TMS9918 on the H8 bus .... alive and well

119 views
Skip to first unread message

Terry Smedley

unread,
Sep 10, 2021, 10:34:15 PM9/10/21
to SEBHC
I assembled the RC2014 Z80 version of a TMS9918 video card (https://github.com/jblang/TMS9918A) and hung it on the H8 bus with a little help from the H8-7 breadboard card.

It passes the NOGDS HA-8-3 video diagnostics. This is a picture of the "KALEIDO" test/demo program running.  

This establishes that the static VRAM method will work fine on a HA-8-3 clone card.  So it's on to the AY-3-8910 section of the card.  

Terry


RC2014BD.jpg
KALEIDO.jpg

Dave McGuire

unread,
Sep 10, 2021, 10:50:46 PM9/10/21
to se...@googlegroups.com
I've been watching this with interest. Very nice work!!

-Dave

--
Dave McGuire, AK4HZ
New Kensington, PA

Norberto Collado

unread,
Sep 11, 2021, 1:20:52 AM9/11/21
to se...@googlegroups.com

WOW! Very nice and I’m impressed. After seen the H-7 breadboard, it gave me the idea of building one out. I just tested the breadboard I ordered on a spare H8 board and it sizes nicely.

 

A picture containing text, electronics, circuit

Description automatically generated

 

Thanks,

Norberto

--
You received this message because you are subscribed to the Google Groups "SEBHC" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sebhc+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sebhc/5aad11cb-eb66-4054-bcbf-d8b5289a16c8n%40googlegroups.com.

Terry Smedley

unread,
Sep 11, 2021, 3:14:31 AM9/11/21
to SEBHC
And we have bombs, lasers, and race cars from the AY-3-8910, which was very easy to hang on the bus compared to the video chip. 

Norberto,  I am sending you a package of parts so you can duplicate my experiments.

Terry

Glenn Roberts

unread,
Sep 11, 2021, 5:07:19 AM9/11/21
to se...@googlegroups.com
I’ve got an unbuilt board, just the pc board, no parts. Would you like that? I’ve also got the documents if we don’t have those scanned…

Sent from my iPad

On Sep 11, 2021, at 1:20 AM, Norberto Collado <norberto...@koyado.com> wrote:



WOW! Very nice and I’m impressed. After seen the H-7 breadboard, it gave me the idea of building one out. I just tested the breadboard I ordered on a spare H8 board and it sizes nicely.

 

<image001.jpg>

 

Thanks,

Norberto

 

I assembled the RC2014 Z80 version of a TMS9918 video card (https://github.com/jblang/TMS9918A) and hung it on the H8 bus with a little help from the H8-7 breadboard card.

 

It passes the NOGDS HA-8-3 video diagnostics. This is a picture of the "KALEIDO" test/demo program running.  

 

This establishes that the static VRAM method will work fine on a HA-8-3 clone card.  So it's on to the AY-3-8910 section of the card.  

 

Terry

 

 

--
You received this message because you are subscribed to the Google Groups "SEBHC" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sebhc+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sebhc/5aad11cb-eb66-4054-bcbf-d8b5289a16c8n%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "SEBHC" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sebhc+un...@googlegroups.com.

Les Bird

unread,
Sep 11, 2021, 8:00:27 AM9/11/21
to SEBHC
Wow Terry! This is fantastic news.

Les

Terry Smedley

unread,
Sep 11, 2021, 8:26:59 AM9/11/21
to SEBHC
Les:

Are your games built using the .ACMs that were supplied with the NOGDS card?  Because of the port conflict with the Z67 at 274Q-277Q, it would be nice to move the HA-8-3 ADC and APU to other ports.  I'm just wondering how easy (or difficult) it will be for your games to be reassembled to reference another port address.

Terry

Les Bird

unread,
Sep 11, 2021, 11:18:29 AM9/11/21
to SEBHC
Terry,

Looks like from the source that I'm hitting those ports directly, not using the ACMs. They could certainly be re-assembled but it's the games that we don't have source for that would not work. Can you make the board port configurable with a jumper or dip switch?

Les

Terry Smedley

unread,
Sep 11, 2021, 11:36:59 AM9/11/21
to se...@googlegroups.com
Les:

Sure, we can make the board so that it can be quickly reconfigured from one set of ports to another.  I'm guessing there are a fair number of users who have a Z67 at 274Q, and we'll want to have some way to avoid having to pull the controller to run with the HA-8-3.

I assume you are not using the APU in your games?

Terry





You received this message because you are subscribed to a topic in the Google Groups "SEBHC" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sebhc/4hlzmOiWoRk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sebhc+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sebhc/2709271b-8e65-433c-8313-ef3ce43f8fb4n%40googlegroups.com.

Les Bird

unread,
Sep 11, 2021, 1:14:43 PM9/11/21
to SEBHC
Terry,

That is correct, I am not using the APU in any of my games. I'm guessing the APU was seldom used due to the fact that it did not come pre-installed.

Les

Norby

unread,
Sep 11, 2021, 5:23:55 PM9/11/21
to se...@googlegroups.com
The APU is on the SSI board and port address can be changed if needed.

Sent from my iPhone

Terry Smedley

unread,
Sep 13, 2021, 2:35:06 PM9/13/21
to SEBHC
Here is a short clip of the "MUSICK" program from the HUG 885-1098 Color Graphics disk.  That program reads a .PLA file originally produced for the HA-8-2 Music Card and translates it into notes for the AY-3-8910 programmable sound generator used on the HA-8-3 card.  This was my test of the TMS9918 and AY-3-8910 working together without glitching.

The "MUSICK" program is interesting because it illustrates the clear difference in the way the HA-8-2 and HA-8-3 produce music.  The HA-8-2 music card is essentially a beautifully filtered pair of DACs.  The H8 has to generate every point on the complex waveforms for 4 channels.  It is extremely compute intensive - interrupts are shut off, and the PLAY program takes over all the available CPU resources.  But it is capable of producing very high quality waveforms and is not limited to pure sine waves.

Contrast that to the AY-3-8910 on the HA-8-3 card.  Once given a frequency, it will generate a waveform at that frequency until it is told to do something else.  There is little control over the waveform produced - everything played through the AY-3-8910 sounds a little like a video arcade - but it is very thrifty with the CPU.  The MUSICK program runs with interrupts on, updates the H8 front panel in time with the music and has time left over to draw a Kaleidoscope using the TMS9918A.  It is limited to 3 channels instead of 4 on the HA-8-2.  As implemented on the HA-8-3, all three channels are wired together for mono sound.  On the clone, those channels will be split so they can be run through a stereo amplifier if desired (or combined into a single mono signal for compatibility).  In a video game, separating the sound channels would allow synchronization of the video movement with the sound effects.

Terry
MUSICK.mp4

Les Bird

unread,
Sep 13, 2021, 2:55:47 PM9/13/21
to SEBHC
Awesome work Terry!

norberto.collado koyado.com

unread,
Sep 13, 2021, 3:09:11 PM9/13/21
to SEBHC
Very nice! Do we have the PACMAN game?

From: se...@googlegroups.com <se...@googlegroups.com> on behalf of Les Bird <lesb...@gmail.com>
Sent: Monday, September 13, 2021 11:55 AM
To: SEBHC <se...@googlegroups.com>
Subject: Re: [sebhc] TMS9918 on the H8 bus .... alive and well
 

Glenn Roberts

unread,
Sep 13, 2021, 4:09:25 PM9/13/21
to se...@googlegroups.com

The video quality I’m getting out of my original HA-8-3 is rather disappointing.  Probably in 1980 this was acceptable. On the clone board can we do anything to crispen up the image? 

 

As for crispening up the HA-8-3 output:

 

At previous VCF shows I’ve seen some very nice quality output from old composite video devices.  I think most of these used scan converters. They’re not cheap.  pretty sure one I saw that did a great job was the OSSC:

https://videogameperfection.com/products/open-source-scan-converter/

 

although that specifically says it doesn’t take in composite.

 

this one’s a little cheaper:

https://www.retrotink.com/

 

the really cheap ones don’t produce any better resolution than what’s built into many modern TVs and monitors that accept composite.

 

Anyone have direct experience and recommendations?

 

 

As for PacMan… I’ve found disappointingly little software for the HA-8-3.  Les’ games are some of the nicest ones I’ve come across.. 

Dave McGuire

unread,
Sep 13, 2021, 5:46:39 PM9/13/21
to se...@googlegroups.com
On 9/13/21 4:09 PM, Glenn Roberts wrote:
> As for PacMan… I’ve found disappointingly little software for the
> HA-8-3.  Les’ games are some of the nicest ones I’ve come across..

Games, games, games. Give me SCHEMATIC CAPTURE!! 8-)

Or maybe some nice plotting and graphing libraries. I've been toying
with the idea of designing a data acquisition board.

Terry Smedley

unread,
Sep 13, 2021, 6:20:21 PM9/13/21
to SEBHC

I agree that the composite video quality looks poor, at least by todays standards.   And I think Glenn's assessment of software for the HA-8-3 is also spot-on.  There's just not a lot of it there.

It's been noted that there are far more capable video controllers available today.  I do wonder, though, whether a 64K H8 can really feed at sufficient speed the enormous (in 64K terms) quantities of data necessary to do higher-res graphics.  And a bigger question is .... what 8-bit software is there for such a beast, and who will want to write anything new for it?

Seems worth asking what the objective is.  My personal goal was to experiment with a card that I hadn't previously had in my H8.  I'm OK living with its limitations because I wanted to see first hand and appreciate what creators like Les were able to accomplish with such limited hardware capabilities.   Maybe I could explain by saying that I know a '57 Chevy leaks oil, runs rough until it warms up, gets lousy gas mileage, handles like a very large boat, and has brakes that take a mile to stop.  But if I want to experience the thrill of riding in an iconic '57 Chevy, well,  those are the things I have to be prepared to live with.

If an objective is to be able resurrect and run, without change, whatever software there might be for an HA-8-3, then we are pretty much stuck with the TMS9918A and all that it does, and doesn't, do.  While there's not much software for the HA-8-3, there is *some*, and what there is seems to have significant historical value.

I plan to continue down the prototype/experimental path to come as close to the HA-8-3 as I can, but it's worth considering if there's any value in investing time in a full-blown production card that would have limited audience appeal.  It's quite possible that everyone who wants the thrill of riding in that '57 Chevy already has their own HA-8-3!

Terry

glenn.f...@gmail.com

unread,
Sep 13, 2021, 6:55:21 PM9/13/21
to se...@googlegroups.com

My objective is to display the TMS9918A output as crisply as possible.  I guess I’d be putting modern Michelin radials on my old Chevy.  I’m not looking to turn the H8 into a screaming CAD/CAM machine, just “de-fuzz” the lousy screen image I’m getting out of the card.

Les Bird

unread,
Sep 13, 2021, 8:20:47 PM9/13/21
to SEBHC
Glenn, I'm seeing the same thing here. When viewed live it's not that great but when taking video/photos the quality is fantastic.

I agree with Terry, practically no software for the HA-8-3 and probably not worth making a production board of it but nice that the option is there.

Norberto, the closest you'll get to Pacman is Munchkin on the H19 although at one time I was writing my own version of Pacman for the H19 that I called Heathman. I thought Munchkin was poorly done so was trying to make a better version. The game was never finished unfortunately but at one time I had a really good looking prototype. It's on one of my CPM Assembler disks (I think AssemblerB2.H8D) although this version crashes for some reason after it draws the maze. Here's a screenshot of the maze. I tried to mimic as closely as possible the original Pacman maze. Maybe one day I will finish it :)

Screen Shot 2021-09-13 at 8.09.39 PM.png

glenn.f...@gmail.com

unread,
Sep 13, 2021, 8:37:03 PM9/13/21
to se...@googlegroups.com

Led: I see what you mean. Maybe I’m too close to the screen (or I should take off my glasses!).  I’ve added some pix to my vcf album (scroll down…)

 

https://photos.app.goo.gl/Bqqr9Cn1ohJaCPyC7

 

the flag and the globe look great.  The kaleidoscope shows a lot of issues at the rectangle boundaries.  Is this about what you’re seeing?

 

From: se...@googlegroups.com <se...@googlegroups.com> On Behalf Of Les Bird
Sent: Monday, September 13, 2021 8:21 PM
To: SEBHC <se...@googlegroups.com>
Subject: Re: [sebhc] TMS9918 on the H8 bus .... alive and well

 

Glenn, I'm seeing the same thing here. When viewed live it's not that great but when taking video/photos the quality is fantastic.

 

I agree with Terry, practically no software for the HA-8-3 and probably not worth making a production board of it but nice that the option is there.

 

Norberto, the closest you'll get to Pacman is Munchkin on the H19 although at one time I was writing my own version of Pacman for the H19 that I called Heathman. I thought Munchkin was poorly done so was trying to make a better version. The game was never finished unfortunately but at one time I had a really good looking prototype. It's on one of my CPM Assembler disks (I think AssemblerB2.H8D) although this version crashes for some reason after it draws the maze. Here's a screenshot of the maze. I tried to mimic as closely as possible the original Pacman maze. Maybe one day I will finish it :)

 

image001.png

Dave McGuire

unread,
Sep 13, 2021, 9:23:29 PM9/13/21
to se...@googlegroups.com
On 9/13/21 8:36 PM, glenn.f...@gmail.com wrote:
> Led: I see what you mean. Maybe I’m too close to the screen (or I should
> take off my glasses!).  I’ve added some pix to my vcf album (scroll down…)
>
> https://photos.app.goo.gl/Bqqr9Cn1ohJaCPyC7
> <https://photos.app.goo.gl/Bqqr9Cn1ohJaCPyC7>
>
> the flag and the globe look great.  The kaleidoscope shows a lot of
> issues at the rectangle boundaries.  Is this about what you’re seeing?

Those "echoes" surrounding vertical lines sure look like reflections
on an unterminated transmission line. How are you feeding the video
into the monitor?

Les Bird

unread,
Sep 13, 2021, 9:24:24 PM9/13/21
to SEBHC
Glenn, yes but also my Warlords game shows a lot of fuzziness when viewed live. I don’t know why videos and pictures are so crisp though.

Les

Dave McGuire

unread,
Sep 13, 2021, 9:33:00 PM9/13/21
to se...@googlegroups.com
On 9/13/21 6:20 PM, Terry Smedley wrote:
> I agree that the composite video quality looks poor, at least by todays
> standards.   And I think Glenn's assessment of software for the HA-8-3
> is also spot-on.  There's just not a lot of it there.
>
> It's been noted that there are far more capable video controllers
> available today.  I do wonder, though, whether a 64K H8 can really feed
> at sufficient speed the enormous (in 64K terms) quantities of data
> necessary to do higher-res graphics.  And a bigger question is .... what
> 8-bit software is there for such a beast, and who will want to write
> anything new for it?

If there's not much software out there for the original board, it
would seem that supporting existing software isn't a very strong
motivation. But, there is some, and the argument can be made that it is
historically important.

But you're right that there are much better video controllers
available than the TMS9918. The S-100 crowd has built some nice boards
with the NEC D7220 chip, which is very capable. One of those designs
could be adapted to the H-8 bus.

If we'd be starting at least mostly from scratch with software anyway..

I myself have zero interest in games of any kind, on any platform,
but I love to see some technical graphics capability on vintage
hardware. As I mentioned earlier, I've been thinking off-and-on for
awhile about putting together a data acquisition board for the H-8.
With decent high-res graphics output, and a simple library of primitives
to drive it, some neat things could be done.

Other systems from the same era have no issues "pushing the bits" for
a functional display; I don't see why the H-8 would.

Terry Smedley

unread,
Sep 13, 2021, 9:42:54 PM9/13/21
to SEBHC
The images I see with the clone have the same edge artifacting that Glenn noted.  Has anyone looked at this on a vintage analog display?  Some of the edge artifacts look like they could be due to sharpening after upscaling to a digital display.  I experimented a bit by turning down the sharpness on the monitor, and at least when viewed from a few feet away, the picture looks better to me.  Instead of the edge ringing, there's a smoother transition which is easier on my eye.  But razor sharp, it's not.

Terry

Terry Smedley

unread,
Sep 13, 2021, 9:46:50 PM9/13/21
to SEBHC
Dave:

I'll pickup a D7220 for experimentation.

Are you aware of (the location of) a Z80 library of primitives?  (I haven't searched yet ... just looking for a starting point).

Terry

Dave McGuire

unread,
Sep 13, 2021, 9:47:53 PM9/13/21
to se...@googlegroups.com

It really does look like ringing. If the sharpness is increased,
there will be a faster rise time on the waveform, resulting in more
ringing. But something is causing those reflections in the first place.

I've seen this in the past using NTSC video monitors that had
switchable 75-ohm termination on their inputs, and the termination was
disabled. A 75-ohm driver sending fast edges into a high-impedance
receiver will certainly do this, due to the impedance mismatch.

-Dave
> --
> You received this message because you are subscribed to the Google
> Groups "SEBHC" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to sebhc+un...@googlegroups.com
> <mailto:sebhc+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sebhc/1640d909-3e4d-4202-99fc-6665f9d15e92n%40googlegroups.com
> <https://groups.google.com/d/msgid/sebhc/1640d909-3e4d-4202-99fc-6665f9d15e92n%40googlegroups.com?utm_medium=email&utm_source=footer>.

Dave McGuire

unread,
Sep 13, 2021, 9:48:57 PM9/13/21
to se...@googlegroups.com

Not offhand, no. I would like to write one myself, but that'd be
Another Project, sigh.

Maybe the S-100 or N8VEM/RBC crowd has some software that could be
ported.

-Dave
> --
> You received this message because you are subscribed to the Google
> Groups "SEBHC" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to sebhc+un...@googlegroups.com
> <mailto:sebhc+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/sebhc/4cbabdc0-0e41-4ed6-b503-4e0f706678b6n%40googlegroups.com
> <https://groups.google.com/d/msgid/sebhc/4cbabdc0-0e41-4ed6-b503-4e0f706678b6n%40googlegroups.com?utm_medium=email&utm_source=footer>.

Glenn Roberts

unread,
Sep 13, 2021, 9:52:15 PM9/13/21
to se...@googlegroups.com
The board has a 3-pin header that outputs the composite signal. It came with a coax cable that has a 3-pin Molex on one end and an RCA jack on the other. I simply plugged the RCA jack into the composite input on a Samsung LED tv.

To Terry's question: I have run this card using the old Zenith color monitor. My recollection is it's not much better.

I wonder if a better cable would help? impedance issue?

-----Original Message-----
From: se...@googlegroups.com <se...@googlegroups.com> On Behalf Of Dave McGuire
Sent: Monday, September 13, 2021 9:48 PM
To: se...@googlegroups.com
Subject: Re: [sebhc] TMS9918 on the H8 bus .... alive and well


To unsubscribe from this group and stop receiving emails from it, send an email to sebhc+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sebhc/927d88f8-a8f9-3135-f7f7-f8d6d0d36b2c%40neurotica.com.

Terry Smedley

unread,
Sep 13, 2021, 10:07:10 PM9/13/21
to SEBHC
Glenn:

The clone goes direct to a PCB mounted RCA jack, so the monitor signal is coaxial all the way.  I swapped between a ultra cheap 10' audio cable and a  2' RG59U coax and I can't honestly say that the image was any better one way or the other.

Both the clone and the HA-8-3 have 75 ohms across their emitter feeds.   I adjusted the video gain on the board from almost dark to eye scorching, and the edges look about the same to me.

Terry

norberto...@koyado.com

unread,
Sep 14, 2021, 12:23:43 AM9/14/21
to se...@googlegroups.com

Les,

 

As you are the expert on gaming, If you can port most of your games to work on this board + my Pacman and my wife’s Galactica, then we have a strong motivation to make a copy of it as Terry S. has done the ground breaking work already and everything else should be easy to put together. We can make a copy of the original board with additional improvements.

 

Terry S. can work on the joystick design and implementation if you provide information on such requirements to control your games.

 

I think we have a team in place to make this work well and don’t want to see this opportunity unexploited as we are getting more elderly as time passes by.

 

Thanks,

Norby

 

From: se...@googlegroups.com <se...@googlegroups.com> On Behalf Of Les Bird
Sent: Monday, September 13, 2021 5:21 PM
To: SEBHC <se...@googlegroups.com>
Subject: Re: [sebhc] TMS9918 on the H8 bus .... alive and well

 

Glenn, I'm seeing the same thing here. When viewed live it's not that great but when taking video/photos the quality is fantastic.

 

I agree with Terry, practically no software for the HA-8-3 and probably not worth making a production board of it but nice that the option is there.

 

Norberto, the closest you'll get to Pacman is Munchkin on the H19 although at one time I was writing my own version of Pacman for the H19 that I called Heathman. I thought Munchkin was poorly done so was trying to make a better version. The game was never finished unfortunately but at one time I had a really good looking prototype. It's on one of my CPM Assembler disks (I think AssemblerB2.H8D) although this version crashes for some reason after it draws the maze. Here's a screenshot of the maze. I tried to mimic as closely as possible the original Pacman maze. Maybe one day I will finish it :)

 

image001.png

Terry Smedley

unread,
Sep 14, 2021, 1:37:37 AM9/14/21
to SEBHC
Glenn:

I found a couple of tweaks that seemed to help the edge ringing.

The TMS9918 has an "External Video" input that is mixed with the internally generated video to produce the composite output.  That external input isn't used, and on the clone and I believe the HA-8-3 that pin is simply left unconnected.  I found an observable and reproducible improvement in video quality by grounding that external video input.  It is pin 35 on the TMS9918 (adjacent to the composite output at pin 36).  You might try jumpering pin 35 to ground and see if you see similar improvement.

With that improvement, what I observed is that edges now appeared "soft" or "smeared", rather than "ringy".  I had previously turned the monitor sharpness all the way down to try and relive the ringing edges.  Since I was now observing soft edges, I returne the monitor's sharpness adjustment to midrange.  The soft edges got considerably sharper, but with much less ringing that I had previously observed.

Finally, I saw in the TMS9918 datasheet a recommended interface to a video monitor that used no external components except for a 1K pull-down resistor.  Using patch cables that are ridiculously long given the frequencies of the video signal, I bypassed the external video amplifier transistor and routed the video signal straight from pin 36 of the TMS9918 to the monitor, with the 1K pull-down resistor (to ground).  I believe that improved things, but I can't gauge that for certain until I can patch the circuit board properly.

All of this is to say that I think there are some improvements possible by tweaking the video amplifier circuitry (Q1 through Q4 on the HA-8-3). What I know about high frequency amplifiers would only partially fill a very small thimble, so if there are any video experts out there who can look at the HA-8-3 schematic and offer suggestions - I'm all ears!  I'm looking for the ColecoVision schematic to see what was done on that high-volume commercial design.  

Terry

On Monday, September 13, 2021 at 6:52:15 PM UTC-7 Glenn wrote:

glenn.f...@gmail.com

unread,
Sep 14, 2021, 6:32:59 AM9/14/21
to se...@googlegroups.com

Thanks Terry. I applied the patch

https://photos.app.goo.gl/x81KA4RSvrqGDQUZ7

 

not sure this makes much of a discernable difference.  The main application that stresses the video is the MUSICK (kaleidoscope) program.  As I watch the kaleidoscope change on the screen it will go from an image that looks quite good to one that shows a lot of ringing simply based on the change of only a few cells. Perhaps the pattern at the edges determines the “ringing” for the whole image?

 

My recollection is that this was always a challenging graphic.  For my purposes I’m good with what we’ve got.  Will focus on having a few games and demos to show.

 

back in the 80s I was interested in high level language bindings for the HA-8-3.  I have a copy of Lucidata Pascal, which had full support.  It produces pseudocode which must be run with a PRUN executive.  I’ll take a look at that and see if they had any interesting demos on that disk.

Terry Smedley

unread,
Sep 14, 2021, 8:58:33 AM9/14/21
to SEBHC
Glenn:

I see the same shifting you are describing - from an image that's pretty good, to one that's got really tough edges.  

Makes me think about things like: AC hum on the output, an unstable crystal oscillator for the TMS9918.  

I'll continue to investigate.  

Terry

glenn.f...@gmail.com

unread,
Sep 14, 2021, 5:09:59 PM9/14/21
to se...@googlegroups.com

Getting some interesting graphics demos from the Lucidata pascal package

https://photos.app.goo.gl/HHpA6eqJguRfrtjB9

https://photos.app.goo.gl/FvdR3a5XqdV6eoAx9

 

pretty nice implementation.  good resolution and quality.

 

I don’t think I used it much back then as all I had was a dual (100K) floppy system.

 

Compiling and running at 2Mhz sure seems slow though!...

 

Is there an HDOS version of SPDSBC out there somewhere that I can grab?  This is for the Rev 3.1 CPU board.  If not I suppose I can create one from Douglas’ CP/M version…

 

Also do we know how fast we can push the HA-8-3 in terms of CPU speed?  I guess I’ll find out soon enough…

 

  • Glenn

terry.smedley

unread,
Sep 14, 2021, 5:24:58 PM9/14/21
to se...@googlegroups.com
Glenn:

I am away from home until Friday...but I'm sure someone can provide an HDOS link.  If you do from scratch, you'll need to write the changes to the shadow memory byte for the control port and the port will be updated at the next 2ms tick.

The HA-8-3 makes extensive use of wait states to meet timing constraints.  Unfortunately those wait states are based on counting bus clock cycles with shift registers so as the bus speed goes up, the delay goes down and will likely fail at higher speeds.  An HA-8-3 clone will use fixed timing waits that are independent of bus or CPU clock and higher speeds will be possible.  When running the experimental board without revised wait state circuitry, I could run video at 16MHz without crashing, but at any speed above 4MHz there were obvious display problems.  Let us know how far you are able to push the HA-8-3.

Terry

Sent from my Verizon, Samsung Galaxy smartphone
You received this message because you are subscribed to a topic in the Google Groups "SEBHC" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/sebhc/4hlzmOiWoRk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to sebhc+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sebhc/042e01d7a9ac%24de8053a0%249b80fae0%24%40gmail.com.

norberto.collado koyado.com

unread,
Sep 14, 2021, 5:29:57 PM9/14/21
to se...@googlegroups.com
HDOS Speed Application is here:




Let me know if any issues in case it is not the latest version.

Norberto
QSH89SBC Software Update (CPM BIOS with H17/H37 Support) - by Ken



From: se...@googlegroups.com <se...@googlegroups.com> on behalf of terry.smedley <terry....@gmail.com>
Sent: Tuesday, September 14, 2021 2:24 PM
To: se...@googlegroups.com <se...@googlegroups.com>

Les Bird

unread,
Sep 14, 2021, 6:21:10 PM9/14/21
to SEBHC
Glenn,

Those demos look awesome. Is this disk in the SEBHC library?

Les

Glenn Roberts

unread,
Sep 14, 2021, 9:39:51 PM9/14/21
to se...@googlegroups.com

No but I can upload these at some point.  Part of my own collection of disks. Pretty sure I don’t have anything else that’s not captured somewhere.

 

But Alex Bodnar recently scanned the disks from the Frazer (PA) HUG group. 40+ disks.  Looks like it’s time for Volume 10 of the archive?

glenn.f...@gmail.com

unread,
Sep 14, 2021, 10:05:21 PM9/14/21
to se...@googlegroups.com

When I try to increase the speed above 2Mhz the system just hangs.  I presume this is because of the presence of the HA-8-3 but there’s a simple way to test that – I’ll pull it out at some point to see if that’s the issue.  Too bad. These demos would look much nicer at higher speed.  Then again we are trying to demonstrate what life was like back in the 80s!

image001.png

terry.smedley

unread,
Sep 14, 2021, 11:54:58 PM9/14/21
to se...@googlegroups.com
Glenn:

The HA-8-3 does its wait states by counting clock cycles, so it's got the right timing only at 2MHz.  I was hoping you'd be able to eek 4 out of it.  

Small consolation, but the clone will not have that limitation.

FWIW, I see there is quite a market demand for old analog monitors by "classic gamers" who are trying to avoid all the digital artifacts associated with upscaling a composite video image to a HD display.  I believe you said you didn't recall the HA-8-3 looking any better on the Heath analog display.  A family member still has a Trinitron (!).  If I can get somebody to help me lift it into and out of my car, I will try to have a look at the HA-8-3 graphics on it.

Glenn Roberts

unread,
Sep 15, 2021, 3:22:08 AM9/15/21
to se...@googlegroups.com
I have a small 9” Sony trinitron and a zenith color (ZVM-135?) crt . I will take pictures of both but I think the Samsung flat screen I’ve been using is probably the best quality of the three.

Sent from my iPad

On Sep 14, 2021, at 11:54 PM, terry.smedley <terry....@gmail.com> wrote:



Kenneth L. Owen tx836519

unread,
Sep 15, 2021, 8:17:08 AM9/15/21
to se...@googlegroups.com

Glenn,

Possibly, if you have a spare pin, bring in the 2 msec clock and let that feed the counter by direct connection.

 

Sent from Mail for Windows

 

Norberto Collado

unread,
Sep 15, 2021, 10:49:38 AM9/15/21
to se...@googlegroups.com

Great idea Ken!

 

On Z80 4.0, the 2MHz clock can be added to the H8 bus by placing jumper JP30 and then the clock is on the H8 bus on pin 9. Lift pin U12 pin 11 on HA-8-3 and then solder a wire from pin 9 on the HA-8-3 H8 bus to U12 pin 11. This will keep a constant 2MHz clock regardless of CPU frequency.

 

Glenn Roberts

unread,
Sep 15, 2021, 11:26:38 AM9/15/21
to <sebhc@googlegroups.com>
Im using the 3.1 rev z80 on this machine. Wonder if that board can do that?

norberto.collado koyado.com

unread,
Sep 15, 2021, 1:11:18 PM9/15/21
to <sebhc@googlegroups.com>
The gate for Z80 V3.1 is not connected. For Z80 3.1 board do the following. Please use coaxial cable if possible. If not, wire-wrap cable attached with hot glue is fine on side 2 of the board.






From: se...@googlegroups.com <se...@googlegroups.com> on behalf of Glenn Roberts <glenn.f...@gmail.com>
Sent: Wednesday, September 15, 2021 8:26 AM
To: <se...@googlegroups.com> <se...@googlegroups.com>

glenn.f...@gmail.com

unread,
Sep 16, 2021, 8:59:07 PM9/16/21
to se...@googlegroups.com

Back to my graphics demo machine…:

 

So this version of SPDSBC (see email below) was originally designed for the H89 and includes code to force to 2ms clock on.  that was what was hanging my system.  I actually seem to have quite a few versions of “speed” utilities hanging around.  I’ve actually written my own version which is similar to ken’s but avoids setting the bit.  I’m happy to share it but worried about Yet Another Speed program propagating in our space!

 

I’m perpetually perplexed about the 2ms clock and the Rev 3 and Rev 4 CPU boards – no doubt my own fault for not sitting down and trying to fully understand.

 

Anyway as long as you don’t set the 2ms mask this is now working. Great (also works great  on my new H8-2020 setup under HDOS 3!)

 

 

The GOOD NEWS is I am now able to run the HA-8-3 graphics card just fine at 4 mhz, at least with everything I’ve tried so far. This makes a very noticeable difference.  I can also push the CPU all the way to 10Mhz to do compiles, which is a great  savings.  Lucidata is no Turbo Pascal!!

 

When I try to run the graphics demos at 8Mhz or 10Mhz I do get odd artifacts on the screen.  I have not implemented the suggested change in Norberto’s email (but wouldn’t that keep the graphics card running “slow”?)

 

So I’m a happy camper. Thanks all!

 

  • Glenn
image001.png

Terry Smedley

unread,
Sep 16, 2021, 9:47:15 PM9/16/21
to SEBHC
Glenn:

A portion of this I can answer for you:

HA-8-3 and the bus/CPU clock

The only device on the HA-8-3 that uses the bus clock to determine its speed of operation is the AM9511 APU.   The other devices are separately clocked.   The ADC has an internal clock that is controlled by external RC components.  The TMS9918 has an external crystal to control its clock, and the AY-3-8911 is clocked by a derivative of the TMS9918 clock.

What the bus clock does determine on the HA-8-3 is the setup and hold times for devices to access the H8 bus.  You will start to see intermittent failures, and eventually complete failures of some of the devices as the CPU speed increases, because the setup and hold times will become too short to meet the device's timing requirements.  The AY-3-8911 sound chip is particularly sensitive to setup, hold, and release times, and I'd expect you to have a lot of trouble with that device at higher speeds.  Best case scenario - the sound is garbled.  Worst case scenario - it doesn't release the bus properly, and the system crashes, most likely intermittently (like in the middle of a demo when you are surrounded by a group of legacy computer fans).

Unfortunately, the TMS9918 has no mechanism to signal the outside world that it is "busy" and needs more time to complete whatever it's doing.  It is completely dependent upon software delays, and of course those are in turn completely dependent upon the CPU speed.  At speeds above 4MHz (and sometimes even at 4MHz), the CPU can throw commands at the TMS9918 faster than it can process them, and you start to see garbage on the display.  The AM9511 and the ADC used on the HA-8-3 both use hardware signaling to force the CPU to enter a wait state until they have completed whatever was asked of them.

These issues will be resolved on an HA-8-3 clone by making bus access timing waits independent of the bus or CPU clock.  Instead of a shift register counting bus clock cycles, a one-shot will be used to provide the required delay regardless of CPU or bus clock speed.  The clock for the AM9511 APU will be an independent can oscillator to make that device completely independent of CPU or bus clock speed.  These things were implemented on Norberto's System Support I card, and it allows running the Z80 at a full 16MHz - even with very slow boards like the H8-4 serial card.  The caveat for the HA-8-3 or its clone is that software delays must still be used to keep from overrunning the TMS9811.

Terry

glenn.f...@gmail.com

unread,
Sep 17, 2021, 6:18:42 AM9/17/21
to se...@googlegroups.com

Well exactly as you predicted I am unable to get sound to work at 4mhz.  I (so far) have had good luck with any of the graphics programs running at 4mhz….

Reply all
Reply to author
Forward
0 new messages