   Author 3046 expo temp.compensation with different voltages
Dimitree
 this is the scheme used on the Moog Prodigy to compensate the temperature drift on the expo.converter. I'd like to adapt this to +/-15V usage and also to +/-10V usage for some different projects that I'm making that don't use +/-12V. can anyone help me understand how should I adapt the voltage (some resistor changes, or some other mods?). I'd like to avoid 10V regulators for example
Synthiq
 First you have to adjust R11 for changes in the negative supply voltage. Transistor Q4 is not used, but the emitter, pin 13, is also the substrate voltage of the chip and should have the most negative voltage of all pins. This voltage is not super critical, so just set R11 to 1kohm per volt so R11=10kohm for -10V and R11=15kohm for -15V. It is fine to choose a somewhat smaller value as well. Q3 is the temperature sensor and you may want to keep the bias current from R12 constant when you change the supply voltage. The voltage across Q3 is ~0.6V, so keep VDD-0.6/R12 constant and this gives R12=82.5kohm for VDD=10V and R12=127kohm for VDD=15V. The current in the resistor divider R13-R15 should also be kept constant by changing R13 when the positive supply changes so the trim range to the inverting input of U2B stays the same. The total resistance is 67.22kohm at 12V and should scale with the supply voltage and should be 56.02kohm at 10V and 84.03kohm at 15V. Subtracting R14 and R15 from these values gives R13=52.2kohm at 10V and 80.2kohm at 15V. (It is also possible to keep R12 constant but it will change the Vbe of Q3 in a non-linear way when the supply voltage changes. It can still be compensated for by changing the R13-R15 resistor divider but the math gets more complicated.) R16 generates the base current to Q6 and must also be adjusted when the supply voltages change so the maximum current stays the same. The LF353 is not guaranteed to generate higher voltages than VDD-3V and base of Q6 is at ~VSS+0.85V worst case. This gives 20.15V across R22 at 12V supplies. This changes to 16.15V at +/-10V and 26.15V at +/-15V. This sets R22=17.6kohm at 10V and R22=28.55kohm at 15V. Round down when looking for standard values. R18 limits the maximum current in Q5 to less than 50mA, the maximum collector current for a transistor in 4046. There is about a 1V drop across Q5 and Q6, leaving VDD-VSS-1V across R18, So for +/-10V, R18 should be reduced to 19V/50mA or 380ohm and would increase to 580ohm at +/-15V. Round up when looking for standard values.
Dimitree
wow, many thanks!!

 Quote: The total resistance is 67.22kohm at 12V and should scale with the supply voltage and should be 56.02kohm at 10V and 84.03kohm at 15V. Subtracting R14 and R15 from these values gives R13=52.2kohm at 10V and 80.2kohm at 15V.

I noticed that those are not standard values resistors, so my question is, do I really need to keep the total resistance equal to the original design, or close is good enough? isn't this a voltage divider?

 Quote: The LF353 is not guaranteed to generate higher voltages than VDD-3V and base of Q6 is at ~VSS+0.85V worst case.

does these values depends on the opamp choise? must it be LF353?

I guess the only values that really need to match the calculations in order to work as intended, are those marked with 1% on the original schematic, right?
fitzgreyve
 Dimitree wrote: must it be LF353?

I've used a derivative of this circuit for my add on for the System-X VCO - that used 1/2 of a TL072 and works fine.
Synthiq
 Quote: I noticed that those are not standard values resistors, so my question is, do I really need to keep the total resistance equal to the original design, or close is good enough? isn't this a voltage divider?

It's fine to round to nearest standard value.

 Quote: must it be LF353?

The amplifier is not critical and you can use pretty much any opamp you have available as long as the input bias current is insignificant compared to the current s in R12 and R13-R15.

 Quote: I guess the only values that really need to match the calculations in order to work as intended, are those marked with 1% on the original schematic, right?

Yes, the 1% resistors are the most critical, especially R13 and R15 as they control the trim range. But you should still keep the R11/R10 ratio accurate and don't make R16 larger or R18 smaller.
Dimitree
 thanks, it's clear I guess I don't have to change the value of R17 since I've changed R16, right?
Synthiq
 R17 can stay the same. It only exists to pull the base down and turn Q6 off when the opamp output goes low. Since the base-emitter voltage is independent of the supply voltage, so is R17. When you turn on Q6 fully, the current in R16 is around 900uA while the current in R17 is 50uA, so small changes up or down in the latter will have minimal effect on the base current in Q6.
 Following the idea of using the transistors in the CA3046 for temperature compensation I wonder if would be possible to use the design above, or that in the Doepfer A-110 VCO Doepfer A-110 VCO schematic, as a direct replacement for the tempco containing parts in those VCOs that use a current sink (e.g. Thomas Henry's XR-2206 and X-4046). (Puh, long sentence.) As far as I see, it all comes down to get the correct voltage jumps at the emitter of the expo transistor pair.
cygmu
 dot matrix madness wrote: Following the idea of using the transistors in the CA3046 for temperature compensation I wonder if would be possible to use the design above, or that in the Doepfer A-110 VCO Doepfer A-110 VCO schematic, as a direct replacement for the tempco containing parts in those VCOs that use a current sink (e.g. Thomas Henry's XR-2206 and X-4046). (Puh, long sentence.) As far as I see, it all comes down to get the correct voltage jumps at the emitter of the expo transistor pair.

I can't see any reason that you couldn't deploy the same circuit more or less unchanged. The design here is heating the 3046 to a steady temperature so that temperature compensation is not needed. It leaves a pair of NPNs on the chip which you can then use in the Thomas Henry circuit without a problem, I think.
 cygmu wrote: I can't see any reason that you couldn't deploy the same circuit more or less unchanged. The design here is heating the 3046 to a steady temperature so that temperature compensation is not needed. It leaves a pair of NPNs on the chip which you can then use in the Thomas Henry circuit without a problem, I think.

Thanks.
Besides minor changes (heating current is limited by a resistor, and the temp trim goes to +10V in the A110), the higher frequency trimming is implemented differently in the A-110. I will have to try out on the breadboard.
sicpaul
 Hi wigglers, my System-X VCO with fitzgreyve's 3046 temp. compensation is my most stabile built, so i wish to modify several Thomas Henry VCOs. Sorry, my abilities in deeply circuit understanding are rather poor, so following questiones occure: 1. Is it that simple to replace Q1/Q2 (LM394/matched 3904) of XR2206 or X-4046 VCO with Q1/Q2 of 3046, tempco with 2k resistor and adding components to 3046 following the circuit of the first post? 2. What to do when expo converter uses matched 2N3906 like TH-8038/TH-555/TH-LM or TH-VCO1 oscillators? (By the way - why do some circuits use pnps and other npns in expo converter?) Thanks for helping hints.
guest
 yes, you can replace this circuit for any NPN expo converter, just remember to eliminate whatever temperature compensation they were using orginally (usually involves taking out the tempco resistor and replacing it with a regular one). NPN vs PNP has to do with which direction the current needs to flow into the circuit. the NPN sinks current, so it pulls current into itself from a higher voltage, whereas the PNP sources current, so it pushes current out into a lower voltage. PNP is usually used for LM13700 or 3080 circuits where a current needs to flow into the 13700 or 3080 down at -Vcc.
sicpaul
 guest wrote: NPN vs PNP has to do with which direction the current needs to flow into the circuit. the NPN sinks current, so it pulls current into itself from a higher voltage, whereas the PNP sources current, so it pushes current out into a lower voltage. PNP is usually used for LM13700 or 3080 circuits where a current needs to flow into the 13700 or 3080 down at -Vcc.

Thanks mate, that's the level i understand Now i wonder if there is a simple solution (means without too much calculations) to invert current flow 1:1 so a 3046 can be used for 13700/3080... (3046 is corrected, i wrote 3080 before editing)

Has anybody tried to use that kind of 1V/Oct control to an OTA driven (TH-) VCO who is willing to show the schematic of it?
Jarno
 I always thought 3080's and 13700/13600 are interchangeable, apart from the fact that the 13700's have two OTA's per package rather than a single. I recently did a board for two TH VCO's (from one of his books) and that originally used 3080's, plonked in a SMT LM13700 (but those boards are not ordered yet, let alone verified).
sicpaul
 Sorry Jarno, i had an error in my former question (corredted now)., i want to control all TH VCOs using LM13700 or CA3080 with a heatetd 3046 expo converter and do look for a schematic for that.
Jarno
 there's a PNP transistor pair in CA3096, slightly harder to get though. The rest of the transistors are NPN, so you could use those for the heating circuit, I presume.
sicpaul
 Thanks a lot. i do have a smd 3096 in stock. i'll give it a try.
 Page 1 of 1