Powertran MCS1 (Part 4)
Part 4: Testing, Testing
Tim Orr on how to test your MIDI Controlled Sampler once you've put it together: tips and diagrams aplenty.
If you're buying an MCS in kit form, you'll need to know how to test it to make sure your unit's working as it should do once it's built. Here's how to do just that.
Now that the previous articles in this series have convinced you that the MCS1 is perhaps the greatest technological advance in musical history and you've built one from a kit, the last thing you want is for the unit to work incorrectly or, Heaven forbid, not at all.
But first things first. Make sure you inspect all the PCBs thoroughly, checking that all the components are in their correct places, that all joints are properly soldered, and that there are no open or short circuits caused by solder splashes, for example. The experienced constructors amongst you will probably undertake all these tasks as a matter of course anyway, but it's worth remembering that these all-too-common faults are easier to spot at the inspection stage prior to final assembly.
Whatever you do, don't fit the ICs yet. Make the wire connections between the PCBs and clean the solder connections: the MCS1 is best tested outside its box, as shown in Figure 1.
With no socketed ICs fitted, power up the unit. Be on your guard for smoke, small fires and minor explosions. Yes, I know it sounds alarmist, but the chances of a power supply smoothing capacitor that's been fitted back to front blowing up in front of you are unsettlingly high.
Using a digital voltmeter or a scope, test the power supply rails for their correct operating values, which are as follows. Unregulated inputs for the IC115 and 116 are +21V and -21V respectively, while that for IC117, 226 and 322 is +8.7V. Ripple (on load) for unregulated rails should be 300mVpp for IC115 & 116 and 700mVpp for IC117, 226 and 322. Current consumption on load should be 150mA for 12V rails and 1.8A for the 5V one, while other voltage regulators worth checking out are IC100, 306, 419 and 420.
In their present condition, the voltage regulators shouldn't even be warm: shorts across the rails don't usually destroy these regulators, as they incorporate their own thermal shutdown mechanism.
Next, turn the power off, fit all the op-amps and retest the ±12V rails, having already turned the power back on, of course. Turn it off again, and insert the logic chips in lots of ten at a time. Retest the 5V rails and continue this procedure until all the ICs have been inserted.
Your MCS1 is now fully powered and should operate without generating too much in the way of heat. The voltage regulators (with heatsinks), the microprocessor and the PSU power diodes may well be warm or even hot, but most of the remaining components should be only slightly warm. Retest the power supply rails: this is vital for the simple reason that the MCS1's electronics will not operate unless the power supply rails are as they should be.
The machine should now be capable of a safe power-up, and if everything else is OK, your MCS1 will operate first time. One encouraging sign of intelligent life is the processor going through a start-up sequence, during which it flashes all the display LEDs on the MCS1's front panel. If by some unlucky chance this doesn't happen on power-up, you've hit trouble, and a further visual inspection is called for. Typical faults are inter-track short circuits, ICs not being where they should be or fitted backwards, IC pins being folded over underneath the chip instead of going into the IC sockets, non-soldered pins, broken tracks, all the LEDs being inserted backwards or the displays being installed upside down. And if you think some of these eventualities are unlikely - not to say mildly amusing - don't laugh until you're sure you haven't made any of them yourself: it's very easy to do.
Now, even if your unit seems to come to life straight away on first power-up, it's still advisable to test everything. Thoroughly.
Using a DVM or a scope, test for correct power on all the ICs: refer to the power pin chart shown.
Ideally, it's this section that should be tested first. If you're working with an oscilloscope, things won't be quite as easy as you'd probably like, but there are a few simple tests you can employ to locate faults, most of which involve connecting the scope to an IC pin or two and having a look-see at the results.
Have a look initially at pins 38 and 39 on IC309, the crystal oscillator. Figure 2 illustrates its correct operation, while Figure 3 does the same for the E signal on pin 37.
The data and address buses should all be busy for normal operation. Note that the data bus is buffered by IC321. Both the reset signal and the IRQ should be high. One thing that mustn't escape your attention is to look out for illegal logic levels on the data and address buses. Levels of between 1V and 1.5V are generally a sign that something is amiss - a bus clash caused by a short between two logic signals, for instance.
Next, check out the CE signal on the EPROM (IC308, pin 20). This should be relatively active, indicating that the software is running. ICs 310 to 312 generate a range of address codes, many of which are continually active, so if a particular machine function doesn't operate, it may well be the result of a decoder fault.
Now we come to the MIDI keyboard side of things. If nothing happens when you play a note on the keyboard, it could be because the keyboard itself is not being scanned. Have a look at IC315: the Q output pins generate active low row scans. Take a look also at the KBI signal, which clocks the latch. One device certainly worthy of your attention in this department is IC316, the tristate buffer. Check its A inputs first of all, then look at pin 2 and check that switches 1, 5, 9, 13, 17 and 21 all pull this point low when pressed (low is only 0.8V in this case). Don't forget to test the other three column pins, too, and look for the KBO signal.
IC324 is used to divide the 1MHz clock signal by two, and pin 5 should be a 500kHz square wave - look for this signal at IC325, pin 8. Note that the M/S signal must be low for normal operation.
Finally, test the MIDI opto-isolator by injecting either a MIDI or a TTL signal into SKT300. This signal should be repeated at IC327, pin 6, and just in case you're stuck for a source of the latter kind, Figure 4 shows a simple circuit for a DIY TTL drive simulator.
Onward ever onward, in this case to the MCS1's spin-wheel controller. Have a look at IC317, pins 13 and 14: the two outputs should be 90° out of phase with each other when the controller is spun. Now spin the wheel in the other direction and check the phase reversal.
Moving to the audio trigger, you should inject a high-level signal into the audio input and take 1 a look at IC317, pins 1 and 2: you should see a TTL signal of the same frequency as the audio one. Check to see that ICs 318 and 319 are cleared regularly by the CLINT signal.
Lastly in this section, we come to the LED display. Each time a controller function is altered, the display should be updated: check that the CKDIS signal is active at IC304, pin 21.
Just because a component is in full view of everybody and not tucked away inside the MCS1's box, doesn't mean to say it shouldn't be checked for correct operation. This is really quite a straightforward procedure, and I'll run through each control in turn.
To kick things off, press the Delay Line and Voice Mode switches: the LEDs should follow the switch selections. In Delay Line mode, the Freeze and Click Track should both have an independent toggle action. Select Sample Speed - the display can now be varied via the controller. Now select RAM Size and try out the Coarse, Medium and Fine sensitivities; the Bypass has a toggle action.
Press Record in Voice Mode. The LED will flash until the system receives an audio trigger, and it will then remain on until the MCS1's maximum recording time has been exhausted. Once this has happened, the Play LED should illuminate if Play is pressed or if an external gate signal occurs. Once again, the Gate Trig and MIDI CV switches both have independent toggle actions - make sure yours have.
The Pitch Shift is a controller function, and since only one controller function can be selected at a time, the currently selected function takes over from that previously in operation. Loop Start and Loop Length are two further examples of controller functions.
It probably won't have escaped your attention that the MCS1's alphanumeric display is four digits long, while some parameters (such as memory address) are actually five digits in length. In these instances, the display simply loses the last digit, so that a memory address of 65535 (the maximum, incidentally) is represented on the front panel as simply 6553.
Moving back to the controller functions, Filter Shift, Sweep Range and Sweep Speed are all examples of these, with display ranges of 0-12, 0-100, and 1-100 respectively. Finally, both the Sweep On/Off and NR (noise reduction) switches are intended to have toggle actions.
...and the master clock generator (IC101), to be precise. Adjust C104 for an oscillation frequency of 2-5MHz, and have a look at pin 14 (C104 will be aligned later), making certain that the M/S signal is low. The next step is to follow the CK signal through to IC109, pin 2, and then on to the same pin of ICs 102, 103 and 104: this is the CK/N generator circuit. If you select Sample Speed mode, you'll be able to vary the value of N via the spin-wheel controller. Have a look at the CK/N signal (IC104, pin 9) and vary the value of N: Figure 5 shows what effect this variation should have. It's worth noting that this circuit should work fine at frequencies of up to 13MHz, but that above this speed, delays in the counters will cause the divide-by-N circuit to crash... Fortunately, normal MCS1 operation avoids going this high.
Let's turn our attention to the timing generator circuit, ICs 110-112. Sync from A4 (IC112, pin 14), and check that all the timing signals are as shown in the timing diagram reproduced in E&MM December 83. Note that if N is large, all 32 portions of ROM are used as a result of extra refresh periods. As N is reduced, the extended refresh is curtailed.
Now for the software sweep. Set the Sweep Range and Sweep Speed to 100 and activate the Sweep On/Off function. Take a look at IC120, pin 7, and you should see a crunchy sinewave, but a filtered sinewave should be produced at pin 1 (Figure 6 illustrates the difference between these two). Try various Range and Speed settings. CKCV clocks the latch that stores the sinewave data that feeds the DAC (very poetic - Ed).
The log converter circuit (IC121, T100, and T101) converts the input from a one-volt-per-octave keyboard into a log control voltage capable of controlling the master clock generator's frequency (see later for alignment notes on this). DC test points are at pins 1 and 3 of IC121, and should both be -0.6V.
To test the memory address counter (ICs 207, 208, 210 and 211) select the highest sampling speed and the largest RAM size possible. Once you've gone into Delay Line mode, the CK/N signal should be the same as the ACK signal, and the counters will all count up. Look at MA0 through to MA15, a 16-stage binary counter.
The end and return address circuits are best tested by recording a sound in Voice mode. You may decide to record something along the lines of the classic phrase 'One, Two, Testing' - then again, you may not. Once you've recorded your speech, set the Loop Length to 0 and the Loop Start to 6553. Press Play and your words will be replayed as if by magic (or something).
Try taking the Loop Start value from 6553 down to 0 - less of the speech will be heard each time Play is activated. Now increase the Loop Length with Loop Start set at 6553 - a loop of increasing length should be audible. Lastly, try the effect of implementing Fine, Medium and Coarse sensitivities.
The 16-bit words generated by the two looping functions are held in latches. The Loop Start (end address) is held in IC200 and 202, which form one 16-bit input for the address comparator, IC201/3. Varying the Loop Start point obviously varies the data held in these latches. The Loop Length function is a computed return address, so that the end address minus the Loop Length equals the return address: this is held in IC206, 209. When the end address is reached, the counters are loaded with the return address. Varying the Loop Length alters the data held at the Q outputs of IC206, 209.
Buffers IC212 and 213 multiplex the memory address into DRAMs. Check the control signals LAE and MAE: again, refer to the timing diagram. IC224 is the refresh counter, so check CKRA and RAE against the timing diagram, too.
ICs 214 to 221 are the DRAM memory. To test this, go into Delay Line mode and inject an audio signal: look for multiplexed memory addresses on A0-A7 on the DRAMs. The data inputs (DIN) are driven directly from the ADC - check to see that they are all busy.
The data outputs are usually tristate, but are occasionally active. Check the RAS, CAS and WE signals against the timing diagram. Pressing Freeze at this point should result in the WE signal going high, preventing any further writes into the memory from taking place. The external Freeze at the rear of the MCS1 should have a similar effect.
Our last port of call for this section is the unit's built-in Click Track. To test the operation of this, set RAM Size to maximum and select Click Track. The click signal itself is generated at pin 1 of IC227: switch the Click Track off and T100 should sort out the signal.
If you have access to a sinewave sweep generator, the MCS1's active filter responses can be analysed very quickly, as Figure 7 shows.
The first thing to do is to inject an input signal into the MCS1. Test sensitivity selector SW400 by looking at pins 1 and 7 of IC400 - the signal should be unfiltered. However, as we saw in E&MM December, the MCS1 incorporates several filter stages, and some of these are illustrated in Figure 8. Pin 7 of IC401 is a 12kHz four-pole lowpass response (a), pin 8 of IC403 is a 12kHz elliptic response with a notch at 24kHz (b), while IC403 is the mobile filter as a whole (c). If the phase-locked loop is working, a filtered output should be visible at pin 5. Look at pin 2 to see the MFCK signal - this is a square wave with a frequency range of 15kHz to 700kHz, and the break frequency in Figure 8(c) is MFCK divided by 50. Varying the Filter Shift should move the break frequency of the mobile filter over a ± 1 octave range. Maximum signal output level is 8Vpp. IC404, pin 1 is another 12kHz lowpass response, incidentally.
Set the Filter Shift to 12 and select the high sampling rate: the MCS1's audio bandwidth is now at its maximum. Take at look at IC404, pin 7, the noise reduction (pre-emphasis) circuit. With NR on, the circuit adds a treble lift to the signal. Beware of clipping when testing this circuit, and use a 3Vpp signal.
To test the MCS1's sample & hold device (IC405), inject a 1kHz sinewave, vary the Sample Rate control, then take a look at pin 5. Figure 9 gives some idea of the sort of waveshapes you can expect to see.
The next checkpoint is IC408, where pins 9 and 10 generate two signals (ADCK and SC) that must be checked against the timing diagram.
The ADC performs eight tests, going from the MSB to the LSB, and these data bits can be seen stabilising as the analogue-to-digital conversion proceeds, if you know where to look. I'll give you a clue - it's IC409, pins 2 to 9.
Moving logically from the ADC to the DAC, the latter can be tested by following this procedure. Select Delay Line mode, set RAM Size to 0, Freeze and Sweep off, and select the highest sampling speed. Use a 1 kHz sinewave input. This signal should be converted by the ADC into binary code, which is stored in memory and subsequently converted back into an analogue signal by the DAC, ICs 412 to 414. Have a look at pin 6 of IC414, then at pin 7 of IC415 - Figure 10 shows two typical outputs. In fact, the DAC's output is further filtered by the mobile filter IC421 and fixed lowpass filter IC422 (pin 7) in turn.
Now have a look at IC422, pin 1. This is the de-emphasis circuit, and as already implied, can be used to make a sound appear bright simply by recording it with noise reduction switched in and playing it back with NR switched out, so that you're recording with a treble lift and playing back with a flat response.
Turning now to the Mute circuit, this is made up of IC424, T402 and 403. Select Voice mode, record a tone and then Play it. When the Play switch is released, the output signal is attenuated by about 90dB and is effectively off (Figure 11).
Our last place of interest on this whistle-stop audio section tour takes us back to the MCS1's front panel. Go into Delay Line mode, set the Pan control to centre and input a microphone signal: the MCS1's output will be a direct signal with echo. Press the Bypass switch, and the echo should disappear, leaving just the direct signal.
To check that the Click Track is working as it should be, select maximum memory size and activate the click signal. Assuming the metronome is now audible, you're free to lower the memory size, at which point a corresponding shortfall in the Click Track sequence's audibility will take place.
The phase locked loop is made up of ICs 410, 416, 417 and 418. If you select Filter Shift and vary it from 0 to 12, this will alter the division number loaded into IC417 by IC416. Sync from the CK/N input, pin 14 of IC418: pin 3 should be a square wave at the same frequency, while pin 4 will be a square wave with a frequency of CK/N x the PLL division ratio. To check this out, look at pin 4 and alter the Filter shift value.
The PLL is designed to limit at about 700kHz, whereupon it is no longer in lock.
Select Delay Line mode. Use a short memory length and turn the Repeat control up to maximum. Adjust P401 so that echoes are not quite self-sustaining. Experiment with other Filter Shift values, different Sample Speeds and with NR on and off. Readjust P401 so that it's as stable as possible with all configurations.
PLL SOT Capacitor (C449)
As already mentioned, the maximum frequency of the PLL is typically 700kHz. Set the Filter Shift to 12 and the Sample Speed to maximum. If the MFCK frequency exceeds 1MHz, add extra capacitor C449 (22pF) to reduce said frequency. However, this is not considered a very likely eventuality...
Sample & Hold DC Offset (P400)
The sample & hold device (IC405) has a DC offset that varies with the sample frequency. If recording with a frequency sweep is undertaken, this effect can move the quiescent DC offset past quantisation levels, and this process often generates small puffs of noise as it happens.
The solution? Select Delay Line mode with no audio input, no NR, and a short delay time. Look at IC414 pin 6 and listen to the delayed audio output. Vary P400, and you'll notice that as you do so, the sample & hold DC offset moves and the above-mentioned aural effects make themselves apparent. Adjust P400 for the most quiet position, which should be equidistant between two quantisation levels. Turn on the Sweep and select a 1 Hz full-range sinewave. If this causes some further noise generation, readjust P400 accordingly.
Log Converter (P100, 101, 102 & C104)
This circuit has already been discussed and enables an analogue synthesiser to control the pitch of the MCS1's sound output. The normal control range is two octaves, but the MCS1 can add a further five octaves of pitch transposition via its CK/N divider circuit. Unfortunately, the log converter is not the easiest device in the world to set up, but the best method is described below and outlined in Figure 12.
First off, select CV Mode: this allows external voltage control of the master clock generator, IC101. Turn off the Sweep.
Our first job is to determine the voltage/transfer frequency function of the VCO (IC101). Figure 13 shows a typical graph for this function, but obviously the exact plot will vary from unit to unit, which is why your MCS1 came with a blank graph pad (don't tell me you used it as a bin-liner by mistake...). In order to plot your own function, adopt the following procedure. Measure the control voltage at IC108, pin 1, against the output frequency at pin 14 of IC101. Adjust P101 and P102 so that the external CV can move the voltage at IC108 fully over the 0V to +5V range. Set this voltage up to be +5V, and adjust C104 for an output at 11 MHz. Now plot the transfer function on the graph pad, using 0.5V steps going from 0V to +5V.
Measure the voltages needed to generate frequencies of 10MHz, 5MHz and 2.5MHz (otherwise known as V10M, V5M, and V2.5M respectively). This is the two-octave range over which the MCS1 will operate in CV mode.
P100 is the volts/octave preset that adjusts the tuning of the keyboard voltage span. Log transistor T100 must generate a current that doubles for every volt increase in the input control voltage.
The next step is to cut the link between T100 and IC121 and insert a current meter to bridge the gap. Set the meter to a 0-2000pA range. For an input of 0V, set P101 to give an approximate current reading of 75pA and P100 to a central position. Make a note of the measured currents for input voltages of 0V, +1V, +2V (those CVs will generate the required two-octave swing). In order for everything to work properly, the current levels must form a ratio of 1:2:4, or in other words, a musical log law of an octave increase per volt. If the ratio is less than two per volt, rotate P100 clockwise; if it's more, rotate it anti-clockwise. Continue to adjust P100 in this way until the ratio is exactly 2.00: typical final currents should be in the order of 75, 150, and 300pA. Once these levels have been attained, it makes sense to increase the chances of P100 remaining in its proper position by daubing it with something indelible and instantly recognisable - a blob of Tipp-Ex fluid should do the trick. Once that's done, you can remove the current meter and solder the link back into position.
However, that's not the end of the procedure, because the log bias preset, P101, must also be aligned for this aspect of the MCS1's performance to be properly exploited. Using Figure 13 in conjunction with your own function plot, subtract V2.5M from V10M to give you the voltage needed for a two-octave shift (we'll give it the theoretical value of V2oct). Once this has been calculated, apply two external voltages, one of 0V and one of +2V. Adjust P101 so that the voltage change at pin 1 of IC108 is equal to V2oct for the input two-volt change. Clockwise rotation of P101 increases the size of the voltage change, anti-clockwise rotation decreases it. Beware of the fact that P102 may also have to be adjusted during this process so as to maintain the DC position on the graph.
Once you've calibrated P101, set the external control voltage to 0V and adjust the linear offset correction (via P101) to give a measured voltage at IC108's pin 1 of V2.5M. If you now apply an external CV of +2V, this voltage measurement should change to V10M.
If, by some miracle, this is precisely what happens, then both P101 and P102 are aligned, and the Tipp-Ex must be brought into action again. The log converter is now working and full aligned.
Just to make sure the theory works out in practice, try sampling a pitched sound (in Voice mode) and replaying it via a one-volt-per-octave keyboard in CV mode. Don't forget to make the Gate connections between synth and MCS1: the latter's gate-on is +2V or more, while gate-off is 0V or negative (the Play LED illuminates when the former condition is prevalent).
The recorded sound can now be played and pitched over a two-octave range, and pitch transposed over a five-octave range via the Pitch Shift function. However, as I mentioned in December, the tuning in this mode is not entirely perfect, due mainly to the bent VCO transfer function. If you have the option of using the now almost ubiquitous MIDI as a control source, then do so - the results are undoubtedly superior.
Feature by Tim Orr
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!