channel voice board.
One of the design goals was to build the board without wire straps. Since all original voice boards had at least one strap (see Voice Board Revisions), some deviation from the original layout was required.
In addition, while going through the layout information from the original board, it was found that a 'ground loop' existed in the sample and hold portion of the circuitry. This (generally undesired) situation was not reproduced on the replacement board (i.e., all ground and power rails were routed to ensure only one path back to their source.)
During schematic creation I did not make provision for the two test points found on the board. To accommodate these in the board after-the-fact, two pads were manually placed and routed to to provide the original test/setup functionality. In the same vein, some manual pad/hole size adjustments were made to ensure both a good mechanical component fit and the appropriate amount of clearance to other tracks/parts.
In an effort to keep costs down, no silk-screen information was placed on the board.
OrCAD PCB was then used to produce Gerber plotter files for both sides of the board, as well as drill/aperture information. (Gerber Format is an industry standard photo plotting language commonly used to describe printed circuit boards.)
Most board development houses can produce a PCB directly from Gerber files.
For low-run prototyping, I use APCircuits in Alberta, Canada. This company will accept files electronically, provide automatic status reports on the progress of the job and generally have a finished product within a day.
To completely specify a circuit board to APCircuits, 6 files are required:
- The Gerber Top Layer File;
- The Gerber Bottom Layer File;
- The Gerber Aperture Table File;
- An NC Drill Name File;
- An NC Drill Size File; and,
- A Text-Based order form.
voce_ord.zip contains the exact files I used to produce the board (I removed my address/payment information from the order form - but other than that, this package would be ready for submission to APCircuits.)
Gerber files can also be viewed directly via one of the many generally available programs (shareware and otherwise). One such program is GC-PREVUE (http://www.graphicode.com/).
The following PDF file is a graphical representations of the front and the back of the PCB.
People who have OrCAD PCB will be able to use the attached dualvoce.brd file directly, and can search for specific reference designations (helping to easily locate items on the board.)
Bill of Materials
Once the raw PCB is available, it needs to be 'stuffed' with components.
The first step to this to the get a complete list of all the parts needed and the quantities required.
OrCAD SDT will automatically generate a Bill of Materials to outline what is called for for a given project. dualvoce.bom is a text-file summarizing the materials for the dual channel voice card. It's not only relevant for the board I've created but also for standard/stock Rhodes voice card as I maintained the same reference designators.
- dualvoce.bom (plain text file)
Sourcing of Parts
It is one thing to have a list of the parts required - it is quite another thing to actually find and purchase them.
I wanted to attempt to purchase all the parts on-line so that I could generate a resource list that others could use to find and purchase replacement parts for their dual channel voice cards.
The more common parts were sourced through Jameco, as they tend to have a good selection of older, standard parts at a very reasonable price. Other places carry these parts too (such as Digikey or Future-Active Electronics); however, they're often more expensive.
Digikey is a good supplier of newer parts, and they have very quick turnaround for shipping.
Very common components - such as standard resistors - can sometimes be tricky. Most larger companies only want to sell large quantities (1000's), while most smaller companies will sell smaller quantities, but at larger prices. A happy medium was found with RP Electronics. They sell standard resistors in packages of 200 for approx $1.
Most components on the board are still readily available. As was noted in the Chroma Notes post, the hardest components to replace are the two CEM chips (CEM3350, CEM3360) and the temperature compensating resistors (KRLP tempco resistors).
As I had had a good experience with Paul in the past, all the CEM's for the project were purchased through him. Looking again now (2005), CEMs are a bit harder to get. Paul lists the CEM3350 as available - but not the 3360. The SynthFool list [gone as of 2009.07] and Wine Country still show the CEM3360. The parts also occasionally appear on eBay.
The KRLP tempco resistors are a bit of another story. There are some people who still have these available, but these are generally expensive (approx $10US each), and possibly not available in large quantities.
Considering that 4 resistors are needed per voice board, paying $10 each did not seem to be an ideal situation. Instead, I contacted the original component manufacturer (KRL/Bantry Components) to see if they could provide a replacement.
The "P" in KRLP stands for "plastic" and reflects the type of packaging of the resistor itself. KRL indicated that they could do a production run of KRLC ("C" standing for conformally coated) if at least 50 pieces were ordered.
Since 30+ parts were going to be needed, this seemed like a much better way to go, especially considering that the price quoted was $3 per item.
The following table outlines the components used, where they can be purchased, and the associated vendor's part number for the item.
|Quantity||Reference||BOM Part Designation||Vendor||Vendor P/N|
|4||C1, C2, C3, C4||0.001UF 10%||Digikey||P3497-ND|
|4||C5, C6, C7, C8||100PF 20%||Digikey||P4116-ND|
|4||C9, C10, C11, C12||0.022UF 10%||Digikey||EF2223-ND|
|5||C13, C14, C16, C17, C18||1UF 35V||Jameco||29831|
|5||C20, C21, C22, C23, C40||0.01UF 20%||Jameco||15229|
|8||C24, C25, C26, C27, C28, C29, C30, C31||0.033UF 10%||Digikey||EF2333-ND|
|8||C32, C33, C34, C35, C36, C37, C38, C39||0.0068UF 10%||Digikey||P4795-ND|
|2||C41, C42||150PF 20%||Digikey||P10826-ND|
|2||C43, C44||0.1UF 10%||Jameco||33486|
|4||P5, P6, P7, P8||CON8||Digikey||WM3006-ND|
|5||Q1, Q4, Q5, Q8, Q9||3904||Jameco||38359|
|4||Q2, Q3, Q6, Q7||3906||Jameco||38375|
|1||R1||10K POT 1/5W 30%||Digikey||X262R103B-ND|
|1||R2||5K POT 1/5W 30%||Digikey||X262R502B-ND|
|4||R3, R4, R5, R6||1.87K 1W 3% 3300PPM/C||KRL||KRLC-1C 1.87K|
|2||R7, R8||43.2K 1%||Digikey||43.2KXBK-ND|
|12||R9, R10, R29, R30, R31, R32, R35, R36, R39, R40, R81, R82||100K||RPE||CF-1/4-100K|
|8||R11, R12, R61, R62, R63, R64, R77, R78||22K||RPE||CF-1/4-22K|
|8||R17, R18, R25, R26, R73, R74, R112, R113||1K||RPE||CF-1/4-1.0K|
|4||R19, R20, R21, R22||100K 1%||Digikey||100KXBK-ND|
|3||R23, R24, R119||4.7K||RPE||CF-1/4-4.7K|
|6||R27, R28, R85, R86, R89, R90||2.7K||RPE||CF-1/4-2.7K|
|8||R37, R38, R43, R44, R55, R56, R94, R95||51K||Digikey||51KQBK-ND|
|2||R45, R46||38.3K 1%||Digikey||38.3KXBK-ND|
|8||R49, R50, R96, R97, R98, R99, R120, R121||27K||RPE||CF-1/4-27K|
|4||R57, R58, R59, R60||10||RPE||CF-1/4-10|
|4||R65, R66, R67, R68||2.2M||RPE||CF-1/4-2.2M|
|7||R75, R76, R114, R115, R116, R117, R118||33K||RPE||CF-1/4-33K|
|3||R84, R87, R91||1.2K||RPE||CF-1/4-1.2K|
|8||R100, R101, R102, R103, R104, R105, R106, R107||1M||RPE||CF-1/4-1.0M|
|2||R108, R109||10K 1%||Digikey||10.0KXBK-ND|
|3||Z4, Z10, Z29||MC4558||Digikey||296-1414-5-ND|
|6||Z5, Z14, Z21, Z22, Z23, Z24||TL082||Jameco||33241|
|6||Z8, Z9, Z11, Z15, Z17, Z18||4052||Jameco||13119|
|3||Z26, Z27, Z28||14174||Jameco||44329*|
NOTE: The connector parts are exact replacements for the original Chroma parts. So, if you have a set of voice boards that have either had the connectors bent away or break free - the specific part number provided should provide a drop-in, form-fit-and-function replacement (same manufacturer, same colour, etc.)
In January 2002 I had estimated that it might cost $130US to create a board.
I wasn't too far off on that estimate, as the actual cost to build was close to $140US (or approx. $230CAD) per board.
By far the single most expensive items were the two different CEM chips per board (one at $40US, the other at $10US), the temperature compensating resistors ($12.60US per board) and the PCB itself (at approximately $50US). These items alone constitute more than 80% of the board's entire cost.
The following spreadsheet outlines the component costs, and cost distribution for one board.
Note: Were a build to be performed again, less expensive suppliers might be able to be used (for instance, Digikey tends to be a bit expensive with capacitors - yet many of them were purchased from there).
Also, in retrospect, it might have been preferred to use multi-turn potentiometers for the pulsewidth adjustment instead of using single-turn devices.
*2005 Update: Part cost and availability was rechecked in June 2005.
All parts were still available, and the great majority of the parts have decreased in price. The CEM3360 has increased in price by approx. $30.
Because of a favourable change in exchange rate, in Canadian funds it would actually cost slightly less to build a board today than it did in 2002. In US funds, the cost to build a new board would be about $25 more - driven mostly by the increase in CEM cost.
Jameco part number 13725 (CD4556) is now listed as part number 354618. Jameco part number 44329 (74C174) is no longer listed on the Web site.
Following the steps outlined above, I created a total of 8 dual channel voice cards, 7 to replace the ones missing from the 'carcass' and one spare.
See Voice Board Revisions: Clarke for pictures of one of the finished boards (completed for approx. $140US).
The material generated and information gathered are provided here as reference to others who may wish to build replacement dual channel voice cards or to simply repair existing cards. Note: While I can not guarantee that the information provided is error-free, it has allowed me to create total of 8 cards for my own use. These cards have been running without problem since the summer of 2002. (At present, I do not have any plans to create additional boards, for my own use or for sale.)