Cheetah Sweet Talker doc/test?

[Moggy]

Moggy,
Can you tell if the RD or WAIT lines are connected anywhere on the Sweet Talker board?

BTW,
I'd love to see pictures of your circuit board.

Yes will take pictures of everything but please don't ask me what's going on as my tech knowledge is zero!

[stroebeljc]

I would be surprised if they are using the /WAIT control line for any length of time. If the Z80 is held for too long, any DRAM that needs refresh via the Z80 refresh cycle will not be refreshed. And all other execution is also “paused”.

The /WAIT input is only intended to pause the Z80 for one, two or maybe three T-states (clock periods) so that a slow memory or slow I/O device can complete a data bus cycle.

Either, a register is able to be read that reflects if the operation has completed or the software has another method of timing how long the current operation takes.

As I don’t have the hardware, I can only speculate...

Mark

That's another reason I doubt the WAIT is being used on the ZX81, but is on the Spectrum. On the Speccy, the refresh is managed by the ULA, not by the Z80.

[1024MAK]

That's another reason I doubt the WAIT is being used on the ZX81, but is on the Spectrum. On the Speccy, the refresh is managed by the ULA, not by the Z80.

The ZX Spectrum - for the "lower" RAM, yes, the ULA performs refresh as it reads screen data. But only the Z80 refreshes the "upper"/expansion RAM.

Mark

[stroebeljc]

The ZX Spectrum - for the "lower" RAM, yes, the ULA performs refresh as it reads screen data. But only the Z80 refreshes the "upper"/expansion RAM.

Mark

Maybe not for all versions. This 4AB version of the Spectrum (from here) refreshes the upper 32k within the expansion logic itself. The /RFSH line from the z80 doesn't even enter the expansion area.

ZX Spectrum version 4A&B Schematic

[Moggy]

A couple pictures as promised.

This board had been butchered, I assume by the previous owner with ripped tracks and missing components and some soldering interconnecting some of the pins of the chip on the lower left of the first picture. I restored it best I could to original and it now works Ok.

The resistor and transistor is missing and this necessitates a a large PAUSE of some type either M/C or BASIC or the speech is too fast and unintelligible. It was whilst messing with an identical unit I discovered the delay applied by the resistor/transistor combo, substituting differing values off resistor altered the speed of the unit. I have left this one "as is" I don't want to disturb the tracks any more than I have to and as long as it works I'm happy.

On the first picture the red circles indicate where the transistor was, the red arrows/blue circles, the resistor (sorry can't remember its value.)

The yellow circles at the top of the picture is where the 10uf bias cap was for the 386 power amp chip. With the cap the chip was biased at X200 which was far too loud, removed it leaving the 386, which is biased internally at X20, more than adequate level wise.

The second picture shows the daughter board fitted to alter the units port setting, I can't show the underside as it's held down with strong double sided foam tape and the wiring is very delicate but I've posted it just to give an idea of what was done.

[Moggy]

Forgot to add that with this mod a simple BASIC program running concurrently is possible which I think is what Sir Morris was aiming at.

[stroebeljc]

Thanks Moggy!
This confirms that the Sweet Talker controls the /WAIT line to introduce delay during speech processing. The yellow circle contains the trace that is connected between your resistor-transistor network and the /WAIT input to the ZX81 (pin A19 on the connector).

The network likely uses an RC circuit to hold /WAIT low for a period of time. Changing the resistance will change the rate of voltage decay on the network, causing the pulses on /WAIT to be shorter or longer, speeding speech up and slowing it down.
Although using /WAIT works on the ZX Spectrum, it's a dangerous game on the ZX81 due to the negative effects on both video synchronization and DRAM refreshing of introducing wait states.
In contrast, the Parrot doesn't touch /WAIT. Instead, it allows the software to poll the I/O port to detect the completion of a speech command, which works well on the ZX81.

[Moggy]

To be honest it's all Greek to me all I know is that it works!

I do know Wilf Rigter designed a speech unit of his own which in all honesty was a close copy of the Sweet Talker but with a volume control added.
I only mention this because no doubt his schematic will be very detailed and should be lurking in a repository somewhere and may be worth seeking out.

[Moggy]

Here it is, Rigters version of the speech unit.

[1024MAK]

This 4AB version of the Spectrum refreshes the upper 32k within the expansion logic itself. The /RFSH line from the z80 doesn't even enter the expansion area.

Yes, you will not find the /RFSH line from the Z80 going to the logic for the "upper" / expansion RAM. In fact, on most 16K/48K boards, this line is not used internally at all. It just goes to the expansion port / edge-connector.

The Z80 activates /MREQ during the refresh cycle and puts a refresh address on the address bus regardless. The logic for the "upper" / expansion RAM activates the /RAS for the DRAM in response to the /MREQ line going low. Hence during the refresh cycle, when the /MREQ line goes low, the seven bit refresh address to the DRAM is strobed in to the DRAM. This is enough to refresh the type of DRAM used (which does not need a separate refresh control signal).

I hope that clears up any misunderstanding.

Mark