Microdrive Mysteries - unanswered questions...

Nagging hardware related question? Post here!
Post Reply
martyn_hill
Aurora
Posts: 909
Joined: Sat Oct 25, 2014 9:53 am

Microdrive Mysteries - unanswered questions...

Post by martyn_hill »

Hi everyone

Prompted by recent testing of a batch of MDV units and inspired by Dave's recent announcement of a potential replacement MDV board being made available some time soon, I thought to gather together some of my own observations and unanswered questions that have accrued over time around the operation and circuitry of our (least?) favourite storage technology...

To be up-front - I still hold a flame to this Sinclair technology of the 80's - beyond all reason and after many hours of frustration and lost files over the years. What I learn about their design and operation has only served to fan the flame of my fascination for the microdrive!

All/any comments, additional questions - and answers - most welcome :-)

+++
Much like the humble bumble-bee, which physics suggests simply shouldn't be able to fly and yet it does - the design of the microdrive - both cassette and circuitry shouldn't on the face of it ever work at all.

1. A continuous spool of tape has a finite difference in radius around the spool from the point where it enters the spool to where it exists. This must create a difference in linear speed at entry/exit, albeit small, yet the tape is pulled at a single linear speed (barring fluctuations). Tape stretch might account for this - but only if it both stretches AND contracts as it passes round the loop. Can any one explain this to me?

2. The ERASE function in the QL MDV design (which differs from the original ZX SPectrum approach) is rather funky in that, with the WP tab open, Vin (c. 9V) is routed not to the Erase coil, but instead to the READ/WRITE input of the ULA - effectively switching ULA mode from WR to RD. Atleast the 9v is divided across those 2x 2k2 resistors that are placed in the MDRDWL line, thus supplying c. 4.5V to the MDV ULA pin. Clever, except that now both MDV ULA _and_ the ZX8302 are both driving the RAW1/2 lines simultaneously with different signal levels. Why does this not hurt each ULA?

3. Perhaps related to the above, but just how is tieing RAW1/2 down to -12V via that pair of 33k Ohm resistors (as recommended in the mandatory modification list) supposed to improve MDV reliability? Is this something to do with ULAs that 'latch-up' their outputs (and that the Service Manual suggests - 'never left the factory'...)?

4. The Service Manual provides a very simple SuperBasic routine to effectively generate a constant 100 KHz square-wave on the tape. I know it does this successfully because I've measured the output with an oscilloscope - on both tracks. However, after enabling the selected MDV unit in WR mode, it simply writes a single 0x01 to the ULA Transmit register (and then pauses for about 8 seconds to cover the entire tape-loop.) I can understand that, if you only write once, the data for the second track is effectively taken from the same single ULA register (normal writing would write two bytes to this register spaced slightly apart in time), so both tracks get to see the same data - but then why does a 0x01 pattern translate after DME (differential Manchester encoding) into a constant 100 KHz waveform?

5. Does the video circuitry (PAL encoder) affect MDV reliability, or the other way around? One can definitely see the display distort slightly (wavey edges) whenever the MDV is in use - especially during FORMATing. Or are both negatively impacted?

M.


User avatar
Dave
SandySuperQDave
Posts: 2765
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Microdrive Mysteries - unanswered questions...

Post by Dave »

martyn_hill wrote:1. A continuous spool of tape has a finite difference in radius around the spool from the point where it enters the spool to where it exists. This must create a difference in linear speed at entry/exit, albeit small, yet the tape is pulled at a single linear speed (barring fluctuations). Tape stretch might account for this - but only if it both stretches AND contracts as it passes round the loop. Can any one explain this to me?
When the cartridges were assembled, they spun a fixed length of tape onto the reel. They left a couple of inches of the inside of the loop hanging out, and attached the end of the loop to it. In operation, the capstan and pinch roller pull tape from the inside of the loop. Note the capstan and pinch roller are after the combo head. This supplies just the correct amount of tension. After the capstan, nothing is driving the tape except the pull of the reel. This is a self-adjusting mechanism. If the loop is slightly tighter, it pulls a little more firmly and the reel is wound a little tighter. If the tape is slightly more loose, it is wound more gently and the wound reel expands slightly. It's completely self-regulating and has a good margin for error.
martyn_hill wrote:2. The ERASE function in the QL MDV design (which differs from the original ZX SPectrum approach) is rather funky in that, with the WP tab open, Vin (c. 9V) is routed not to the Erase coil, but instead to the READ/WRITE input of the ULA - effectively switching ULA mode from WR to RD. Atleast the 9v is divided across those 2x 2k2 resistors that are placed in the MDRDWL line, thus supplying c. 4.5V to the MDV ULA pin. Clever, except that now both MDV ULA _and_ the ZX8302 are both driving the RAW1/2 lines simultaneously with different signal levels. Why does this not hurt each ULA?
The erase function is operated across pins 8 and 11 of the ribbon cable. If the erase allow tab is in place, the switch is in position 1->3 allowing current, if present, to activate the high frequency circuit. This feed is conditional on the state of pin 14 and pin 9, via two transistors.
martyn_hill wrote:3. Perhaps related to the above, but just how is tieing RAW1/2 down to -12V via that pair of 33k Ohm resistors (as recommended in the mandatory modification list) supposed to improve MDV reliability? Is this something to do with ULAs that 'latch-up' their outputs (and that the Service Manual suggests - 'never left the factory'...)?
On my Samsung microdrive units, this is not present. Can you point me to the mandatory mod info?
martyn_hill wrote:4. The Service Manual provides a very simple SuperBasic routine to effectively generate a constant 100 KHz square-wave on the tape. I know it does this successfully because I've measured the output with an oscilloscope - on both tracks. However, after enabling the selected MDV unit in WR mode, it simply writes a single 0x01 to the ULA Transmit register (and then pauses for about 8 seconds to cover the entire tape-loop.) I can understand that, if you only write once, the data for the second track is effectively taken from the same single ULA register (normal writing would write two bytes to this register spaced slightly apart in time), so both tracks get to see the same data - but then why does a 0x01 pattern translate after DME (differential Manchester encoding) into a constant 100 KHz waveform?
Pass. I've used the code for testing, but didn't look into its operation.
martyn_hill wrote:5. Does the video circuitry (PAL encoder) affect MDV reliability, or the other way around? One can definitely see the display distort slightly (wavey edges) whenever the MDV is in use - especially during FORMATing. Or are both negatively impacted?
From my testing on Samsung units and one UK unit, the video circuitry if no shield is present induces around .15v p-p at the head. That's significant. With the shield in place that drops to .06v p-p on the UK unit, and 0.03v p-p on the Samsung units. On my revised board 0.015-0.018v p-p. Much of this is induced directly into the head, not into any particular circuitry on the mdv board itself. The ribbon cable is also a large receiving antenna. The inducted noise is "mostly" from the video circuitry and is highly dependent on the screen content. Solid white vertical bars seem to be worst. Black or single color screen seems to be best. There's also an AC component (50Hz in the UK, 60Hz in the US) but this is minimal. I added an electrolytic on the input side of the voltage regulator to completely eliminate this.

My solution to this noise issue was to go mostly SMD, place the components on the "bottom" of the PCB, and make the entire top of the PCB a ground layer. It's a 4-layer board and I used three layers in the previous version. The new version is also 4-layer, but the tracks are moved to an inner layer as they also picked up a small amount of noise from the motor. Wrapping the ribbon cable with a shield tied to pin 7 (GND) helps too.


martyn_hill
Aurora
Posts: 909
Joined: Sat Oct 25, 2014 9:53 am

Re: Microdrive Mysteries - unanswered questions...

Post by martyn_hill »

Thanks Dave!

I was hoping you'd take the bait!

I still don't feel I grasp items 1 & 2, but thanks for offering an explanation!
Dave wrote:
martyn_hill wrote:3. Perhaps related to the above, but just how is tieing RAW1/2 down to -12V via that pair of 33k Ohm resistors (as recommended in the mandatory modification list) supposed to improve MDV reliability? Is this something to do with ULAs that 'latch-up' their outputs (and that the Service Manual suggests - 'never left the factory'...)?
On my Samsung microdrive units, this is not present. Can you point me to the mandatory mod info?
Sure - Section 3 "RETROSPECTIVE (MANDATORY) MODIFICATIONS", item (5): R102/R103 (33 k ohm) introduced between IC23 pins 21 and 19 and VM12 (-12 V rail).

Any other takers?


User avatar
Dave
SandySuperQDave
Posts: 2765
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Microdrive Mysteries - unanswered questions...

Post by Dave »

1. Tape loop

I was answering a slightly different question than the one you asked. I explained how the system self-tensions, and how it controls and manages slack as the tape stretches over time.

As the tape is drawn from the inside, it is pulled tight against the middle of the reel, but the tape loops moves freely against it, so you can ignore the reel itself. You seem to think that as the outer diameter of the reel is larger than the inner diameter, 1 revolution of the reel at the inside would equal almost 2 revolutions at the outside, which is twice the distance. However, because the tape is being PULLED from the inside, you have to think where the tension within the loop is. It's not to the inside, as that is where the pulling comes from. That's a loosening force.

So, if the inside turns and releases 1cm of tape, and the outside has 2x the diameter, doesn't it have to pull in 2x the tape? Or (some magic math that surely involves PI somewhere?).... Nope. The tape in the loop has very low friction between the layers, so the loop continually slides against itself. There isn't a 1:1 relationship of contact points between the layers.

2. Erase function

The pin 9 element of the R/W function simply sets the ULA status as READ or /WRITE. The tab detect switch can either allow or disallow write capability (pos 1-3 is allow, 1-2 is disallow, which works as a pull-up to assert read status, even if write is attempted. The erase circuit is activated by the switch being in the 1-2 position.

So, if the erase allow tab is present, READ/WRITE *can* go low, and the erase head *can* be energized. If the erase allow tab is missing, neither /WRITE can be taken low, nor can power be provided to the erase head.

3.Mandatory mod.

Ah, I see the confusion. R102/R103 0 ohm (wire links) are in series from RAW1/RAW2 on 8301 to the outside world. On Issue 6+, they added R107/R108 33K to -12v. This mod was for issue 5 only, and was incorporated on Iss.6+ boards. It looks to me that it may be just to bias logic 0 negative so +ve noise can't be present on the lines when they aren't asserted. Nasta may have a more nuanced (or completely different) understanding.


martyn_hill
Aurora
Posts: 909
Joined: Sat Oct 25, 2014 9:53 am

Re: Microdrive Mysteries - unanswered questions...

Post by martyn_hill »

Thanks again Dave

Yes, I think I get it now re. the tape-loop - my (admittedly naive) assumption was that the loop itself was a constant here and that any given point on the tape would remain in contact with the same point on an adjacent loop. But as you point out very clearly, the tape actually 'slides' against itself within the loop and thus makes up the (small) difference in linear speeds between entry and exit that would otherwise result if my assumption had been correct! Phew, that's one I can put to bed, finally!

As for the ERASE question - actually, my point was not the ERASE signalling itself, but rather that fact that, with the WP tab detached the switch will redirect Vin (divided by 2) to the READ/WRITE pin on the ULA (as you confirm), effectively engaging READ mode _and_ should the ZX8302 actually be trying to Write to the tape (as the 8302/OS have no way of knowing the cartridge is WP'd due to the 8049 IPC not connecting to this line, unlike on the Spectrum Int#1), both ZX8302 and the MDV ULA will both be driving RAW1 & 2 simultaneously? Aka - bus-contention...

Might'nt this cause some of the heat build-up in the MDV ULA that we know can cause failure over time?

As for the Mandatory Mod (RAW1/2 pulled low to VM12), whilst it was already applied to later boards, it still lives there - that resistor spider sitting on-top of the ZX8302. Whilst
the official Iss5 schematic shows the 0-Ohm R102/R103, I've never actually found them on the un-modded Iss5 boards I've seen and assumed it was an error in the schematic. I'll have to look again!

I'm still a little unclear why it would help reduce noise and why, for example, a pull-down just to GND would not be sufficient instead. I wonder what your fellow gurus have to say?


Post Reply