ReproductionsComments Off on Seibu ‘HB-45A’ reproduction
Jun082018
The ‘HB-45A’ is a custom IC used on some PCBs manufactured by Seibu like Raiden II/DX, Zero Team.It can be considered an evolution of the ‘HB-41’ that we already treated in a past article:
It shares the same 20 pin SIL package embedding most of the analog sound system.Its function is to mix and pre-amplify both music and sound FXs :
Here’s how it look “naked” of its external coating:
Design is almost identical to its predecessor with the presence of two OP-AMPs (a dual and a quad one) although they operate from a single supply (+12V) and not dual (+5V/-5V) like the ‘HB-41’.All other components are ceramic/tantalum capacitors and resistors (surface mounted and printed film ones) which are part of the application circuit of the two OP-AMPs
It took some time to draw schematics not without some difficulty due the presence of some hidden printed film resistors located under the quad OP-AMP:
Schematics were then routed to a PCB which looked pretty good on a 3D preview :
The real thing assembled in all its parts :
For a while I was not able to test this reproduction due the lack of a working PCB (the “decapped” original part was taken from a dead board) until the user ‘opt2not’ from Arcade Projects and KLOV forums kindly volunteered.The test platform was his working Raiden DX PCB.He first removed the original ‘HB-45A’ in a very clean way (he used an Hakko FR-300 desoldering gun)
Installed some female 2.54mm round machined headers:
Lastly the reproduction fitted onto the board:
Then he made recordings of audio of both original custom and reproduction redirecting it to a PC from a HAS supergun using a SPDIF breakout from the SCART connection.
Here ‘s capture using the original HB-45A:
Here’s the one using my reproduction:
Comparing the music and FXs at the beginning and end of the recordings sound identical.No differences in quality can be noticed (the slight background noise is normal when there is silence).
For this successful reproduction project I would like to thank again ‘opt2not’ for his precious testing and all material (pictures and audio captures) he provided me.I recommend you to check out his blog :
ROM_REGION( 0x040000, “layer2”, 0 ) /* Text / Character Layer */
ROM_LOAD( “text.u58”, 0x000000, 0x040000, CRC(5dba06e7) SHA1(f9dab7f6c732a683fddb4cae090a875b3962332b) )
I translated this info on hardware and figured out the relevant circuit of each layer:
Each circuit consists in a QFP custom ASIC (maked ’38’) which addresses an 8Mbit MASK ROM (or a 2Mbit EPROM for the text layer) reading back data that then are written to two 8K x 8-bit static RAMs.After succesfully checked connection between ASIC, ROM and RAMs my suspicions fell on the 8k x 8-bit SRAMs, they were all manufactured by Sanyo so in my experience not a great guarantee of reliability.Probing the ones @U50 and U51 (which lie in the ‘layer 0’ circuit) revealed weak or stuck signals on many data lines:
I removed both:
Actually only the one @U51 failed the out-of-circuit testing:
Now all graphics were visibile.Backgrounds and sprites were fine but text was corrupted throwing garbage over the screen :
The ‘layer 2’ identified in MAME source as ‘ Text / Character Layer’ was obvioulsy the involved one.Checking the two Sanyo 8K x 8-bit SRAMs @U55 and U56 revealed again weak signals on data lines:
I removed them and installed machine sockets:
Actually both RAM chips successfully passed the out-of-circuit testing of my different programmers.Anyway, replacing them fixed completely the graphics:
But sound was horribly scratchy and corrupted as you can hear in the above video.Here it comes again to help my audio probe for checking the relevant circutit:
“Listening” to various points revealed the sound was clear on the analog output of the two OKI MSM6295 and still good on outputs of the LM324 OP-AMP and input of the LA4460 amplifer.It was fine too on one output (pin 7) of the LA4460 amplifier:
But corrupted on the other output (pin 9)
I ruled out all electrolytic capacitors checking them in-circuit with my ESR meter (bad ones can affect sound in this way) so I decided to remove the LA4460 amplifer:
Put back a good one and some thermal compound for a better heat dissipation:
This restored a crystal-clear sound.Mission DonPachi accomplished!
PCB Repair LogsComments Off on Toaplan (Knuckle Bash – FixEight) double repair log
May252018
I’ve been sent from Germany some faulty PCBs for repair.There were among them two PCBs manufactured by Toaplan :
Knuckle Bash
Fixeight:
The first was throwing a “A VRAM ERR” message on boot:
The video RAMs (VRAM stands for this) are two 8k x 8-bit static RAM located @42 and U51:
They are accessed by the QFP ASIC marked ‘GP9001’ which is the graphics controller of the system:
For first I did a visual inspection of it and found some pins not really firm (common issue on QFP with this fine pitch)
I reflowed them but this didn’t led to any improvement so I moved to check the VIDEO RAM ICs and found weak signal on data lines:
I pulled both RAMs, only the one @U42 failed the out-of-circuit testing:
The board successfully booted into game now but the colors were wrong, they were “bleeding”:
This is another common issue on this board so I knew where to look exactly.The data bits from the two 6116 (2K x 8-bit) palette SRAMs are latched by two 74HC273:
Probing with a scope the 74HC273 @U9 revealed weak signal on outputs (inputs on the left of the below snapshot, outputs on the right)
I pulled the IC:
It actually passed the out-of-circuit testing on all my programmers so I think IC was not really bad but it had altered voltage thresholds.Anyway, I replaced it and this fixed the board completely.
Now the FixEight troubleshooting.
Board gave me a steady black screen.This board uses the GP9001 graphic controller too:
Found some lifted pins on it:
I reflowed them as well as replaced the main program code 4Mbit EPROM since it had some rebuilt pins that gave me trouble when dumping it:
But still no boot.But I noticed that pressing the PAL16V8 @U51 :
produced a couple of different RAM error messages on screen:
Solderside of the PAL socket had some dry joints:
So I replaced it and finally the board initialized but it quickly went straight into SERVICE mode and then reset in an endless loop :
From a still image I noticed that some inputs were stuck in I/O check:
The inputs are handled by the custom ‘HK-1000’, you can find more info about it in a past article from mine:
On my PCB a replacement board had been used in place of the original part:
Some inputs of a 74LS240 on this replacement board were shorted to ground.Instead of repairing it, I opted for a reproduction of mine (the blue jumper wire was already there to patch a trace previoulsy broken during the removal of the original ‘HK-1000’)
Finally the board booted into game with no further issues.Toplan double repair log accomplished.
And it was so, all I got was a steady black screen but this was not a suprise to me since the boardset (a three stack one) was fully populated by Fujitsu TTLs.First thing I checked was the presence of clock on CPUs, it was missing on both main and sub Z80 (but present on sound one)
I traced the signal from CPU board back to a 74LS367 @10A on bottom board:
Input pin 6 was properly receiving the 16MHz clock from the nearby oscillator but output pin 7 was floating:
As expected the TTL obviously failed the out-of-circuit testing:
Board sprang to life but colors were mostly incorrect and also sprites wrong and garbled during moviments :
I focused my troubleshooting on the bottom board as it contains sprites and background data and circuitry :
For first I checked the EPROMs, some of them were oxidized and corroded:
Afer some cleaning I dumped them, they were all good.So I fired up my logic comparator to test TTLs in circuit and found a couple of them with floating outputs (that’s the way Fujitsu TTLs are used to fail)
This restored the correct sprites :
But colors were still wrong.This was confirmed also in SERVICE mode, none of the them were correctly displayed:
The RGB lookup tables are stored in three 1k x 4-bit bipolar PROMs (Fujitsu MB7122 used here)
Probing the devices revealed that two address line were stuck high, I traced them back to a nearby 74LS174 @1M whose inputs were missing.These came from the outputs of a 75LS153 @1P-The TTL didn’t look very good with rust and corrosion on its pins:
The scope confirmed that both outputs had no valid voltage logic levels:
I pulled it:
Obviously it failed the out-of-circuit testing:
Colors were finally good and game played fine but I noticed there were some horizontal lines on left of the screen during scolling (see from seconds 25 of the below video)
At first glance I thought it was a fault but then I’ve been told by the owner and my friend Corrado that it’s actually a bug of the PCB.MAME Testers site confirmed it too:
I was about to close this repair when during testing suddenly the board failed again.Colors went bad again and lastly most of graphics disappeared:
The affected area was again the colors circuitry.The 74LS174 @1M which was previously tested as good showed floating outputs :
It failed when tested out-of-circuit:
At this point I took the chance to replace all other Fujitsu TTLs in this area :
Some passed the out-of-circuit test but two 74LS153 failed miserably:
Hard work always pays off and finally I could restore this board in all its glory (game was the first to use a fully digitized, full-length vocal sountrack)
