Low-distortion Audio-range Oscillator

Measured or sim? Measuring noise that low is not easy. I also think the model for the TL431 doesn't reflect how it operates in this configuration.

I'll set up one of the production units and measure it.

Simmed of course. I'm not that fast at putting circuits together. What is it in reality?

I added a bit to it. I can't get a spectrum from it 10 seconds after settling.
The four LM317 are because of a current limit in the model.
 

Attachments

  • Demian's.JPG
    Demian's.JPG
    108 KB · Views: 299
  • Demian's2.JPG
    Demian's2.JPG
    84.2 KB · Views: 298
Last edited:
Member
Joined 2004
Paid Member
I have measured it. I had to build a transformer input "pre-preamp" to be able to measure something. It was in the 1 nV/rtHz range but I'll redo the measurement and document it.

Let me sketch a negative variation. I also did a shunt variant but have not built it. Do you need tracking? I think that would be important in some applications.
 
I have measured it. I had to build a transformer input "pre-preamp" to be able to measure something. It was in the 1 nV/rtHz range but I'll redo the measurement and document it.

Let me sketch a negative variation. I also did a shunt variant but have not built it. Do you need tracking? I think that would be important in some applications.

May as well add the tracking. I plan to put the power supply in it own enclosure.
I want to keep that transformer as far away from the oscillator as possible.

Also see what the influence on noise of C2 is. It not costing much to make it a high value.
In sim the settling is a few hundred uV so who cares.
 
Member
Joined 2004
Paid Member
Split Supply

Here is a first pass at it. Making the cap on the TL431 larger really slows down the startup. I did some tweaking but still room for more. The opamp can be pretty much any generic if the tracking doesn't need to be perfect.

I will "noodle" on an alternative solution that may work better. can you run the unbalanced loads from a different supply (Relays, ADC/DAC etc)? That will allow for almost zero ground current.
 

Attachments

  • Split supply.PNG
    Split supply.PNG
    34.6 KB · Views: 293
  • Split low noise tracking supply.asc
    5.9 KB · Views: 48
In the "widget" thread you had mentioned that LTSpice doesn't have an TL431 model. Christophe Basso put one together for his book: "Switch-Mode Power Supplies":

.SUBCKT TL431 1 2 3
C1 3 20 43p
Q1 3 20 10 QN3904
R1 10 7 240
R2 7 2 6600
Q2 3 7 2 QN3904 50
D1 1 3 D4mod
D2 2 3 D2mod
V1 5 2 DC=2.49
R4 3 0 1e12
D3 2 20 D3mod
C2 2 7 1n
G2 2 20 5 1 -250e-6
*
.model QN3904 NPN AF=1.0 BF=300 BR=7.5 CJC=3.5pf
+IKF=0.025 IS=1.4e-14 ISE=3e-13 KF=9e-16 NE=1.5
+RC=2.4 TF=4E-10 TR=21E-9 VAF=100 XTB=1.5
.model D3mod D N=0.01
.model D2mod D BV=36 CJO=4E-12 IS=7E-9 M=0.45 N=1.3
+RS=40 TT=6E-9 VJ=0.6V
.model D4mod D BV=100V CJO=4E-12 IS=7E-9 M=0.45
+N=2 RS=0.8 TT=6E-9 VJ=0.6V
*
 
Here is a first pass at it. Making the cap on the TL431 larger really slows down the startup. I did some tweaking but still room for more. The opamp can be pretty much any generic if the tracking doesn't need to be perfect.

I will "noodle" on an alternative solution that may work better. can you run the unbalanced loads from a different supply (Relays, ADC/DAC etc)? That will allow for almost zero ground current.

The unbalance load is not quite so unbalanced. There is a strong negative current, almost equal on the one of the -7V legs which should balance thing up. The shunts can be adjusted
to make up the difference. I don't care if a bit of current gets pissed away doing this.

But to answer the question I can do what ever I want.
I can separate the power rails but I can't separate the grounds because of how the currents resolve. Everything is cascaded digital to analog. The best thing to do is keep the current localized and short looped. I'm using compression terminals on the board. I can add extra contacts for the proto board. This way I can ties the legs together or power them separately. All power is routed to the same entry on the board. I don't want to get too crazy with thing unless the need arises. The only part of the digital that can be separate is the relays and drivers and micro controller to a point. The Micro is on it's own board anyway but it also has to connect to the tuning dacs which have only one ground pin. But like I said the tuning is only active when there is a change.

This is tricky business.

It doesn't have to be tracking if that's easier to do. Ground current will only balance on average anyway.

Cheers,
 
Last edited:
Here is a first pass at it. Making the cap on the TL431 larger really slows down the startup. I did some tweaking but still room for more. The opamp can be pretty much any generic if the tracking doesn't need to be perfect.

I will "noodle" on an alternative solution that may work better. can you run the unbalanced loads from a different supply (Relays, ADC/DAC etc)? That will allow for almost zero ground current.

I like the addition of the 10k to reference and smaller C2. It's here the noise gets amplified.
 
Member
Joined 2004
Paid Member
It also reduces a parallel path for noise. Remove the cap across R1 and change the cap on the TL431 to 1 uF. In the simulation the turn on transient is much better behaved.

I substituted some larger transistors and increased the load current here. I think it needs more analysis before going live. I can add current limiting with some simple parts I think.

If the unbalanced currents are small I think an active center tap and a floating supply will be lower noise from external stuff and easier to make work well.
 

Attachments

  • Split low noise tracking supply-1A.asc
    5.8 KB · Views: 57
It also reduces a parallel path for noise. Remove the cap across R1 and change the cap on the TL431 to 1 uF. In the simulation the turn on transient is much better behaved.

I substituted some larger transistors and increased the load current here. I think it needs more analysis before going live. I can add current limiting with some simple parts I think.

If the unbalanced currents are small I think an active center tap and a floating supply will be lower noise from external stuff and easier to make work well.


Don't worry about current limiting. I'm going to pre regulate with 317/337 connected in a tracking manner as in the sim posted. The 317 keeps the voltage across the pass transistor constant at just few volts above. This lowers the dissipation in the pass transistor and dumps current into the output. In sim I input a 5Vpp ripple and the two regulators in tandem removed it completely.

The 317/337 provide current limiting as well as thermal shutdown without raising the noise. So no need for this.
Together they provide better line regulation.

It needs some work on the negative side noise profile.
 
Last edited:
Member
Joined 2004
Paid Member
Too many active parts. That extra transistor could cost $0.10.

Actually tis not bad but the lower TL431 needs a resistor to ground from its V+ to work properly I think. And the balance will be questionable on startup, because of the bias/reference requirements of the TL431.

Here is a shunt regulator based on the same core-
 

Attachments

  • Shunt Regulator.asc
    2.1 KB · Views: 42
  • Shunt Reg.PNG
    Shunt Reg.PNG
    16.1 KB · Views: 271
Too many active parts. That extra transistor could cost $0.10.

Actually tis not bad but the lower TL431 needs a resistor to ground from its V+ to work properly I think. And the balance will be questionable on startup, because of the bias/reference requirements of the TL431.

Here is a shunt regulator based on the same core-

"Too many active parts. That extra transistor could cost $0.10."

LOL.

But your are right it is to noisy. Too much fuss.
You're also right about the resistor to ground. Somehow it disappeared between edits and copying the screen.

I've seen your shunt version before.
 
Last edited:
Disabled Account
Joined 2012
One of the interests i have since playing with a range of oscillators and distortion analyzers is to compare their results starting at -100dBr and below. Check for harmonic level accuracy and consistancy.

My Audio-Precision 2722 just showed up and I have to order some special interface cables to be compatible with the my new PC i7/WIN8 and software etc to get it going. I'll have some surplus distortion equipment to unload now.

Thx-RNMarsh