A few months ago I scored a used VM1 for 20 euros, looked like new.
The 'playing a strat into an AC30' youtube video comparing it with a DMM sounded so promising.
Mine sounded nothing like it, but I have no strat nor ac30 and all my guitars have rather hot and high output pickups.
I greatly prefer my Memory Toy, which a lot of people think is rather dark, but compared to the VM1 it is bright!
The VM1 I have is a murky, muddy, boomy, distorted mess of repeats.
So after playing it a couple of hours it ended up in a drawer until a few days ago, when I decided to 'fix' this thing.
I knew very little about bbd delay circuits but after spending countless hours comparing the circuits of different DMM's and of those that are based on it and reading what every part does, I finally managed to get some awesome tones out of it.
Too bad there's so much smt in there which makes extensive modding pretty much impossible for a diy person.
So, most of my mods were chosen because they did not require removing or replacing smd components.
So here it goes:
1) Input impedance:
100k input impedance? And on top of that, it's left hanging there in bypass?
I unsoldered one leg of C6 ( a greenie, non smd), so I could place a 1M Ohm resistor in series with the 100k (R2) input resistor.
This brings the input impedance up to 1,1M Ohm.
While you're in there you could also add a 47nF cap in parallel to the greenie, or replace it with a 100nF cap to raise the input capacitance. This will let some more bass in, but eventually I removed it and kept it at 47nF. Lower notes contain lots of energy and since this thing already distorts so quickly, I'm keeping them out for now.
The gain of the preamp has now dropped to less than 1: 22k (R5) + 1M (VR1) / 100k (R2) + 1 MOhm (new) = 1,022/ 1,1 = 0,93.
With the dry/wet mix between 60:40 and 50:50 I found, I needed a gain between 2 and 3 to compensate for the volume loss.
The best solution would be a 3M pot instead of the 1M (VR1) but I don't think that exists so I choose for the easy way: solder a 2,2M resistor in series with the 1M pot (VR1)
Not wanting to unsolder where not needed I simply clipped one outer lug of the pot in the middle and bent the two parts (one sticking in the pcb, the other in the pot) a little outwards and soldered the 2,2MOhm resistor in between. I know, it's a hack, but this thing cost 20 euros and it's an easy fix if you want to revert to stock.
The only downside of this fix/mod is that the minimum gain (level pot at 0) is now 2 instead of 0,2: 22k(R5) + 2,2 Ohm (new) / 100k + 1,1MOhm (new) = 2,222/1,1 = 2.
The max gain is 3 with the level pot at 10.
It's a compromise and you might want to experiment with slightly lower input resistors in series with the 100k instead of my 1MOhm. That would reduce the input impedance and increase tone suck in bypass though. Or go for a new pot to have more gain range.
I find the gain range perfect for my needs. I'm not putting a delay pedal on my board for the 10db boost anyway...
Result:
1)Bright guitar tone going in and consequently, brighter repeats coming out. I might have stopped there.
2)True bypass is no longer absolutely necessary. Having that 1,1M Ohm input resistor 'hanging' there in bypass is not perfect but it's like plugging your guitar into a stompbox or amp with a 500kOhm input impedance. Acceptable.
2)Pre and post filtering and make up gain in between bdd stages:
This is where (according to the experts) the VM1 fails compared to the DMM.
I plotted/simmed the simpler filters and couldn't find anything out of the ordinary compared to the DMM.
Yeah, some filters are passive instead of active through an opamp, but the corner freqs seem the same.
The 3 pole post filtering is done with transistors instead of opamps, but that shouldn't matter if done correctly.
Since most of the filter components are smd there was little to try out there.
Then there's the 'oh so important' make-up gain in between bbd stages. I fail to see the importance or why it's called that way. It looks more like a gain trimmer than a gain make upper to me...
It took a lot of hours of googling to find out its purpose or why it's so important.
Eventually I found a very old EHX service note and also some calibration tips for a DMM based delay on the GGG site.
They both said the same thing: It's there to match the output of bbd1 to the input of bbd1. So it seems bbd can add gain which could need to be trimmed back.
I checked with an audio probe right before bbd1 (C52) and directly after bbd1 (TP1).
If there had been a trimmer I probably couldn't have set it more equally.
If there's any other purpose (bias?) I'd like to know but afaic it's not needed here. And if it were, I guess the Behringer dude wouldn't have left the other half of IC4 unused. It's not as if it cost them another IC.
One thing I did notice however, when probing the input at C52, was that there was already a distorted signal coming in. No wonder that turning the bbd bias trimpots didn't give me any cleaner repeats than the factory setting.
Result:
I'm no longer worrying that the Behringer dude messed up the filtering and overlooked the importance of the make up gain. As it stands, afaic, nothing to do here.
PS: The example Dirk_Hendrik posted above for adding the makeup gain is a cut and paste copy from the 'official'/factory EHX DMM schematic.
Unfortunately Mr Dang from EHX made a mistake, so the copied example above contains the same mistake. Rin should be outside the NFB loop.
No disrespect for Dirk_Hendrik, as he's probably the only person on the planet that has ever crammed a DMM in a 1590.
3)Compander:
With the raised input impedance and a bright guitar signal going in and brighter repeats coming out most people would have been happy about it.
But when investigating and debugging, one tends to listen for every tiny nuance and it's at those times your ears tune into a little thingy and then you're doomed. You can never again 'unhear' it.
The white noise was 'dirty'. It didn't go ssssshhhh but ssRRhhhsRRRhhhSS as if the white noise itself was distorted. From that point on it was the only thing I could hear with every repeat.
After a lot of Googling I finally stumbled upon a forum post by Mark Hammer, on the 'other' forum', and a blogpost about 'ripple' in companders that could explain what I was hearing.
Here's the blogpost or article:
https://electricdruid.net/noise-reducti ... ompanders/
And more specifically:
The C1 and C7 1μF values are a compromise between fast response and low ripple. If you want a faster response, try 470nF. If you want less ripple, try 2.2μF.
Checking all the dmm (and pedal derivatives) schematics I had and comparing them to the VM1 schematic, I notice that the VM1 is the only one using such low values: 200nF for C32 and C33. Others used at least 470nF and the DMM uses 1uF.
So, VM1 opts for fast response at the expensive of more ripple. Others go for slow(er) response and less ripple.
I soldered extra caps parallel to C32 and C33. I tried several values: 470nF, 1uF, 2,2uF, 3,3uF.
I ended up using what the DMM uses: 1 uF.
Result:
Dirty white noise is completely gone. It does a smooth sssshhhh. The trails of the repeats are much cleaner now.
Higher values removed even more white noise, besides eliminating the ripple, but it also seemed to slightly dull the repeats. 1uF was perfect.
I won't deny it's a subtle thing, but as said before, once you know it's there, it's the only thing you can hear.
As for the slower response, that's a subtle thing as well. Similar to dialing in a compressor. You want to make it do something, but not so much that it becomes obvious or overbearing.
It's a 'you'll only notice it was there, when you turn it off' kinda thing.
Big improvement afaic as this really bugged me.
I got the slight impression that the max delay time became a tiny bit longer as well, but it's probably an illusion. I'll have to measure it again.
Before all the modding, I measured it to be 499ms. My Memory Toy does 650, but is only completely clock whine free up to 500ms.
4)Distorted repeats, boomy, murky mess and overeager self oscillation.
Going back to investigate the already distorted signal entering the first bbd and the way over the top oscillation I decided to temporarily turn the gain of IC3A down from 2 to 1. Calculation: 1+ 200k (R10) / 200k (R9) = 2
I soldered two wires at each end of that tiny smd R10 resistor. The ends of the wires were then soldered together, actually jumpering R10 and turning IC3A into a unity gain buffer (gain = 1+0/200k)
Afterwards I'll maybe solder a resistor between the wire ends to have it parallel to R10, or maybe a 1 MOhm pot, to reduce its gain but not all the way down to 1.
Having a unity gain buffer there does impact the gain of the repeats, with a good chance of loosing the possibility to make it oscillate.
However..., while doing that I stumble upon the feedback loop and more importantly R26 and C36.
While every DMM schematic I have, uses:
100k + 22n
VM1 uses:
100n + 22k
Coincidence? Or did the Behringer dude mess up and accidentally swapped values?
If EHX had used a value for the resistor that wasn't available as cap or vice versa, he might have noticed.
Is that the reason for trying a 50k FB pot instead of 10k?
Anyway, time for another snip and tuck as I'm still not unsoldering smd parts.
The good thing is that for both resistor and cap we need additional series components: 78k series for that 22k resistor to obtain 100k and some cap between 22nF and 30nF to obtain 22nF (100nF || 28nF = 22nF).
I did the same thing as with the level pot: cut the middle lug of V3 in two, bent the two lug parts a bit outwards and soldered wires to lug parts. The wire ends are soldered to a 100k pot (instead of a 78k fixed resistor) and that in series with a 22nF cap.
Result:
Gone is the boomy murky mess of repeats, welcome bright repeats.
Gone is oscillation as well but that's probably because I reduced the gain of IC3A and turned it into a buffer. I will need to up the gain a bit again there.
Coming out of the expander, the repeat goes to both the output IC4B (and eventually to the mix/blend pot) but the same repeat also goes back to the start of the circuit, so that the repeat itself can be repeated.
With the faulty(?) 22k resistor the repeat-signal going out was (too) soft and going back in it was too strong, resulting in runaway feedback.
On top of that, due to the faulty(?) 100nF cap the feedback repeat retains a lot of the low frequencies, turning it into a muddy, farty mess.
With the correct(?) 100k resistor the repeat going to the output is much stronger and the same repeat going back in is softer.
With the correct(?) 22nF cap a lot of the lows of the feedback repeats are removed resulting in brighter repeats and less oscillation.
In stead of the fixed cap I installed a switch with a couple of caps to choose from.
Together with the 100k pot instead of a fixed 100k resistor it gives great control over the way the feedback loop sounds and behaves.
At the moment IC3A is still at unity gain and with the feedback knob at 10 I have infinite repeats. It does turn into some kind of droning, but because IC3A is at unity gain it never gets out of hand like every other delay does. Once it's saturated it just keeps going but at the same loudness level.
Maybe hard to believe, but I'm wondering if the repeats have become too bright now...
DMM does roll off some highs in the feedback loop. LP filter with R35(1k) and C24 (47nF) in the factory schematic. 3db down around 3,5kHz. That could easily be added to the resistor and cap we added in the FB.
Additional notes:
I hated the chorus on this thing but now with the brighter repeats, for some odd reason, it suddenly sounds great.
I never understood how people could use delay as (faux) reverb. On my delay units it was either a rockabilly slapback which to me does not sound like reverb or I had the flying saucer sounds because when I wanted to dial more repeats to simulate longer reverberation the feedback just ran away. But now that I'm able to precisely control how long it takes for the oscillation to come down and not runaway at the slightest strum, this pedal makes for a great reverb. Especially with a hint of chorus.