Tuesday, 6 June 2017

Electro-Harmonix Freeze

The Freeze launched in 2010(?) and has been very popular. It's essentially a very short looper with some windowing to hide the "jumps", giving a smooth drone of whatever was captured. There is a decent claim for this being a genuinely new effect, even though there have been reverbs, delays and granular synth effects that can get similar results there is nothing that competes with it directly. I have never played with one so I've been looking out for cheap broken units.

This one passes no signal, crackles a little bit and I probably overpaid for it. The insides are more simple than I expected, a 56k-series DSP (DSP56374), a PCM3052A ADC/DAC, a serial EEPROM for the DSP code, an LM317, a TLC2272 opamp and a tiny SOT23-5 package that I am fairly sure is a NC7SZ66M5X SPST switch. It's surprising to see the DSP56374 as these are fairly long obsolete now, and other EHX products are using Analog Devices DSPs that are still in production. This might have been a 2010 purchasing decision, or they may have large stock as the 56k series is also used in other older EHX products. The PCM3052A is also fairly old but I have seen pictures of other Freeze revisions that used different codecs.

This Freeze passed no signal in bypass or effected mode, just some crackling noises. The LEDs did light up according to the Slow/Fast/Latch switch, and as there is no microcontroller I have to assume that the DSP handles this and that it is probably working correctly.

I fed in a test signal and traced it with an oscilloscope. I could see it at the input jack, through the ferrite bead and then at one of the input pins on the opamp (U2). It appeared to be wired as a buffer (makes sense) and there was nothing on the output pin. So the digital end was probably fine, it was just not being fed an input signal. Replacing this opamp fixed it.

In process of replacing U2

 To be honest, sometimes the process of fixing stuff is more fun than playing with it afterwards. The appeal for me is a mixture of curiosity about the technical stuff and the musical/creative aspect, and it can be easy to switch the things off after testing if it's "just" another delay, overdrive or whatever. The Freeze has consistently been a lot of fun every time I plugged into it. On the day I got it I spent a lot more time playing with it than I did repairing it (that has to count for something). I have used loopers before and found that they need a bit of planning and practice to use well, the Freeze is very immediate and it lends itself to noodly bullshit in the first couple of minutes of use, especially in latching mode. Thumbs up.

I expected to see some kind of microcontroller as the DSP 56k series have usually been paired with one in the previous pedals I've looked at, usually some kind of 8051 core (Line 6, Digitech examples). As the Freeze is a single processor design it might make this an interesting one to reverse engineer, there should only be code for a single architecture in the EPROM. I would guess that the effect is fairly simple, but the magic is in the delay length and the filtering to avoid clicking or popping. There is an assembly language manual available from NXP, I've looked for a disassembler and have found that it is supported by Ida Pro (great, but the full version is out of my price range) and some pretty old 90s tools that I haven't tried yet. If there is a decent free tool please let me know.

First step is dumping the EPROM and figuring out if the code is on there or if the DSP is pre-programmed. OpenOCD does support the DSP56374, so the programming header may be another option.

Monday, 15 May 2017

Electro-Harmonix Pog2

Much like the Nano Pog repair, this one is also written up after the fact and I have probably forgotten a lot of important details. I sold this one on, so this is mostly based on blurry pictures I took.

I got this with one of the slide pots broken off, and dead on powering up. The picture below is out of order and was taken after it was repaired.

All good.

Inside, the problem was immediately obvious. FB3 near the DC jack is missing but the pads have visible solder on them. It's very common for high-speed digital pedals to have ferrite beads installed on the 9V input and ground lines, to keep high-frequency noise out of the power wiring and comply with EMI/EMC regulations.

FB3 is gone.
My guess is that someone broke off one of the slider pots and tried to remove the board for a repair. It looks like an attempt was made to desolder the 9V jack and FB3 was lost in the process. I added an 0805 ferrite bead and the Pog powered up again, everything working.

 The architecture seems to be similar to the Nano Pog, with a larger Analog Devices ADSP-BF531 DSP, and additional controls for filters, volume swells and presets.

After desoldering the broken pot.

Replacement slider pots are available from Small Bear, replacements should be 10 kOhm. They also sell the rubber slider tips.

Update: I came across a similar repair at this blog that states the pots are 5k. My order confirmation email says 10k, I'm certain I matched the new part to the old one. It's likely that these are just used as voltage dividers and both values may be used in different batches or versions.

Compared to the Nano, I much prefer this version. The additional octaves are nice but attack/filter controls really take this from being a novelty to something pretty inspiring. Fun!

Monday, 1 May 2017

Rozz R8 Repair

Another analog delay with no delayed signal. I had never heard of Rozz (not Ross), apparently it was a brand applied by a Japanese OEM that released pedals under many different names. Effects database found quite a few variants, it seems that the Ampeg rebrand is sought after nowadays.


Freestompboxes has a schematic for the Guyatone variant. It looks like it was drawn using the official Boss DM-2 schematic as a starting point.


The narrow English font looks very Japanese.

Inside is a Mitsubishi MN3005 BBD and MN3101 clock generator, which is nice. The only other IC is an NE570 compander, all buffering and filtering is done with discrete amplifiers. Otherwise it's similar to the DM-2/AD80 style delays of the time, except that there is no "cancel" trim pot to dial out any clock noise, and the wet and dry signals are passively mixed with no output buffers. This might have been a lower-cost design.

PCB is the usual single-sided type used by Boss in the 80s.

Looking around the board I could see clock signals at the 3101 and BBD, but I was confused as they seemed to disappear when I flipped the board over to see the traces on the solder side. Pushing down on the MN3005 would give a momentary burst of delayed noise, so the delay circuit appeared to work, only with a broken connection somewhere. I moved all the wiring from lap-soldered pads to run through the PCB in case a bad joint was causing the problem. No dice.

Eventually I realised that the trimpot for setting clock rates was broken, it only made contact when the pedal was upside down! After desoldering it literally fell part, the wiper only contacts the conductive track when held in place by gravity.

Original timing trimpot.

With a 100k replacement pot the delayed signal was back. The available delay times were not quite right, it would happily go to ~1/2 second at one end of the pot but the other was not much shorter. Fast short delays were not possible. Clock bleed-through was also really bad, it would whistle at the clock frequency even in bypass mode.

100k temporary replacement.

After finding a 50k replacement things were a lot better. Clock rates were manageable from ~8 kHz up to 15 kz or so, allowing long and short delays. Clock noise dropped a lot but is still audible at long delay times - I think this will need a trimpot added to the BBD outputs to dial this out.

With 50k trimmer. Much better.

This sounds very nice, it does DM-2 style delays on the verge of oscillation very well. With some minor mods to remove any clock sounds I think these would be really nice analog delays.

Saturday, 1 April 2017

T Rex NeoComp

Another T Rex pedal with problems. This is a VCA-based compressor, and it was really noisy, like white noise with some low-frequency popping and crackling. Fortunately the noise was there in both effected and bypassed modes - this meant that everything was more or less working and that the source of the problem could be isolated to parts of the circuit common to both.

On the inside there are some similarities with the previous T Rex pedal, there is a CD4013B and momentary footswitch for latching bypass, so there are probably some JFETs nearby. MC33174 quad opamp. The compression is done with a THAT 4316 which seems like a nice chip for pedal applications. It's sold as an "analog engine" and contains a VCA and the RMS detector. The QSOP package will probably turn people off. An LM393 comparator is used for something, but I have no idea what for.

Main PCB.

Noise problems like this are usually caused by a bad semiconductor or a missing or failed capacitor. I like to find the source of the noise with an oscilloscope and then remove parts to see if the source disappears. Opening the circuits isolates things into two sub-circuits and narrows things down. In this case I ended up removing nearly all the ICs to figure it out. I started tracing out the circuit as well but solved the problem before I got very far with that. The SOT23 transistors in the corner of the board are driven by the 4013 and are connected together in the same way as the switching JFETs in other T Rex pedals ADD LINKS One of them showed a lot of noise when probed with the scope, so I replaced it with a J201 and that killed the noise. The originals are marked "G5   7" which I can't identify, but for switching it seems a J201 will work just fine.

Switching FETs are in the top right corner, faulty FET has been removed.

Apparently this noise problem is a common failure with T Rex pedals. Swap the JFETs.

Saturday, 11 March 2017

T Rex Mudhoney II

This is something like two slightly different Proco RAT circuits in one box. As received it only seemed to work with very heavily attenuated output volume. Maybe it was something to do with an electronic switching system.

Shown here without footswitches

The PCB photo came out very noisy, but the inside is just a CD4013 dual flip flop IC and 2 OP7 opamps driving some diodes to ground - I believe this is standard for modern reissue RATs. The red wire and replacement capacitor was a previous repair. The switches are held by rectangular cut-outs in the PCB, I like this method.

As well as being very quiet, the LEDs underneath the tone pots were not lighting up. I guessed that the 4013 was driving some switching FETs and it was dead.

First attempt at a fix was replacing the CD4013, as these are cheap and I can replace SOIC parts pretty quickly nowadays. No change. Probing the new chip I saw it was getting no power - I should have checked this first. I followed the traces from the 9V input outwards and they were hidden underneath the input jack. That had to be removed.

Input jack removed
Underneath was a trace carrying 9V that ran right next to the edge of the board and had been cut right through. I added a wire from to restore 9V to a power bypass cap at the 4013 and reinstalled the jack. Now the pedal lit up when engaged, and the left side (channel 1) was loud and sounded pretty great. The right side still had a very low output volume, much less than unity.

I hooked up a signal gen and tried to follow with an oscilloscope to see where signal was lost. Thi schematic was helpful. The opamp has outputting a 6Vpp signal but this was lost at the tone control, so the problem was probably somewhere in between. Testing the clipping diodes showed that one was shorted. I guessed and lifted one off.

One clipping diode removed.

This was lucky, that diode tested as a short out of circuit. This side of the pedal was now much louder, but less distorted than the other side as only half of the signal is getting clipped. These diodes are MELF packages and I can't find any identifying markings on them. The remaining diode measured around 0.5-0.6V on a DMM, so probably something silicon. Internet consensus seems to indicate that these pedals used 1n4148, and I had MELF 4148s at hand, so I used that.

My hand soldered replacement looks different to the reflowed diode. The good news is that it sounds the same as the other channel, and now both will get fairly loud. I forgot how nice a RAT can sound.

Sunday, 5 March 2017

Dunlop Cry Baby 105Q Bass Wah

Unlike the traditional Crybaby, the Dunlop bass wah doesn't use a stompswitch to switch the effect. Instead the pedal is spring-loaded to return to the heel position, where it engages a small momentary switch which bypasses the pedal. You only have to lift a foot off to disengage the pedal, which makes using it for just a few notes much easier (maybe this was thought to be more useful for bass players). The downside is that it can't do cocked-wah sounds without a foot holding it in place.

This one doesn't switch at all, the output signal is always in bypass mode. It does seem to work in that it buffers the input signal, and when power is removed no signal is passed. This is likely to be a problem with the switching system.

The insides are very different to the Crybaby. This is an inductor-less wah, with 2 quad opamps doing the filtering, presumably with a gyrator design. The normal Dunlop Hot Potz is used.

Momentary button desoldered
The switching system uses a small momentary button soldered on to the hidden side of the PCB and poking through the top of the enclosure. One side of the switch looked like it was always closed, so this was replaced with a new one - without fixing the problem. The transistor is just buffering the switching signal, with the new switch installed this could be seen changing from 0 to 9V as the pedal was lifted and dropped, so this part of the PCB is working correctly.

Main board.
All of the rest is done with 4 ICs. The MC33174 opamps are almost certainly doing the filtering, the CD4066 analog switch is routing signals when changing from bypassed to effected modes. The signal from the momentary footswitch is running into a CD4049 buffer. There are some Rs and Cs around the buffer, I would guess it is de-bouncing the switching signal to cut out any popping. There was no activity on any of the CD4049 pins.

4098 buffer desoldered.
I took a picture after I removed U3 but apparently forgot to take one after I soldered a new part. The new chip correctly sent on the switching signals and the pedal wahs once again.

This is a pretty nice design, I only tried it with a guitar as I didn't have a bass around at the time. I would have guessed that the pushbutton is the weak point, but in this case it seems an IC died first.

Saturday, 4 March 2017

Empress ParaEQ Repair

This is another case of getting to look at something  nicer than I would normally get to play with. When it comes to guitar gear I am pretty cheap, and it's hard to justify spending a lot on "simple" pedals (this is a dumb distinction to make). Empress' ParaEq seems like it's probably worth it, it's a good idea executed really well.

This is basically a 3 band EQ (low, mid & high) with adjustable centre frequencies and independent Q settings. There is an input pad and a clean boost as well. The build quality is very high - the toggle switches are positioned behind a row of knobs to protect them from stomping feet, and the pots all have nice metal knobs. Empress seem to do pedal graphics with enough personality to not be completely boring to the eye.

Revision 7, Jan 6 2014. Copyright date is 2013...

Inside we have:
- OPA4134 & 2274A quad opamps, buffering, amplifying & filtering.
- an ADG442 analogue switch for the buffered bypass
- TC7760HE charge pump controller for a negative voltage rail.
- 16F630 Pic microcontroller, monitoring footswitches to change between relay and buffered bypass.
- a SOT223 package, marked "4576". I'm certain this is an LDO and I think I had the part number and looked over a datasheet but I have lost it again.

Plugged in, the pedal was completely dead. My little test amp was humming, which usually means a power supply is shorted out somewhere and some filter caps can't do their job. Measuring inside the pedal, the DC jack read at ~2.5V (using a 200mA Boss power supply).

Flush mounted DC jack

No through-hole anchors for the jack.
Something is pulling power to ground, usually this is a bad power filtering capacitor, bad IC or a shorted protection diode. Empress are using a surface mount DC jack which must be desoldered to remove the PCB from the enclosure. I guess they really wanted the power jack to sit flush with the edge of the case. I took it off to see if there are more parts or any damage on the backside.

There is nothing on the backside. I took pictures of pots while I was here.

Gain pots are B5K.
Frequency pots are marked "B1 P1416C"

As there are not too many ICs on the board I tried to order them in likelihood of failure and then removing them to see if the short cleared. Pulling these chips off is pretty fast with hot air.
The 7660 was first, as I have seen these fail before. No change. Then I removed U1, the sot-223 LDO and the short was gone - but as this chip is powering all the opamps I couldn't say of this was the problem or not. U101, a 2274A quad opamp had a mark on the casing which may have indicated some kind of failure, so I removed it and replaced U1. Short was gone, I now measured +8V at the opamp positive rail and -6V at the negative. It passed a signal, the LEDs now lit up when the footswitches were pressed, and the boost worked. The EQ controls did nothing, so obviously this opamp was handling some of the filtering.

U1 & U25(?) removed.
U101 removed, short cleared.
After soldering in a new opamp everything worked as expected. Re-soldering the power jack is easy with some kapton tape holding it flush to outside face of the enclosure.

This is a more useful design than I had expected, parametric EQ solves a different problem than a graphic EQ. It's much easier to sweep the frequency control with a high Q and identify exactly what will be boosted or cut, it always allows adjustment of frequencies "between" the bands available on a graphic EQ.

The only downside for longevity is the PIC - again, this is a programmed part. If it was bad I would not have been able to replace it and would have had to re-wire for true bypass.