Polysix Deeper Bass - Properly Jumpering C61

Following this post and this one, I've decided that I like the sound of my Korg Polysix when bypassing C61.  So, I've decided to remove the clip leads that were shown in the pictures in those previous posts and replace them with a proper jumper wire.  So, I cut one to length and soldered it in.

The Red Wire Jumpers Around C61 by Connecting the Left Leg of Q15 with the Right Leg of R115.

Again, the Red Wire Jumps from the Left Leg of Q15 to the Right Leg of R115.

Now, I agree that the proper thing to do would have been to remove C61 and to solder a jumper wire into the holes of C61, but I didn't want to do that.  I just soldered in the red jumper wire shown above.  I don't perceive any additional added noise by using this flying lead, so I think that it's probably OK.  Plus, if I choose to un-do this mod, removing the red wire is easier than finding and installing a replacement for C61.

Smell the solder!

Polysix - Frequency Response with Deeper Bass

In my previous post, I modified my Korg Polysix to strengthen the deepest bass frequencies.  The key is to bypass (or remove) C61 on the KLM-368 Effects PCB.  In my previous post, I attempted to show the frequency response due to this modification, but the graph was pretty poor.  Today, I have taken new measurements and made a much better graph.  Now we can clearly see the effect of bypassing C61.

Lower Cutoff Frequency:  This graph clearly shows that the low-frequency cutoff for the synth drops substantially by bypassing C61.  As measured at the -3dB point, removing C61 drops the cutoff from about 62 Hz down to about 20 Hz.  This means that removing C61 extends the deepest bass frequencies.  Whether or not this is a good idea is up to you.  For me, after living with it for a few more days since my original post, I like it.  I think that I will keep it.

Let's talk about some details of the measurement technique...

Measurement Approach:  By treating the Polysix as a "black box" system, I evaluated the frequency response by measuring the transfer function of the "black box".  I did this using a standard technique -- I injected a known broadband signal into the system and I recorded the output signal that was generated by the system.  Comparing the output to the input yields the transfer function.  By looking at the transfer function in the frequency domain, you get the frequency response of the system.  In this case, of course, the "system" is my Polysix.

Injecting the Test Signal:  All of the circuits that interest me at the moment are on the KLM-368 Effects PCB.  To measure its frequency response, I need to inject my signal before the audio pathway gets to KLM-368.  I chose to inject my signal at the end of Voice 1 on KLM-366, just before it is mixed with the other voices and sent off to KLM-368.  As seen in the picture below, I injected my signal at R133.  To allow my signal to mix properly into the synth's audio path at this point, I used a 10K resistor in series between my computer (which is playing the signal) and the green clip lead shown in the picture.  For the "output" of KLM-368, I simply recorded the main output of the synth because there is very little circuitry after KLM-368.

Injecting my Signal on the Lower Leg of R133 (the Green Clip).  Not shown is the 10K Resistor Between my Signal Source and the Green Clip.

Processing with Matlab:  To produce the frequency response graph shown at the top, I processed the audio recording of the input signal and of the output signal using Matlab, which is unfortunately not cheap nor readily accessible.  It is a very good programming environment for doing this kind of signal processing, but there are other choices.  The Matlab functions that I needed are the FFT function (which converts time-domain signals into frequency-domain signals) and Matlab's plotting functions.  As an alternative to Matlab, I believe that this analysis could be easily done in Octave (which is free) because it has a perfectly fine FFT function, as well as, perfectly fine plotting functions.

Compute the Transfer Function:  Whatever computational tool you use, the core of the calculation is to take the FFT of the output audio and divide it by the FFT of the input audio.  This division operation in the frequency domain yields the output/input transfer function of the system being measured (in my case, KLM-368).  Take the magnitude of the transfer function, plot as "dB", and you've got the amplitude response as a function of frequency.  This is what I show in my graph.

Chosing the Input Signal:  For anyone who has made these kinds of measurements before, you know that there are several different choices for "broadband" input signals that one can use.   Ideally, the input signal is flat in the frequency domain, so that any deviation from a flat output is most easily assessed.  The typical choice is to use either a linear frequency sweep or some random white noise.  Personally, I like to use noise.

Maximum Length Sequence:  In the category of "random white noise", I chose to try something new...instead of traditional Gaussian white noise, today I tried using a Maximum Length Sequence.  Unlike traditional white noise, which is only truly flat in the frequency domain after lots and lots of averaging, an MLS sequence is designed to be perfectly flat within whatever fixed period of time that you'd like.  As a result, you get much smoother results in a much shorter recording.

Smooth MLS Results:  I generated a sample of MLS using the "MLS.m" routine downloaded from the Matlab File Exchange. I generated the sequence and saved it out as WAV file, just like I would do for any other noise sample.  After running it through the synth and processing the results, I get the very nice graph seen at the top of this post.  This is the first time that I've used MLS and, given the smoothness of the graph (copied again below, but with different annotations), I like how the results turned out.

One More Look at the Graph:  OK, sorry for the digression about transfer functions and maximum length sequences.  Let's get back to the results at hand.  However I got there, this new graph shows the frequency response of the synth much better than my old one.  It shows that the effect of bypassing C61 is substantial, but only at the deepest frequencies.  As a secondary result, I also see that the KLM-368 PCB (with or without C61) produces a sizable boost seen in the treble frequencies.  I believe that this is the effect of Korg's built-in treble boost that was discussed in this older post.  I'm going to address this "feature" in another post later.

Update: I decided to properly bypass C61 using a jumper wire instead of my clip leads. See here.

Polysix Deeper Bass - Bypass C61

Ever since this post back in February, I've been interested in optimizing the frequency response Korg Polysix.  In that post, I discussed how both the highest treble frequencies and the lowest bass frequencies diverged from a pure sawtooth wave.  Later posts (especially this one) discussed how I modified the treble response to get what I wanted.  By contrast, the bass response seemed to diverge from the ideal only by a little bit, so I didn't put much thought into it.  As part of his post on Yahoo Groups, though, Tony from Oakley Sound remarked that C61 on KLM-368 appeared unnecessary and couple probably be removed.  Being an AC coupling cap, C61 naturally acts as a high-pass filter, which could possibly affect the bass response.  So, if I want more bass, maybe I should remove it.  Here, I discuss how I bypassed that cap.  To get straight to the good stuff, below are audio comparisons of the "before" and "after" (you'll want good speakers or headphones...we're talking about changes mainly to the deepest bass frequencies!):



The modification that I performed is located entirely on the KLM-368 PCB ("Effects").  An excerpt of the schematic is shown below.  C61 is a 1 uF electroltyic capacitor that the uneffected must audio pass through.  If you want to remove this cap, you could simply de-solder it and put a jumper in its place.  If you merely want to try the mod without removing any components, you could instead use clip leads to jumper from one leg of C61 to the other leg.  Unfortunately, my C61 was soldered too tightly to the PCB, so I chose to jumper between the easily-accessible points on R115 and Q15.

On KLM-368, Bypass C61 by Using a Jumper Wire from R115 to Q15

A photograph of my modified circuit is below.  The green wire is my added jumper.  Also in the frame, but unrelated to this modification, is the yellow wire and the empty IC socket.  Those changes were performed as part of a different mod to remove the post-effects VCF, which improves the apparent attack of the synth and which alters the high frequency response of the synth.  More info on that mod can be found here.  Today's mod, though, is just about the addition of the green wire.

The Green Jumper Connects the Right Side of R116 to the Left Leg of Q15

To assess the impact of today's modification, I recorded the audio at the synth's main output.  I recorded it with my trusty M-Audio Microtrack, which I used for all my previous assessments of the frequency response of my Polysix and Mono/Poly.  For this test, I set the Polysix for a simple sawtooth waveform, using the lowest octave.  I set the VCF to about "5" with no resonance.  Below is a visual comparison of the raw waveforms output by the Polysix when playing the lowest "C" on the keyboard.  The top plot is the unmodified Polysix, which has its C61 in place.  The bottom plot is the modified Polysix, which has C61 bypassed using my green jumper wire.

Recorded Output of the Lowest "C" with a Sawtooth Waveform (VCF at "5")

In this plot, you can see that the unmodified Polysix decays back to zero (which is the red horizontal line) more quickly than with the modified Polysix.  Remember, the synth is trying to do a pure sawtooth wave, so any "decay back to zero" is an indication of some amount of low frequency attenuation in the synth.  The fact that the modified Polysix decays back to zero less quickly means that it has better low frequency performance.

If we're discussing frequency response, we really should be looking at the signals in the frequency domain.  So, below, I do a frequency analysis of two 5 second audio samples of the output of the synth.  The blue trace is the unmodified Polysix, which has the C61 in place.  The red trace is the modified Polysix where C61 has been bypassed.  Again, this if for the lowest "C" on the keyboard, which has a fundamental frequency of about 32.8 Hz.  That is a *very* low frequency.  Comparing the two traces, we see that bypassing C61 seems to increase the synth's response at this frequency by about 6 dB.  That's a pretty big change!

Comparing the Frequency Content of the Lowest "C" Using a Sawtooth Waveform (VCF at "5")

In the real-world, 32.8 Hz is too low for our loudspeakers or headphones to reproduce accurately (especially for hobbyists like me).  So I'm not sure that I'm able to hear the impact on these deepest bass frequencies with my equipment.  But, you'll see that the next couple of harmonics (66 Hz, 99 Hz) are also slightly stronger after the modification.  My system can easily reproduce these frequencies.  So, when I'm playing my synth (or when I'm playing the Soundcloud demos at the top of this post), I do hear a difference between the unmodified and modified conditions, I'm just not sure its the change at 32.8 Hz that I'm detecting.  That is some seriously deep bass..

Given that I do hear some difference in tone (whether at 66 Hz or at 32.8 Hz), which version do I prefer?  Certainly, for raw visceral excitement, I like the added thickness and rumble of the full bass experience resulting from bypassing C61.  But, the Polysix isn't intended to be a deep-bass rumble machine -- it isn't supposed to be a Minimoog, or even a Mono/Poly.  Instead, it's a polysynth meant for chords and pads and strings and such.  So, when used for these purposes, perhaps my modified Polysix now as too much bass.  I think that it might sound too thick, too bloated.

It's interesting (to me) to note that the Polysix's main competition back in the day -- the Roland Juno 6/60 -- includes a high-pass filter as part of its architecture.  One use for the HP filter is to cut the low-end rumble to purposely make the sound more skinny.  For chord stabs, a skinnier sound can often sit better in the mix, especially when you've got other instruments providing a deep and punchy bass line (such as for dance music).  Perhaps this ability to control the low end bloat to sit better in the mix is why the Juno's continue to be more popular than the Polysix.

But that's a digression...

Back on the topic of the increased deep bass on my Polysix, I'm still deciding whether or not I like the modification.  I'm going to have to live with it for a while to see.  Thoughts?

Update: Better frequency response graph here along with more discussion of how it's done.

Update: I decided to properly bypass C61 using a jumper wire instead of my clip leads. See here.

High Frequencies -- Polysix Adjustments

Continuing from my last post, I've been exploring the high frequency performance of my Korg Polysix and Korg Mono/Poly.  I'm trying to add more sizzle to the Mono/Poly and I'm trying to reduce a bit of upper-treble harshness in the Polysix.  My latest attempt at improving the Polysix was to follow the re-calibration procedure from the service manual, particularly regarding the re-calibration of the filters.  Sadly, it didn't affect the high frequency performance of he synth.  It did, however, bring my resonance and filter frequency control into better consistency between the voices.  As a result, I had fun discovering the unique vibe that comes with actually trying to play the self-oscillating SSM2044 filters...

But, back to the beginning.  After my last post, I received an interesting comment by "terjewinther" from the Polysix Yahoo Group.  He suggested that the calibration can have a strong effect on the sound of the synth.  So, I opened her up again, brought out the multi-meter and oscilloscope, and started in on the calibration procedure as listed in the Service Manual.  I made it through the filter tuning including offset, filter frequency, resonance, and EG intensity.  What I found was that my DAC was a bit off and my EG intensity was way off.  My filter cutoff and resonance were pretty close, but there was some variation from voice to voice.  So, I'd say that the biggest impact of the calibration was to make the 6 voices more consistent with each other, especially at higher resonance settings.

After completing this portion of the calibration (I stopped just before the calibration of the Keyboard Tracking), I closed the lid, plugged in the audio recorder, and took some new measurements of the trusty sawtooth wave.  Did I smooth out the high end harshness??? Did I get rid of that weird bump that I was seeing around 7 kHz???  

Well, the graph below has the answer....and the answer is "no".  There appears to be no change in the frequency response.  I doubt, therefore, that the calibration got rid of the harshness.  (Sure, when I play with the synth over the next few days, my ears will tell me things that this graph can't...but you can't really trust your own initial impression because we humans are so easily prey to confirmation bias...hence, objective measures are best for immediate trouble-shooting and feedback).

Frequency Response After Following the Tune-Up Procedure in the Service Manual

So, calibration did not appear to affect the high frequencies.  I would not, though, consider my effort to be wasted on the calibration.  For example, the EG Intensity range on the filter is so much more usable now.  And, as I mentioned, the 6 voices are much more consistent at high resonance...and this has actually been a bit of an inspiration.  For the first time ever, I found myself actually trying to play the self-oscillating filters in a musical way.  The video at the top of this post shows some of my results.  Sure, the pitches from the self-oscillating filters are not perfectly in tune (that's really really hard to do), but their out-of-tuneness is what engaged me.  Their tone (a fairly pure sine wave) is also strangely engaging to me.  It's a whole type of sound that I didn't know was inside the Polysix.  Now I know.  Thank you calibration!

High Frequencies - Polysix vs Mono/Poly

As you know, I have both a Korg Mono/Poly and a Korg Polysix.  I've had the Mono/Poly for longer and my visceral response to its sound is what motivated my purchase of the Polysix.  Being from the same vintage, and having many of the same components (like the SSM filter chip), made me assume that they would sound as similar as two analog synths could.  Well, once I'm got the Polysix home, I found that it didn't sound the same as the Mono/Poly.  In some ways the Polysix was better and in some ways the Mono/Poly was better.

Recording Tones from my Korg Polysix (Left) for Comparison to my Korg Mono/Poly (Right)

Specifically, I felt that the Mono/Poly had better bass and that the Mono/Poly had a much more engaging (less harsh, more smooth) lead sound in the upper octaves.  The Polysix, on the other hand, had a bit more sizzle and, on chords, the upper mids / low treble felt more liquid and present.  Being a bit geeky, I immediately wondered if I could measure and quantify the difference.  If I could quantify it, then maybe I could understand it, which means that maybe I could control and command it at will.  And, therein, lies the power of synth hacking.

OK, let's do some recordings, analyze them, and see what we can find...

For recordings, I simply plugged the output of the Polysix into my trusty portable audio recorder (M-Audio Microtrak II, shown in pictures above).  I recorded raw WAV files at 44.1 kHz at 16 bits.  I recorded 4 seconds of C1, then four seconds of C2, and so on up through all the octaves.  The synth was configured to play one voice, it was set to sawtooth, and the filter was wide open with no resonance.  I repeated the same process for the Mono/Poly (again, just one voice).  I brought all the files into Matlab for some plotting and analysis.  Below is plot comparing the raw time-domain audio of a sawtooth waveform at the lowest note, C1.

Polysix and Mono/Poly when playing a sawtooth waveform down at C1.

Clearly, there is a difference in the shape of these two waveforms.  To those used to looking at oscilloscope traces, this plot will be familiar.  You'll note that neither plot is as much like a sawtooth as one might like -- in both cases, the "ramp" portion of the sawtooth is not as straight as one might expect.  This is due to some roll-off in the bass frequencies in the output.  The Polysix waveform (blue) is more rounded than the Mono/Poly (green), so it appears to have a little less bass.  As for high frequencies, note that the Polysix (blue) has a very sharp downward spike at each vertical transition in the sawtooth.  It actually appears to overshoot.  This kind of sharp, narrow spike requires very high frequency response, which suggests that the Polysix does indeed have more "sizzle" than the Mono/Poly.

Now that we've seen some interesting features, let's try to quantify them.  I've chosen to take the FFT of each recording in order to assess the frequency content of each note.  After a bit of normalization to equalize the volume of each recording, I would get plots like the one below.  Note that frequency is now on the horizontal axis instead of time.

Spectrum recorded from the Polysix and the Mono/Poly when playing the lowest C ("C1").

The first thing to notice is that the spectrum of the note is composed of a large number of "spikes" in the frequency domain.  This is what one should expect for a sawtooth wave.  If I had used a sine wave instead of a sawtooth, I would have just gotten one spike...the one at the fundamental pitch.  What differentiates a sine wave from a sawtooth wave from a square wave is the number of harmonics and their relative magnitude.  Therefore, this plot is very normal.

Looking now at the Polysix spectrum (blue) compared to the Mono/Poly spectrum green), we see that they are largely similar until we get to the highest frequencies.  Above ~3 kHz, we see that the Polysix has stronger high frequencies than the Mono/Poly.  This could be part of the difference in "sizzle" that I'm hearing.  Let's dig in a little deeper.

The plot above is confusing because it shows the signal energy at important frequencies (the fundamental and all the harmonics) and at unimportant frequencies (everything between the harmonics).  Let's extract just the energy at the fundamental and at each harmonic and plot just those values.  The resulting plot (below) is much simpler and easier to see what's going on.  For reference, I even include a line that shows the spectrum for the ideal sawtooth waveform.  You'll see that both the Polysix (blue) and Mono/Poly (green) have pretty good sawtooths.  Only above ~3 kHz do they begin to diverge in any significant way.

Spectrum Assessed at Just the Fundamental and Harmonic Frequencies.

Let's further simplify this plot by taking the difference of each spectrum relative to the ideal sawtooth spectrum.  The result is shown below.  Now we're getting somewhere.  

Spectrum Compared to the Spectrum of an Ideal Sawtooth Waveform

First, look at the left-most side of the graph. Here are the low frequencies. The fundamental at C1 is abut 33 Hz. That's where this graph starts. Note that it shows that the Polysix (blue) is a little below the Mono/Poly (green). At this very low frequency, the Polysix is showing 1.6 dB less bass than the Mono/Poly. That's not much of a difference. So, maybe my subjective assessment that the Mono/Poly has more bass than the Polysix is supported by this data, or maybe not.  Lots of other factors can affect the psychoacoustics of bass perception besides just the literal amount of bass in the air.  So, I'm going to refrain on making any conclusions about bass.

Looking at the right-most side of the graph, we see huge differences in the treble response.  Unlike the assessment of bass, this difference in treble is very clear.  Comparing the Polysix (blue) to the Mono/Poly (green) we see a 3 dB difference by 4 kHz.  That's definitely audible.  By the time you get out to 10 kHz, we've got a 10 dB difference.  That's a big difference in "sizzle".  This definitely confirms what I was hearing.  This doesn't tell us *why* they're different, but it's an objective measure that we can now use to probe within each synth to find where the difference occurs.  That's an exercise for later.

Another behavior that catches my eye in the figure above is that the Polysix diverges from the ideal sawtooth by first going *up* before going down.  It's as if there is a treble knob within the synth and that it is turned up a bit in the 3 kHz to 10 kHz region.  The peak response is at 6.8 kHz and is 2.2 dB above the ideal sawtooth.  What's the cause of this apparent enhancement of the treble?  Well, I'm not sure, but to my eye, it appears that the resonance of the VCF on the Polysix might be a bit active, even though I turned it down to "zero".  I'll have to open her up and check it out at a later time.  

The more important question is whether this boost in treble is the cause of the "upper mids / low treble feel more liquid and present" perception that I mentioned earlier.  Maybe.  If I'm able to tune the boost out of the system (through adjustments to the resonance or whatever), we'll see if the "liquid and present" feel goes away.  Given that I like the "liquid and present" feel, I might choose to keep it the way that it is.

A downside of this enhanced upper-mids is reflected in my original comment that I preferred the sound of the Mono/Poly for lead work in the upper octaves.  I felt that it was more engaging and less harsh.  Excessive upper-mids could be the source of "harshness".  When I examine the data for a high note (C6) many of the conclusions drawn from the graphs above still hold for C6...

For example, below, the time domain plot shows that the Polysix still has its over-shooting downward spike suggested lots of high-treble.

Time-Domain Plot of Raw Waveform of High Note, C6

The FFT output comparing the two C6 notes shows that the Polysix definitely has more more treble.

Frequency-Domain Plot of High Note, C6

And, comparison of the amplitude of the fundamental and harmonics to the ideal sawtooth waveform (below) shows the same high-frequency roll-off in the Mono/Poly and the same slight high-frequency boost in the Polysix.

Comparison of Harmonic Content of High Notes (C6) to Ideal Sawtooth

So, while "harshness" is also a complicated psychoacoustic phenomenon, these plots confirm that there is a substantial difference in the amount of treble above 3 kHz between the Polysix and the Monopoly.  This is true for both low notes (the C1 analyzed first) and for high notes (the C6 analyzed second).  In my opinion, this extra treble is at least one of the factors of the apparent harshness (for lead work) of the Polysix compared to the Mono/Poly.

Edit: Follow-up on the Polysix is here.
Edit: Follow-up on the Mono/Poly is here.
Edit: Modification of the Polysix to restore the deep bass is here.