A Tale of Two Racal 9917A Frequency Counters.
Posted: Sun Oct 15, 2023 7:05 pm
These two were bought along with a Racal 9009 modulation meter for £10. The seller said they didn't work but said they might be OK for the cases. Given the cost of new cases, a mains transformer and IEC socket, it's always worth bearing in mind that junk equipment may be worth buying for that. They both looked neglected and showed signs of being kept in the damp.
I'll call the one with the later serial number Newer, and the other Older.
Newer was missing its frequency reference. Internally it looked quite clean. Older was rather more tatty. There were cobwebs on the back and some of the screws were rusty. There were water stains on the PCB. The bodies of the slide switches were rusty. I think it must have been kept in a shed with a leaky roof.
Older is the one on the bottom.
I looked for the manual on the WWW. The 9917 manual is easy to find but not the 9917A manual. I eventually I found the circuit diagrams, parts list and PCB layout on radiomuseum.org. There are differences between the 9917 and 9917A particularly in the board layout and parts numbering. The quality of the scans I found isn't great. Some pin numberings on ICs are hard to read. The quality of the original manual isn't top notch either, eg several ICs have the same number.
I put a spare OCXO into Newer and tried it. Nothing on the display. After a time it started to smell warm and a hybrid circuit seemed to be very hot.
I tried Older. It did show a row of zeroes on the display. Nothing was getting hot. It didn't work.
These things have roughly five parts:
A low frequency 10Hz to 60MHz input channel. This consists of a FET/bipolar buffer which feeds a video amplifier and it outputs a TTL level version of the waveform the instrument is fed
.
A high frequency 40 to 560MHz input channel. This has overload protection circuitry to operate a cut off relay. A PIN diode attenuator operated by an AGC system. An OM335 hybrid UHF amplifier. AGC circuitry. Comparators to stop the instrument operating on too low a signal level. A 600MHz ECL decade counter. The ECL counter does the heavy lifting on the high frequency signals this channel handles, produces a BCD output and overflows to the TTL counters.
TTL/CMOS control logic which directs the output of the selected channel to the counter and makes operational changes depending on switch settings.
The counting circuitry which consists of two TTL decade counters and a special IC which discharges most of the functions of a 60MHz frequency counter.
The display and supporting circuitry.
Older seemed the more promising. I checked the supply lines, Two +5V lines, a -5V line and a +24V line. The three 5V lines were spot on with no ripple. The 24V line is surprisingly hard to get to. The -5V and +24V lines are only used in the high frequency input channel, for the ECL counter and the hybrid UHF amplifier, and I'd decided to concentrate on getting it to work with the low frequency channel first as that's simpler. I cleaned the switches with contact cleaner.
I fed in a 10MHz signal and traced it through the B channel with a scope. It produced the expected TTL level version. The signal is input to a 3 channel NAND gate to select it, if the other two inputs are high. One was never high. Channel selection is controlled by a morass of TTL and CMOS.
I traced through the logic, which took a long time and some head scratching, and found a dual CMOS one-shot, one output of which was always low regardless of the input conditions. The chip in this unit is a CD4098, the parts list says it's an MC14528. There are a few similar chips with the same pinout, recommended for different pulse width ranges. They are still obtainable and not very dear. Anyway, there should be one in Newer. Newer has a 74LS221 in that position and a slightly different PCB as the 74LS221 has a different pinout. I removed the old chip and fitted a socket. I suspected the reason the 74LS221 was used in later version was that the CMOS monostables proved unreliable.
I could see no reason why a 74LS221 shouldn't be used and I have some on hand. It would take a time for an order for the CMOS chip to go through. I made this beautiful dead bug adapter for a 74LS221.
I ordered a couple of MC14538 chips anyway. The dead bug 74LS221 worked but It was a bit wobbly and the timing must have been slightly off, as after a reset there were a few low end digits, rather than a row of zeroes. When the MC14538 came I fitted it and the problem disappeared. The counter now worked properly on channel B.
Channel A also worked but was very insensitive. It needed a minimum 3V p-t-p signal. I traced the signal through and the OM335 UHF amp produced an attenuated version. It was probably capacitative pass through. I made the effort to measure the voltage on the 24V line. It was 300mV at the regulator and the regulator was hot. I took out the regulator and checked it. It was good. On the PCB there was a short from the output side of the regulator to earth. I followed the circuit through a there was a low pass filter consisting of two tants and an inductor. One tant had a broken lead so was out of the picture. The short went away then the second tant was removed. It had a dead short, but it showed no signs of overheating. I replaced the tants and the regulator and channel A now worked properly with a 20mV p-t-p signal. I should been onto it quicker because i've come across this problem with tants in the 24V supply line of these counters before, and the tant in that position in Newer has exploded.
On the left there are two white DIL packages in line, then a black one in line, then a DIL package perpendicular to the line. That's the MC14538 in an IC holder. On the right about halfway down are the two orange tants I put in.
Older is now a working 9917A counter. It's been running for 24 hours and everything seems OK. The OCXO is fairly good, a few parts in 10^7, which is OK for a 40 plus year old Racal OCXO. Some are useless and can't be adjusted to their proper frequency, others have a frequency which wanders. Here's Older fed by 10 MHz from a GPSDO.
I'm wondering what to do about Newer. It does have a working custom counter chip which is hard to find and costs as much as a working counter. I could put its rear panel on older although after cleaning it doesn't look too bad. It has the low frequency option which some of my other counters don't have and I could put in one of them I suppose I could raise the steam to investigate it. I'm torn between striking while the iron is hot, and having had enough of fixing Racal counters for now.
General thoughts:
Hours of innocent amusement at a low cost, resulting in a working Racal 9917A counter, plus at least a spares mule for the others. Another thing saved from landfill.
The counter works well enough. I'd recommend that if you were in the market for a working Racal counter and you had a choice of several at around the same price, avoid the 9917 and 9917A. The 9917 must have been replaced by the 9917A because it had problems. The documentation for the 9917A is hard to get and the quality of what's there is dubious. There are obvious changes which the documentation which has turned up doesn't reflect, such as the change to a 74LS221.
I'll call the one with the later serial number Newer, and the other Older.
Newer was missing its frequency reference. Internally it looked quite clean. Older was rather more tatty. There were cobwebs on the back and some of the screws were rusty. There were water stains on the PCB. The bodies of the slide switches were rusty. I think it must have been kept in a shed with a leaky roof.
Older is the one on the bottom.
I looked for the manual on the WWW. The 9917 manual is easy to find but not the 9917A manual. I eventually I found the circuit diagrams, parts list and PCB layout on radiomuseum.org. There are differences between the 9917 and 9917A particularly in the board layout and parts numbering. The quality of the scans I found isn't great. Some pin numberings on ICs are hard to read. The quality of the original manual isn't top notch either, eg several ICs have the same number.
I put a spare OCXO into Newer and tried it. Nothing on the display. After a time it started to smell warm and a hybrid circuit seemed to be very hot.
I tried Older. It did show a row of zeroes on the display. Nothing was getting hot. It didn't work.
These things have roughly five parts:
A low frequency 10Hz to 60MHz input channel. This consists of a FET/bipolar buffer which feeds a video amplifier and it outputs a TTL level version of the waveform the instrument is fed
.
A high frequency 40 to 560MHz input channel. This has overload protection circuitry to operate a cut off relay. A PIN diode attenuator operated by an AGC system. An OM335 hybrid UHF amplifier. AGC circuitry. Comparators to stop the instrument operating on too low a signal level. A 600MHz ECL decade counter. The ECL counter does the heavy lifting on the high frequency signals this channel handles, produces a BCD output and overflows to the TTL counters.
TTL/CMOS control logic which directs the output of the selected channel to the counter and makes operational changes depending on switch settings.
The counting circuitry which consists of two TTL decade counters and a special IC which discharges most of the functions of a 60MHz frequency counter.
The display and supporting circuitry.
Older seemed the more promising. I checked the supply lines, Two +5V lines, a -5V line and a +24V line. The three 5V lines were spot on with no ripple. The 24V line is surprisingly hard to get to. The -5V and +24V lines are only used in the high frequency input channel, for the ECL counter and the hybrid UHF amplifier, and I'd decided to concentrate on getting it to work with the low frequency channel first as that's simpler. I cleaned the switches with contact cleaner.
I fed in a 10MHz signal and traced it through the B channel with a scope. It produced the expected TTL level version. The signal is input to a 3 channel NAND gate to select it, if the other two inputs are high. One was never high. Channel selection is controlled by a morass of TTL and CMOS.
I traced through the logic, which took a long time and some head scratching, and found a dual CMOS one-shot, one output of which was always low regardless of the input conditions. The chip in this unit is a CD4098, the parts list says it's an MC14528. There are a few similar chips with the same pinout, recommended for different pulse width ranges. They are still obtainable and not very dear. Anyway, there should be one in Newer. Newer has a 74LS221 in that position and a slightly different PCB as the 74LS221 has a different pinout. I removed the old chip and fitted a socket. I suspected the reason the 74LS221 was used in later version was that the CMOS monostables proved unreliable.
I could see no reason why a 74LS221 shouldn't be used and I have some on hand. It would take a time for an order for the CMOS chip to go through. I made this beautiful dead bug adapter for a 74LS221.
I ordered a couple of MC14538 chips anyway. The dead bug 74LS221 worked but It was a bit wobbly and the timing must have been slightly off, as after a reset there were a few low end digits, rather than a row of zeroes. When the MC14538 came I fitted it and the problem disappeared. The counter now worked properly on channel B.
Channel A also worked but was very insensitive. It needed a minimum 3V p-t-p signal. I traced the signal through and the OM335 UHF amp produced an attenuated version. It was probably capacitative pass through. I made the effort to measure the voltage on the 24V line. It was 300mV at the regulator and the regulator was hot. I took out the regulator and checked it. It was good. On the PCB there was a short from the output side of the regulator to earth. I followed the circuit through a there was a low pass filter consisting of two tants and an inductor. One tant had a broken lead so was out of the picture. The short went away then the second tant was removed. It had a dead short, but it showed no signs of overheating. I replaced the tants and the regulator and channel A now worked properly with a 20mV p-t-p signal. I should been onto it quicker because i've come across this problem with tants in the 24V supply line of these counters before, and the tant in that position in Newer has exploded.
On the left there are two white DIL packages in line, then a black one in line, then a DIL package perpendicular to the line. That's the MC14538 in an IC holder. On the right about halfway down are the two orange tants I put in.
Older is now a working 9917A counter. It's been running for 24 hours and everything seems OK. The OCXO is fairly good, a few parts in 10^7, which is OK for a 40 plus year old Racal OCXO. Some are useless and can't be adjusted to their proper frequency, others have a frequency which wanders. Here's Older fed by 10 MHz from a GPSDO.
I'm wondering what to do about Newer. It does have a working custom counter chip which is hard to find and costs as much as a working counter. I could put its rear panel on older although after cleaning it doesn't look too bad. It has the low frequency option which some of my other counters don't have and I could put in one of them I suppose I could raise the steam to investigate it. I'm torn between striking while the iron is hot, and having had enough of fixing Racal counters for now.
General thoughts:
Hours of innocent amusement at a low cost, resulting in a working Racal 9917A counter, plus at least a spares mule for the others. Another thing saved from landfill.
The counter works well enough. I'd recommend that if you were in the market for a working Racal counter and you had a choice of several at around the same price, avoid the 9917 and 9917A. The 9917 must have been replaced by the 9917A because it had problems. The documentation for the 9917A is hard to get and the quality of what's there is dubious. There are obvious changes which the documentation which has turned up doesn't reflect, such as the change to a 74LS221.