AM26C31: Usage of output and inverted output

In reply to Max Robertson:

Hello Max,

Thanks for your quick response.
The background of my question is an erroneous behavior that we rarely observed in conjunction with a stepper motor optical encoder.
I reengineered the circuit of the encoder and found out that the encoder manufacturer use the AM26C31I as output line driver.

I connect the motor encoder (channels A and B -> as I mentioned before not the inverted signals /A and /B) via a cable to a pcb where I merge the encoder signal with some other signals and use a different cable (SMC-ribbon cable) to bring the signals to the mainboard. On the mainboard the encoder signals have a pull up resistor of 2k7 to 3.3 V and then go via a low passes (100R/1n) to inputs of a SN74LVC3G17 (VCC=3.3 V). The output signals of the SN74LVC3G17 are processed by the FPGA then.

On the encoder electronics the AM26C31I is supplied with 5 V. The high level output voltage of the AM26C31I is approx. 3.4 V as I expected according to the data sheet 6.5 Electrical Characteristics: AM26C31C and AM26C31I.

The erroneous behavior that I rarely observe is that the low level is not driven correctly below the low level trigger voltage of the SN74LVC3G17. This behavior lasts from several 100 microsecondes to some milliseconds and then disappears. I saw that on one channel but our FPGA signal processing flags us error where both channels A and B of the encoder are influences at the same time. But this behavior I have not seen with my oscilloscope.

To make the situation clearer, it would be possible to support you with schematic excerpts and scope plots, if you are interrested.

Thank you in advance for your efforts.

Regards, Burkhard

In reply to Burkhard Seyferth:

Hi Burkhard,

Thank you for summarizing your set-up so clearly. Just to make sure I understand, what you refer to as A and /A would correspond to the Y and Z differential output pair one channel of the AM26C32I (to use the datasheet's nomenclature), and B and /B signals would correspond to the Y and Z output pair of another channel, correct? If so, I would be interested to better understand the behavior you report regarding the output not reaching a proper "low" level for short periods of time. Would it be possible for you to share a waveform capture from an oscilloscope that shows this issue?

When this issue occurs, are there any potential sources of pull-up current on the line besides the 2.7-kOhm resistor that you mentioned? Or, is it possible that the transceiver circuit has been disabled during this period? If that is the case, and if the output were low prior to becoming disabled, I would expect to see a relatively slowly-rising voltage on the output as the 1-nF capacitance is charged by the 2.7-kOhm pull-up resistance. (The transceiver output will be high impedance when disabled, and so it will not be able to quickly charge or discharge this capacitance as it would during normal operation.)

Regards,
Max

In reply to Max Robertson:

Hi Max,

You are right with the comparsion between A = Y and /A = Z  to the data sheet's nomencla-ture. I attached you first a photo from the encoder pcb with the AM26C31I and my circuit analysis sketch of the encoder circuit. I give you also some oscilloscope waveform captures of the erroneous behavior. There you can see, what it means when I describe it with "no proper low level". I'll hope this can clarify the problem I am facing. Regards, Burkhard

C1= Encoder signal from channel A of the motor encoder

C1= Encoder signal from channel A of the motor encoder

C1= A, C2= B Encoder signals from channels A and B of the motor encoder

C1= A, C2= B Encoder signals from channels A and B of the motor encoder

In reply to Burkhard Seyferth:

In reply to Burkhard Seyferth:

Burkhard,

This waveform is indeed strange. It looks to me as if the Channel A signal is periodically shorted to another signal (something that is driven to the VCC level, e.g., the power rail itself or the complementary output of the transceiver). I believe that it is an external short rather than something generated by the device since it seems that the abnormal behavior begins and ends with timing that is unrelated to the data being output from the transceiver. (This is best shown in your third waveform, which shows the "short" being introduced shortly after the first high-to-low transition and disappearing ~1.3 ms later in the middle of a low period.)

Is there anything else in this set-up that is turning on and off coincidentally with these period of abnormal signals? Or, is it possible that there are some intermittent shorts present in the cabling/connectors (perhaps exacerbated by the vibration introduced by the motor)? To simplify the debug, if not done already it may be good to try directly probing the encoder outputs with no other circuitry connected to see if this behavior is still present.

Regards,
Max

In reply to Max Robertson:

Max,
unfortunately the occurring of the error is very seldom, so investigations are very difficult. If you look to the zoom of the third waveform from my former answer, it seems to me, that the "distorted" driving low is not so strong (different time constants) as in normal case. What could be the reason for that? But I share your thoughts and if a periodically short to a low impedance source, like VCC, would be the cause, in this case the output of the AM26C31 should try to switch this source to low. For what reason ever the waveform plot shows, it cannot drive the low level in this case. If the reason would be the limited drive capability ot the AM26C31, then an excessive current should flow into the AM26C31. To exclude this, I measured the VCC current to the motor encoder and the GND current too. I would expect if the AM26C31 shorts a low impedance source with its output to GND and the capability of its output would exceeded, the current should flow into the output of the AM26C31 and back via encoder GND to the source. I'll hope you can follow my thoughts. But neither the VCC nor the GND current of the encoder showed abnormal values under normal and erroneous conditions. That leads me to the conclusion, that an "additional" resistor between the output of the AM26C31 and my pull-up resistor (2k7) would lead to the recorded waveforms. The value of the additional resistor could have the magnitude of 5k5. I added you an overview sketch of my arrangement. The error is unfortunately not limited to one single device. I am facing several error messages from the field, where some devices seem affected with this failure. At the moment I can't do anything because the error on may setup is not present. Devices delivered back from the field, don't show the failure when I try to do further measurements. Any further suggestions to find the root cause of the behavior are welcome. Regards, Burkhard

C1= Mainboard Signal TABLE_ENC_B @ R118

C4= GND current flowed back from NOE1 to BExtension Board, Cable NOE1 to BExtension Board -> wire1,8 and10 in current clamp (20 mA/Div)

C1= Mainboard Signal TABLE_ENC_B @ R118

C4= VCC current flowed to NOE1 from BExtension Board, Cable NOE1 to BExtension Board -> wire 9 in current clamp (20 mA/Div)

In reply to Burkhard Seyferth:

NOE1 is the motor encoder which uses the AM26C31 as line driver.

In reply to Burkhard Seyferth:

Burkhard,

Checking the VCC and ground currents was a good idea, and if there were an issue related to a periodic short to a low-impedance source then I agree some deviation in one of these currents should have been observed between normal and abnormal operational states.

I notice in some of your screenshots that the abnormal behavior differs from the normal behavior in a few ways other than the higher "low" level. It seems there is more noise in general, perhaps a slightly slower rise/fall time (as you pointed out with respect to the time constant), and in some cases it appears there is more cross-coupling from the adjacent channel (i.e., the transitions on the adjacent channel result in spikes on the problem channel, most likely due to AC coupling of the relatively fast edges). I'm unfortunately not sure yet what exactly this may imply, but I thought I would point it out in case it brings you any ideas. Perhaps these signals are being somehow subject to a higher series impedance rather than a low shunt impedance (partial short) as we had previously thought. If it were a persistent issue it would be straightforward to debug (most likely as physical damage to the IC or PCB or an assembly issue), but given that it only appears periodically it is hard to point to a likely root cause.

In these failing systems, have you noticed the incidence of this abnormal behavior correlating to anything else? For example, is it coincident with switching of the other four signals on the extension board? Or, have you tried inducing the failure through mechanical manipulation, e.g., by stressing different cables and connectors, varying the speed of the motor, etc.?

Is it possible to probe the complement to the problem signal at the same time? I am curious if it is behaving normally, if it is showing higher low levels, or if it is behaving "complementary" by having lower high levels.

Another test may be to try removing R118 or increasing its resistance, then seeing the effect on the abnormal low level. That would give us some idea of the source impedance that is driving this ~2-V level.

Are you able to check the problem signal when the main board is disconnected, and then with the extension board disconnected? That may better help contain the debug efforts as well.

I'm sorry at this point I don't have a stronger theory on the root cause, but hopefully with a little more debug things will become clearer.

Regards,
Max