Headphones for an F18A console

[Asmusr]

On my F18A I only use the 5 pin DIN connector for the audio, so I tried to modify some old headphones to connect to this. However, the sound is EXTREMELY load. Should I just add a resistor to the circuit?

[JasonACT]

Add 2:

Audio>----^^^^^--+-->Earphones
            R1   |
                 Z
                 Z R2
                 |
                 V
                GND

A guess (but what I've used before for audio) would be R1=6.8K & R2=2.7K but R2 being smaller still, will lower the volume more.  A potentiometer for R2 is even better of course.

[matthew180]

The TI's audio output is "hot" line-level, and designed to drive some sort of coupled amplifier stage, etc..  Audio line-levels are typically terminated with a 10K resistor to ground to develop the signal, then AC or DC coupled (lots of circuits out in the wild, they are all different) into the input stage for volume control, mixing, or whatever.

 

I'm not sure about driving headphones, but at the very least you will want to provide the 10K impedance resistor, and probably AC couple the signal, then attenuate with another resistor.  See attached circuit.

 

I measured the TI's output, as well as a typical PC audio out, and the ColecoVision's audio output.  Additional info on WikiPedia:

 

Audio In levels: https://en.wikipedia.org/wiki/Line_level

 

0 dBu ref level: 2.19Vpp
-10 dBV line level: 0.894Vpp

 

Measured at full volume:
PC line out: 1.8Vpp, +/- 900mV
ColecoVision at C88 output: 1.68Vpp, +900mV, -780mV
TI-99/4A at L201 output: 1.8Vpp

 

The 1.8V is *hot* for an audio line level, yet sampled sources regularly output higher than the consumer -10 dBV level.  C28 in the attached circuit helps reduce high frequency noise on the input.  Measured inputs (TI and CV) had noticeable "fuzz" on the input signals.

 

TI's audio output is notoriously noisy.  I did some digging into the board with my O-scope and concluded the noise is on the input power to the SN76489 IC, probably due to its proximity to the also noisy system GROMs.  All the filtering of the output audio will never remove the noise, it needs to be eliminated at the power input of the audio IC.  Basically just really bad design / placement of the audio chip.

 

Usually the speakers attached to the system become a natural low-pass filter that help remove the noise, but headphones...  I'm not sure how your audio experience will be in that case.

 

 

[+arcadeshopper]

On 10/1/2023 at 9:40 PM, matthew180 said:

The TI's audio output is "hot" line-level, and designed to drive some sort of coupled amplifier stage, etc..  Audio line-levels are typically terminated with a 10K resistor to ground to develop the signal, then AC or DC coupled (lots of circuits out in the wild, they are all different) into the input stage for volume control, mixing, or whatever.

 

I'm not sure about driving headphones, but at the very least you will want to provide the 10K impedance resistor, and probably AC couple the signal, then attenuate with another resistor.  See attached circuit.

 

I measured the TI's output, as well as a typical PC audio out, and the ColecoVision's audio output.  Additional info on WikiPedia:

 

Audio In levels: https://en.wikipedia.org/wiki/Line_level

 

0 dBu ref level: 2.19Vpp
-10 dBV line level: 0.894Vpp

 

Measured at full volume:
PC line out: 1.8Vpp, +/- 900mV
ColecoVision at C88 output: 1.68Vpp, +900mV, -780mV
TI-99/4A at L201 output: 1.8Vpp

 

The 1.8V is *hot* for an audio line level, yet sampled sources regularly output higher than the consumer -10 dBV level.  C28 in the attached circuit helps reduce high frequency noise on the input.  Measured inputs (TI and CV) had noticeable "fuzz" on the input signals.

 

TI's audio output is notoriously noisy.  I did some digging into the board with my O-scope and concluded the noise is on the input power to the SN76489 IC, probably due to its proximity to the also noisy system GROMs.  All the filtering of the output audio will never remove the noise, it needs to be eliminated at the power input of the audio IC.  Basically just really bad design / placement of the audio chip.

 

Usually the speakers attached to the system become a natural low-pass filter that help remove the noise, but headphones...  I'm not sure how your audio experience will be in that case.

 

 

 

so if I were to take my console apart, lift that power pin and put something there to filter the noise..what would be the something?

 

[HOME AUTOMATION]

I'd start with a battery, in order to determine the best case scenario.

[+FarmerPotato]

First step: decouple  chips by adding a 0.1 uF cap between the chip's VCC pin and the nearest ground. I would go under the PCB, locate the 76489 and solder the cap legs to pins 8 and 16. This provides electrons (current) when the chip needs it most. That is, when signals are changing high-low.

 

That said, I haven't measured on the console. 

[Tursi]

Not sure I'd like a properly silent TI anymore, though.

 

 

[matthew180]

A battery test might be a good way to test the theory.  More decoupling of the power pin of all the chips in the vicinity might help.  Noise is a hard problem, and the best solution would be to move the audio circuits to and isolated part of the PCB, well away from the digital parts.  The placement of the audio IC would suggest that the TI designers were not even remotely concerned with audio noise in the 4A, or they did not know better...

 

Since the GROMs can be anywhere in the system, one test I thought of would be to move the system GROMs to a cartridge.  That removes those chips from being close to the audio IC, so it would be a good data point.

 

 

[Tursi]

I have run the console with no GROMs installed in it back when I was testing the UberGROM, and I don't remember things being very different. But, I wasn't paying attention to sound so that's a weak assertion.

 

I'm pretty sure that the noise is a combination of most of the TTL logic. While the Classic99 reproduction is not 100% accurate, just ticking on the vertical blank and the keyboard column scan is very reminiscent. I'm 100% convinced on the keyboard part, since it nicely reproduces the noise during key and program processing.

 

Though now that I think about it, both of those run via the 9901. Maybe it's the noisy part. I figured the keyboard noise was the 1-of-8 used to decode the column, but maybe not...

 

 

[Tursi]

 it occurred to me I could use a signal tracer to see where the console was noisiest. So I took a video:

 

I'm back to believing it's the TTL circuitry.  The keyboard makes a lot of noise, too (but, it's being fed from a TTL chip). The worst offenders arguably seem to be two ICs near the edge port - the 74LS244, which is just an output buffer for the 8-bit data bus, and the two 74LS367s, which I hadn't previously paid much attention to. Looks like they are also output buffers, though these ones can tri-state. These appear to be all the control and address lines. 

 

I was surprised to hear 60hz continuing past the CPU into the DC power lines. I wonder how noisy they are on a scope?

 

I did not have a 9918A in there, I was running the F18A MK1, so I don't have a read on how much noise it adds.

[matthew180]

Ha, I forgot about the 9901 and the keyboard scanning.  Very noisy circuits, and clearly affecting a lot of the other chips including the audio IC.  The output audio trace also has to run from the audio IC all the way over to the output jack, which means passing close to the 9901 as well.  TI certainly had noise issues with the board, hence the requirement for the metal top and bottom shields.  I don't think there is any simple fix to this problem.

[+arcadeshopper]

On 10/3/2023 at 2:13 AM, Tursi said:

Not sure I'd like a properly silent TI anymore, though.

 

 

this would probably just be for the obnoxiously loud ones

 

[+Vorticon]

I recently connected a small speaker to my spare console and was getting all sorts of noise output which varied depending on what the program was doing. I was planning on trying Matt's circuit above but then I found out that I kind of liked being able to hear the program activity as it ran. Go figure...

[+FarmerPotato]

3 hours ago, Vorticon said:

 but then I found out that I kind of liked being able to hear the program activity as it ran. Go figure...

Hearing the noise output on the TV was fascinating to me as a kid. You could "hear" the cursor blinking, the keyboard, and the processor shuffling as you did a RESequence. 
 

@Tursi thanks for the thought-provoking noise tour! Did you do more with the scope, to confirm if the bus drivers are the  worst offenders? What about measuring the noise with the external bus drivers removed completely?
 

There's a TI Application Note in the Advanced CMOS Designer's Guide, explaining what TI did to tackle switching noise from bus drivers. The primary symptom to treat was undershoot during fast switching edges. While our LS and NMOS console is not in the same league as 74HC chips, I thought there might be some lessons here. 

 

TI found that inductance inside the chip carrier frame was the biggest contributor that could be eliminated.  

 

The so-called "TI Solution" was to move VCC and GND pins to the center, which shortens their internal leads to the chip, minimizing the inductance! Second, add more ground pins, to put a return path next to more signals.  Smaller packages help too.

 

 

 

Hence the pinout of chips like 16245 with lots of grounds. 

 

 

 

 

I wonder if it would help to replace the 244s and 367 with newer chips? Less power consumption should help right? (But switching time..) Maybe investigate using the 74HCT equivalents (I don't know) or 74AHCT  (Too pricey). At an extreme, hack  up a solution where a tiny PCB replaces the drivers with available 16245 (might defeat the purpose). 

 

I'm just barely grasping this stuff so I could be wildly wrong, but I just wanted to share the CMOS Designer's Handbook insight. 

 

[Tursi]

Neat stuff! No, I don't have my scope unpacked at the moment, so testing is a hassle