CCS + Voltage regulator to supply DAC and ADC

[domingo]

So here is a circuit I came out with, one of my first ones. Its purpose is to supply with 5.2v a DAC (TDA1387) and an ADC (PCM1802) simultaneously.

The challenge is that, for specific reasons that are not in the drawing, the DAC has to work in a voltage above ground. Therefore zener D38 (+3v). Meanwhile the ADC has to be grounded at 0v, so I hang on the 8.2v zener circuit of the DAC, which is constant voltage and constant current, to reference the opamp via a resistor divider. I suppose that using a voltage divider before the opamp, rather than resistors within the feedback loop, the opamp will work in unity gain and therefore not dissipate any power (unlike opamp-based linear regulator circuits).

The output current of the opamp is increased with a transistor. This is mostly because in simulations this was the only way for me to get the wanted 5.2v. The transistor might also help in case the opamp runs short to power the 25-30mA ADC, but I'm not sure if it will produce unwanted power dissipation in exchange. This is for a device powered on 12v battery supply, which never goes under 9v following discharge.

–Some caps are there as per datasheet default (C75, C40; C78, C81), some other because the circuit takes part on a major one (C83).

Any comments on the design will be highly appreciated! Before trying it out.

 

[domingo]

I came to realise that the linear voltage regulator section was lacking the feedback loop, so it was unstable to load changes. This thread can be considered irrelevant.

 

[abbey road d enfer]

I came to realise that the linear voltage regulator section was lacking the feedback loop, so it was unstable to load changes. This thread can be considered irrelevant.

That could be easily overcome by connecting the opamp's inverting output to the emitter of the output transistor.
However, the 8V rail would be quite noisy because the npise of the 3V zener is not filtered at all.
Now I'm curious as to why you need this strange power rail offset...

 

[domingo]

Hi Abbey,

Do you mean connecting the opamp's 'inverting input' directly to emitter follower would work, even without feedback resistors (unity gain)? Saving that pair of transistors would indeed help me.

On the 3v zener, I thought it wouldn't produce noise to Vref, since it only comprises the DAC circuit (linear regulator is grounded directly to 0v). But I trust you and thanks for the observation. What would you recommend as filter? Or maybe a different approach is needed?

Context: The DAC's DC output voltage compliance is max. 3.5v (TD1387), but its output IV opamp is set at 6v (midpoint of 12v). By increasing the DAC's ground in 3v, its output voltage will be 3v instead of 6v, therefore within VDCC. The 8.2v zener with constant current PNPs (to avoid battery drain) produce, together with the 3v zener, the 5.2v DAC's VDD.
Since I need to supply a 5.2v ADC in parallel, which nonetheless needs to be grounded at 0v, I build another (linear) voltage regulator using a spare opamp. Instead of introducing another zener (more noise), I thought of using the stable 8.2v of the DAC (unaffected by battery drain) with a voltage divider to reach the 5.2v needed for the ADC's voltage regulator's Vref.

 

[abbey road d enfer]

Hi Abbey,

Do you mean connecting the opamp's 'inverting input' directly to emitter follower would work, even without feedback resistors (unity gain)?

Yes. You need to know that the sourcing capacity (i.e. capacity of providing positive current) is higher to the sourcing capacity (capacity at drawing negative current) because the transistor conducts in only one direction. that may or may not be an issue.

Context: The DAC's DC output voltage compliance is max. 3.5v (TD1387), but its output IV opamp is set at 6v (midpoint of 12v). By increasing the DAC's ground in 3v, its output voltage will be 3v instead of 6v, therefore within VDCC. The 8.2v zener with constant current PNPs (to avoid battery drain) produce, together with the 3v zener, the 5.2v DAC's VDD.
Since I need to supply a 5.2v ADC in parallel, which nonetheless needs to be grounded at 0v, I build another (linear) voltage regulator using a spare opamp. Instead of introducing another zener (more noise), I thought of using the stable 8.2v of the DAC (unaffected by battery drain) with a voltage divider to reach the 5.2v needed for the ADC's voltage regulator's Vref.

Instead pof making a mess of the power rails, I would address the offset issue by biasing the opamp. Why is it set at 6V?
A schemo would help.

 

[domingo]

The DAC's output opamp is set at 6v because its voltage reference is taken from a +/-6v rail splitter that supplies some EQs and mixers. You opened my mind with your idea of biasing the opamp instead, I'll work on it. Meanwhile please see full diagram, sorry if it looks like a mess.

 

[abbey road d enfer]

You should use the REF pin of the DAC to bias the opamps.

 

[domingo]

I thought REF was intended as an input voltage (case different from VCC), not as an output voltage meant to bypass the IV opamp. Should I expect VDCC voltage coming out from there (0-3.5v)? I'll breadboard it soon.

What would you recommend as 5v regulator to be shared by both DAC and ADC? Noise free particularly for the ADC and hopefully not so inefficient. Single zener with filter caps? Discrete linear voltage regulator with BJTs or opamp? TL431?

 

[abbey road d enfer]

I thought REF was intended as an input voltage (case different from VCC), not as an output voltage meant to bypass the IV opamp. Should I expect VDCC voltage coming out from there (0-3.5v)?

The TDA1387 datasheet does not mention an output offset. The block dgm shows the opamps referenced to 0V. Maybe the normal operation is with the opamps output offset negatively, which would require AC-coupling caps.
The REF pin is indeed a "reference decoupling"; it is certainly not capable of delivering significant current, but in order to bias the opamps, the current requirement is negligible, so, on teh condition that the voltage there is teh same as the output offset, it should work.

What would you recommend as 5v regulator to be shared by both DAC and ADC? Noise free particularly for the ADC and hopefully not so inefficient. Single zener with filter caps? Discrete linear voltage regulator with BJTs or opamp? TL431?

If you're ready to spend significant money, you may use one of these ready-made solutions.
https://www.ldovr.com/product-p/lt3045-a.htmThere are other possibilities, such as https://www.mouser.fr/c/?marcom=155262253 but I don't think it's really DIY friendly.
Why don't you try with a good old 317?
Zener with filter is a no-go. Not stable enough, too noisy, energy guzzler.
TL431 is a shunt regulator, cannot be applied directly to your needs.
A hybrip opamp+BJT is potentially excellent, but proper tuning may take very long.
HiFi enthousiasts who have endeavoured such tasks found out it takes a long time from the idea to a finished practical solution.
Anyway I would recommand two separate regulators, so there would be no interference between their points of reference.

 

[domingo]

Thank you a lot. I'm thinking mostly to use low-tech and not so highly top of the line packed solutions, both to reduce costs and keep it standard for parts replacement/availability.

An opamp-based linear regulator with a clean 5v voltage reference instead of zener seems the best to me at the moment (LT1029 comes in TO-92 and can buy it here). I take your advice on separate regulators, but I only have one spare opamp in the board which I leave for the ADC. DAC regulated only with BJT's and a voltage divider to bias opamps in the diagram below (case direct connection to REF pin doesn't work). Is this making sense?

Appreciate a lot your mind opening.

 

[abbey road d enfer]

Thank you a lot. I'm thinking mostly to use low-tech and not so highly top of the line packed solutions, both to reduce costs and keep it standard for parts replacement/availability.

An opamp-based linear regulator with a clean 5v voltage reference instead of zener seems the best to me at the moment (LT1029 comes in TO-92 and can buy it here). I take your advice on separate regulators, but I only have one spare opamp in the board which I leave for the ADC. DAC regulated only with BJT's and a voltage divider to bias opamps in the diagram below (case direct connection to REF pin doesn't work). Is this making sense?

There's no bias current for the zener and base of T1.
The hybrid regulator just cannot work.
Why don't you use 78L05's?

 

[domingo]

I think you are right. I didn't know the 78L05 but apparently I can get it around the corner and no way I can beat its 50mV noise figure @ 12v and 6mA quiescent current. Especially considering that I'm just an enthusiast and the main thing is that this sounds good.

Thanks for the advice. I'll try it tomorrow and research how to implement it. Probably the regulator will need some caps as well.

 

[abbey road d enfer]

I'll try it tomorrow and research how to implement it. Probably the regulator will need some caps as well.

You need a capacitor at the input, close to the regulator. Actually it is frequent to use a ceramic cap (0.1-1uF) in parallels with a 'lytic, which can be any value.
At the output, you need a ceramic cap for stability, 0.01uF is a minimum. The higher the value, the better the noise. I commonly use a 0.1uF ceramic in parallels with a 100uF. Again, it must be placed as close as possible to the regulator.

 

[domingo]

Elliott recommends the use of a 10ohm series resistor at the output plus a 1000u cap to ground.

https://sound-au.com/articles/vi-regulators.html#s3
The main reason of the resistor is to compensate for low ESR of newer caps. Would you back that up for the current application (both for ADC and DAC)? ADC is a pcm1802 btw, input impedance is 20k.

 

[abbey road d enfer]

Elliott recommends the use of a 10ohm series resistor at the output plus a 1000u cap to ground.

I haven't seen that in the link.
I must say I haven't read it all.
At some point he says that increasing the value of the output capacitor doesn't improve ripple; he's right. However, the capacitor takes care of HF noise, when the regulator's output impedance increases, as he mentions. It also absorbs current surges. So I maintain that large output caps are a benefit.

The main reason of the resistor is to compensate for low ESR of newer caps.

Seems odd to me. ESR of "older" caps was typically between 0.1 and 1 ohm. 10 ohms seems a lot.

Would you back that up for the current application (both for ADC and DAC)?

Can you post a schemo?

ADC is a pcm1802 btw, input impedance is 20k.

Input impedance of the circuit has no correlation with power rail requirement.

 

[domingo]

Elliott writes in section 3 (on 7815):

"Typical output noise is claimed to be 90µV. An easy way to improve the noise and ripple voltages is to add a simple resistor/capacitor filter at the output of the regulator. For output currents of 100mA or less, a 10 ohm resistor and 1,000µF cap will reduce the output voltage by 1V at 100mA, but will reduce 100Hz ripple by another 16dB (minimum). It will also reduce wideband noise. At 1kHz, any regulator noise is reduced by 36dB, and 56dB at 10kHz."

In a discussion here contributors also mention that low ESR can produce noise with the LM317 and recommend also a series resistor (100ohm!) at the output before a large cap to ground.

I'm worrying only because I want to grant very low noise to the ADC, since it is recording audio in 16bits and the noise floor might be heard when I raise the volume of the sampled files. PCM1802 board I'm using comes with its own bypass 10u elco + 0.1u ceramic embedded. The 10u elco included seems low ESR, therefore part of my worry. The other general worry is that the ADC SNR is 105dB and I wouldn't like avoid downgrading that as much as possible from the power supply's side.

In attachment goes diagram so far.

 

[abbey road d enfer]

Elliott writes in section 3 (on 7815):

"Typical output noise is claimed to be 90µV. An easy way to improve the noise and ripple voltages is to add a simple resistor/capacitor filter at the output of the regulator. For output currents of 100mA or less, a 10 ohm resistor and 1,000µF cap will reduce the output voltage by 1V at 100mA, but will reduce 100Hz ripple by another 16dB (minimum). It will also reduce wideband noise. At 1kHz, any regulator noise is reduced by 36dB, and 56dB at 10kHz."

The problem there is that load regulation is totally screwed up.

In a discussion here contributors also mention that low ESR can produce noise with the LM317 and recommend also a series resistor (100ohm!) at the output before a large cap to ground.

THis is nuts.

In attachment goes diagram so far.

It's good.

 

[JohnRoberts]

typical trade off, one parameter for another...

Modern regulators are generally better for all parameters.

JR