Flesher TU-170 Terminal Unit

This is an interesting design. On the transmit side, a 555 timer chip is used as a multivibrator square wave generator. The low mark frequency (2125 Hz) is set by R64, P8, and C26. When switched to space (2295 Hz), Q8 places P7 and R63 in parallel with P8 and R64, increasing the frequency. The square wave output of IC8 is converted to a sine wave by passing through low pass filter formed by R69 and C29, and a bandpass filter formed by IC7A and associated resistors and capacitors. The tuning of the bandpass filter can be adjusted to balance the tone levels. This filter was adjusted while watching the RF output power while switching between mark and space so that the power was the same.

The image at the right shows a waterfall graph of rhe received FSK transmitted by the TU-170 through the SEA-245. Note the components above the FSK signal (between the FSK and the suppressed SSB carrier. This image is with the TU-170 transmitting repeated space characters from the Teletype model 15. The keyboard and selector magnets are in series. Note that a carrier from another station is at 3.5498 MHz.

The next image shows the waterfall graph with the selector magnets out of the circuit (shorted). Again, there is a carrier from another station at 3.5498 MHz. However, a large portion of the splatter has been removed. The RF is largely confined to the mark and space frequencies. It appears the inductive spike when the loop current is interrupted may be causing excessive bandwidth.

The TU-170 has an internal 84 V loop supply. The printer selector magnets and keyboard are placed between the positive end of the loop supply and the collector of Q6, an MJE340 NPN transistor. Q6 acts as an active current sink, setting the loop current to 60 mA for mark and 0 mA for space. In addition, emitter resistor R59 provides a voltage proportional to the loop current. This voltage (marked Loop Monitor on the rear panel) drives the AFSK keying input of the modulator. The AFSK oscillator is keyed by both received signals and the keyboard.

When Q6 interrupts the loop current, the selector magnets generate a high voltage (the flyback voltage). It is possible this voltage could go high enough to damage Q6, which has a maximum collector voltage rating of 300 V. The image at the right shows the flyback voltage when Q6 interrupts the loop current and the selector magnet armature is released. When the armature is released, there is a gap in the magnetic circuit, and the inductance is considerably less than when the armature is against the selector magnet pole pieces. Under these conditions, we see a 280 V flyback voltage followed by some ringing. Note that there is a 10 nF capacitor (C40) between the Q6 collector and ground to limit the speed of loop current changes. This, along with the selector magnet inductance, may be causing the ringing.

The TU-170 supports an external oscilloscope as a tuning indicator. Traditionally, a CRT wired as an XY display has been used. The horizontal deflection would be driven with the audio out of the mark filter and the vertical deflection with the audio out of the space filter (or vice versa). Another possibility would be to use a dual trace scope with mark on one trace and space on another. This would, however, require more circuitry since a timebase and dual trace circuitry would be required (typically a high frequency chopper between the two inputs or drawing the traces alternately). It is, of course, much simpler with digital scopes since the only difference is in software.

The top image shows a dual trace view of the scope outputs. The bottom image is an XY view. The XY view appears more intuitive.

Single channel oscilloscope kits are available for under $20. Dual channel scopes that support XY mode are less than $100.

The shape of the keying signal to the AFSK generator affects the RF bandwidth. A very square waveform with a short rise and fall time has more high frequency harmonics of the keying frequency. These generate sidebands in the FSK signal that are farther from the center frequency. They are the FM equivalent of "key clicks". With a bit time of 22 ms, the square wave frequency is 22.7 Hz. We should includes at least the third harmonic to allow recovery of the data. The output resistance of the loop monitor is 82 ohms or less (82 ohms for space and about 45 ohms for mark). With 82 ohms and a cutoff frequency of 68 Hz, a 29 uF capacitor would be required. A 20 uF capacitor (two 10 uF capacitors in parallel) was added between the Loop Monitor terminal and ground on the rear panel. The image at the right shows the resulting waveform at the Loop Monitor output (and the 170 Hz Shift Input). As expected, the rise time is shorter than the fall time. The actual speed of the shift between the mark and space frequencies will be faster than these times due to the gain of Q7 and Q8.

The image at the right shows the waterfall spectrum from the modified system as received at KFS, Half Moon Bay, CA. It looks considreably better than when we started. I believe the loop clamp circuit and the low pass filter on the AFSK keying input helped. I was unable to decode the RTTY using the FSK decoder on KiwiSDR. However, acoustically coupling the received audio from KiwiSDR or the KFS websdr to DroidRTTY did result in proper decoding.

Some slight modifications are required to get the TU-170 to work with my SEA 245 transceiver. The SEA 245 has a tri-state balanced audio bus between the transceiver and the control head(s). Either transmit or receive audio is on this bus (transmit audio when PTT is low, receive audio when PTT is high). The TU-170 is designed to drive the microphone input of a transceiver, so there's a voltage divider on the AFSK output. The modification consists of adding a 600 ohm to 600 ohm audio transformer. One side of this transformer is across the audio bus. The other side has one end grounded, and the other going to the demodulator input (TB2 or PCB D). The contacts of a reed relay connect the high level AFSK output (top of R72) with a series 470 ohm resistor to the ungrounded end of the transformer when the contacts close. The series resistor causes about 500 mV on the audio bus when transmitting. The coil of the reed relay is connected between PCB J and ground so the AFSK output drives the transformer when transmitting. The reed relay and audio transformer are visible at the bottom of the photo. They are held in place by RTV silicone adhesive. The unit I have had a cable strain relief mounted on the rear panel between the 4 pin transceiver jack and the CW ID jack. I ran 2 pair shielded cable through this strain relief with one pair carrying the bidirectional tristate audio and the other pair carrying PTT.

It's interesting to listen to the transmitted signal and look at its spectrum. The following recordings were from the KFS websdr. For comparison, here's a recording of a CW carrier (audio. The CW carrier is at about 2 kHz with a level of about -59 dB. The surrounding noise is at about -77 dB. The low frequency noise starts to roll off at about 200 Hz with some high level components from 2 Hz down. This may be the fading of the signal. The top image to the right is the spectrum as shown by Audacity.

The second image is the spectrum with a 300 Hz HPF applied.

The third image shows the spectrum as the receiver was adjusted a couple hundred Hz to see how flat it is with the 170 Hz FSK shift.

The received audio does not sound that great. There are some pretty strong clicks that the capacitor on the AFSK keying input was supposed to help.

The second image at the right is the spectrum of the received signal with a linear frequency axis making the mark and space frequencies more visible. The higher frequency peak is the space frequency (space is low on RF, but this is an LSB receiver). The amplitude is lower since the signal spends less time on space than it spends on mark (since it idles on mark between characters and the stop bit is longer than the data bits and start bit). However, as shown in the next image, it appears the space amplitude is indeed quite a bit lower than the mark amplitude.

The bottom image is a zoomed in version of the received AFSK waveform. It appears the tones are substantially different im amplituede. The tone level balance (P9) was set for equal output power for mark and space. This may not be appropriate.

Looking more closely at the clicks and inconsitent levels, we find the following. The first image is the AFSK output as measured on the TU-170 side of the added audio transformer. There are some pretty good transients when the loop is keyed. These are probably causing the compressor in the transceiver to reduce gain on the mark to space transition.

The second image is a close up view of the same waveform.

The third image is the AFSK output with the loop being keyed by the keyboard, but the selector magnet is shorted. A very small transient is visible. The issue seems to be the selector magnet flyback spike getting into the AFSK.

The issue appears to be keyboard arcing getting into everything. A snubber consisting of 100 ohms in series with 470 nF was put across the keyboard contacts. That cleaned up the AFSK and RF. Here is s recording from the KFS SDR in Half Moon Bay, CA. The top image at the right is the waterfall view of the RF spectrum. This looks and sounds a lot better. The audio properly decodes on DroidRTTY.

The second image is the audio spectrum with a linear frequency axis from Audacity.

The third image shows the audio waveform (resampled from 8 kHz to 48 kHz sample rate). We can see that there is an amplitude shift between the mark and space frequencies, but the large transients on the mark to space shift are gone.

The amplitude shift appears to be due to receiver bandwidth. In the image at right, the websdr bandwidth was increased to 3.68 kHz for -6 dB. On the transmit side, no difference in the output power meter is observed between mark and space. The default LSB Wide bandwidth is 2.8 kHz for -6 dB.

KE3BK reported difficulty receiving my signal. Here are some scope photos he sent of the TU filter outputs.

The first image is ITTY being received on a Doveton TU. The vertical scale is 5V/div (not 50). Channel 3 (pink) is mark, while channel 4 (blue) is space. Loos pretty good!

The second image is my received signal on the Dovetron. It appears that the space has an excessively long rise time, perhaps close to 10ms or 1/2 bit time.

The third image is my signal on a HAL CWR-6850.

In each case, there seems to be a notch in the rise of the space envelope (blue trace).

Recall above that 20 uF was added between the AFSK keying input and ground to slow the edges to avoid "key clicks". Perhaps this was unnecessary since a lot of problems appear to be due to keyboard arcing and were resolved with a snubber.

The images at the right show the TU-170 filter outputs with locally generated AFSK. The top image has an added 20uF between the AFSK keying input and ground, the second has 10uF, and the third has no added capacity. Channel 1 (yellow) is the output of the mark filter. Channel 2 (blue) is the output of the space filter. In the first image, the notch in the rise of the space envelope is visible. At the same time, it appears that the mark filter output starts going up a bit, then continues dropping. If anything, the middle image, with 10 uF, appears the best.

As part of these tests, the AFSK tone frequencies were rechecked:

• Mark - 2127.4 Hz, 2.4 Hz high, 0.113% high
• Space - 2299.7 Hz, 4.7 Hz high, 0.205% high

The trim pots used in the TU=170 are single turn and pretty finicky. On 7/2/21, I replaced them all with Bourns 3296W-1 series pots. These make adjustment of the TU-170 much easier.

The top image shows the original trim pots in the receive filter section. The bottom image shows the same section with the new trimpots.