I really like your comparisons and waveforms. They are great and shows the characteristics of the synth.
The filters, amp, wave shapers do all kind of stuff to the original waveform depending on pitch, amplitude, temperature and even what has been played before(ringing filter for example). That is what makes the different synths different, and musical, in different ways.
The “triangle” (it’s not very triangle) of the A4 is also a really interesting subject.
When I look to the oscilloscope, A4 square isn’t so different from SE1 square thru A4 filter… I think one problem comes from the filter.
Yes and precisely why others don’t as much.
It’s like any synth though. They all have a certain character or sound.
You either like it or not. Just accept it for what it is or sell it. Pretty simple really
p.s. for interests sake Elektron can you please put a filter bypass switch on A4 / AK mk2 so we can confirm once and for all 
Could you reply to me in your support case and we will test this together.
With pleasure but I don’t know where. Did you send me any message I didn’t answered?
I have read this thread 3 times and I own 2x A4’s
I am I correct in interpreting that using EXT IN does not remove bass frequencies from the source?
I didn’t notice audio difference when comparing a square and the same signal thru EXT IN. Anyway, I couldn’t cancel signal with phase reverse.
Here are oscilloscope freeze (1st is the original SE1, second one is SE1 thru EXT IN) :
One could think it isn’t equally calibrated, but that’s the best gain setting to get same Vu meter level and audio sensation.
Not exactly the same… but I think it is OK to use EXT IN.
This oscillator “problem” is as old as the A4 itself. There’s even a thread on the old elektron-users site.
What is a “proper square wave,” anyway?
What is “normal?”
Fitter, better, happier.
The TB-303’s square wave looks more like a melted candle than an actual square.
A4 is what A4 is; I can’t imagine it any other way.
It took me a week to accept the A4 as an instrument with its own character. I may not get the OSC behaviour that I am used to, but then I have some VSTs that cover that.
What did help me was the fact that the A4 can be used as a monosynth with up to 4 layered but independent voices. Whenever I need more power in bass, I can layer a second voice, either a subby triangle or a detuned copy of the original sound, or a neighbor track with a resonant high pass. There are plenty of options to create deep bass when you go beyond one voice.
^Yea!
Unison use track sounds is massif!!!
I agree with all of you. It’s possible to get perfect waves with other gear. Anyway, Analog 4 would be better if it could respect input waves thru the filters before tweaking. I miss the “normal” option. And about HiPass res to get more bass…that could be even better. Other HiPass filter do that job in a more accurate way. I would prefer a HiPass filter that doesn’t filter any low frequency if it is set to its minimum level (other HiPass filter have the same “problem”).
I kept the Analog 4 for all the valuables functions in despite of several weaknesses. I was very disappointed at the beginning, because I thought I could get Prophet 6 pure sounds. At the same time I had a lot of pleasure with mangling ability, or even sweet sounds. I love to use it with 0 Coast. It controls 4 CV, and adds overdrive, AM, feedback, filters… very funny. I also love the fact overdrive is part of each voice sound path. It’s the only analog polyphonic synth I know that has dual filter with overdrive and feedback per voice… It’s wonderful to make strange organs for example. That inspired me new Nord G2 patches. Sure, they don’t sound exactly the same in the Virtual Analog way… but that’s also interesting. Everything leads to use G2 
maybe try a keytracked peak filter then
In what way more accurate ?
thankyou very much for this!
Have a look here fwiw
I hadn’t seen any of the tail end of this interesting thread when I posted just above here, there’s lots to ponder
One thing that has popped up here again is the lack of key track as a modulation source, it could be handy for adjusting the pulsewidth of all shapes to achieve consistency or otherwise amongst many other things
Okay - I was somewhat curious, i needed to satisfy myself that some of @guiltyrollmops’s findings were the same here, so i set up a little test
I’ll present the findings in a similar manner, this is not necessarily the full story, but it’s certainly a fascinating insight into the character of the A4 filter, and crucially a lesson into how to set it to be ‘neutral’-ish
Please note, this was looked at with no key tracking, normal default res ‘boost’ and for an audition/viewing friendly frequency 65Hz-ish
I fed the A4 a Nord Modular G1 signal with three different shapes for info, having been amazed by the influence of the filter (and variations) on the square - for the square there is a recommended ‘flat’ filter setting (25 res on LPF ) and a value which gives better visual correlation (a pulse width varied shape was thrown in as it wasn’t always easy to see where the duty cycle of the post filter square was) for the saw and tri i have added in a zero res setting to show just how much the filter will impact the waveform, so they’re both shown at 0,25,64 (where 64 gives a closer approximate visual representation of that wave at that frequency) (Note no filter key tracking)
Two identical signals were sent into the A4, one direct via Ext In roughly level matched to a signal pulled clean through the Oscillator Injection of Osc 1 (so the A4 oscillators are all off) with max amp level (ext in level lower)
Both pulled through OB/Live to a pair of waveform monitors running simultaneously on Left and Right respectively, that way you can view the source and the impact on the source in real time
Note this doesn’t tell the whole story and it may be a flawed analysis, but it’s an interesting insight nonetheless - those high res settings may colour the sound somewhat, especially in other frequencies, so this is a partial insight into a complex picture
The filters are certainly not transparent, it does indeed beg the question, could they be bypassed for an extra flavour of sound, it’d be interesting to see what shape the A4 sends out pre filter, i bet they are more ‘regular’ looking than we all assumed
I’m glad i caught this part of the discussion above, i missed most of it since it got revisited a week or so ago, interesting … i didn’t try with resonance boost off fwiw, but it’s clear that the filter pair are a big player in the overall sound
It’s fascinating to see the impact of varying the peak filter full range, it inverts the signal when you go from freqmin/res0 to freqmax/res0
Anyway - for all the figures, they are presented as a snapshot of the filter settings and the A4 inject via filter is the upper waveform, the lower waveform is the direct one via Ext in - both are sent through main output via OB (i was monitoring pre OB) - not sure if i should have been doing anything about the DC offset
It would have been better to include a spectroscope
res 25
res 25 (reduced pulse width)
res 70 !
res 0
res 25
res 64
res 0
res 25
res 64
I meant… A HiPass that only filters under cutoff frequency.
You:re right. Peak filter can be useful.
Very interesting!
Res 65 usually looks better, no?
It gets closer visually (at those osc settings), very strange, but it was an interesting learning exercise
In the interest of fairness, i noted that the input level can also make quite a big difference, there’s a threshold after which the input level will not be matched 1:1 by a throughput level but the shape will distort as the disparity between levels gets more noticeable, even on a sine wave … it can be very accurate up to a certain point and the level is increasingly constrained, but the shape will change
So I ran these in Max to overlay them, but with a reduced input level to the filter stage (reso boost was off, but it seems to make no difference to the shape at these input/filter settings anyway) … i tried to match the energy at the nominally flat resonance 25 setting so that a tri/saw were roughly level, then left all settings and levels alone, except for resonance and i varied the shapes, this time presented showing the influence of resonance on the apparent transparency of the filter block, it gets more analogous with higher resonance at these osc/filter settings, but it may be problematic at other frequencies
attached for completeness - square is still suffering the most visually, there is some phase shifting due to the odd post-filter waveform and the detection algorithm struggles to find the rising edge
seeing this has not changed my view of how to use the machine, i still use my ears, but if you’re striving for a certain sound, it may be good to appreciate some of these nuances … resonance really plays a big part in this
So, sin/tri/saw/square Nord shapes (only varying the A4 redo, nothing else) (all else as per post above, including caveats) Note : no A4 oscillators
In all shapes, the increasing resonance is possibly having the effect of reducing the energy levels (away from the resonant freq) though it’s harder to read for the saw/sqr
various resonances
various resonances
various resonances
various resonances
