Music 500 -> Music 5000 and Music 500 -> Music 3000

[maniacminer]

I've got some more fixes to apply to the emulation, but with my M500 being broken I'm a little stuck. I've fixed some of the issues I've found to the Pi1MHz project, but need to port them to beebem.

I thought I had a dead AM6070 DAC in a Music 500 that I converted into a M3000 - the DAC was giving out nothing despite all the signals arriving to it. Eventually, I worked out the 12V reference zener diode had died, 15V on one side and zippity-doodah-zilch on the other. I ordered a replacement, fingersX that's the only thing stopping that from working. I have ordered all of the ICs inside the Music 5000 except the DAC, luckily they're still manufactured by Texas Instruments.

[dp11]

Mine has definitely got a dead DAC

[maniacminer]

I had a free electricity morning and for once, a bit of free time, so I decided to strip the working M500(0) to get to the bottom of its weird (partially working) behaviour.

The choir has left the cathedral

Just milling around waiting to sit down

A seating upgrade

The 24 pin SRAMs really didn't want to come out of the board I had a tough battle with those - but not as bad as the IDC connector on the previous M500...

I tested each IC (except the TL072's and the AM6090) and found three damaged 74LS374's octal flip-flops, each with a faulty bit in different places. As this design basically runs whatever it is given (there is no program) then faults like these are going to make the strange sounds that I am experiencing. Had it been a microprocessor controlled device, then the program would have crashed way before hearing a single note.

[maniacminer]

I replaced the dead zener diode on the M3000 and it now makes sound, yay! Unfortunately, it's making the wrong sounds I guess there are at least three dead 74LS374s I'll remove all of them and put in new ones. On the plus side the DAC is working (phew)

[maniacminer]

On the M5000 I am still getting some strange sound when playing music demos. I found on this forum, a piece of music that I think sounds really terrible on my Music 5000, but sounds OK on a working unit. viewtopic.php?p=114780#p114780 has "Pachelbel" and it sounds totally different (and much better) than mine!

With this playback in mind, does anyone know what part of the circuitry I should pay attention to?

[dp11]

I think first of all I'd play a full volume sinewave and check the outputs of the dac and the opamps for distortion. It does sound like a bit error.

[maniacminer]

I think first of all I'd play a full volume sinewave and check the outputs of the dac and the opamps for distortion. It does sound like a bit error.

OK, I put the sinewave through two DACs and I get the same weird output. I also tried bending the sign bit pin up on one DAC to see what difference it made, so unless there's something very wrong with the op-amp chain that has destroyed both of the TL072's, then both these DACs have failed in the same way, or perhaps there's something else going on in the digital chain. Not really sure where to start tbh.

DAC#1, then DAC#1 with sign pin bent up, then DAC#2

DAC#1.mp3

DAC#1 sine wave output

(675.84 KiB) Downloaded 4 times

DAC#1 sign pin bent up.mp3

DAC#1 with sign pin bent up, sine wave output

(506.16 KiB) Downloaded 4 times

DAC#2.mp3

DAC#2 sine wave output

(418.31 KiB) Downloaded 4 times

[maniacminer]

After watching ladies and gentlemen perilously avoiding a nasty bout of E-coli infection, I spent a bit of time figuring out why two different DACs and two different sets of op-amps would give the "wrong" result.

As it turns out the test program that puts a sine wave out, doesn't actually do so.

The shape that comes out isn't a sine wave, it looks as if the sign bit in the M5000 isn't as straightforward as making a memory location negative...

I tried a different technique and went into AMPLE and made a simple test using an enveloped sine wave "Panflute" and yup, perfect sine waves. So the DAC and op-amp is fine. The problem lies somewhere else inside the digital guts of the M5000. I am suspecting the RAM, even though it tests fine in the chip tester, I'm not so confident that it works at the 6MHz cycle time of the M5000.

Playing the same program in AMPLE on BeebEm and M5000

BeebEm Test - Panflute.mp3

BeebEm MP3 file

(69.94 KiB) Downloaded 5 times

M5000 Test - Panflute.mp3

M5000 MP3 file

(64.31 KiB) Downloaded 4 times

[dp11]

I would also guess a ram issue, but It might be worth checking the drive levels of the logic devices.

To help me debug my M5000, could you possible measure the voltage on Pin 11 and 12 of the DAC ?

[maniacminer]

I would also guess a ram issue, but It might be worth checking the drive levels of the logic devices.

To help me debug my M5000, could you possible measure the voltage on Pin 11 and 12 of the DAC ?

Using pin#10 of the DAC as ground reference, I get 15mV on pin #11 and 4mV on pin #12.

[dp11]

I would also guess a ram issue, but It might be worth checking the drive levels of the logic devices.

To help me debug my M5000, could you possible measure the voltage on Pin 11 and 12 of the DAC ?

Using pin#10 of the DAC as ground reference, I get 15mV on pin #11 and 4mV on pin #12.

Thanks

[dp11]

I would also guess a ram issue, but It might be worth checking the drive levels of the logic devices.

To help me debug my M5000, could you possible measure the voltage on Pin 11 and 12 of the DAC ?

Using pin#10 of the DAC as ground reference, I get 15mV on pin #11 and 4mV on pin #12.

Thanks

Just in case anyone else is interested . My 5v rail is a little low @ 4.55v Pin11 I get 22.3mV and on pin 12 5.6mV. -15v is at -14.3v and +15v is at 14.3v

[maniacminer]

My 5V rail is 4.96V - good enough - the colder the 7805 regulator is, the lower the output voltage my "15V" rail is 13.86V and the "-15V" rail is around -13.65V - a bit low, but that's down to the transformer output, that varies with input voltage.

[maniacminer]

The SRAMs in this are 2k x 8 100ns, I've got some but they're skinny DIP so I'll need to do some bodging... In the meantime I swapped the SRAMs around and there was no change in behaviour. So perhaps it is a marginal logic problem? I'm going to have to find the pieces of music that sound bad and reverse engineer them. I can't figure out how to load something that sounds bad and pick it apart. There's no AMPLE editor that I can load performances into and edit them...

[hoglet]

On the M5000 I am still getting some strange sound when playing music demos. I found on this forum, a piece of music that I think sounds really terrible on my Music 5000, but sounds OK on a working unit. viewtopic.php?p=114780#p114780 has "Pachelbel" and it sounds totally different (and much better) than mine!

Can you post a disk image containing "Pachelbel" track you are testing with?

Dave

[maniacminer]

On the M5000 I am still getting some strange sound when playing music demos. I found on this forum, a piece of music that I think sounds really terrible on my Music 5000, but sounds OK on a working unit. viewtopic.php?p=114780#p114780 has "Pachelbel" and it sounds totally different (and much better) than mine!

Can you post a disk image containing "Pachelbel" track you are testing with?

Dave

Here is the disc image you requested

M500 - Canon - Pachelbel.dsd

M500 Pachelbel's "Canon" with Ample BCE

(71.75 KiB) Downloaded 2 times

I had a bit of frustration making this disc as using wget on the Pi with the PiBridge running was turning all my files to shit and making BeebEm crash etc. took me a little while to figure it out, started another ssh on a different Pi and restarted the wget-ting and all was well again. To relax, the real thing sounds so much better https://www.youtube.com/watch?v=JvNQLJ1_HQ0

[maniacminer]

I got a bit further with sorting out the M3000 and finally got a M3000 demo working-ish on it, yay! I don't think it is working 100% though as when I listen to the M3000 solo, it is barely making any extra sounds.

Does anyone have a working combo and is willing to record solo and combined for comparison purposes?

[hoglet]

On the M5000 I am still getting some strange sound when playing music demos. I found on this forum, a piece of music that I think sounds really terrible on my Music 5000, but sounds OK on a working unit. viewtopic.php?p=114780#p114780 has "Pachelbel" and it sounds totally different (and much better) than mine!

With this playback in mind, does anyone know what part of the circuitry I should pay attention to?

It sounds to me like the modulation feature is not working correctly, but everything else is. In Pachelbel's Canon, modulation is used (I think) in the solo bass instrument at the start, but not in the second higher pitched instrument.

I would focus on this part of the circuitry:

I suspect the INDEX signal (which selects between the normal and alternative register sets in a modulated channel) is not being generated correctly.

I've just tried introducing a few different faults into my Music 5000 FPGA, and bypassing IC3, or clocking it more frequently, generates something vaguely like the MP3 you posted. Your MP3 does sound like the modulation has more of a random element to it though.

In case you are interested, here's a recording with IC3 being clocked more frequently (i.e. ignoring nS7):

canon_pachelbel_broken.mp3

(5.99 MiB) Downloaded 6 times

I would start with a scope on the INDEX signal with Pachbel's Canon playing and work backwards looking for evidence that one of the upstream devices is misbehaving. Also, pin 3 of IC3 should be a regular 750KHz clock (low for 83ns, high for 1250ns).

Dave

[maniacminer]

Thanks for that. The odd music is similar, but not quite the same. I traced around the area and couldn't see anything strange, I got the same periodicity of signal as you. I was fiddling with the SRAM and now I've got barely any sound coming out at all no idea what I did, there was some sound like a DAT machine trying its best to error correct dropout and then a high pitched squeal and then silence. It still plays, but only if I really crank the amplifier up, then it is audible but very distorted. I swapped out the DAC but that made no difference. So I'm back to square one with the whole thing, I must have accidentally shorted something, or a cracked track has decided now is the time to lose contact, or an IC has started to wander around in the bracken... I'll need to have a think on what to do next. I really fell down the rabbit hole with this, wondered if it was worth bothering with whilst having a clear out. I wish I had one of those special DeLorean's EDIT:just realised I put the wrong op-amps in, single vs. double

[maniacminer]

I noticed a serious difference between these M5000's - the one on the bench plays a very boring rendition of Chris Jordan's "Syncron" vs. the M3000 (temporarily modified back to a M5000) the only problem is how to use that track to locate the fault. There's so much going on inside the M5000, it's like a DSP made from discrete TTL logic

[dp11]

Modify them both to be the Same e.g. M3000 then you can run them both together and compare

[maniacminer]

Modify them both to be the Same e.g. M3000 then you can run them both together and compare

It is very difficult to find something that drives the M5000+M3000 and produces the same result. The most telling problem is the "Syncron" demo, it's like night and day. I did a bit of Yoga on some SRAM to see if that was the issue with the dodgy M5000 but it made no difference.

Morning yoga for an unsuspecting 2k SRAM

[hoglet]

It is very difficult to find something that drives the M5000+M3000 and produces the same result.

You don't need to do that

You should be able to set both units up as M5000s and connect them to the same host.

This should work because the M5000 is a write-only device, it's not able to drive the data bus, nor does it generate any interrupts.

[maniacminer]

It is very difficult to find something that drives the M5000+M3000 and produces the same result.

You don't need to do that

You should be able to set both units up as M5000s and connect them to the same host.

This should work because the M5000 is a write-only device, it's not able to drive the data bus, nor does it generate any interrupts.

Oh yes, cool, I forgot about that! Thanks!

[maniacminer]

Both M5000's together at the same time sounds strangely pleasing

[dp11]

Now using a scope you can check the signals are the same give or take the 24MHz phasing. You could of course temporary use the same 24MHz clock source.

[maniacminer]

Now using a scope you can check the signals are the same give or take the 24MHz phasing. You could of course temporary use the same 24MHz clock source.

I did that and then realised the M5000 has the same problem as the BBC Micro video ULA with it's very similar clock divider chain - the lack of reset on power up... So there's a random delay between signals even though the two M5000's are driven with the same clock. So next thing to do is run another set of bodge wires to connect the reset line from the 1MHz bus to both M5000s and then to both halves of the 74LS393. There's some kind of breakable link somewhere on the PCB according to the schematic, maybe be it is there for this very reason?

[maniacminer]

Having poked about with two synchronised M5000's I found that the address bus on IC29 was garbage on the bad M5000 and the only way I could remove the garbage was to pull IC18 and IC23 (74LS257) then the address bus was quiet. I checked the two control signals to these ICs nPROC and nS6 and these were identical to the working M5000. The INDEX pulse is almost random

INDEX pulse (B)ad and (G)ood

It would be good to know what things can create addresses, I think only the BBC via the 1MHz bus and this pair of 74LS257's do, but I'm not 100% sure, anyone know?

My current lash-up

[hoglet]

It would be good to know what things can create addresses, I think only the BBC via the 1MHz bus and this pair of 74LS257's do, but I'm not 100% sure, anyone know?

Yes, the address bus which feeds IC25, the Wave RAM, is timesliced (over 8 cycles) between:
- the ADDR bus (which comes from the LS393/LS374 sequencer)
- the 1MHz bus address (if a write is pending)
- the SUM bus (which is the output of the ALU)

Anything that depends on the ALU output will be complex/hard to predect, and indeed if the two systems sound different, then very likely the ALU will be outputting different values.

I would step back and methodically check the the following signal groups are identical between the two systems.

Group 1 - these are the control signals from the sequencer, and don't depend on the data path (i.e. what's actually playing):
- nS0/nS1/nS4/nS6/nS7 (on IC10)
- ADDR[6:0] (on IC13)
- nADDR0 (on IC5 pin 8 )
- PA1 (on IC13 pin 19)
- PA2 (on IC27 pin 9)

Group 2 - these are the control signals for the 1MHz bus interface, and will only be active when AMPLE is writing to the wave RAM (i.e. when music is playing):
- nREQ (on IC35 pin 3)
- SEL (on IC34 pin 8 )
- nPROC (on IC27 pin 6)
- PROC (on IC15 pin 11)
- nWR (on IC27 pin 19)

Once you have confirmed all of these are good, the next thing might be to confirm writes to the Wave RAM look correct (i.e. there are no stuck address or data bits on the RAM). This will need a few lines of BASIC.

Dave

[maniacminer]

I would step back and methodically check the the following signal groups are identical between the two systems.

Group 1 - these are the control signals from the sequencer, and don't depend on the data path (i.e. what's actually playing):
- nS0/nS1/nS4/nS6/nS7 (on IC10) matched
- ADDR[6:0] (on IC13) matched
- nADDR0 (on IC5 pin 8 ) matched
- PA1 (on IC13 pin 19) matched
- PA2 (on IC27 pin 9) matched

Group 2 - these are the control signals for the 1MHz bus interface, and will only be active when AMPLE is writing to the wave RAM (i.e. when music is playing):
- nREQ (on IC35 pin 3) matched
- SEL (on IC34 pin 8 ) matched
- nPROC (on IC27 pin 6) matched
- PROC (on IC15 pin 11) matched
- nWR (on IC27 pin 19) matched

Thanks for the help. The ALU seems to wander off into the bracken quite quickly from reset, yet despite changing all the logic ICs, I am still getting the same sound as I had before. Maybe there's a fault on the board somewhere that is causing this issue?