Failed Crystal in Heavy Sixer, Anyone?

[+DrVenkman]

This past week, I took possession of an interesting Taiwan-made Heavy Sixer. On powering up with two known good PSUs and any cart I care to test with, I get a black screen. I have verified good and steady power from the voltage regulator, and stable Vcc and Vss at each chip. I have removed and tested each main IC in another system and they are all working fine.

 

When the basic steps above failed, I pulled out the scope. Checking the OSC input at pin 11 of TIA, I see nothing except a steady 1.14V. No signal at all. I've pulled out the oscillator schematic from the Field Service Manual for reference below.

 

 

I've verified that none of the caps or diodes are shorted, and verified continuity between adjacent components in the circuit. All the resistors save one measure well within spec. The only out of spec resistor is R203, which is supposed to be 1.8K but measures 635 ohms or so. I don't know off-hand if that resistor being that far out of spec would cause the oscillator circuit to fail to start, and I don't seem to have any 1.8K resistors. But I will have replacements in hand tomorrow and swap it just in case. I do have spare 2N3906 transistors, so I have removed and replaced both of them in the oscillator circuit, but it made no difference. For what it's worth, both old ones tested fine in my tester once I got them out of the board. 

 

That seems to leave the crystal itself. That strikes me as a bit odd. I've sent and received many retro systems by mail for decades now and never had a failed crystal but who knows? Has anyone else run into a failed crystal in an Atari before? 

[+DrVenkman]

I received a replacement crystal today from Console5 and sure enough, the Heavy has come back to life. 

 

OSC signal into pin 11 of TIA:

 

Pih0 at pin 39 of the 6507:

 

And Phi2 coming out of pin 40 of the 6507:

 

Unfortunately, I am not getting a good image from the system:

 

 

Here's a quick video clip:

 

NOTES:

1. The system would not boot at all prior to crystal replacement.

2. CPU, TIA and 6532 all work in other systems.

3. This TV usually accepts almost any RF signal; I use it for vintage console repairs routinely.

4. I've cleaned and cycled the color pot. and reflowed the solder joints.

5. I've cleaned and cycled the channel select switch and it does work; same results for Channels 2 and 3.

5. I've reflowed all the solder joints for the RF coil, the audio coil and the all the pins on the RF modulator, and RF coax jack.

 

[+batari]

From what I learned recently, the actual spec for TV colorburst is 3.579545 MHz +/- 10Hz, but most TVs can handle a few hundred Hz either way. You are at -900 Hz, and loss of colorburst is clearly caused by the vastly incorrect frequency.

 

Incorrect frequency can be caused by incorrect circuitry around the crystal. Crystals are pretty robust, so I am sure it didn't just go bad, but another failing part helped it to go. Obviously, the first thing to do is replace that resistor since it is clearly wrong, and go from there. I'd at least try a larger value if you do not have 1.8k.

[+DrVenkman]

14 minutes ago, batari said:

From what I learned recently, the actual spec for TV colorburst is 3.579545 MHz +/- 10Hz, but most TVs can handle a few hundred Hz either way. You are at -900 Hz, and loss of colorburst is clearly caused by the vastly incorrect frequency.

 

Incorrect frequency can be caused by incorrect circuitry around the crystal. Crystals are pretty robust, so I am sure it didn't just go bad, but another failing part helped it to go. Obviously, the first thing to do is replace that resistor since it is clearly wrong, and go from there. I'd at least try a larger value if you do not have 1.8k.

Hmm. You're right. Here's the replacement crystal I bought: 

 

https://console5.com/store/atari-2600-ntsc-color-crystal-oscillator-c015510-replaces-3-579575-mhz.html

 

What I'm getting on the scope is obviously quite a bit off. I do have 1.8K replacements now (and the first one I pulled out measured at 1,796 ohms - so WELL within spec!).  

 

I'm going out of town for an extended weekend trip for a couple days so I'm looking to seeing what the results are once that resistor is replaced next week. 

[+-^CrossBow^-]

Okay, now that I've seen a video of this and not just a still screen of the black n white only image, I think it could actually be the trimmer itself. I've had to replace 2 of these now in the past year. One in a 7800 that was only doing black n white on 2600 games, and another in a light sixer recently that was doing something similar to what you have shown. Actually it had colors that were very wrong and touching the trimmer would cause them to go nuts and sometimes go away completely. 

 

 

[+DrVenkman]

32 minutes ago, -^CrossBow^- said:

Okay, now that I've seen a video of this and not just a still screen of the black n white only image, I think it could actually be the trimmer itself. I've had to replace 2 of these now in the past year. One in a 7800 that was only doing black n white on 2600 games, and another in a light sixer recently that was doing something similar to what you have shown. Actually it had colors that were very wrong and touching the trimmer would cause them to go nuts and sometimes go away completely. 

 

 

I had that thought initially but it's clean and touching and adjusting it has no effect at all on the video or what's going on on the screen. With power off, I measured about 225K on one end of the scale up to about 360K on the other. I'm not opposed to changing it out if it proves necessary, but as Fred points out, even with a new (working!) crystal, the actual frequency still seems to be off. So once I get back from out of town, I'll keep investigating the discrete components around the clock circuit. 

[+DrVenkman]

Back from a few days out of town, I thought I'd take a look at this confusing Heavy Sixer of mine. First thing I did was remove R203 (the resistor in the clock circuit that seemed so out of spec). Out of circuit, it measures 1,793 ohms. WELL within spec for a 1.8K 1% resistor. So clearly that's not the issue. I reinstalled it since there's no point in swapping out a part that isn't bad. So next I tried replacing the new crystal I installed last week with another new one in case it's faulty (I bought 5 of them - spares are always good). This gives me the same exact reading. Huh.
 

Unfortunately, while I am an engineer (among other things), I am not an electrical engineer. What remains in my head from my minimal formal electrical engineering education of three and a half decades ago begins and ends with "V=IR." I'm certainly not capable of analyzing a full clock circuit and figuring out if it's possible to have a damaged component such that it alters the fundamental frequency the quartz crystal itself. 

 

So now my conundrum: how likely is it that more than one crystal would be bad out of the five I bought? Not very, in my experience in electronics repair, but who knows? The Field Service Manual flowchart doesn't even account for other possibilities if the crystal isn't in spec - it just says "Replace X200." Well, I've tried that but both times I get exactly the same reading. Could it be these are mis-labeled by the manufactuer? Dunno. I am going to ponder the schematics a bit more, but I may have to remove a vintage clock crystal from one of my spare 2600's and try it out. 

 

Anyone who knows anything about crystal oscillators and clock circuits, I'm all ears. 

[+DrVenkman]

I have now replaced every component in the clock circuit except the color pot (R211) and the rest of the resistors: the crystal (which was actually dead), the two transistors, the two diodes and all the ceramic caps. And the color pot functions exactly correctly per the voltage readings on pin 10 of TIA (the color phase DEL line) when rotated. Still no change. I even went so far as to remove the crystal from my daily driver 4-switch and install it in this Heavy. STILL no change - so we can rule out the idea of both new crystals being off-spec. SOMETHING about this Sixer is pulling the clock low and I have no idea what it could be.

 

So I am at a loss at the moment.  

[+nanochess]

I never have found a failed crystal.

 

I would suspect the TIA is having an "almost" short-circuit.

[+DrVenkman]

34 minutes ago, nanochess said:

I never have found a failed crystal.

 

I would suspect the TIA is having an "almost" short-circuit.

Already tested just fine in another 2600, as has the 6507 and RIOT. The original crystal was stone dead. Pin 11 of TIA was showing a steady 1.14V high level with no oscillation at all. 

[+DrVenkman]

I am completely flummoxed here.

 

I have replaced the dead crystal, but SOMETHING somewhere on this board is slowing down the clock by about 860 Hz and I cannot for the life of me figure out what's going on. All three main chips test fine in other systems and replacements into this board reflect the same symptoms. I realized I haven't done any chip-swap tests after rebuilding the entire clock circuit, so just be to sure it's 100% not the TIA, I pulled the TIA from a socketed 7800 that I *know* is working perfectly and tested it in this board again - same exact low clock frequency. So I put it back in the 7800 and before I buttoned up that console, I tested functionality while also measuring OSC at pin 11. Perfect. But in this Heavy Sixer, 860 Hz low. Remember, I've already tested a known-good vintage original Atari crystal in this same Sixer board and it, too, runs slow.

 

Just to be 100% the 4050 didn't have some kind of internal fault developing that might be putting increased load on the system somehow, I removed it and installed a socket, then replaced it. No change. And the original 4050 tests good in my MiniPro so that wasn't an issue anyway. 

 

At this point, I am going to put it in the bin for a few days until I have any other ideas.

[+splendidnut]

Is the crystal you bought the correct Resonance type for the circuit?  https://en.wikipedia.org/wiki/Crystal_oscillator#Resonance_modes

 

I once ran into a similar issue of wrong oscillating frequency for a crystal based oscillator circuit a few years back when experimenting with developing my own NTSC video display circuit.  I used the wrong circuit for the crystal I had gotten and it ran at the wrong frequency, similar to what you're seeing here.

 

Maybe you were sent the wrong crystal.  My knowledge is limited, I'm not an electrical engineer, so I'm unsure how you would identify which type you have.  Maybe someone here with electrical engineering knowledge could help you with that.

 

[+DrVenkman]

7 minutes ago, splendidnut said:

Maybe you were sent the wrong crystal

There's a link up-thread to the crystals I bought from Console5, which advertises them as replacements for the 2600. 

 

But aside from that, I even went so far as to remove the vintage crystal from my daily-driver 4-switch and transplant it into this Heavy, but it ALSO runs ~860 Hz slow. So there's something about this motherboard that is dragging down the oscillation rate. From what I vaguely remember from my pair of required EE classes in college, plus what I've been reading up on in the last few days, the resistor/capacitor values are the real "tuning" of current loads on the crystal such that it settles into its natural resonant frequency. I've replaced literally everything else in the clock circuit except the 6 fixed value resistors and the color pot, so it might come down to pulling those too, if for nothing else than to verify none of them are greatly out of spec. 

[+mytek]

I know people have been suggesting that your oscillator is not running at the correct frequency, but when I took a good look at your earlier scope readings it looks pretty darn close to me. I modified your photo and added some details in blue that show that it's on frequency.

 

I don't know about your scope, but the ones I have are not always accurate when they show the actual frequency reading vs. the time period on the scope. So I wouldn't put a lot of faith in that F= reading.

[+DrVenkman]

1 hour ago, mytek said:

I know people have been suggesting that your oscillator is not running at the correct frequency, but when I took a good look at your earlier scope readings it looks pretty darn close to me. I modified your photo and added some details in blue that show that it's on frequency.

 

 

I don't know about your scope, but the ones I have are not always accurate when they show the actual frequency reading vs. the time period on the scope. So I wouldn't put a lot of faith in that F= reading.

 

So here's my reading from Pin 11 of TIA with a working test system (a working, conveniently socketed 7800):

 

 

By contrast, with that system's TIA transplanted into my troublesome Heavy:

 

Both pictures were taken this morning as I continued trying to figure out why the clock is running slow.

 

 

[+batari]

Capacitance is what changes the frequency. But active elements like transistors can add variable capacitance so it nay not be as obvious as the capacitors c202,203 or 209, but you could try putting another cap in parallel to

them to see how frequency changes. 
 

But looking at the circuit - and noting that while I do have an EE degree, analog electronics was not my specialty, but I will examine the circuit later when I have a chance. 

[+batari]

Ok, I think this circuit is a variant of a voltage controlled crystal oscillator. An applied voltage exerts a small amount of control over the frequency of the crystal.
 

this being said check the voltage output of the 7805 under load and at the various points at the circuit marked 5v and also check the grounds. 
 

if this doesn’t help I’ll look at it later. 

[+mytek]

5 hours ago, DrVenkman said:

Both pictures were taken this morning as I continued trying to figure out why the clock is running slow.

Neither one is running slow. They both have the same time period. They also both have the same peak to peak voltage. What is different is the waveform shape, with one being a sawtooth pattern, and the other being closer to a sine wave. The sawtooth being more digital in nature is easier for the scope to count, thus the reason that its F= number is pretty much spot on, whereas the more sinusoidal wave form being more analog in nature also makes it more difficult to accurately count.

 

Anyway what I also see is that the sinusoidal wave form isn't perfectly analog, and you can still see it's somewhat buried sawtooth origin. However I think that's more likely being caused by in circuit capacitive loading across the oscillator output on the heavy sixer. As to where that is coming from I'm not exactly sure. But I think that should be the focus, and not whether the clock is slow or fast in reality. The time period doesn't lie in this case -- a repeating 279ns cycle is 3.57954 Mhz whether it's a square wave, triangle wave, sawtooth wave, or sine wave. The scope's frequency counter just happens to be more accurate with a more digital waveform and not so good with an analog one (my scope is like that).

 

When I get a chance I'll take a look at the schematic to see what else could potentially cause a video problem, and/or explain the waveform difference. If the oscillator circuits are different in design, it could just come down one having a lower drive capability and the capacitive loading is coming from the TIA itself, with the end result being to round off the edges of a sawtooth waveform making it look like a sine wave.

 

One last thought is that the scope probe could be the source of capacitive loading. Unless I'm mistaken it looks like the probe might be set to 1X. if this is true, then the probe is adding quite a bit of capacitance in parallel. Normally it is better to use the 10X setting to minimize the probe's capacitive loading of the circuit it is attached to.

 

Well I lied I have one more thought. No matter the shape of that waveform, if it's 3.57954 Mhz and has good amplitude which they both do, then the TIA should be happy. And just for a bit more info, the clock on an 8-bit is pretty much a square wave, with some being a 50/50 duty cycle, and some not. But for the Heavy Sixer I'm thinking that the oscillator is not the problem that is causing the bad video. But I could be wrong

[+mytek]

On 6/2/2023 at 1:36 PM, DrVenkman said:

I had that thought initially but it's clean and touching and adjusting it has no effect at all on the video or what's going on on the screen. With power off, I measured about 225K on one end of the scale up to about 360K on the other.

Here are snippets from the two schematics I've found thus far.

 

 

In both cases and just like on an 8-bit, the color delay trim pot is 500K. Your readings added up to 585K which is probably pretty normal for a 40 year old 500K trim pot. But I believe you mentioned not seeing any affect on the video when you turned it. I wonder if it's truly going from zero to 500K when you do that? Be easy enough to check. Also might be very informative to take a voltage reading on TIA's pin-10 on both your working and non-working 2600. If those readings are radically different that could suggest a problem with the color delay trim circuit, and it might not necessarily be a problem with the trim pot itself. It could simply be that the voltage being fed to one end of the trim pot is not correct due to some other component being out of spec, open, or shorted.

 

BTW, the oscillator circuits are different on both of those schematics, with one only being a single transistor and the other a more conventional 2 transistor circuit. So I really wouldn't expect identical waveforms, or the same drive capability. In fact the 2nd circuit with the single transistor would probably be very easily influenced by capacitance added to pin-10 since one side of the crystal appears to have less isolation from that pin (EDIT: but I could be wrong about that, and the opposite might actually be the case).

[+batari]

4 hours ago, batari said:

Ok, I think this circuit is a variant of a voltage controlled crystal oscillator. An applied voltage exerts a small amount of control over the frequency of the crystal.
 

this being said check the voltage output of the 7805 under load and at the various points at the circuit marked 5v and also check the grounds. 
 

if this doesn’t help I’ll look at it later. 

Thought I would follow up.

 

I still contend that the frequency is off too much to make NTSC happy. A few hundred Hz difference can cause a loss of color. I don't think the pot adjustment is the problem - I believe all the pot does is control the phase delay of the colorburst signal.

 

So the left half appears to be a variant of a Butler crystal oscillator circuit. The left transistor looks to be configured as an emitter follower.

 

The right transistor is not strictly a part of the Butler circuit, and looks like an emitter follower, which helps amplify the output of the crystal to TTL levels. That seems to be working, but the waveform shape difference is odd as mentioned earlier. The one on the broken console looks sinusoidal, while the one on the working console appears to me like a RC charge/discharge profile of a resistor/capacitor. The capacitors also can pull the frequency off which we are seeing. So I will repeat the advice to check these capacitors - even if just to shunt them with other values to see if the frequency of waveform changes.

 

Also, the load resistance on the transistors (100 ohm and 220 ohm) is pretty important for emitter followers - I would check those.

 

Lastly, the diode CH200 - I wouldn't just check it for shorts, check it with the diode tester on your multimeter.

[+DrVenkman]

I appreciate your thoughts, Michael and Fred. Great to hear from folks who may have some insights here. 

 

@mytek - the first post above has the Heavy Sixer clock circuit from the Atari VCS Domestic Field Service Manual and it matches the board I see in front of me. I’ve measured the voltages at pin 10 (DEL) on TIA as recommended by one of the steps in the FSM troubleshooting steps and didn’t uncover any obvious issues, so I think the color pot itself isn’t a problem. Neither end is shorted or failed open, though I don’t recall if I replaced C208 (the filter cap between TIA pin 10 and R211 (the color pot). FWIW, I will switch my scope and probe to 10X to see if it makes any difference in the waveforms measured. 

 

@batari - I’ve already replaced both diodes in the clock circuit with modern 4148s in place of the vintage 1N914’s. Both old diodes tested good out of circuit. I’ve also replaced both transistors with new, though the old ones tested good once removed. I’ve also replaced C202, C203, and C209. 

 

***

Anyway, I’m kind of out of ideas at this point. As I said, I don’t recall right now if I’ve replaced C208 but if I haven’t, I will do so after some coffee. Other than that, there’s nothing else I can think of except to lift each resistor out of circuit to get accurate resistance readings. I know none of them are failed open or shorted at least.

 

Fundamentally, SOMETHING killed that original crystal so it is not out of the question that something else may have been damaged at the same time - I just can’t figure out what it might be. 

[+mytek]

I don't think the difference in wave form shape really has any significance, since

14 hours ago, DrVenkman said:

So here's my reading from Pin 11 of TIA with a working test system (a working, conveniently socketed 7800):

It was late when I was reading the posts, and I missed the fact that one scope image was taken from a 7800. And after having my breakfast today and feeling a bit more awake, I looked at the schematic for a 7800 and see that the TIA oscillator input on pin-11 is being derived from the Maria chip's pin-7 which is labeled as TIACLK on the schematic via a 100 ohm series resistor.

 

Whereas the 2600 H6 reading is coming out of a 2 transistor crystal oscillator circuit.

 

When I said that an 8-bit oscillator would be a square wave I was mistaken. Basically any crystal oscillator built from one or two transistors is going to give you a sine wave output or pretty close to it. So the wave form that was seen for the 2600 H6 is perfectly normal.

 

The 7800 Maria derived TIACLK output could be generated any number of ways, so it's difficult to know what its wave form should look like. However if it's more logic gate created, it very well could be triangle or even square wave in shape -- something more digital. For instance in the 576NUC+, schmitt-trigger inverters are used in it's crystal oscillator circuit, so we get a nice square wave clock output. Since a stock 8-Bit has a sine wave clock, it just goes to show you that the GTIA doesn't care about what the wave form is, so long as the frequency is correct, and I suspect this holds true for the TIA as well.

 

As I mentioned earlier, not all oscilloscope based frequency counter measurements are going to be spot on accurate. this depends a lot on the method the O-Scope uses to get this reading. An expensive Tektronix or Keysight scope will have dedicated frequency counter hardware built-in which can be trusted, whereas less expensive scopes rely on some kind of software interpolation. Anyway based on the time period that was seen, it really doesn't appear that the 2600 H6 clock is all that different than the 7800's TIACLK. Just an observation based on the photos you took. If your scope has a screen capture feature, which it appears it does since I see a USB port, could you do a screen shot of both the 7800 and the 2600 and post them?

[+DrVenkman]

1 hour ago, mytek said:

If your scope has a screen capture feature, which it appears it does since I see a USB port, could you do a screen shot of both the 7800 and the 2600 and post them?

It does but they're fairly low-resolution, just the actual physical dimensions of the LCD screen. I'll try to make some later this afternoon/evening if I have time to get back to the bench. If not I can do it tomorrow after work. I should also mention that I have also scoped OSC on my 4-switch (which works perfectly). The only reason that hasn't been my go-to test system for everything is because it was the first console I installed a UAV into (one of Bryan's first rev C boards) and I tacked all the COL, BLK, and LUM signal wired to TIA's pin shoulders. But I can capture a waveform from that clock too (which measures the correct 3.57945 MHz or close enough. And to reiterate, the vintage crystal from THAT system results in the same low clock transplanted into this Heavy. 

 

EDITED: This is the clock circuit on a 4-switch, btw. Substantially different than that of the Sixers, and I noticed the color DEL line isn't tied to the clock directly at all.

***

 

***

Anyway, for the sake of clarity about what I've done to the Heavy, I had already replaced C208 (the filter cap on the DEL line to TIA pin 10. And voltage readings at pin 10, with the system running, range from 0.000V at one extreme to 7.97V at the other - the color pot is working fine. 

 

The only thing I noticed today is that compared to yesterday, my waveforms are much closer to square waves, at least on the bottom half of the oscillation. I haven't changed anything on the board but I did go over a lot of dodgy 45 year old solder - this very early Taiwan-made board is rough compared to later-made Light Sixers and 4-switches. They clearly were working on production quality initially. Most compoent legs untrimmed, lots of solder joints with gaps behind component legs, and the empty pads for the planned ROM chip (empty on all production Sixers) are only about half-filled - very poor wave-soldering. On my Light and (Sears) Heavy, this kind of thing is done perfectly, as it is on my later Taiwan-made Ataris.  So I did spend time this morning reflowing nearly every solder joint on the back of the board, just to rule out any possible dodgy connections or cold solder joints as a contributing factor.

 

I also noticed I'm getting a bit of static on the audio, and touching/pressing the silver audio caps makes it worse. Those have nothing to do with the clock, but I guess they can affect the overall capacitance of the board. I'll end up changing them too if I have spares - I should, since I replaced them on a Vader of mine several years ago but I'll have to look.

 

***

Much as I hate to do it, I'm halfway tempted to open up my Light Sixer for some closer comparisons. The boards are almost identical anyway, and at least I could make 100% apples-to-apples comparisons. 

[+bent_pin]

I'd just make a new clock circuit on a daughter board, then pull the components for the existing clock so you can tap it in.

[+mytek]

1 hour ago, DrVenkman said:

This is the clock circuit on a 4-switch, btw. Substantially different than that of the Sixers, and I noticed the color DEL line isn't tied to the clock directly at all.

1 hour ago, DrVenkman said:

And voltage readings at pin 10, with the system running, range from 0.000V at one extreme to 7.97V at the other - the color pot is working fine. 

It's actually quite ingenious what they did, combining the charge pump with the system oscillator. On the 8-Bit they kept the two aspects separate, however requiring more components and complexity. that is also why you see an upper range of the color delay being nearly 8 volts and not 5 VDC. This is pretty close to what the similar pin on the GTIA would see. So I agree, nothing wrong with that part of the circuit.

 

11 minutes ago, bent_pin said:

I'd just make a new clock circuit on a daughter board, then pull the components for the existing clock so you can tap it in.

I was thinking something similar, but maybe using a clock oscillator in a can. Something that you know is outputting the precise frequency desired. However once i saw that the color delay trimmer was integrated, that could be tricky. However per your suggestion, creating the exact same circuit off board and then patching it in place of the original might help to figure out what's going on.