Tuesday, 5 September 2017

Yamaha DDS-20M Delay/Sampler

I wouldn't really think of Yamaha as a guitar pedal company. They have made a lot of classical instruments, drum machines, e-drums and studio gear (I have an SPX90 I really need to repair) so I guess the pedigree is there. This is a mid-80s 2 second digital delay/sampler, from that period when putting "sampler" on products must have excited guitar players.

The mechanical design is a little different - all of the controls are recessed at a height below the footswitch and there is a "unibrow" along the top with the logos and LED.

There are no schematics and not a huge amount of info on these that I could find online, other than there was a range of pedals in this form factor. No signs of life.



After looking on the inside, I got the impression that someone else had attempted a fix. There was a resistor between the +9V and ground pads of the DC jack (?), the ground trace had been cut in two places for some reason and a wire had been soldered in to replace it. I can't see why the resistor between 9V and ground would be needed for anything other than wasting power, so I removed it. I also found that the ground from the DC jack was not connected to circuit ground. Adding a jumper was enough to get this to fire up with a DC supply, but not with a battery. It's possible that the resistor was originally connected between jack ground and circuit ground but as it was several kilo-ohms a wire would be a better solution.

The pedal uses a 2 board design, connected together with wires. Both are single-sided PCBs that look like a phenolic paper material and no through-hole plating anywhere. This is pretty typical for 80s Japanese electronics but it can mean that any board-mounted hardware can be fragile as there is only solder on one side of the board anchoring it place. In this delay, the output jack had taken some force (probably a cable was stepped on) and the pad to the jack tip had broken, this had to be re-soldered to get an output signal. One of the PCBs has most of the analog parts, the inner one has the digital controller and the pots and rotary switch.

2 wires added and a diode replaced

It still didn't work from a battery. Following the (broken) ground trace to the other side of the board I could see that a diode was used to switch out the battery when a power supply was used, replacing this diode fixed everything.

Analog board, component side

The analog board has a 4066 for bypass/effected switching, 5534 opamps, a Motorola M5218 audio pre-amp (I have only seen this part in Japanese pedals, I think the original Boss DS-1 used something like this), ╬╝pc1571 compander (NEC's NE571 clone) and the chip in the bottom left corner is an LM311 comparator. The comparator is forming part of a sigma-delta ADC with the digital controller, much like the Boss DD-2 and the Digitech PDS series of delays.

Bizarre vertical component assemblies

I thought the parts marked "< RIVER < RIVER" were SIP ICs, but they are actually components standing vertically! They seem to be clamped together at the top. I don't know if there was automated machines to produce these assemblies or if they were re-used in other designs but it's surprising that this was cheap enough to be worth the effort.


Digital board

The digital controller is an "M654131". No info on this, probably a Yamaha custom. Maybe they used it somewhere else in a different product. I don't have any pictures of the back side of this board so I don't know what type of memory is used. It looks like there are two 9-pin SIPs connected straight to the controller so I would guess 2 64k DRAMs, assuming this is an 8-bit delay.

It sounds great. I like that these early digital designs use companders and have simple master clock signals. It's like a holdover from the days of analog delay designs. This one can pitch-shift the delay signal smoothly, unlike some more modern delays.

Sunday, 27 August 2017

Boss DD-2/DSD-2 repairs

Once again, I have a backlog of photos of things with very poor notes - hopefully I haven't forgotten everything.

Early Boss Delays

I've played a few of the early Boss delays, but oddly enough I've had none of the more common modern versions (DD-5/6/7 etc). The DD-2 is regarded as the first digital delay in a stompbox, using some of the tech from Roland's larger rackmount delays (see Boss's article on the history of delays). The architecture is similar to early Japanese delays, with one large custom controller IC doing all the digital work - a sigma delta ADC (with an external comparator), DRAM interfacing & parallel outputs for an resistor network DAC. This is the same RDD63H101 custom used in the famous Roland SDE-3000, often referred to as the "long chip" as it barely fits in the Boss stompbox case. Everything is synchronised off a single master clock which is directly controlled by the delay time knob - changing the delay time smoothly pitch-shifts whatever audio is memory without any glitching.

Digital section, from Boss Service Notes

Unlike many other 80s delay pedals, the audio is quantised to 12-bit samples (8 is more typical) for lower noise and there is also more memory than I would expected, 3 64k DRAM chips (a single 64k IC is common in some 1 second digital delays from the era).

Parts of the schematic look more like what would be expected from an analog delay, there is a NE570 compressing on the way into the digital delay line and expanding on the way out to further help with noise, and pre-emphasis and de-emphasis filters at 7 kHz to hide any sampling bandwidth limitations.

The custom chip is also a weak point in repairing these. If it's dead then the only way to find a replacement is to pull one from another pedal. I am hoping that most of these are still working and that faulty pedals just have power or memory issues.

DD-2 #1

This was another eBay purchase. It powered up, but only produced a whining noise. The noise would change in volume with the Level knob and change in pitch with Delay Time knob, so the digital end was the first suspect. Power from the 5V regulator looked good.

DD-2 #1

I had seen something like this before on an 80s Digitech PDS delay (write-up to come at some point in the future). In that case I was getting massively distorted repeats and I narrowed things down to bad memory, data was beings written in but garbage was being read out. I also knew that DRAM failure was very common in the late 70s/early 80s so I decided to try swapping out new memory chips on this DD-2.

All memory ICs desoldered
The 3 DRAM ICs are directly below the main controller. These pedals are absolutely crammed with through-hole parts, it's impressive that so much could be fit into the standard BOSS housing when you consider how much more simpler the earlier Boss pedals were in comparison.

Socketed memory ICs - a dead end.

Initially I tried to install sockets to make troubleshooting easier. After a struggle to get socket into the board I realised this wouldn't work, the flying leads over ICs would need to replaced with longer wires and it was possible the PCB wouldn't fit back inside the case.

Replacement memory installed.

I removed the sockets and installed 3 MK4164 64K DRAMs with a compatible pinouts. Switched the pedal on - absolutely nothing had changed. Shit.

Going over the each pin of the controller IC with an oscilloscope, I found that even though 5V was present, when I had a test signal connected to the input (a triangle wave in this case) I could see the triangle wave riding on top of the 5V supply! I traced this back to a cracked solder joint on the 5V regulator, which was then reflowed. The power filtering electrolytics were replaced at the same time. Another delay working again.

DD-2 #2


This much dirtier DD-2 was not mine, but was bought broken at a market for €5(!) and eventually found it's way to me to take a look.

DD-2 #2

This was a lot easier to figure out, the pedal wasn't lighting up and was shorting out my power supply. The reverse polarity protection zener diode (D6) was burnt out and failed as a short. It looked like someone had been in here before and had tried to jumper over a PCB track with a piece of wire and some cold-looking solder joints.

Previous attempt at a fix
 After removing the jumper and the diode (which tested as a short out-of-circuit) the board looked a little rough. I added a 1n4001 as a replacement and tried it out.

Some PCB foil damage
This time the LED would only come on when the pedal button was held down, it wouldn't latch and stay on. Usually Boss use a discrete flip-flop for switching (an excellent Geofex article on this) but in the DD-2 they use a BA634 flip-flop in a SIP package, probably just to save some real estate on the PCB. I thought this IC was bad but it just had a broken ground trace - another small jumper wire fixed it.

Both of these pedals sounded identical, and when I checked the calibration routines in the service manuals they were both still perfectly dialled, with max delay coming in at 800ms on both.

Bonus Repair: Boss DSD-2

This pedal is a couple of years newer. The problem was pretty obvious, all of the 1/4" jacks and the DC power jack were broken and had to be replaced. I can only guess someone was putting the pedals into a bag without unplugging the cables and they managed to break them all.

DSD-2

Guts - note broken jacks everywhere

What is interesting is how close the DSD-2 is to the DD-2. There are a couple of part changes for cost and space savings (the RAM chips have been replaced with SIP packages that are a better use of space) but it's basically the DD-2 without the hold mode and with an external trigger input instead as a "sampler" mode. The service notes even shows that the only differences in the digital side is that a couple of pins are wired differently on the controller to trigger a delay from the Trig In jack instead of the footswitch.

The DSD-2 should perform identically to the DD-2 in delay modes, and to my ears they sound identical. If you like this sound it may be worth finding a DSD-2/3 as they don't command the same prices as the DD-2.


Monday, 14 August 2017

Electro-Harmonix Pog (Big Box)

Following the Nano POG & POG2, here is the original big box Electro-Harmonix POG. The big POG has mostly the same set of features as the POG2 - it doesn't have presets and the filter has fewer settings, but there are separate detune sliders for the 1 & 2 octave up harmonies.

No slide pots were placed this time
In keeping with a trend I've noticed with EHX products, it's completely different on the inside than the other POG pedals. The POG2 and Nano POG use Analog Devices BlackFin DSPs, the OG POG is much more late 90s and has a DSP56364 and a 8051-style P87C52 MCU, just like some Line 6 products, or the Digitech Whammy. I have no idea why this cheap microcontroller family is so commonly found alongside the DSP56K series but I'm starting to suspect there is an application note somewhere that shows how to use these parts together. The AK4552 codec and 2272 & 5532 opamps are also very familiar by now. A TLV1543 ADC reads out the slider positions.



DSP56364

This pedal is also an outlier in that the complete factory schematic is online at Freestompboxes. The P87* series MCUs are OTP (but could be replaced with a compatible programmable part), so I have dumped this one. There is no other programmable part on the board so this should hold firmware for both the MCU and the DSP. The binary is here.
There are a lot of unpopulated parts in the PCB layout, some may be used for development and some look abandoned. There are footprints with silkscreen labels that indicate that they are for MIDI control and the schematic also indicates MIDI was planned at some point. I didn't experiment.

Trashed slide switch

No complicated fix for this one, once I had rigged up an 18V supply I found that the LPF mode switch was broken and a couple of wires looked like they were about to come off the board. After some solder touch ups everything worked fine. It sounds very much like the POG2, but as I don't own one anymore

I couldn't do a direct comparison.

Saturday, 15 July 2017

Ibanez FL9

A quick one. I got this original 80's 9-series flanger that didn't seem to be "flange-ing".

Ibanez Fl9

When plugged in it gave out that tell-tale "whirr" sound when power was applied that some kind of delay was happening. It sounded like the signal was delayed correctly and mixed in with the clean signal, but nothing was happening with the LFO, so there was no "movement". First thing to check is if the Speed pot is wired up or broken.


Back of speed pot

Hey, are those two lugs of the speed pot shorting together? I removed the pot and separated them with some needle nose pliers. The speed pot now varies the speed of the LFO and everything works as expected.

PCB topside

A quick shot of the board while I have this open. The BBD and clock generator ICs are MN3207 and MN3102, manufactured by Mitsubishi. Build quality is pretty high for what was standard at the time. All the electrolytic capacitors are Nichicon brands, some have "8138" datecodes so I think this is 1981 vintage.

The flanger sounds great, but is much more subtle than the MXR I looked at recently. This does chorus-style sounds much better rather than screaming airplane sounds. 9V operation is a bonus.

Line 6 Echo Park

The other Line 6 green delay. The ToneCore series was released around 2004, after the larger 4x4 stompbox modeller series. The series looks like they were intended to be small factor versions of models from the previous series, the Echo Park is something like a refined DL4, smaller and cheaper and shares a lot of the same delay models. It has the advantage of running off a standard 9V supply but omitted the looper that made the DL4 such a success.


Echo Park

In an unusual design choice, the pedals are split into two components, the "dock" (containing the main chassis, DSP, switches, jacks, power supplies and input/output buffers & amplifiers) and the "module" (holding the knobs, switches, a microcontroller and the program code for the DSP). This is the same idea as the DL4/FM4/MM4 series using the same PCBs with different program code, but allows different pedals to have different numbers or types of knobs and switches instead of shoe-horning everything into one shared format.

Modules and docks were sold separately and marketed as interchangeable. Red Panda even made a third-party module. Line6 released a programmable Developer's Kit module in 2008 that allowed hobbyists to create DSP effects that ran on the platform. This is a very cool idea but it doesn't look like it ever took off, there are very few examples online of anyone actually building anything.

This Echo Park was bought used and needing repair. The LED would flash once on when power was applied, but it was otherwise completely dead. I was hoping that I could easly isolate the problem to either the module or dock.

Unlike the DL4, there are no service manuals online. However, Line 6 did publish a ToneCore SDK Hardware Guide which has block diagrams and some slightly blurry schematics for the dock and the programmable Developer Kit module. The Developer module is not the same as the product modules, but it should be close enough to make repairs possible.

 ToneCore Dock



Dock, backside. Note test points.

Dock PCB, topside.

The insides of the dock look familiar, there is a DSP56364 dsp and some power and analog stuff. I managed to connect the dock and module together lying flying flat on my bench using some right-angle headers and some jumper wires so I could probe the signals between them. I was getting power and some of the clock signals were present, but I couldn't see any signs of the module programming the DSP when applying power (as the module contains the "effect", the DSP code must be stored there and transferred to the dock on startup).
There are testpoints on the backside of the PCB for MCU, DSP and ADC/DAC clocks, this is the place to look with an oscilloscope if you are debugging one of these.

Negative voltage generator

One interesting feature of this series is that they were the first (possibly only?) Line 6 pedal to run with a standard 9V DC supply instead of shipping with a bulky AC supply. The dock uses the 9V power and one of the clock signals from the DSP to form a curde charge pump to create a negative voltage rail of about -7V. This is pretty smart! I'm surprised it isn't more common in digital pedals.

Echo Park Module



Module PCB, as found

After opening the module it was pretty clear that someone had been in here before. Some traces were damaged and a SOT23 package device (Q1) was missing. A red jumper wire has been added. There are a lot of similarities with the schematic in the SDK manuals, the MCU is different (it's a P89LPC935F, an 80C51 derivative like in the 4x4 pedals) and the product modules add an SST25BF512 SPI flash memory chip but otherwise it looks close enough to use as a reference for repair. The manual schematic shows that the missing SOT23 package is a PMBT4401, I didn't have one at hand but I did have some BC847C that should be a good alternative. The schematic also had enough information to replace the damaged traces with some jumper wires after beeping things out with a DMM.

After repair attempts. One of the switches was removed temporarily to help with following hidden traces.

So I added transistor and the jumpers, and... nothing. Just one short LED flash when it's powered up and nothing else. I suspected either the flash memory or the MCU in the module were bad, but the MCU was now expensive to replace and difficult to source, and I had no copy of the contents of the memory so there wasn't much that could be done. So this went back into the box and into the "fix" pile and stayed there for a year.

Second attempt:

 

The line-up


Eventually I got two more non-working Echo Parks to look at from Moose Electronics. Maybe I could mix and match modules and docks to figure out which were good, maybe I could even fix all three. I labelled these to keep track, mine is #1 and the new ones are #2 and #3.

Starting out, none of the pedals worked. The immediate good news from swapping parts was that my dock (#1) worked with the modules from #2 & #3. No other permutations gave me a working pedal, so I knew my module was bad and docks #2 & #3 were bad. I opened the bad docks and one of them had a battery cable pinched between some header pins and sockets. Nothing else looked obviously bad. When I tested it again it worked perfectly. Either moving that cable or reseating the header pin/socket connection must have restored power.
The other bad dock had no obvious damage, and looked like it had a dead DSP. None of the test signals on the backside of the PCB were present, and they should be generated by the DSP, which did have power. In this case it's actually cheaper to buy a less desirable ToneCore pedal just for the dock than attempting to swap out the DSP, so this Echo Park will be getting a replacement instead of a repair.

This just leaves my bad module, #1. Suspecting the serial flash or the MCU, I decided to desolder the serial flash and see if it could be read in my programmer, and what data was on there. The chip ID was correctly identified, but the contents were all 0xFF! Effectively this was blank and there was no program code for the pedal to run. I pulled one of the flash chips from one of the working modules and dumped it, and re-programmed the empty chip. After resoldering it, the pedal lit up and worked just like the other one. I can only assume that the soldering "accident" I saw either caused the memory to be erased or was an attempt to fix it, I can't be sure.

Desoldering U1

Dumping U1

So, if I had had a memory dump of this a year ago I could have fixed my pedal then. In the interest of helping someone out, here is the contents of the flash.

Thursday, 29 June 2017

T Rex Mudhoney II, Again

I'm in two minds about posting about repairs on stuff I've already covered, but maybe highlighting common faults will help someone else troubleshoot. I got another Mudhoney II to look at, and this had the exact same issue as the last one.


Bypass didn't work. No power at the flip-flop IC that drives the switching JFETs. The trace to this IC was broken in the exact same way as the last one. It runs right along the edge of the PCB and underneath the jacks so this is probably a common fault.


Before
A jumper to the jack restores power and everything works.

After

This is kind of a shame, as it sounds great (dual RAT). If you have one with this problem then check this connection. If I had one that was working I would probably add a wire as insurance.




MXR EVH Flanger

Another eBay pickup. This is a slightly modified re-issue of the MXR M117 with an "EVH" button that adds a fixed preset for a Van Halen sound. Dead, as usual.



This pedal uses an 18V supply or two 9V batteries, like some of the original M117s. The insides however are the standard modern MXR stye - almost entirely surface mount, a red PCB and board-mounted jacks and pots. It's a very tight fit to the enclosure, but it looks fairly robust as everything lines up very well. Turning it on lights up the LED but there is no signal.


Before soldering the new DC jack


The fix for this one was fairly simple. The DC jack was broken, so that was replaced. Still no power at any ICs. The traces from the DC jack & battery harness run to a 15V linear regulator, which feeds a 9V linear regulator downstream. Replacing the 15V regulator (78L15) brought it back to life. The LED is powered from the unregulated 18V supply, everything else needs working regulators. Drop-out voltage on a 78L15 is usually only ~1.5 - 2.0V so if you use batteries you probably can't drain them below ~8.3V each before the pedal will stop working.

It's possibly that this was killed by an incorrect power supply, but there is a reverse polarity diode which is still intact and the 15V regulator should handle up to 30V

The insides are pretty packed but this is more-or-less a standard Flanger design. Here is an overview of the ICs used:

V3204. A Bucket-Brigade Delay chip, almost definitely made by Coolaudio, but they don't list it on their site. As the original MXR Flangers used a Reticon SAD1024 (dual 512 stages) this is probably a 1024 stage BBD. Coolaudio make a V3207 chip with 1024 stages though, so I can't be certain.
SA572D compandor. Compressing and expanding before and after the BBD for better SNR, Electricdruid has an article on how this works.
MC33178, MC33179, TL072: opamps.
MC14504b level shifter. I was surprised to see this, I think it's translating low voltage clock signals to the higher voltage ranges that the BBD wants. This could be done with discrete FETs but maybe this solution was cheaper.
HEF4013 dual D flip-flop. I would have guessed that this used for bypass, but looking at older 117 schematic this is used with an opamp LFO for generating the complementary BBD clock signals.
HEF4053 triple SPDT analog switch. This switches out some of the pots for fixed-value resistors when the EVH button is pressed. This is a neat solution, I think more pedals could use this for "channel switching" by having two sets of pots.

It sounds great, it's a classic design and the controls are broad enough to go from subtle to over the top which I always enjoy. I can't hear any clock noise or bleed-through. The EVH button may be a bit of a gimmick as it's not easy to activate by foot, if you bend down to push the button you could just turn the knobs.