teach me exponential converters & tempco

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: luketeaford, Joe., lisa, Kent

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

teach me exponential converters & tempco

Post by EATyourGUITAR » Fri Jun 14, 2013 9:22 am

I have been looking at various exponential converters and they all seem very similar.

I will post the schematic of the VCO I am working on so we can have some context to this discussion.

Image

T3, T4 is a current mirror that sets the charge time for the capacitor in the core of the ramp up vco.

I used R9 11K6 to set the upper limit on vco frequency by limiting the maximum current through the mirror.

T1, T2 are the exponential converter.

some exponential converters have R8 as 100k but then others have 1M. how do you select the best value?

can I buy a complementary transistor pair on one IC that has advantages in temperature stability?

if I already have an IC that has 2 NPN and 2 PNP, is there a way to do an exponential converter using just the 2 NPN's that are left over on the IC? if that is even possible it would all be thermally coupled by default.

after I figure out this exponential converter, I also need tempco compensation. are there any tutorials online?

just for fun. here is what it looks like with a scope probe placed on the timing cap

Image

frijitz
Super Deluxe Wiggler
Posts: 1637
Joined: Mon Aug 31, 2009 3:37 pm

Post by frijitz » Fri Jun 14, 2013 9:44 am

Could you say something about why you are using that particular design?

User avatar
gwaidan
Super Deluxe Wiggler
Posts: 1938
Joined: Sat Mar 07, 2009 7:26 pm
Location: Sinney, Orstraya
Contact:

Post by gwaidan » Fri Jun 14, 2013 10:07 am

Why not have the input resistors and pot go into an inverting op-amp configuration with the 2k resistor in the feedback path, then to a trimmer which can attenuate its output? That takes care of scale tracking adjustment and will then enable you to "flip" T1 to NPN, flip T2 to PNP and thus drive the timing cap directly from it, eliminating the current mirror.

The 2k above should be a tempco.

AFAIK UJTs haven't been commonly used for audio VCOs since the 1960s-what inspired you to give it a shot?

User avatar
The Real MC
Veteran Wiggler
Posts: 670
Joined: Sun Jun 02, 2013 8:36 pm
Location: In front of computer

Post by The Real MC » Fri Jun 14, 2013 10:23 am

I haz a wrench, to fix da synth.
http://www.analoguediehard.com/

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Fri Jun 14, 2013 10:37 am

frijitz wrote:Could you say something about why you are using that particular design?
gwaidan wrote:AFAIK UJTs haven't been commonly used for audio VCOs since the 1960s-what inspired you to give it a shot?
I found a 3 transistor ramp oscillator in a book I borrowed from my local library. I got one vintage UJT sealed in the plastic for free in a trade. it seemed like a fun project with no money invested. I just got an oscilloscope so I wanted to look at waveforms.
gwaidan wrote:Why not have the input resistors and pot go into an inverting op-amp configuration with the 2k resistor in the feedback path, then to a trimmer which can attenuate its output?
I have seen this before in some schematics online. some people put the 2K tempco in the feedback path of the opamp and others put a 1k tempco on the bottom of a voltage divider on the output of the opamp. I think I understand it but I'm not sure if there are math formulas available to plot the tempco curves against the temperature drift of my circuit. I have never done anything like this before so I'm not sure when I am over or under compensating for temperature.
gwaidan wrote:That takes care of scale tracking adjustment and will then enable you to "flip" T1 to NPN, flip T2 to PNP and thus drive the timing cap directly from it, eliminating the current mirror.
good idea. I will try it.

modulus in moduli
Learning to Wiggle
Posts: 7
Joined: Thu Dec 09, 2010 11:26 am

Post by modulus in moduli » Fri Jun 14, 2013 12:44 pm

I always like discrete VCO's (or anything). OPAMPS are great for low distortion and easy to design whith and discrete VCO's might have less performance, but they have more character. and they are more fun to design :)

I tune my guitar every time I play it and do the same with my VCO's (of course I did not eat my guitar)

about your design:

Expo converter: disadvantage of this topology is that T1 and T2 cannot be matched because they are different types.

T5 does not seem to do anything

T6: I don't follow your design choices here, seems a voltage follower but why the R at the base? then anything connected to the timing cap better have a high imput impedance so a MOSFET such as bs170 might be a good option.

I am missing a reset circuit

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Fri Jun 14, 2013 2:20 pm

modulus in moduli wrote:I always like discrete VCO's (or anything). OPAMPS are great for low distortion and easy to design whith and discrete VCO's might have less performance, but they have more character. and they are more fun to design :)
and lower parts count 8_)
modulus in moduli wrote:of course I did not eat my guitar
neither did I. not yet. it is there in case of emergency.
guitar = money = food
modulus in moduli wrote:about your design:

Expo converter: disadvantage of this topology is that T1 and T2 cannot be matched because they are different types.
well yes and no. you can have matched gain and you can have them thermally coupled on the same IC. but because of the physics involved in semiconductor construction they will always make the NPN BJT a bit differently than the PNP BJT. in that way they cannot be matched because they have different construction, materials, and doping.
modulus in moduli wrote:T5 does not seem to do anything
ok let me first say that T5 is the core of the oscillator. that symbol is for a unijunction transistor. it will discharge the cap very fast if the voltage gets close to B2v. when the voltage gets down to about 1v above ground it will go back into high impedance mode to allow charging of the cap again.
modulus in moduli wrote:T6: I don't follow your design choices here, seems a voltage follower but why the R at the base? then anything connected to the timing cap better have a high imput impedance so a MOSFET such as bs170 might be a good option.

I am missing a reset circuit
this is good observation. I also had that idea about using a BS170 for the extremely high input impedance but the linearity might not be as good as BJT from what I have read online about fet voltage followers. from my experiments I found that the impedance of T6 is not an issue since it will always be there and it will always be following the same range of voltages. there is an issue with cascading output impedance changes backwards when patching the output of T6 to a load. my frequency went from 68Hz to 47Hz! that is when I decided to add the 10k to the output. I will have an opamp output buffer in the finished design that should keep the impedance of T6 constant for both base and collector.

the 10K at the base of T6 is only there to keep T6 in the linear region. I need to do more testing to get the best value. without the base resistor I get a slightly rounded top. at the output of T6.

if you have some experience with MOSFET voltage followers I would be interested to know more about it.

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Fri Jun 14, 2013 2:31 pm

modulus in moduli wrote:I am missing a reset circuit
I have it but I am going to test it before I post a schematic. I do not like the sync circuits I see online because they all just send a narrow pulse of a predetermined length. my method simply charges the cap as fast as possible and then stops charging when the UJT resets automatically. basically your sync signal runs through a comparator to a flip flop clock input. then the flip flop gets a reset signal everytime the UJT resets. in this way it will start syncing on a rising edge regardless of trigger length and it will not have a limited frequency range.

User avatar
gwaidan
Super Deluxe Wiggler
Posts: 1938
Joined: Sat Mar 07, 2009 7:26 pm
Location: Sinney, Orstraya
Contact:

Post by gwaidan » Fri Jun 14, 2013 6:02 pm

Formula for the expo converter is it needs to attenuate 1V/oct to approximately 18mV/oct at 25 degrees C to drive the transistors, hence most converters have about about a 54:1 ratio in a resistive divider. Transistors have about 3300ppm temperature drift at that temperature, so the :1 resistor in the divider is usually has a +3300ppm tempco to compensate. Using a 100k/2k pair as you did will "undershoot" the division ratio, hence you will need the attenuator I suggested to get you to 18mV/oct output.

The NPN/PNP pair approach is a worthwhile one-Moog used it in Minimoogs and Aries used it in the 70s. Just make sure that both transistors and the 2k resistor are in as close contact as possible, preferably with nonconductive heaatsink compound gooping it all together

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Mon Jun 17, 2013 11:31 am

thanks to everyone for your help. I have tested the current mirror with resistors to ground so I have some data on the range of this vco using various charging currents and timing caps. I had some problems with leakage and an unmatched current mirror preventing me from doing anything longer than 6 minute cycles. it also resulted in non-linear ramp using very low charging currents. 100k or less on the left side of the current mirror results in a mostly linear ramp. this is even without the T6 base resistor or the opamp buffer out. I used 100nf timing cap for VCO range and 2200uf for LFO range.

here is the new proposed schematic using what I have learned in this thread. there might be ways to do it on less parts. maybe the output buffer can be just one dc mixer with a trimmer in the feedback path. I used two opamps on the output for now for the best chance at linearity with my limited understanding. tell me what you think.

Image

User avatar
daverj
Vintage Video Wiggler
Posts: 8514
Joined: Fri Jun 19, 2009 11:09 am

Post by daverj » Mon Jun 17, 2013 2:13 pm

I would suggest getting rid of T6 and it's resistors and using IC1B in a non-inverting configuration with the plus pin tied directly to the timing cap. Then do your gain and bias adjustments at IC2A. That way the op amp provides a very high impedance load to the timing cap and won't effect the signal as much as T6 does.

While you're at it I would also swap IC1A and IC2A so that the two output op amps are in the same package. It'll make the wiring less messy.

User avatar
Peake
I'm in ur DIY. Filling cases with Buchla
Posts: 7061
Joined: Fri Nov 07, 2008 10:00 pm
Location: Loss Angeles

Post by Peake » Mon Jun 17, 2013 4:57 pm

2N2646 FTW!

http://www.freeinfosociety.com/media.php?id=4142

Note: No expo converter in that pic. That's in the "A". Best luck on your project!
This is not the place I'd imagined it to be.

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Mon Jun 17, 2013 5:48 pm

the output is inverted but that is an easy fix. all I need is one more opamp if I decide I want both saw and ramp output jacks.

I was planning on using AD648 for the 1v/oct input but since you made the suggestion about using two opamps from the same IC for the output, I switched the input to a single version AD548. the AD548 is more money for less stuff. but I was reading an article today about how unused opamps can consume current or cause damage :deadbanana:

is this what you had in mind?

Image

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Mon Jun 17, 2013 5:54 pm

Peake wrote:2N2646 FTW!

http://www.freeinfosociety.com/media.php?id=4142

Note: No expo converter in that pic. That's in the "A". Best luck on your project!
yes I have been all over this schematic. my opinion is that there are too many trimmers, too many transistors on the cv input (adding up the A + B) and redundant octave switches. I also think the sin looks like shit on the scope. love it or hate it I guess. I'm trying to make this project just a little more practical to build while adding improved performance. the only rare part I use is a $3 UJT. anyway thanks for the link, I'm sure there are people on the forum that are seeing it for the first time.

paults
mice nuts
Posts: 4472
Joined: Fri Jan 16, 2009 12:46 am

Post by paults » Mon Jun 17, 2013 9:03 pm

Rare? Here is a shitpile from a real vendor:

http://www.newark.com/jsp/search/produc ... KU=35C0693

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Tue Jun 18, 2013 12:26 am

paults wrote:Rare? Here is a shitpile from a real vendor:

http://www.newark.com/jsp/search/produc ... KU=35C0693
thanks for the link. I thought there was no new production of any 2646 but I guess I was wrong.

User avatar
Peake
I'm in ur DIY. Filling cases with Buchla
Posts: 7061
Joined: Fri Nov 07, 2008 10:00 pm
Location: Loss Angeles

Post by Peake » Tue Jun 18, 2013 1:36 pm

EATyourGUITAR wrote:
Peake wrote:2N2646 FTW!

http://www.freeinfosociety.com/media.php?id=4142

Note: No expo converter in that pic. That's in the "A". Best luck on your project!
yes I have been all over this schematic. my opinion is that there are too many trimmers, too many transistors on the cv input (adding up the A + B) and redundant octave switches. I also think the sin looks like shit on the scope. love it or hate it I guess. I'm trying to make this project just a little more practical to build while adding improved performance. the only rare part I use is a $3 UJT. anyway thanks for the link, I'm sure there are people on the forum that are seeing it for the first time.
Some of the waveforms on the 901 series -are- "not perfect", and some people like the slight difference, the same as with the Buchla 258 oscillators, which have extremely imperfect saw and square waves. Follow your ears! Best luck on your design!
This is not the place I'd imagined it to be.

User avatar
Peake
I'm in ur DIY. Filling cases with Buchla
Posts: 7061
Joined: Fri Nov 07, 2008 10:00 pm
Location: Loss Angeles

Post by Peake » Tue Jun 18, 2013 1:37 pm

2nd post is the same as the last, but slightly detuned.
This is not the place I'd imagined it to be.

User avatar
daverj
Vintage Video Wiggler
Posts: 8514
Joined: Fri Jun 19, 2009 11:09 am

Post by daverj » Tue Jun 18, 2013 3:01 pm

EATyourGUITAR wrote:the output is inverted but that is an easy fix. all I need is one more opamp if I decide I want both saw and ramp output jacks.

I was planning on using AD648 for the 1v/oct input but since you made the suggestion about using two opamps from the same IC for the output, I switched the input to a single version AD548. the AD548 is more money for less stuff. but I was reading an article today about how unused opamps can consume current or cause damage :deadbanana:

is this what you had in mind?

Image
The 15K to ground on the plus input of IC1A (R20) isn't needed with a TL072. It is something that is used with bipolar op amps but not FET op amps like the TL072. (and when it is used with bipolar op amps it's value must match the parallel resistance on the other input). Just tie that pin directly to ground.

The 100K feedback resistor on IC1B is also not needed. You can tie the output of IC1B directly to it's minus input.

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Fri Jun 21, 2013 9:33 am

thanks Dave. I also dropped the redundant attenuator at the input. I might not get to testing this right away but I will post when I do.
Image

User avatar
Dr. Sketch-n-Etch
Super Deluxe Wiggler
Posts: 7831
Joined: Thu Jan 21, 2010 1:46 pm
Location: Vancouver, BC, Canada

Post by Dr. Sketch-n-Etch » Fri Jun 21, 2013 11:58 am

This is an entertaining thread.

I have tried to understand where to put tempcos, how to put tempcos in unconventional places, and how to get tempco without tempcos.

In every case, I started with a current node balance model of the circuit in question. Without some basic algebra, all of these questions are just stabbing in the dark. IMHO, if you don't know how to develop a basic mathematical model of the circuit in question, then you have almost no chance of finding the correct placement and value of a tempco resistor.
this night wounds time,

User avatar
EATyourGUITAR
has no life
Posts: 5398
Joined: Tue Aug 31, 2010 12:24 am
Location: Providence, RI, USA

Post by EATyourGUITAR » Sun Aug 18, 2013 8:28 pm

I am waiting on a power cable to test out the saw core. anyway, here is a picture. this one has a SSM2220 matched PNP pair. I can swap out one of the resistors for tempco later. it is temco compatible.
Image

tubehead
Learning to Wiggle
Posts: 18
Joined: Sat Mar 14, 2015 7:42 pm
Location: United States

Post by tubehead » Tue Mar 31, 2015 2:21 pm

the first generation of moog modular oscillators used UJTs and they didn't have the precision needed for proper pitch tracking for musical oscillators, so he redid the design and the rest is history. in the 60s UJTs were a cheap and cheerful solution for non-precise oscillators but they died out quick when better, cheaper BJTs were available and CMOS crushed them both to make cheap, accurate electronic oscillators.

Jarno
Super Deluxe Wiggler
Posts: 2772
Joined: Mon Nov 15, 2010 6:34 am
Location: Rosmalen, NL

Post by Jarno » Fri Apr 22, 2016 3:53 pm

Old thread I came across, but interesting!
So, I have some PUT's in my parts box, and these are similar components, right?
Can they be used in the schematic above, does anything need changing?

User avatar
Graham Hinton
Super Deluxe Wiggler
Posts: 3284
Joined: Fri Jul 02, 2010 6:28 pm
Location: England
Contact:

Post by Graham Hinton » Sun Apr 24, 2016 8:50 am

UJTs are temperature sensitive devices so compensating an exponential converter won't help a lot as the oscillator still won't be stable. They are also sensitive to magnetic fields.

Ignoring that "feature", in the last schematic one of VR1 and VR2 is redundant, they both adjust the scale.
There are two ways of summing and compensating:
1) put the tempco resistor in the adder feedback to reduce the 1V/oct by a factor of 50 to 20mV/oct and have a trimmer to get 18.5mV/oct at the transistor base and take up prior errors. In this case the TC should be exactly the same as the transistor: 3353ppm @ 25 deg C.
2) put the tempco resistor at the transistor base and the trimmer in the adder feedback. In this case the TC is changed by being in a potential divider with a resistor having a much smaller TC and will be reduced so a larger TC is needed. If R1 is the larger resistor with negligible TC and R2 the tempco the composite TC is given by:

New TC = (R1 + R2)(1 + TC) / (R1 + R2(1 + TC)) - 1

e.g. If R2 is a 1k 3600ppm tempco then R1 needs to be 13.6k.

In both cases TCs of 3000 and 3300 ppm are lower than needed, but as they vary +/-10% some rated at 3300 ppm will be correct for case 1. The problem is selecting them...

Lots of classic VCO designs, e.g. ARP, EMS, sum the CVs into the tempco resistor through different value resistors. This gives every input a different interacting TC! Be careful what you clone.

Post Reply

Return to “Music Tech DIY”