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

An in depth look at the Roland Jupiter 8's VCO
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3, 4, 5, 6, 7  Next [all]
Author An in depth look at the Roland Jupiter 8's VCO
Jackdamery
So I took the liberty of copying out the VCO-1 from the Jupiter 8 schematics and cleaning it up a bit in MS Paint.



It's a saw core, based on integrator/schmitt trigger design. Frequency is controlled by a voltage to current sink expo converter.

Original Parts List
Semiconductors:
1x CA3046 NPN array //Available surplus/ebay
1x 4558 dual opamp //Still manufactured
2x TL080 single opamp //Available surplus/ebay
1x NF510 (FET) //Impossible to find, suggested replacement 2N4393
1x 2SA1015- GR (PNP) //Still available via surplus/ebay
2x 1S2473 diodes. //No longer made, sometimes available via surplus, read online that it can be subbed with a 1N4148, although might not match for this purpose.

Other:
1x 1K TSP102J temperature sensing 'posistor' //Available as pulls/NOS from Littlediode.

Resistors:
It seems that MF stands for 'metal film' ie 1% tolerance resistors where stability is critical in the circuit.

Trimmers:
BMF ? Not sure what B stands for but do metal film trimmers exist?
1x 100k - Tune
1x 2k - Width
1x 5k - Linearity

Capacitors:
Either listed as 'p' for picofarad or 'c' for microfarad. C6A is a 1000pf polystyrene for timing stability.

New Parts List
Semiconductors:
1x LS318 NPN array
1x 4558 dual opamp //Still manufactured
2x TL071 single opamp
1x 2N4393 (FET)
1x 2N3906
2x 1N4148
1X 1K Tempco
How the VCO core works



As I understand it, the integrator charges the capacitor in the feedback loop, so the output is a rising ramp. When the voltage reaches ~10v the schmitt trigger/comparator goes fully positive with a narrow pulse at 26v p-p. This is more than the breakdown voltage of the diode in its path so the pulse hits the gate of the FET, opening it as an easier path for the current to travel than through the charging capacitor. It also hits the second diode opening the path for ground, discharging the capacitor and reseting the comparator. The cycle then repeats.

Could this be made with currently manufactured components?

The CA3046 NPN array is no longer manufactured. However, other NPN pairs are still available (The LS318, for instance). The Jupiter doesn't seem to make use of the remaining parts of the 3046 as a temperature sensor and heater as some other synths do, so really all that is needed is 2 NPNs.

The 'posistor' as far as I can tell is a resistor that increases in impedance relative to rising temperature. You can still buy these, mouser sells 560 ohm ones which are used in the VCF of the Jupiter. I'm not sure what the TCR of these is though as there is no info provided. I believe the 3300ppm TCR tempcos usually used in SDIY are negative coefficient, but perhaps the circuit could be adapted to use them instead.

The TL080 is a JFET input opamp. It's no longer manufactured, but National Semiconductor do make one, the LF411. It looks like the 10pf capacitor between pins 1 and 8 set the bandwith of the integrator TL080, this might be important, I'm not sure. Maybe someone here could suggest a better replacement.

It seems like this design could indeed be resurrected at least fairly closely to the original. Comments/corrections very welcome.

TL:DR - here's a JP8 schematic, let's make one.
roglok
Quote:
Other:
1x 1K temperature sensing 'posistor' //Impossible to find, not sure exact properties and spec.


Little Diode stock the original thermistors:
http://www.littlediode.com/components/TSP102J.html
filterstein
Thonk has the tempco's: http://www.thonk.co.uk/product-category/electronic-component/tempco/

About the NF510 substitute: never found out what the recommended fet is.
Heard 4092 and mpf102 as succesfull replacements too.

Remember that the JP8 is microprocessor controlled, the autotune could correct when it drifted out of tune.
2thick4uni
I think I'd replace the 4558 with a dual precision op amp such as LT1013, won't affect the sound at all but will reduce drift in the summer and expo converter. Likewise replace the 3046 with LS318 or similar or even match a pair of BC547's or similar. The transistors in the 3046 are only matched to about 3mv, you can hand match transistors closer than this. Any 1k 3300ppm or 3500ppm ptc tempco should be fine. Multi turn cermet trimmers will be ok too. I'm sure there will be a few fet's that will work ok, best way is to model the circuit in Ltspice and see what works.
Jackdamery
roglok wrote:
Quote:
Other:
1x 1K temperature sensing 'posistor' //Impossible to find, not sure exact properties and spec.


Little Diode stock the original thermistors:
http://www.littlediode.com/components/TSP102J.html


Cool thanks. Since your post I ordered and received two of them. Just tested on in the office with the DMM. It's starting impedance was 980Ω. As I placed my finger on it, it increased incrementally to 985Ω. Then took it off and it lowered back down to 980Ω.

This 980/1k is in series with a 390Ω in the circuit, making 1.390KΩ which is of course scaled down, as it's in parallel with a 3.3kΩ.
Peake
2thick4uni wrote:
I think I'd replace the 4558 with a dual precision op amp such as LT1013, won't affect the sound at all but will reduce drift in the summer and expo converter. Likewise replace the 3046 with LS318 or similar or even match a pair of BC547's or similar. The transistors in the 3046 are only matched to about 3mv, you can hand match transistors closer than this. Any 1k 3300ppm or 3500ppm ptc tempco should be fine. Multi turn cermet trimmers will be ok too. I'm sure there will be a few fet's that will work ok, best way is to model the circuit in Ltspice and see what works.


"Stability" =does= affect sound in that the beats and tracking become more predictable and less slightly randomized. The resultant sound is more mechanical, especially in a polysynth. This is why Ensoniq etc. put a "Humanize" randomization feature in their LFOs, for example. (Repeating if it's not "gotten"..."sound" means what you hear and how you respond to it.)

Don't fall into the trap of "improving" something which already interests you for exactly what it is. Don't "fix" it if it isn't broken.
thetwlo
didn't know the 3046 was discontinued.
fwiw: Tayda still sells them:
http://www.taydaelectronics.com/catalogsearch/result/?q=ca3046
2thick4uni
Peake wrote:
2thick4uni wrote:
I think I'd replace the 4558 with a dual precision op amp such as LT1013, won't affect the sound at all but will reduce drift in the summer and expo converter. Likewise replace the 3046 with LS318 or similar or even match a pair of BC547's or similar. The transistors in the 3046 are only matched to about 3mv, you can hand match transistors closer than this. Any 1k 3300ppm or 3500ppm ptc tempco should be fine. Multi turn cermet trimmers will be ok too. I'm sure there will be a few fet's that will work ok, best way is to model the circuit in Ltspice and see what works.


"Stability" =does= affect sound in that the beats and tracking become more predictable and less slightly randomized. The resultant sound is more mechanical, especially in a polysynth. This is why Ensoniq etc. put a "Humanize" randomization feature in their LFOs, for example. (Repeating if it's not "gotten"..."sound" means what you hear and how you respond to it.)

Don't fall into the trap of "improving" something which already interests you for exactly what it is. Don't "fix" it if it isn't broken.


Yep, a good point and could be interesting to pursue this further. I was just trying to clean it up a bit as you'd be losing the benefit of autotune if using it as a stand alone vco, perhaps precision op amps will make it too accurate and sterile. I think that, as you say, instability is one of the reasons we love old analog gear so much and it needs to be preserved. However, thermal instability isn't particularly musical when it throws you out of tune in the middle of a performance as so many early minimoog users know all too well. Closely matched exponential transistors and correct tempco seem to be the key here, surely its possible to have thermal stability and still keep a bit of organic drift? So, how about keeping the 4558 and just addressing thermal drift by better matched expo transistors?
Jackdamery


Here's a page from the Siloconix transistor reference book suggesting the 2N4393 as a replacement for the N510 FET.

(Original link)
http://www.datasheetarchive.com/dl/Scans-026/ScansUX9527.pdf

Today I showed the schematic to my colleague who has some 40 years of experience as an electronic engineer, a lot of it with analogue circuits. He made a lot of calculations on the spot including one (CV = IT) ? Googling CV & IT turns up a lot of programming jobs though obviously cry A lot of what he said went over my head but seemed to make sense.

He calculated the maximum frequency possible from this oscillator to be 160kHz.

With regards to the FET he said the saturation drain current (IDss) should be 10mA. In this circuit the gate-source cutoff voltage(VGS(off)) should be 10v. So for this purpose a 2N3819 should work too.

He was curious as to the reason for the comparator and the PNP transistor connected to the long tail NPN pair of the 3046 on the expo converter. Any ideas?

Also he was asking why a tempco was necessary when the transistors being monolithic should compensate eachother nicely.



RE: the trimpots. VCO1 tune sets the rough starting tune. In the real synth the computer would then finely 'tune' it by varying the voltage fed to it from the DAC. VCO1 width sets the range of frequency it can cover. The first opamp is a DC signal mixer and this pot sets the gain of it - a larger gain equals a wider frequency. VCO1 linearity is for tuning the 1v/octave response of the expo converter.
filterstein
[quote="Jackdamery"]

Here's a page from the Siloconix transistor reference book suggesting the 2N4393 as a replacement for the N510 FET.

(Original link)
http://www.datasheetarchive.com/dl/Scans-026/ScansUX9527.pdf

Thanks, thats a great find!
2thick4uni
A feedback loop to the op amp is formed by one of the matched pair transistors, R12 and the third transistor emitter base junction. As it has a feedback loop it is not acting as a comparator, it also has a cap across it which acts as an integrator. I think the op amp's purpose is to control the current through the expo convertor, quite common in VCO's. I've seen a transistor at the base of the long tailed pair in there as well (early minimoog oscillator for example) but have no idea what purpose it serves! A tempco is still required even though a monolithic matched pair is used, it compensates for scale drift which is about -3300ppM/C, by using a PTC with 3300ppm/deg c it nulls this out.

Can anyone with a more thorough understanding of this schematic step in and give a full explanation of the circuit? I'm intrigued as to why a 3k3 resistor (R14) is parallel to the tempco, is it because the tempco has a larger compensation coefficient than 3300ppm and this pulls it down?
Jackdamery
2thick4uni wrote:


Can anyone with a more thorough understanding of this schematic step in and give a full explanation of the circuit? I'm intrigued as to why a 3k3 resistor (R14) is parallel to the tempco, is it because the tempco has a larger compensation coefficient than 3300ppm and this pulls it down?


Well yeh, that's what my esteemed colleague said. He reckoned it should scale it down to half.
Jackdamery
Also regarding the TSP102J. I nicked this quote from the Roland Jupiter forum. Here is a mention of it in a patent for a clothes dryer. It says it's coefficient is 0.7% per degree celcius.

Quote:
Although other similar devices may readily be utilized, in the preferred embodiment sensor 24 is a manufacturer's type TSP102 thermistor, obtainable from Texas Instruments Corporation, P.O. Box 225474, Dallas, TX 75265. Sensor 24 is preferably a positive temperature coefficient thermistor having a nominal 1000 ohms resistance at 25.degree. C. and a linear coefficient of 0.7%/C..degree.. Sensor 24 is located so as to be sensitive to drying air temperature, and will exhibit a change in resistance linearly proportional to the sensed temperature.


http://www.freepatentsonline.com/4275508.html

I forgot to mention earlier the TSP102Js I received yesterday are packaged in a TO-92 transistor case with only two legs. One of them seems to have the stub of the middle one left though. They seem to have come from different sources though as the lettering is different on each.
Peake
2N4393 are easily availalble on eBay but PN4393 are much less expensive. Surplussales have 2N for $0.79 each:

http://www.surplussales.com/Semiconductors/Transistors-SCR-2.html
Dr. Sketch-n-Etch
PN4391 is very commonly used for VCO reset duties in the synth DIY world. I have a small pile of them in my garage.
2thick4uni
Jackdamery wrote:
Also regarding the TSP102J. I nicked this quote from the Roland Jupiter forum. Here is a mention of it in a patent for a clothes dryer. It says it's coefficient is 0.7% per degree celcius.

Quote:
Although other similar devices may readily be utilized, in the preferred embodiment sensor 24 is a manufacturer's type TSP102 thermistor, obtainable from Texas Instruments Corporation, P.O. Box 225474, Dallas, TX 75265. Sensor 24 is preferably a positive temperature coefficient thermistor having a nominal 1000 ohms resistance at 25.degree. C. and a linear coefficient of 0.7%/C..degree.. Sensor 24 is located so as to be sensitive to drying air temperature, and will exhibit a change in resistance linearly proportional to the sensed temperature.


http://www.freepatentsonline.com/4275508.html

I forgot to mention earlier the TSP102Js I received yesterday are packaged in a TO-92 transistor case with only two legs. One of them seems to have the stub of the middle one left though. They seem to have come from different sources though as the lettering is different on each.


I've had another look at this and done some simple calculations.

Ok, so we now have the temperature coefficient for the tempco. ppm is 10,000 times smaller than deg c, so the tempco is actually 7,000ppm/deg C. This is more than twice as large as needed, hence R14.

So we have 1000R + 390R /3300R = 0.4212
If we multiply 7000ppm by this it gives 2948ppm/c

And if we assume the 3k3 resistor is metal oxide it will have a coefficient of around 300ppm/c which we need to add in.

So, 2948 + 300 = 3249ppm/degC

Which is pretty close to the ideal 3,300ppm/degC needed to compensate the expo transistors.

The combined parallel/series combination of R13 + R15 / R14 is actually 978R, which is not a bad approximation of 1K.

So, R13, 14 and 15 can be replaced with a single 1K 3300ppm/c tempco! There should be sufficient adjustment on VR2A to get compensate for 1K rather than 978R.

My guess is that when the JP8 was designed these TSP102 tempco's were cheaper and or more readily available than 1K 3300ppm parts.



.
Jackdamery
2thick4uni wrote:
Jackdamery wrote:
Also regarding the TSP102J. I nicked this quote from the Roland Jupiter forum. Here is a mention of it in a patent for a clothes dryer. It says it's coefficient is 0.7% per degree celcius.

Quote:
Although other similar devices may readily be utilized, in the preferred embodiment sensor 24 is a manufacturer's type TSP102 thermistor, obtainable from Texas Instruments Corporation, P.O. Box 225474, Dallas, TX 75265. Sensor 24 is preferably a positive temperature coefficient thermistor having a nominal 1000 ohms resistance at 25.degree. C. and a linear coefficient of 0.7%/C..degree.. Sensor 24 is located so as to be sensitive to drying air temperature, and will exhibit a change in resistance linearly proportional to the sensed temperature.


http://www.freepatentsonline.com/4275508.html

I forgot to mention earlier the TSP102Js I received yesterday are packaged in a TO-92 transistor case with only two legs. One of them seems to have the stub of the middle one left though. They seem to have come from different sources though as the lettering is different on each.


I've had another look at this and done some simple calculations.

Ok, so we now have the temperature coefficient for the tempco. ppm is 10,000 times smaller than deg c, so the tempco is actually 7,000ppm/deg C. This is more than twice as large as needed, hence R14.

So we have 1000R + 390R /3300R = 0.4212
If we multiply 7000ppm by this it gives 2948ppm/c

And if we assume the 3k3 resistor is metal oxide it will have a coefficient of around 300ppm/c which we need to add in.

So, 2948 + 300 = 3249ppm/degC

Which is pretty close to the ideal 3,300ppm/degC needed to compensate the expo transistors.

The combined parallel/series combination of R13 + R15 / R14 is actually 978R, which is not a bad approximation of 1K.

So, R13, 14 and 15 can be replaced with a single 1K 3300ppm/c tempco! There should be sufficient adjustment on VR2A to get compensate for 1K rather than 978R.

My guess is that when the JP8 was designed these TSP102 tempco's were cheaper and or more readily available than 1K 3300ppm parts.



.


Excellent analysis! Are you really too thick for Uni? Also lol at the Viz quotes.

One point to mention though is aren't tempcos usually used in expo converters NTC? The TSP102J is PTC. Is it important which way it compensates as long as it does move relative to temperature?

Tempcos are usually placed in the negative feedback path of the signal mixing opamp of the expo converter aren't they? Whereas this one is acting a voltage divider on the output of the mixer.

So an NTC in the feedback of an opamp would reduce the value of the resistor with an increase in temperature. Gain in an inverting amplifier is R2/R1, so assuming R1 is a 100k resistor: 1/100 = 0.01 gain. If the tempco drops to 950ohm with an increase in temperature then 0.950/100 = 0.0095 Thus lowering the gain with an increase in temperature to keep the scale correct.

Assuming the gain is constant in the JP8 mixer opamp, then our temperature compensation is in the form of a voltage divider.

following the opamp we have 10k resistor as R1 of the divider and then say we follow it with R2 as a 1k tempco Vout R2/(R1+R2).Vin so say our input voltage is 1v : with R2 at 1K we have an output voltage of 0.091V. If the tempco increases with temperature to say 1050 ohms then the output voltage will be 0.095v Thus, in this circuit gain increases with temperature
2thick4uni
Jackdamery wrote:


Excellent analysis! Are you really too thick for Uni? Also lol at the Viz quotes.

One point to mention though is aren't tempcos usually used in expo converters NTC? The TSP102J is PTC. Is it important which way it compensates as long as it does move relative to temperature?

Tempcos are usually placed in the negative feedback path of the signal mixing opamp of the expo converter aren't they? Whereas this one is acting a voltage divider on the output of the mixer.


Yes, I really am too thick for uni, I have the exam results to prove it!

PTC tempco's are the usual type in expo convertors for vco's, the much loved PTC146 is PTC.

http://www.elby-designs.com/contents/en-us/tempco.pdf

Do try one in the circuit and let us know if it works.
2thick4uni
Forgot to answer second part of your question. A PTC works here because it is at the bottom half of the potential divider coming from the output of the op amp, so as its value increases it will increase the voltage from the op amp into the base of the expo transistor (as R7a, the other part of the potential divider, doesn't change). If tempco was at the top of the potential divider or took the output directly from the op amp it would reduce the voltage on an increase in temperature, so would then need to be a NTC component.
Jackdamery
2thick4uni wrote:
Jackdamery wrote:


Excellent analysis! Are you really too thick for Uni? Also lol at the Viz quotes.

One point to mention though is aren't tempcos usually used in expo converters NTC? The TSP102J is PTC. Is it important which way it compensates as long as it does move relative to temperature?

Tempcos are usually placed in the negative feedback path of the signal mixing opamp of the expo converter aren't they? Whereas this one is acting a voltage divider on the output of the mixer.


Yes, I really am too thick for uni, I have the exam results to prove it!

PTC tempco's are the usual type in expo convertors for vco's, the much loved PTC146 is PTC.

http://www.elby-designs.com/contents/en-us/tempco.pdf

Do try one in the circuit and let us know if it works.


Do you think this would work?

http://uk.farnell.com/te-connectivity-citec/lt300014t261k0j/resistor-t emperature-sensing-1k/dp/1174306?whydiditmatch=rel_3&matchedProduct=73 2278

Also for breadboarding, I'm thinking of subbing the TL080 for a TL071, as it's also a JFET input single opamp. Do you think this would be ok?
2thick4uni
That tempco will be fine, not perfect but near enough, I know a few people have used these for vco's and they have performed well. If you are in the UK I can send you a 3300ppm 1k tempco, I have a few spares.

The TL072 should work ok, but do be aware that the 10pf capacitor will not then be connected to anything as the TL080 has a different offset null arrangement to TL081 and TL071. Pin 8 is comp on the TL80 but N/C 0n the TL071 and 81. I think the 10pf cap is just there for stability, I'd guess it will run just fine without it, or it could be wired across the feedback loop instead which may do the same job. Thing to do is wire it up and try it. TL080 is still available from Mouser if all else fails.
2thick4uni
Just had another look at TL080 and 81 datasheet - you don't need to fit the 10pf cap if using the TL071 or 81 as they already have an 18pf cap fitted internally in the same position, whereas it is not included in TL080 and has to be fitted externally. So, the TL081/2 and TL071 or TL072 should work fine and in that case you can omit the 10pf cap from the circuit.
Jackdamery




I got the oscillator part of the circuit working, sans expo converter on +/-9v. I subbed the TL080s for TL071s, the diodes are 1N4148s and the FET is a 2N4393. To get it to work without the expo I connected the input via a 10k resistor to -9V. It was oscillating at a very high frequency.
Jackdamery
Any ideas how I can slow this oscillator down? I've already tried adding larger capacitors.
2thick4uni
A resistor between pin 2 of the op amp (R17A) and ground should do it, or make it a pot to vary the frequency. As I understand it he expo converter sucks current out of the oscillator, the more current removed, the lower the pitch.
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3, 4, 5, 6, 7  Next [all]
Page 1 of 7
Powered by phpBB © phpBB Group