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 Previous  1, 2, 3, 4, 5, 6, 7  Next [all]
Author An in depth look at the Roland Jupiter 8's VCO
Jackdamery
oldcrow wrote:


This is what I did for the crOwBX VCOs. The J112 FET works fine in place of the 2N4302, would probably do the same here. Of course, you now need the LS318 in the SOIC-8 or TO-78 (can) package as they no longer sell the PDIP version. angry


Wow that's cool, I was completely unaware of this. So you already did the same kind of thing with an OBX?

I was perusing the service manual today and noticed that it said the TL080 and TL082s were selected by slew rate. Presumably those with a faster slew.
2thick4uni
Hiya Jack,

I've just done a simulation of the circuit in LTSpice, the oscillator is running way too fast, as you found. I changed C8a from 15p to 15n and it is running nearer the speed it should, with 0volts in I'm getting just under 1khz - much better than the 30khz or more with it at original value but still a bit on the high side. The only other component that varies the oscillator speed is R18a, but I needed to take this up to 680k with C8a at 15pf to get the same result.

Anyone out there want to take their JP8 apart and let us know these actual component values in the VCO??? very frustrating

The expo converter works fine and everything is as expected there.

I've tried many different nfet transistors for and all of them seem to work ok, very little difference between them.

If someone can pm me to let me know how to upload pix I'll put up some pix of the sim running.




.
Nordcore
I haven't done exactly *this* circuit - but from experience with similar cicuits I guess it does work well with the shown component values.
At 0 Volt input the current in the 1nF integrator capacitor is 15V/470k= 32µA[1].
The comparator voltage is 11.5V, so it takes 360µs[2] to charge the capacitor to the comparator switch voltage, giving 2.7kHz.
(... the reset time is (guessed...) around some µs and ignored. Adjust the 15pf to give a pulse long enough to discharge C6A completely. Can be seen on the scope w/o problems. )


[1] Pin 11 (CA3046) is regulated to zero volts. So the current in Pin 11 is exactly 15V/470k. Pin 6 is zero volts, Pin 9 is zero volts (because input=0V). Because both transistors are matched, and they have the same base-emitter voltage, their collector current must be the same. So the current to C8a is the same as what goes through R9a.

[2] C=Q/U, Q=It => C=It/U. Its that dead simple because the collector of the transistor is a good current source - so current keeps constant and we don't have to do integrals. (As long as R17A is ignored=set to zero )
minisystem
2thick4uni wrote:
Hiya Jack,

I've just done a simulation of the circuit in LTSpice, the oscillator is running way too fast, as you found. I changed C8a from 15p to 15n and it is running nearer the speed it should, with 0volts in I'm getting just under 1khz - much better than the 30khz or more with it at original value but still a bit on the high side. The only other component that varies the oscillator speed is R18a, but I needed to take this up to 680k with C8a at 15pf to get the same result.

Anyone out there want to take their JP8 apart and let us know these actual component values in the VCO??? very frustrating

The expo converter works fine and everything is as expected there.

I've tried many different nfet transistors for and all of them seem to work ok, very little difference between them.

If someone can pm me to let me know how to upload pix I'll put up some pix of the sim running.


.


Any chance you'd be willing to zip and share the LTSpice project and models you used? I've successfully simulated filters, but have not had any luck with VCOs in LTSpice (lock up or won't oscillate). I find LTSpice is a great learning tool, but I've hit a wall when it comes to VCO design.

Back on topic: fascinating study Jack! I picked up 10 LS318s a few months ago, but they were bloody hard to get hold of. I built a discrete version of the Hearn Morley VCO using the LS318 and another version using the BCM847. Both worked and sounded great but the LS318 version definitely tracked better. I suppose this was due to the better log conformance of the 318? In the case of the JP8, I guess the auto-tune algorithm takes care of any non-ideal tracking.
Jackdamery
2thick4uni wrote:
Hiya Jack,

I've just done a simulation of the circuit in LTSpice, the oscillator is running way too fast, as you found. I changed C8a from 15p to 15n and it is running nearer the speed it should, with 0volts in I'm getting just under 1khz - much better than the 30khz or more with it at original value but still a bit on the high side. The only other component that varies the oscillator speed is R18a, but I needed to take this up to 680k with C8a at 15pf to get the same result.

Anyone out there want to take their JP8 apart and let us know these actual component values in the VCO??? very frustrating

The expo converter works fine and everything is as expected there.

I've tried many different nfet transistors for and all of them seem to work ok, very little difference between them.

If someone can pm me to let me know how to upload pix I'll put up some pix of the sim running.

.


Nice one. I got it working fine with C8A as 100pf, did you build the expo part as well?

I moaned at everyone on Gearslutz and they'ved provided me some really high res photos of the VCO module board, unfortunately top down so you cant read cap or transistor values, but still:
Gearslutz JP8 photos

.
Jackdamery
minisystem wrote:

Back on topic: fascinating study Jack! I picked up 10 LS318s a few months ago, but they were bloody hard to get hold of. I built a discrete version of the Hearn Morley VCO using the LS318 and another version using the BCM847. Both worked and sounded great but the LS318 version definitely tracked better. I suppose this was due to the better log conformance of the 318? In the case of the JP8, I guess the auto-tune algorithm takes care of any non-ideal tracking.


Thank you! So that was you that built the HMVCO. Is it similar? I have a couple of HM chips and they sound a track similarly to the JP8 one I built.

Did you use the DIP8 LS318s? I got the TO-71 metal can ones as free samples from a distributor in the UK. £3.40 each was the quote. Orders subject to minimum order value £50.00.
2thick4uni
[/img]
Nordcore wrote:
I haven't done exactly *this* circuit - but from experience with similar cicuits I guess it does work well with the shown component values.
At 0 Volt input the current in the 1nF integrator capacitor is 15V/470k= 32µA[1].
The comparator voltage is 11.5V, so it takes 360µs[2] to charge the capacitor to the comparator switch voltage, giving 2.7kHz.
(... the reset time is (guessed...) around some µs and ignored. Adjust the 15pf to give a pulse long enough to discharge C6A completely. Can be seen on the scope w/o problems. )


[1] Pin 11 (CA3046) is regulated to zero volts. So the current in Pin 11 is exactly 15V/470k. Pin 6 is zero volts, Pin 9 is zero volts (because input=0V). Because both transistors are matched, and they have the same base-emitter voltage, their collector current must be the same. So the current to C8a is the same as what goes through R9a.

[2] C=Q/U, Q=It => C=It/U. Its that dead simple because the collector of the transistor is a good current source - so current keeps constant and we don't have to do integrals. (As long as R17A is ignored=set to zero )



Many thanks Nordcore, very helpful indeedl!

I have got to the root of the problem so the sim now works correctly. The slew rate of IC4a was causing the problem. I modeled it with LT devices, I used the LT1013 instead of TL080, I didn't realise it had a much poorer slew rate. It's a better op amp but a poorer comparator d'oh!

So, I tried the LT1057 instead, it has similar slew rate to TL080. With that in circuit I can get C8a down to 220pf and everything runs ok. If I replace it with a comparator instead (I used the LT1017) then it runs fine with C8a at the documented value of 15pf. Could be that Roland had to select TL080's for fastest slew rate. I'm sure that Roland will be relieved to hear that the VCO does actually work! hihi

BTW at zero volts in it shows about 350mv at output of the summing opamp (pin7), this is because of the bias voltage applied by R1a and R5a, if I take these out of circuit it gives 0v at pin 7. With these in circuit it needs about 1.78v in for zero volts to expo convertor. I think that R5a could be for high end tracking?

I've included pix of sim and also zip file containing LTspice files for anyone else interested in further exploration!

2thick4uni
minisystem wrote:


Any chance you'd be willing to zip and share the LTSpice project and models you used? I've successfully simulated filters, but have not had any luck with VCOs in LTSpice (lock up or won't oscillate). I find LTSpice is a great learning tool, but I've hit a wall when it comes to VCO design.


Yep, no problem. Zipped files attached. Pretty sure all of the components are in standard LTspice library so should run without importing any extra component models.

Have fun!
Jackdamery
2thick4uni wrote:
Could be that Roland had to select TL080's for fastest slew rate. I'm sure that Roland will be relieved to hear that the VCO does actually work! hihi



Yeh if you look at page 13 of the service manual on the top right it mentions TL080s and TL082s are selected on slew rate, presumably the fastest.
neil.johnson
A good reference implementation of this kind of oscillator is the ASM-1 VCO.

Oh, and what you amusingly refer to as the "breakdown diode" is nothing more than converting the output of the op-amp-as-comparator into an open-collector output (see ASM-1 VCO use of LM311 comparator), only pulling the gate of the JFET to ground - you don't want the gate going any higher than one diode drop above the D-S terminals or the reverse-bias gate diode will start to conduct.

Neil
Jackdamery
neil.johnson wrote:
A good reference implementation of this kind of oscillator is the ASM-1 VCO.

Oh, and what you amusingly refer to as the "breakdown diode" is nothing more than converting the output of the op-amp-as-comparator into an open-collector output (see ASM-1 VCO use of LM311 comparator), only pulling the gate of the JFET to ground - you don't want the gate going any higher than one diode drop above the D-S terminals or the reverse-bias gate diode will start to conduct.

Neil


Ok, so whats the other diode doing?
neil.johnson
Jackdamery wrote:
neil.johnson wrote:
A good reference implementation of this kind of oscillator is the ASM-1 VCO.

Oh, and what you amusingly refer to as the "breakdown diode" is nothing more than converting the output of the op-amp-as-comparator into an open-collector output (see ASM-1 VCO use of LM311 comparator), only pulling the gate of the JFET to ground - you don't want the gate going any higher than one diode drop above the D-S terminals or the reverse-bias gate diode will start to conduct.

Neil


Ok, so whats the other diode doing?

The one strapped across the output of the integrator op-amp? At a guess I would say clamping that pin to no more than one diode below GND to protect the circuit against start-up conditions when power is applied (op-amps can do funny things while their power rails are coming up, including weird stuff on their outputs).

Neil
minisystem
2thick4uni wrote:
minisystem wrote:


Any chance you'd be willing to zip and share the LTSpice project and models you used? I've successfully simulated filters, but have not had any luck with VCOs in LTSpice (lock up or won't oscillate). I find LTSpice is a great learning tool, but I've hit a wall when it comes to VCO design.


Yep, no problem. Zipped files attached. Pretty sure all of the components are in standard LTspice library so should run without importing any extra component models.

Have fun!


Excellent. Thanks. Having fun now. thumbs up
Jackdamery
Okay so I built up the VCA section of the Jupiter using an LM13700 instead of the BA662s. I added a second DAC to the Arduino and used it to send a CV to the VCA with control only over decay from a pot into the arduino.

So the topology now is:

Farfisa keybed -> Arduino -> DAC for VCO CV & DAC for VCA CV -> Jupiter 8 VCO Saw/Triangle and Jupiter 8 VCA.

[s]http://soundcloud.com/dot/jupiter-8-homebuilt-vca[/s]
minisystem
Jackdamery wrote:
minisystem wrote:

Back on topic: fascinating study Jack! I picked up 10 LS318s a few months ago, but they were bloody hard to get hold of. I built a discrete version of the Hearn Morley VCO using the LS318 and another version using the BCM847. Both worked and sounded great but the LS318 version definitely tracked better. I suppose this was due to the better log conformance of the 318? In the case of the JP8, I guess the auto-tune algorithm takes care of any non-ideal tracking.


Thank you! So that was you that built the HMVCO. Is it similar? I have a couple of HM chips and they sound a track similarly to the JP8 one I built.

Did you use the DIP8 LS318s? I got the TO-71 metal can ones as free samples from a distributor in the UK. £3.40 each was the quote. Orders subject to minimum order value £50.00.



I've got a couple of Hearn Morley VCO chips but I've yet to get around to testing one. I used the SOIC8 LS318 (I used an SMT design as I wanted it as small as possible to keep PCB costs down). I got 10 from Trendsetter electronics but shipping to Canada was astronomical so I had them couriered to a cousin in NYC. Certainly not as easy as Digikey. They were about $4.50 USD each. In another version I used the BCM847 matched pair. At 1/10th the price and readily available, I plan to stick with the BCM847.

Looking forward to seeing how this develops! smile
Jackdamery
I received four TO-18 NF510 FETs from the US today. They have the old style National Semiconductor logo on them. I'm wondering if they are any different to the TO-92 NF510s from the photos of the Jupiter Module board?
filterstein
Would it be inpolite if i asked were you got them from?
Would like to see if they sound different for the system 700 vco clone.
Jackdamery
filterstein wrote:
Would it be inpolite if i asked were you got them from?
Would like to see if they sound different for the system 700 vco clone.


Not at all. They were pulls from apparently working test equipment advertised on ebay.

Here's the auction

How many are used in the system 700 vco?
filterstein
Ah, luck in auction.
I hoped you'd found a more regular source.

There is one for each vco module.
Many early rolands use the ua726/nf510 combination.
Jupiter 8 is different in that others use 2sk30 in designs that are not based on the ua726.
Jackdamery
filterstein wrote:
Ah, luck in auction.
I hoped you'd found a more regular source.

There is one for each vco module.
Many early rolands use the ua726/nf510 combination.
Jupiter 8 is different in that others use 2sk30 in designs that are not based on the ua726.


Well I just messaged him asking what piece of equipment he pulled them from. Once I've done some IDss, Gs, etc tests I can probably send you one to test with. Where are you based?
Jackdamery
Ok so ran some tests and suprise, suprise, the TL080 showed no difference in amplitude or frequency from the TL071 in the comparator section. The only difference is the TL080 worked fine without the C8A flyback capacitor and showed no difference between 15pf and 100pf.

In the integrator there was no difference in amplitude or frequency between the TL080 and the TL071, the TL080 needed the 10pf compensation capacitor as specified, otherwise it went into super-sonic frequencies.

There was no measurable difference between the NF510 and the 2N4393 either.

Regarding the NF510 pulls I bought on ebay the guy came back saying.

Quote:
I'm not positive but I think it was a Hewlett packard function generator. I've already scrapped it for the gold.


Jarno
Gold? The gold plating on cardedges and connectors, really?
Jackdamery
Jarno wrote:
Gold? The gold plating on cardedges and connectors, really?


Well according to some ex-BT coworkers, there aren't many old mechanical telephone exchanges left because the silver in the relays is worth more than the units.
negativspace
Is that red paint on your cutting board, or was this transistor removed from the still-beating heart of some unfortunate beast?
Jackdamery
negativspace wrote:
Is that red paint on your cutting board, or was this transistor removed from the still-beating heart of some unfortunate beast?


Lol, it's red paint I spilled aged 12, which is half my life ago cry
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page Previous  1, 2, 3, 4, 5, 6, 7  Next [all]
Page 3 of 7
Powered by phpBB © phpBB Group