New Regulator

[Bemused]

Either an intermittent board problem or my scope has improved from having an hours use, it is very old and has sat idle for a year.
Needs burning in and perhaps a better mains lead

This morning I removed all output bypass capacitors and got a good scope lock where previously I had wideband noise.
Scope shots with no output bypass capacitors.


Q6 emitter
AC trace, vertical 0.1V / division, horizontal 0.05uS / division.
Frequency 25000kHz

AD797 output pin 6
AC trace, vertical 0.1V / division, horizontal 0.05uS / division.
Frequency 25000kHz

Q6 base
AC trace, vertical 0.1V / division, horizontal 0.05uS / division.
Frequency 25000kHz

Around this point I felt stupid sticking film capacitors on the end of a super regulator I’ve always used Tantalum so I refitted the 22uF Tantalum, ESR perhaps too low at 300mΩ
Scope shot below. Fixed ???
AC trace, vertical 5mV / division, horizontal 0.05uS / division.
Frequency 100MHz

The 100000 kHz 2mV noise is my scopes background noise with the probe shorted so its as close to flat line as my scope gets.

All shots were into 55mA load at 3v37.
Next I rechecked the previously strange DC operating points.

R9 (2R2) AD797 feed line series resistor with 15v7 supply. (2 x LED in the 7812 reference leg)
Volt drop over R9 is 18.6mV so the AD797 is taking 8.45mA.

R12 (18R) AD797 PIN 6 output line series resistor.
Volt drop over R12 is 15.8mV so the AD797 is delivering 0.87mA to the load, slightly in class A.

All looking reasonable other than for the fact that it intermittently craps out and starts oscillating. It can be well behaved for anything from 30 seconds to over 1 hour.

 

[Bemused]

I may have Schrodinger's cat syndrome!

I have been sitting the scope probe in the vacant pad for D18 as a convenient method of holding the probe aon pin 6.
All seems fine, scope is flat line, Then I measure the voltage across the adjacent R12 AD797 pin 6 resistor. Under good conditions I have 18mV, during crap out I have just a few mV and the scope shows oscillation.
I have now found oscillation can start by just touching one multimeter lead to R12 or even opamp pin 1.

So I looks good and leave it alone and dont touch a multimeter probe down.
Most times I attempt to measure it craps out.

I have upped the load to actual DAC load of 85mA and will observe the rest of the day without touching a multimeter probe on.

 

[RyanSoundLab]

Tony,

Even the "fixed?" scope trace shows oscillation so it's not fixed, unfortunately. Yes, there is a similarity between electrons and Schrodinger's cat but I wouldn't wait all day for a miracle. Also, a circuit that is so sensitive that it will break into oscillation (and keep oscillating) with little pokes here or there is just not stable enough for production.

Try putting a 47pf cap between pins 2 and 6 and try that while also removing the 47pf cap on pin 8 (e.g. try both with and without C9). You might also try a 47pf between pin 3 (+ input) of the opamp and ground - could be positive feedback possibly coming in through there.

Good luck and keep us posted.

Kit

 

[Bemused]

The 100 MHz 2mV noise in the last shot is my scopes background noise with the probe shorted so its as close to flat line as my scope gets.

Just waiting to prove its stable if not probed, previously it seemed so frustratingly intermittent.
Yes I know its not real sorted but I feel better with it,

 

[Bemused]

I tried 47pF dabbed around to no obvious effect, but should really return and try it more methodically.
For sure 47pF from C12 R14 junction to ground has no effect, I tested that one thoroughly.

Now R14 is a new visitor to the board which made me think a little. I have had series R in the sense line way back when, to damp an oscillation but only around 10 to 40 ohms worth and was on the physical load termination. (Capacitance in the sense wire etc.) I just tacked a duplicate resistor on top of R14 to reduce it down to 50R and things seem better. It is more reluctant to provocation, oscillation less frequent.
Not got much in the way of SMD flavours in stock but I have an 18R value I could try.

What I did discover was shorting R14 allows any Oscillation to recover, so its the new visitor and is shorting it returns the board to stability. Looks like a bad boy to me.

I added it on Martins suggestion but did I implement it as he intended?
Possibly I goofed

Post 367 IMO the local-loop / C8 trick works better if you define quite clearly the cutoff at which local feedback takes over from the larger loop. You want to keep this above the audio band.

I suggest a resistor in the FB+ connection, something like 100ohms, and then C8 can be say 10nF to define the local loop above about 150Khz. Increase C8 if that isn't enough. (bet it will though - 22nF brings it down to 72khz or so, which is also a good compromise. And these small cap values allow compact high-quality feedback caps (cough, polystyrene or COG). Exactly the approach to what I use in my preamp.

Incidentally this is how Deltec kept the DPA50S power amps stable despite four-wire remote sensing from the speaker terminals - with no output inductor. 1K in the feedback leads and a local cap like c8 inside the amp to limit the feedback loop out side the amp to below ~160Khz. It's exactly like the classic technique for compensating an opamp for capacitative loads.

 

[Bemused]

R14 went 100R, 50R, 18R. Reluctance to oscillate by probing increased each time.
Just fitted 2R2 part and been unable to provoke the oscillation.

Also previously an output short circuit left an odd little and widely spaced bump in the output that would not recover. Now output shorts recover as soon as the short is removed.

Now I’m pretty sure I goofed with Martins implementation of R14 cause R14 looks like the naughty part, I mostly goof and how to say this... Well Martin doesn't.

 

[RyanSoundLab]

Tony,

Just tried to duplicate your circuit. Here's the portion of your circuit I put together on a plug-in breadboard (and yes, I actually have an AD797 and an 2SD669A to use!):

Sure enough, without C2 (100pf feedback cap) it oscillates nicely at about 5MHz with a 50ma load into R2 and 10uf tantalum output cap. Removing the output cap doesn't help - just increases the frequency of oscillation. Neither does changing it to an aluminum electrolytic with a larger capacitance. Changing R1 to various values higher or lower also did nothing to help (tried zero to 500 ohms).

When C2 is added, the circuit instantly settles down and no poking or prodding can make it oscillate. Not sure why this didn't help your circuit because it usually works well. Perhaps I missed some important aspect of your design. For example, with the dual 797 pods I have, I can't get to pin 8 (decompensation pin) to see if that capacitor is part of the problem.

It would help if I could duplicate your exact problem (don't you just hate it when you go to the garage mechanic who says - "well, it didn't make that noise when I drove it"!) but it's close.

Kit

 

[Bemused]

Wow Kit you did all that to help me!
Your most kind for sure.

Ill study more closely tomorrow.

Excluding the de-compensation capacitor there are subtle differences between your circuit and mine though.

My last board has been working in circuit for a year, it had some bug wired modifications and was through hole.
The last board had two issues.
Issue 1
Instability in which you pointed me to a data sheet which gave me my C12 and solved that problem.
Issue 2
Output over shoot on start-up.
I solved that one by accelerating the charging of the gyrator with an extra transistor.
The new board just incorporates the bug wiring modifications, change to mostly SMD components and Martins R14.

Effectively removing R14 by changing from 100R to 2R2 looks to have restored performance.

Regarding your additional capacitor. I had moderate stability, it did not oscillate every time I probed in fact it took me quite a while to realise probing was the trigger. On top of that I was short of hands, scope and multimeter and dabbing a capacitor on was all a bit fraught. I only soldered "your capacitor" in one of the few places you suggested, in that position I did some thorough testing and it didn't help the problem. In other positions I just dabbed it on and testing was not satisfactory having only two hands.

I had intended to return to "your capacitor" in other positions but got this R14 thing in my head and performance seemed to improve every time I reduced the value of the previously absent R14.

In my mind R14 is in the wrong circuit, it should be in series with C12 !

 

[Bemused]

Sure enough, without C2 (100pf feedback cap) it oscillates nicely at about 5MHz with a 50ma load into R2 and 10uf tantalum output cap. Removing the output cap doesn't help - just increases the frequency of oscillation. Neither does changing it to an aluminum electrolytic with a larger capacitance. Changing R1 to various values higher or lower also did nothing to help (tried zero to 500 ohms).

I use your R1 mostly as a convenient method to measure the AD797 output current. Mostly it could be omitted I think.
Others uses are to jack-up the pin 6 voltage when running at voltages much lower than 3v, but in those cases I usually omit the resistor and fit the parallel LED.

Just noting the resistor I have changed is not your R1 position!

When C2 is added, the circuit instantly settles down and no poking or prodding can make it oscillate. Not sure why this didn't help your circuit because it usually works well. Perhaps I missed some important aspect of your design.

Nope I just did not test thoroughly with your C2 on that position.
It was something I was to get back to but hit on my R14 fix.
Just only so many hours for me to bodge and still keep the wife happy.

Tony

 

[RyanSoundLab]

Tony,

Right - I got R14 and R12 mixed up in your circuit. If that fixes the problem for sure then go for it. I can see how the R14/C12 and possibly C9 could introduce poles into the response that would add to the instability, so getting rid of R14 makes sense.

No problem in trying out your circuit over here. There's no substitute for the actual circuit in the actual circuit board. Sometimes circuits that are OK on the breadboard go haywire on the board (and vice versa). Glad to try to be of service. It was fun.

Kit

 

[Bemused]

Thanks Kit,
I have space to include your capacitor so I may add it in as a get out off jail card.

Could just use Martin clarifing what I got wrong with his take on R14.

Regards
Tony

 

[Bemused]

Right nice and stable on the bench but oscillating when fitted to the DAC.
So I did Kits mod of adding a feedback capacitor between op-amp inverting input and output. I used a 47pF COG through hole bug wired over the chip.

Nice and stable now, all DC operating current and voltage points are as exactly as expected. Not scoped it in the DAC yet but I have never had an oscillation without bad DC operating points somewhere so I have a high confidence level.

New board required again! need to add Kits feedback capacitor, fix a screen print error on the back and sort Martins R14 problem.

Regard
Tony

 

[Bemused]

Now thinking about increasing current capacity.

Capacity is always going to be principally limited by the dissipation across the output transistor, it drops 15 volts in this application which is about the thermal limit.

Wondering on the feasibility of paralleling a pair of output transistors for increased current handling, is this prone to problems?

Other options are maybe an additional series output transistor, briefly considered driving both bases by a potential divider but that seems a bad idea so then considered a separate series transistor with fixed biasing, the biasing could be set for an application specific volt drop.

Probably acceptable to loose a little performance with this option just to extend the application envelope.
The old 5V at 1A is at the back of my mind here!

Tony

 

[Bemused]

The wife says the mood of this house largely depends on the state of the regulator.
Christmases holiday has ended on a
Good regulator day

 

[RyanSoundLab]

Tony,

Glad the capacitor worked and saved your marriage (LOL).

Regarding more capacity - I don't recommend paralleling outputs. I found that it just increases stability problems, plus because you have to use emitter resistors to balance the load between output transistors, your output impedance will necessarily go up (but maybe not all that much). There are a lot of alternatives, though, to what you have now which I think is 85ma (1.28w) output over 15v using a small heat sink (looking at a previous post). Here are some ideas:

The spec sheet says you can get 20w out of the 2SD669 if the case is at 25C, thus it has about a 6C/w thermal impedance from junction to case. Even the small heat sink you have now will give you about 20C/w so that you can get 125C (rise)/(20+6)= 5w, or about 300ma output with a 15v drop. Yes, it will get hot to the touch! If you use a chunky heat sink with a 6C/w rating, you can get 10w or 600ma out of it.

A possibly simpler approach is to replace the 2SD669 with a TIP41C (a TO-220 size transistor with the same pin spacing (but reversed leads so just flip the transistor around 180 degrees) as the 2SD669). The TIP will give you 2w dissipation without any heat sink, or about 2w/15v = 130ma (and no board mods needed for it). Of course a TIP with even a small heat sink will give you that much more output. So it just depends on how much more output you want.

If you prefer to use higher frequency TO-220 transistors you might try the 2SC2527. It has a similar power rating to the TIP.

Best regards,

Kit

 

[Bemused]

Great Kit, thanks for carefully explaining my options, nice to know.
We like options

Dropped the scope on briefly with the regulator in the DAC, a little 1MHz noise appeared on the output nothing I'm worried about and sure it can be removed with more attention to your feedback cap. Mine is 47pF with 2 inch legs! Just grasping the cap with my fingers reduces the oscillation by 75%.

Spent the rest of today listening instead of bodging and its sounding rather sublime, best performance yet so I'm more than happy.

Tony

 

[martin clark]

Only just catching-up with this: R14 value does look like my goof

I think the issue with such large values is the effect on phase lag for the decompensation. c9 at 47pf+100 ohms will put a 'zero' c. 35MHz meaning oscillation above v likely ... does that match with the vhf weirdness?

If it's working better without R14 , short it out! checked the freehanded regs in my cd player, used 10 ohms didn't I. Doh. Sorry Tony!

Larger values will work in conjunction with a 'C12' and can be helpful when not also using c9, i.e a different/slower opamp... I've successfully used 100r + 1nF with AD825-based regs similar to this using such remote-sensing (last 10yrs in my preamp.)

 

[Bemused]

Thanks Martin, R14 actually gave me a little fun

Looking at the scope shots I saved oscillation was at 25MHz, not sure what other frequencies I saw.

Possibly best leaving the R14 pads in circuit for any future variations then and building with 0R0 link or do you think best removed.
If R14 was required wouldn’t it be better remote from the board at the load end of the sense wire?

Is there any mileage in adding pads for further series R actually in series with C12 or in Kits new feedback capacitor pin 2 to pin 6.

Tony

[/IMG]

 

[Bemused]

Current schematic marked up with actual components.

 

[Bemused]

Further bodging has led to a slight improvement but still not good.
I defaulted some values and started again.
As usual it is good on the Bench but poor in the DAC.
Just to note adjacent in the DAC are raspberry pi and Arduino.

R14 progressed from 100R down to 0R0 and improved along the way, not improved since dropping below 2R2

C12 was 22pF and increased to 100nF then 1uF, possibly slightly better around 100nF

I then added Kits op-amp feed back capacitor from pin 2 to pin 6.
Started with 47pF and gained some improvement.
Changed to 100pF and best yet, still have noise but sounds very good in the DAC


I need to get C12 back to 100nF
Also wondering if my general decoupling values are goofed.
LM7812 capacitors are 0.33uF and 0.1uF as per data sheet.


Best scope picture below
Scales 5mV and 1uS per division.
R14 0R0, C12 100nF, Ckit 100pF
20mV noise at around 0.625 MHz.
I can see some 5 MHz on top, the scope has a background noise of 100 MHz 2mV

Picture of not so good measurement and implementation.