MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

TH x4046 VCO V/Oct tracking with different chips
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Author TH x4046 VCO V/Oct tracking with different chips
ixtern
Recently I've built fonitronik's TH x4046 VCO and I have measured V/Oct tracking for some chips I have. To get proper response from Motorola chips I had to made some changes to the schematic to get wider range of adjustment:
- R9 tempco was changed from 2k to 3k (2k+1k serial),
- R1 trimpot was changed from 100R to 200R,
- R2 resistor was changed from 390R to 200R,
- R15 resistor was changed from 10k to 47k (to reduce HF tracking).
LM394CN was used as an expo converter chip.
Couldn't get proper tracking from TI chip, so results were not included.

For the simplicity, tracking errors were calculated as if control voltage was absolutely correct (0V, 1V, 2V, etc.)

Here are the results (the best are for Philips chip):

===============================================
MC14046B Motorola SO16L

U[V] f[Hz] fn[Hz] Diff [Hz] Diff[cents]
0.0022 19.13 20 -0.87 -77
0.9993 39.23 40 -0.77 -33.65
2.0000 79.65 80 -0.35 -7.6
2.9991 160.10 160 +0.1 +1.1
3.9998 320.6 320 +0.6 +3.24
4.9991 640.1 640 +0.1 +0.27
6.000 1278.0 1280 -2.0 -2.71
7.000 2557 2560 -3 -2.03
7.994 5187 5120 +67 +22.51
===============================================
HEF4046BT Philips SO16

U[V] f[Hz] fn[Hz] Diff [Hz] Diff[cents]
0.0021 19.96 20 -0.04 -3.5
0.9993 39.92 40 -0.08 -3.5
2.0000 79.97 80 -0.03 -0.7
2.9992 159.94 160 -0.06 -0.7
3.9999 320.0 320 0 0
4.9993 638.9 640 -1.1 -3
6.000 1277.4 1280 -2.6 -3.5
7.000 2559 2560 -1 -0.7
7.994 5145 5120 +25 +8.4
===============================================
Motorola MC14046BCP DIP16

U[V] f[Hz] fn[Hz] Diff [Hz] Diff[cents]
0.0021 19.43 20 -0.57 -50
0.9993 39.43 40 -0.57 -24.8
2.0000 79.63 80 -0.37 -8
2.9993 159.92 160 -0.08 -0.9
3.9999 320.4 320 +0.4 +2.2
4.9992 639.8 640 -0.2 -0.5
6.001 1276.6 1280 -3.4 -4.6
7.000 2557 2560 -3 -2
7.994 5218 5120 +98 +32.8
===============================================
Harris CD4046BE DIP16

U[V] f[Hz] fn[Hz] Diff [Hz] Diff[cents]
0.0021 19.1 20 -0.9 -79.7
0.9993 39.26 40 -0.74 -32.3
2.0000 79.78 80 -0.22 -4.8
2.9992 160.34 160 +0.34 +3.7
3.9998 320.74 320 +0.74 +4
4.9993 639.8 640 -0.2 -0.5
6.001 1278.3 1280 -1.7 -2.3
7.000 2560 2560 0 0
7.994 5147 5120 +27 +9.1
===============================================
Difference in cents was calculated with the help of this page:
http://www.sengpielaudio.com/calculator-centsratio.htm

I have precise voltage source only for 8 octaves so higher were not measured.

Adjustment procedure was as follows:
- set HF tracking to minimum,
- in the range of 20-640 Hz set Scale trimpot to get best tracking,
- set frequency to the next two octaves (1280 Hz/2560 Hz) and adjust HF tracking to get best result for these two octaves.
- higher than 2560 Hz frequencies are not useful for me so tracking was not important.

And here is the photo of measured chips:


Sorry but cannot format results to be more readable. They looks different in the post editor and on the forum page.
guest
thanks for posting up your work. that phillips chip is way better than all the rest.
plushterry
Thanks! This is really useful, I remember having real trouble getting a 4046 that would even work in the one I built. I'll have a look at picking up some of the Philips chips.
forbin
I did a graph of tracking of ti, philips (nxp) and Fairchild 4046 devices on the em forum back when Thomas Henry posted the design. Found the Fairchild part was probably the best and the TI was pretty poor... didn’t have a Motorola part (think they are On semiconductor now).
wackelpeter
As someone who started a similar thread some weeks ago, i can confirm that Texas Instruments only gave very poor results.

Then got some Philips and they were much better... but perhaps it's worth to mention that the range and difference within the same brand can also be of a wider range... So it's always best to order a couple of them and then select for the best tracking...

From the few i tested that came with my order, there were also some that gave just average results... i wasn't messing much around with them and didn't spent too much time but while i had calbrated one properly, is inserted the next chip and of course then the initial frequency was almost off from the previous, some times smaller sometimes nearly an octave... and some of those chips showed a resistent behaviour against proper calibration and didn't got over 3-4 octaves, no matter how much i messed with the V/oct trim around.

...and ah yes, my philips were dip16 chips i got from a asian shop at ebay...
ixtern
My HEF4046 chips were from this source (very reliable supplier):
https://www.ebay.com/itm/10pcs-HEF4046BT-PLL-Loop-ICs-HEF4046-SO16-PHI LIPS/263522591268?hash=item3d5b2ba224:g:QH4AAOSw1jJaliHj
but proper adapter to DIP16 is needed what may be a problem for somebody.
ixtern
As I have got some National 4046 chips, I've made additional measurements but the result were rather disappointing. This time I've converted raw data to the line charts with the help of https://www.rapidtables.com/tools/line-graph.html site.
First is the chart of control voltage vs corresponding frequency:


Next is the summary of tracking for 4 different 4046 chips (raw data for most of them included in previous post, all supplied by +-12V):
- HEF4046BT Philips SO16
- Motorola MC14046BCP DIP16
- Harris CD4046BE DIP16
- National CD4046BCN DIP16


Then I've changed supply voltage from +-12V to +-15V but results were not much better. For National chip,tracking was better in upper frequencies but worse for lower:


For the reference I've measured +-15V HEF4046 tracking:

Tracking was better for 0 to 6V (lower frequencies) but worse for higher range.

I cannot understand how Tomas Henry could achieve such good tracking results with National chips. Maybe older ones were better?
I have checked three National 4046s:

Results were similar, so only for one were included.
elektrouwe
the graphs show clearly that these chips have a leakage current problem at the low end. That's why they behave better with lower VDD.
At the high end there is a "ON resistance" problem, which unfortunately worsens with lower VDD.
IMHO all these 4xxx-series VCOs are not "high end" VCOs. I guess '70s metal gate process tolerances are higher compared to decent analog chips. You get what you pay for - 4046 and 4069 VCOs are easy to build, but when it comes to "well tempered scale" I would not use them.
I also doubt that TH or someone else could achive better results, maybe by chance. Once I designed a '555 VCO which was not as good as expected. TH's 555 VCO was famous for it's linearity. I got a original design to compare and measure the performance. It turned out that it had exactly the same behaviour as my (different) design and it took me many hours to find out, that the frequency dependent delay time of the '555 (CMOS version 7555 is even worse) was the reason and TH's published Hz/V table was "too optimistic".
ixtern
elektrouwe wrote:
Once I designed a '555 VCO which was not as good as expected. TH's 555 VCO was famous for it's linearity. I got a original design to compare and measure the performance. It turned out that it had exactly the same behaviour as my (different) design and it took me many hours to find out, that the frequency dependent delay time of the '555 (CMOS version 7555 is even worse) was the reason and TH's published Hz/V table was "too optimistic".

Recently I have played with my 555 VCO design and my conclusion is that 555 is not bad and 7555 (ICM7555) is great. 7555 can make a sharp clear 1-microsecond pulse for switch discharging capacitor. But 555/7555 has two problems:
- it is a "window comparator" meaning that saw from the core is more dependent on supply voltage and temperature, as it compares two levels, not one like ordinary comparator;
- output pulse voltage translation (npn and pnp transistors) for JFET switch resulting in not so sharp and longer (2-3us) discharging pulse (maybe I've tried not enough).

Initially I have on my stripboard two paralell circuits: one with 7555 and level translators and second with LM311, working with the same core and switched by jumper. But in the end I've stayed with LM311 - simpler and faster solution. And I needed predictable and constant saw voltage level to convert it to triangle and for waveshaper so I couldn't stay with both switchable solutions with different saw levels.

BTW. x4046 VCO with HEF4046 is tracking quite good (5-6 octaves) with my TH VCO Maximus (CEM3340). It's enough to be useful.
elektrouwe
ixtern wrote:
Recently I have played with my 555 VCO design and .... I've stayed with LM311.

yes, for a classic saw core a 555 does not make much sense.You can't benefit from the 2nd (lower) threshold because the reset switch already does the job.
I guess that's why the 555 is mainly used in tri cores.
re. 7555: propagation delay time is really bad in tri core applications. datasheet talks about steep pulse inputs, not about slow triangles :-)
you've mentioned " tracking quite good (5-6 octaves) ". I expected 9..10 oct. tracking with a 555 tri core ( which I would call "very good") but could only get 7..8 oct. ("good")
And even therefore I had to misuse the Rossum style HF compensation (which is a expo error compensation) to cure the 555's switching delay.
With a modern fast push/pullCMOS comparator switching between 0 and 5V in some ns I was able to get 9..10 oct. (which perhaps nobody really needs)
BTW Rosssum HF compensation - the rising tracking error @ the high end looks like the expos bulk resistor error. Could it be the expo and not the 4046 ?
I have measured frequency vs expo collector current with a precision amp-meter to seperate expo and VCO-core errors.
Could be a problem to measure in a hard switching saw core, though...
ixtern
elektrouwe wrote:

BTW Rosssum HF compensation - the rising tracking error @ the high end looks like the expos bulk resistor error. Could it be the expo and not the 4046 ?
I have measured frequency vs expo collector current with a precision amp-meter to seperate expo and VCO-core errors.
Could be a problem to measure in a hard switching saw core, though...

Are you aware that 4046 internals are made from MOSFETS, not bipolar transistors? So it is a wonder that 4046 VCO even has some linear range.
I don't suspect my expo although I didn't measured LM394 used there. It is not an easy thing to get such characteristic, especially in low currents area.
JimY
Taken as a VCO, the 4046 CV (pin9) response is most linear in mid-range. At low CV's they plummet down to cut-off, at High CV it shoots up to maximum according to the control R and C values.

It's always bothered me that the X4046 design simply ties the CV pin to V+. Maybe that's ok, but I wonder if it might be different if tied to a well-regulated reference that's maybe 75% of V+?

On the copy of the schematic I have, the VCO cap (pins 6,7) has no ID or value, but when that gets small like 100pF, then stray capacitance and lead + self-inductance start to really bugger things up, and that x10 on a solderless breadboard. And then the X4046 (very cleverly I think), adds to the pin connections for the ramp wave generation.

Classic National Semiconductor parts have the double wave logo

ixtern
JimY wrote:
Taken as a VCO, the 4046 CV (pin9) response is most linear in mid-range. At low CV's they plummet down to cut-off, at High CV it shoots up to maximum according to the control R and C values.

It's always bothered me that the X4046 design simply ties the CV pin to V+. Maybe that's ok, but I wonder if it might be different if tied to a well-regulated reference that's maybe 75% of V+?

On the copy of the schematic I have, the VCO cap (pins 6,7) has no ID or value, but when that gets small like 100pF, then stray capacitance and lead + self-inductance start to really bugger things up, and that x10 on a solderless breadboard. And then the X4046 (very cleverly I think), adds to the pin connections for the ramp wave generation.

I am not so sure if connecting CV pin to other voltage is a good idea. Look at the internal 4046 VCO schematic (at least such is in TI chip):

V+ at the CV ensures that n1 nMOSFET is fully open. Other way you would get a kind of not-very-linear-resistor (voltage dependent) in series with curent source. I wonder rather why R2 (pin 12) was not used.
JimY
That is interesting, although I think you meant to say that n1 is meant to be saturated on? In which case it's on-resistance does vary some. Vgs can change as Id changes and so Ron changes.

So, if you tie pin 9 (and maybe also pin 11 to be sure) to 0v and use R2 as the control it might be more consistent?
JimY
I just noticed that the minimum recommended resistance of R1 is 10k. This is in connection, I suppose, with VCO CV linearity which we might suppose isn't a concern with the CV tied up to V+. But, maybe it suggests that Ron of N1 is as high as 1k (the chip is full of weak channels for its micropower status), in which case it would create some offset.
guest
i spent some time analyzing the 4046 VCO, and although its cheap and pretty easy (still requires a diffamp at the output), it wasnt good across a large range of input currents, as discussed here. i used the I2 input, and tied the I1 input off. another source of error is the clamping diodes:



when the capacitor polarity is switched, the voltage spikes downwards. to protect the internal comparator transistors, this needs to be clamped. so the voltage that it starts from is ~-0.6V, and then it goes up to the comparator voltage. so, the amount of charge put into and taken out of the capacitor (and therefore the oscillation frequency) will drift with temperature due to both the diode changing is clamp voltage, and the comparator transistors drifting.
JimY
There is a hack for the timing cap, and that is to fit an equal pair of caps from each pin to 0v. The chip doesn't know the difference, but when the logic switches to ground the cap charge out, it stops the negative spike at the opposite side passing through the internals. If you ever wanted to, you can change the squarewave duty cycle with unequal caps.

This makes me wonder, what if one cap in a dual cap setup was very small and the bulk of the timing taken by a larger cap. Would that give a usable sawtooth wave right away (given you don't care that the pin4 wave is now a narrow blip)?
guest
thats a good suggestion. you could probably just use a resistor for the reset pulse. but, if you used 2 capacitors of equal size, you could just sum the two outputs to have a sawtooth wave, or invert and offset it to make a triangle wave. but, at that point, there isnt much difference between this and a 555, which would be more accurate. or, even better, an HC221 which has 2 VCOs on it.
ixtern
I have another HC/HCT4046 VCO diagram taken from Philips' "CMOS phase-locked loops:74HC(T)4046A/7046A & 74HCT9046A HCMOS Designer's Guide" (1995).
I've read somewhere that HC/HCT4046 VCO implementation varies more for different manufacturers than it was with original 4046.


9046 VCO uses comparators and reference voltage to detect VCO's switching levels instead of Schmitt triggers.
JimY
Good info. I'm slowly starting to understand these things better.

I breadboarded one last night. Only for the basic VCO, not an entire X4046!
I used pin12 and grounded pins 9 and 11. With a 1M pot + 4k7 fixed and 47nF cap I got around 80Hz to 8Khz sweep. Obviously, that range can be much wider if the resistance could go over 1M.

A pair of 47nF to 0v from each pin gave similar results but the pin 6 and 7 waveforms are noticeably cleaner. It's just occurred to me that it could keep the usual single timing cap but add Shottky clamping diodes to the pins to bypass the internal clamps.

If I used odd caps, the chip shows an increased inability to fully discharge the ramp on the largest cap fast enough. The pin with the larger cap has a decent sawtooth but the amplitude falls as frequency increases. I suppose a BJT could be used driven off pin4 to fully discharge the cap.

I was using Philips HCF from 2004 if I read the date code right. I have one ancient Nat-Semi CD4046 from 1982 and that ran about 5% faster. That part has the older stylized block capital NS logo.
ixtern
Interesting. But are you testing Philips' HEF4046 or ST's HCF4046?
JimY
Philips HEF sorry!
JimY
I tried a few things to improve the ramp waves at the timing cap.
2 caps to 0v
BJT NPN " helper" transistors to discharge the caps. One driven from pin 4 via 47k to base, the other with the XOR used to invert and drive the other cap from pin 2. Results in marked improvement.
Removed negative spike.
Squares up the ramp trailing edge.
Maintains amplitude to a higher frequency.
Improved matching of pin6 and 7 waveforms.
Bypasses variation in on-chip capacitor cap discharge n-channels

Downside
Reduces upper frequency with same control due to more complete discharge of timing caps.
Can't easily be retrofitted to X4046 because XOR is in use for Sync - another inverter would be needed for an anti-phase transistor drive.
Would be even better if the chips digital outputs had sharper transitions.
JimY
I just got some TI CD4046 to play with.
Testing on breadboard with a local 12v regulator because it needs really good regulation to be stable.

Soon becomes apparent that the TI doesn't like small control currents. It becomes very prone to noise and leakage at the LF end while at the HF end it runs faster than the Phillips part. To be fair, the TI (Harris) datasheet specifies a maximum 1M for timing resistance. With a 10nF cap, the x4046 control must be exceeding an effective resistance of 1M by quite a large factor to get down low.
The easiest way to get a musically useful range with the TI chip is a larger timing cap value (I got to 47nF) and probably accept it can never go super high at the top end if you also want bass. You can extend the range upwards because it doesn't seem to mind if the effective timing resistance goes below the datasheet minimum, but it makes it harder to get a good result at the bass end.
It looks like 8Khz is a reasonable upper limit if you also want it to go down to sub-bass, although I'd forget LFO territory. I was seeing a minimum of around 6Hz before the ramp wave started to wobble. That said, it should do better on a pcb and the frequency control very close to the chip pins.

I found that it made no difference if I kept the x4046 pin11 resistor scheme or used pin12, in fact I could have pins 9,11 and 12 all tied together with the same results
ixtern
I didn't wasted so much time with TI chips but your measurements are interesting. Although not very good for our VCOs, they are still valuable parts for wave multipliers where linearity is not so important and output frequency may be two, three or five times higher than the base one.
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Page 1 of 2
Powered by phpBB © phpBB Group