Jump to content

Sinclair QL RAM Replacement

From RetroTechCollection
Sinclair QL motherboard
A QL with the top case removed, showing the main board (photo: Ewx, CC BY-SA 3.0)

The Sinclair QL has 128 KB of RAM on sixteen 64K × 1 dynamic RAM chips, IC1 to IC16, arranged as two banks: IC1 to IC8 and IC9 to IC16. Within each bank every chip holds one bit of each byte, so each data line is shared by one chip in each bank.[1][2] When a chip fails, the QL's start-up memory test shows it as a white or green screen instead of the copyright message, and Sinclair's Service Manual gives a method for finding the failed chip from the pattern on the screen.[3]

Symptoms

[edit | edit source]
What the screen shows at switch-on
Screen What it means
A fine pattern of green, white, red and black dots ("tweed") runs down the screen, then the copyright message appears The RAM test has run and passed[3]
Plain white screen The first part of the RAM test failed: the RAM could not store ones, zeros and a random pattern[3]
Plain green screen The second part of the test failed: data did not stay in the RAM. Either the RAM is not being refreshed or an address or data line is shorted[3]
White or green with very narrow vertical stripes or a vertical pattern of dots One bit is faulty: a failed RAM chip, or a short on a data or address line[3]
After RESET, the tweed pattern shows for a while and then the screen turns white A failure in the second bank of RAM[3]
A geometric pattern such as red and green vertical bars The ZX8301 is working but the processor is not running the ROM code. The CPU clock may be missing, the CPU may be absent, or the RAM may be faulty[3]
The display becomes erratic after the QL has warmed up Sinclair's fault table says to check the RAMs[4]

A black screen with the power light on has several causes; Sinclair's list starts with the supply voltages and the ZX8301 (IC22) before the RAMs.[4] Tony Firshman of TF Services adds that the QL's own memory test does not find every RAM fault, and recommends the Minerva replacement ROM, whose memory test reports the first faulty address.[5]

Safety

[edit | edit source]

There is no mains voltage inside the QL; the highest voltage on the board is the 44 V peak-to-peak AC feed from the power pack.[6] Use an earthed, low-power soldering iron and do not overheat the pads when desoldering.[7] Take static precautions when handling the chips.[5]

Tools and parts

[edit | edit source]
  • A multimeter.
  • A probe lead that can be connected to ground.[3]
  • A 16-pin IC test clip with a known-good 4164 attached, if you want to try the piggyback test.[3]
  • Replacement 64K × 1 dynamic RAMs. Sinclair's parts list gives the Hitachi HM4864P-2, with the TMS4164-15NL, MCM6665AP15, MK4564N15, UPD4164C-3, MSM3764-15RS and HYB4164P2BD as alternatives.[8]
  • A soldering iron, desoldering braid or pump, and a cross-head screwdriver.[7]
  • Optionally, a Minerva ROM for its memory test.[5]

Procedure

[edit | edit source]

Checking the supplies

[edit | edit source]
  1. Open the case: remove the four 5/16-inch screws along the front edge and the four 1¼-inch screws along the rear edge, but not the two screws below the Microdrives, and lift the top half. It stays connected by the keyboard ribbons and LED leads.[9]
  2. Power the QL and measure the +5 V output at the + side of C42 (5.0 V ± 0.15 V, no discernible ripple) and the +12 V output at the + side of C38 (12.0 V ± 0.25 V).[6] Sinclair's fault finding starts with these voltages when the copyright screen is missing; if a rail is out of limits, fix the supply first (see Sinclair QL Voltage Regulator Replacement).[3]

Reading the screen

[edit | edit source]
  1. Switch on and watch the screen from the start. If the copyright message does not appear, switch off and try two or three times: Sinclair notes that a QL can lock up at start-up and look dead.[3]
  2. Match what you see against the table above. A plain white or green screen means the RAMs are working in general and the fault is a short on an address or data line or a single faulty chip.[3]
  3. On a green screen, look over the RAM area for solder bridges between pins or across the address and data tracks, then further afield, for example around the processor. If you check with an ohmmeter, remember that readings can be misleading where the meter feeds back through the CPU.[3]

Finding the faulty chip

[edit | edit source]
Sinclair QL circuit diagram (Issue 6), RAM data pins
Part of Sinclair's Issue 6 circuit diagram: on each 4164, pins 2 and 14 are joined to one data line (DB0, DB1, DB2 for IC1, IC2, IC3)
Sinclair QL Issue 6 layout, RAM and keyboard connectors
Sinclair's Issue 6 layout, showing where the sixteen RAM chips IC1 to IC16 sit on the board

This works when the screen shows narrow stripes or dots.[3]

  1. Connect one end of a probe lead to ground.[3]
  2. With the QL showing the faulty pattern, touch the probe on the data pin of each RAM in turn and watch the screen. On the QL's 4164s the data input (pin 2) and output (pin 14) are joined to the same data line.[3][2]
  3. Each touch puts a vertical black line or lines on the screen. When the black line falls on the faulty stripes or dots, you have found the data line.[3]
  4. The data line is shared by one chip in each bank, so the fault may be in either: for example IC1 or IC9. Sinclair's example is to change IC1, because it is in the memory-mapped area of the store.[3][2]
  5. As a rule of thumb, a long vertical stripe points to a short on a data line, and dots point to a RAM fault.[3]
Which chips share each data line[2]
Data line First bank Second bank
DB0 IC1 IC9
DB1 IC2 IC10
DB2 IC3 IC11
DB3 IC4 IC12
DB4 IC5 IC13
DB5 IC6 IC14
DB6 IC7 IC15
DB7 IC8 IC16

Confirming without desoldering

[edit | edit source]
  1. Sinclair suggests a "test IC": a known-good 4164 on an IC test clip, clipped over each suspect chip in turn. Sinclair says this method is not guaranteed to work but can save time spent changing chips unnecessarily.[3]
  2. If you have a Minerva ROM fitted, its memory test reports the first faulty address, from which the chip can be identified.[5]

Replacing the chip

[edit | edit source]
  1. To work on the underside of the board, take it out. Release the screw that holds the regulator IC35 to the heatsink and lift the heatsink away, keeping the screw and its washers. Remove both Microdrives (three pan-head screws each; do not confuse them with the countersunk motor screws), then the reset button, the loudspeaker leads, the two pan-head screws near the expansion connector and the loudspeaker, and the push-fit Microdrive extension bung. Ease the board off its locating dowels, lifting it at the expansion connector end.[9]
  2. Desolder the faulty RAM without overheating the pads.[7]
  3. Fit the new chip with its notch the same way round as the others in its row.
  4. Inspect the board for solder splashes; Sinclair warns that hairline splatter can measure several ohms and give misleading symptoms.[7]

Reassembly

[edit | edit source]
  1. Refit the board reset-button end first, then the screws, the Microdrive extension bung, the loudspeaker leads and the reset button.[9]
  2. Refit the heatsink and the IC35 screw with its washers.[9]
  3. Refit the Microdrives. The drive in the left-hand (MDV1) position has an RF shield over its ULA; check that it is in place. Straighten the bare ends of each drive's ribbon cables and push them fully home.[9]
  4. Reconnect the keyboard ribbons and LED leads and close the case.[9]

Testing

[edit | edit source]
  1. Switch on. The tweed pattern should run down the screen and disappear, followed by the copyright message with the F1 (monitor) and F2 (television) options.[3]
  2. Press RESET and check that the screen does not turn white after the tweed pattern, which would point to the second bank.[3]
  3. If a Minerva ROM is available, let its memory test run; Firshman notes that its test data varies, so it finds faults the standard test misses.[5]

Revision notes

[edit | edit source]

Both the pre-Issue 6 and Issue 6 boards have 128 KB of RAM on IC1 to IC16.[1] The data-line table above is taken from the Issue 6 circuit diagram.[2]

[edit | edit source]

References

[edit | edit source]
  1. ↑ 1.0 1.1 Sinclair QL Service Manual (File:Sinclair QL Service Manual.pdf), Sinclair Research, prepared by Thorn (EMI) Datatech, issue 3, October 1985, Section 1 paras 5.1.1, 5.1.3 and 5.2.2 to 5.2.3: 128 KB of RAM in both the pre-Issue 6 and Issue 6 versions; 128 KB of RAM at addresses 20000 to 3FFFF on sixteen 64K-bit dynamic RAMs, IC1 to IC16, organised as two matrices, IC1 to 8 and IC9 to 16, selected by CAS0L and CAS1L from the ZX8301.
  2. ↑ 2.0 2.1 2.2 2.3 2.4 Sinclair QL Service Manual, Figure 1.5, circuit diagram (Issue 6), hosted as File:Sinclair QL circuit diagram (Issue 6).png: IC1 to IC16 are 4164s; in each chip pins 2 and 14 are joined to the data line, IC1 to IC8 to DB0 to DB7 and IC9 to IC16 to DB0 to DB7 in the same order; pin 8 is +5 V and pin 16 is 0 V.
  3. ↑ 3.00 3.01 3.02 3.03 3.04 3.05 3.06 3.07 3.08 3.09 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 3.20 3.21 Sinclair QL Service Manual, Section 4 paras 2.2.1 to 2.2.18, fault finding: the copyright screen; supply voltages; the tweed pattern at the start of the RAM test; white screen (first part of the test failed) and green screen (second part failed); narrow stripes or dots for a single faulty bit; finding the chip with a grounded probe on each RAM's data pin; RESET followed by tweed and then white for the second bank; the order for changing plug-in ICs; and the "test IC" method for non plug-in ICs such as the 4164 RAMs.
  4. ↑ 4.0 4.1 Sinclair QL Service Manual, Section 4, fault-finding guide: "TV/monitor screens black. Power light on. Fails to power up": check voltages, renew IC22, renew IC24, check RAMs, renew IC34, IC18, IC33; "Erratic display after warm-up period": check RAMs.
  5. ↑ 5.0 5.1 5.2 5.3 5.4 Tony Firshman, "QL Troubleshooting", TF Services, last updated 28 October 2003: RAM failures, using Minerva, which flags the first faulty address in RAM, and the standard QL memory test, which does not pick up errors where NUL bits are written; taking static precautions when handling chips.
  6. ↑ 6.0 6.1 Sinclair QL Service Manual, Section 1 para 8 "Power Supplies": 9 V DC and 44 V peak-to-peak AC from the power pack. Section 4 para 2.2.3, supply test points: ±12 V regulator input at the + side of D26, 15.6 V AC ± 2.0 V; +5 V regulator input at the + side of C41, 9.0 V ± 2.0 V; +12 V output at the + side of C38, 12.0 V ± 0.25 V; +5 V output at the + side of C42, 5.0 V ± 0.15 V.
  7. ↑ 7.0 7.1 7.2 7.3 Sinclair QL Service Manual, Section 4 para 3.1 repair notes: desoldering technique, an earthed low-power iron, and solder splashes that can measure several ohms and give misleading symptoms.
  8. ↑ Sinclair QL Service Manual, Section 5, parts lists, integrated circuits: IC1 to 16, HM4864 P2 (Hitachi), with alternatives TMS4164-15NL (Texas), MCM6665AP15 (Motorola), MK4564N15 (Mostek), UPD4164C-3 (NEC), MSM3764-15RS (OKI), TMS4146-12NL (Texas) and HYB4164P2BD (Siemens).
  9. ↑ 9.0 9.1 9.2 9.3 9.4 9.5 Sinclair QL Service Manual, Section 2 "Disassembly/Assembly": paras 1.1 (case), 1.2 (heatsink and Microdrives), 1.3 (circuit board) and 2.3 to 2.4 (assembly).