Phonoclone 3

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

rjm

Member
Joined 2004
Paid Member
Hello everyone. I guess it's to cold weather that does it, but at long last I've spared a little time to review the Phonoclone design.

I've always felt that the weakest link was the voltage regulators, and this is where most of the new stuff is. The phono circuit is unchanged.

Regarding the voltage regulation, to make any improvement one has to first identify what's wrong with the old version. With the standard 3-terminal fixed and adjustable types, I can spot two main areas right from the data sheets, confirmed by measurements:

1. They are noisy. Output noise is measured in mV p-p.
2. They are slow, the bandwidth under which they do their thing is about 2 kHz. Above that they stop being regulators, and let the noise through, too.

As far as their primary functions of removing ripple voltage and keeping the output voltage constant, however, the work just fine. Furthermore, the output capacitor nicely covers for the regulator at higher frequency. However, we have to start from somewhere, so we'll accept that a better regulator will have wider bandwidth and lower noise.

It's been rattling around my head for a while now, and actually I'm pretty sure I floated the idea here some time ago, but I'm actually going to go ahead and try it out this time: namely, try using the same OP27s used in the phonoclone circuit to also provide a low impedance, low noise power source for that circuit.

Since playing around in Eagle is easier than actually soldering stuff on the bench, I've gone ahead and designed the new board. Full size 16x10 cm, and double sided with a full ground plane, just like the old BE boards. The next step is to prototype the regulation unit and measure it to make sure is does actually work as advertised. After that I'll finalize the board and order it from Olimex.

If anyone wants to join me at the guinea pig stage, you are welcome to drop me an email. I'll order extra boards and send them out to you, at cost.

Anyhow, without further ado, here's what the new beasties look like:

both layers shown
An externally hosted image should be here but it was not working when we last tested it.


bottom layer (signal trace)
An externally hosted image should be here but it was not working when we last tested it.


Eagle brd and sch files attached.

regards,

/rjm
 

Attachments

  • phonoclone 30c.zip
    44.3 KB · Views: 1,281

rjm

Member
Joined 2004
Paid Member
The output voltage of the regulator section, Vout, is determined by the resistors R9-R16 and the input voltage Vin.

Vout=Vin * R9/(R9+R11) * R15/R13

Assuming R9 = R13 and R9<<R11

Vout=Vin R15/R11

To adjust the ratio, change R11.

To preempt and answer the inevitable in advance: Yes, you can use the Jung Superreg instead, or any number of other "high end" voltage regulator circuits. I'm not ... because... [Singing] I (wanna say I) did it .... MY WAY. :)
 

rjm

Member
Joined 2004
Paid Member
That would be "C" : excessive noise.

If you have a look at the circuit attached, which is pretty much as I bench-tested it, you'll see the op-amp is run from a single-sided supply. The negative rail is connected to the circuit common... the source of my woes, as it turned out.

All the ripple on V_in feeds into the opamp positive power pin (7). I originally estimated the PSRR figure of ~85 dB from the datasheet PSRR vs frequency plot. This is even better than the attenuation of the RC filter of the voltage divider network R3/R4 so the contribution at the output would, I reasoned, be inconsequential.

No. Connected to a single supply, the PSRR would seem to be less than 20 dB, certainly not anything like 85 dB. For 100 mV ripple on V_in about 10 mV will appear at V_out. Ouch.


Opamps make good voltage amplifiers but poor regulators. When used in a traditional series regulator with reference and pass transistor, its not stated explicitly but the voltage amplifier itself is powered by a separate, heavily filtered or regulated line. Or there's that trick Walt did and have it powered by the output voltage of the regulator. Any which way, it has to be a clean voltage.

So, what options? One is to backtrack and connect the opamp to a split supply, by bringing the V- rail up to the opamp pin (4). This should bring the configuraiton back to teh default datasheet PSRR.

Fine as long as the total voltage doesn't exceed 44V.

Another possibility is to switch tactics and go into a simple shunt regulator mode, either a zener, zener + pass transistor, or something like the TL431.

Finally, we could pre-regulate the input voltage. Either a LM317, or a zener. I don't much care for this, since if you need regulation for the opamp input, you might was well get rid of the opamp entirely and feed into the phonoclone circuit directly, as it hardly seems to be doing anything worthwhile now.

So, split supply connection for the OP27s, or move to zener-style shunt configuration? What should it be?
 

Attachments

  • supply temp.png
    supply temp.png
    6.6 KB · Views: 7,815
rjm said:

No. Connected to a single supply, the PSRR would seem to be less than 20 dB, certainly not anything like 85 dB.

Nope.

The opamp doesn't know its lower supply is at ground level. Opamps are ignorant of ground.

Check your reference voltage generation (or rather the lack
thereof). That's how the ripple gets into the output.

You may want to replace R4 (in the latest schematic) with a
zener or better still LM329, and then adapt R3 for proper biasing. Even so a zener or LM329 has insufficient dynamic impedance to give good filtering with a shunt cap, so try to squeeze a resistor between the reference node and the opamp+/filter cap node.

That's how it is done ;-)

Even so I would expect the OP27 to be unhappy during power-on and during operation as its naked output sees the cap load and has to deliver serious current. Opamps don't like that.
 

rjm

Member
Joined 2004
Paid Member
Werner said:

The opamp doesn't know its lower supply is at ground level. Opamps are ignorant of ground.

Check your reference voltage generation (or rather the lack
thereof). That's how the ripple gets into the output.

Thanks. I was just coming back to the forum to post that after doing some research on the subject this afternoon. Stupid mistake on my part, shortcutting the full calculation the R3/R4/C2 divider. In a nutshell: it's not nearly as good as filtering the 120 Hz ripple as I thought it was.

Back in a bit with the next iteration...
 
I'm afraid that one is not even going to work: the opamp's output can't get above the reg's output, but yet it has to reach the input voltage minus Vbe.

Just feed the opamp from the raw input, possibly with a bit of filtering.

And there's still not a lot of noise filter on that reference diode!

This is something I made for a phonostage:

An externally hosted image should be here but it was not working when we last tested it.


More complex than what you want, but then it performs a little trick in that the output impedance is resistive to beyond 20kHz.

Mind, this was NOT made for the lowest ripple rejection: it's dangling off a pair of emitter followers.
 

rjm

Member
Joined 2004
Paid Member
I'm not going to pretend I follow all of that, but I do see what you are doing with the filtering of the reference voltage.

One thing I think I could do, going back to my original concept, is just add more filtering to the voltage divider. The circuit below should manage about -90 dB or better (really this time I calculated it properly!) at 120 Hz., from both the reference voltage and opamp.

The time constant of the divider should limit the inrush currents through the op amps.

I've got it ready on the bench for testing later today...
 

Attachments

  • rjmreg1.png
    rjmreg1.png
    7.1 KB · Views: 4,005

rjm

Member
Joined 2004
Paid Member
Alright - here's the situation:

Two configurations shown attached. The one on the left, where the op amp is connected to split supply, works exactly like I expected. With 12 V rails, R3 and R4 produce an input voltage of 0.55 V, with ripple below 1 mV (my measurement threshold) due to the capacitor C2. The op amp takes that voltage and amplifies it by R2/R1+1 = 11 for an output voltage of 6.1V.

The ripple on the voltage rails is 1.8 V p-p positive and 0.2 V p-p negative, for a 470 ohm load (13mA output current drawn from the positive rail). The ripple on the output is less than 1 mV, estimated at less than 0.1 mV. Nothing but a bit of HF noise on my scope.

The ripple rejection of this circuit is therefore about 80 dB, certainly in line with what I was expecting.

The circuit on the right, meanwhile, is configured as a single supply. It does not give the same performance. The DC voltages are right: the V+ is now 24 V, the input voltage 1.1 V and the output 12.7V. The ripple on the V+ is again 1.8 V p-p, but now the ripple on the output is 400 mV p-p!

I don't understand why the ripple is so high.
 

Attachments

  • ps-png-sm.png
    ps-png-sm.png
    35.1 KB · Views: 3,972

rjm

Member
Joined 2004
Paid Member
Awesome! Yep, that does the trick. Even added a pass transistor to the mix.

(test point: V++ 24V, input 7V, output 14V)

Though with the lower gain, the op amp seems a little less stable. Needed a capacitor across the test load to settle it.

I have about 1mV p-p of ugly looking diode switching noise that I can't seem to shake, otherwise we really made some progress today.

/rjm
 

rjm

Member
Joined 2004
Paid Member
Here's what the breadboarded test circuit looks like.

I added an additional RC filter to supply the op amp and voltage divider.

The NPN transistor I used is just what happened to be in the drawer.

As shown, the output voltage will be about 3/5th of the input.

In my test configuration, 24 V in, 15 V out. Ripple is 1.2 V p-p at the input, output noise is less than 0.2 mV, for a 470 ohm + 100 uF load.

The configuration is unstable without the 100 uF cap on the load.

There is some residual airborne pickup, if I step away from the desk, it gets worse. The whole circuit is unshielded, however, just sitting on the desk, with long lead wires posted into the breadboard. I'm not sure if I should bother looking into it or not.

Also I have the feeling that the bandwidth of 4 Mhz is a tad high. I'd rather reduce it to 200 kHz to avoid, or at least minimize, stability problems.
 

Attachments

  • pass.png
    pass.png
    8.9 KB · Views: 3,863

rjm

Member
Joined 2004
Paid Member
- a pass transistor doesn't just let more current pass to the load, it also removes the output current from the voltage amplifier entirely. The main benefit is efficient RC filtering can be now be applied to the supply feeding the voltage amplifier.

- having a split supply to begin with, there are some advantages to running the voltage amplifiers from V+ to V-, rather than connecting one side to GND. First, you can have a small input voltage. A small input voltage means you use a high gain in the voltage amplifier to generate the output voltage ... increasing the gain is a good way to improve the stability. Second, the voltage amplifier can be configured inverting. The reference generated from the negative rail can be fed to the input of the op amp that provides the positive output voltage and vice versa. Inverting amplifiers are much less likely to cause problems, in my experience.

I tried it on the breadboard, it worked great. It was stable without an output capacitor, though with some HF noise. Adding 100uF brought the noise down to <0.2 mV.

The latest configuration, then, implements these latest ideas.
 

Attachments

  • pc30h.png
    pc30h.png
    22.6 KB · Views: 3,790
Looking more and more like mine now ;)

Multiloop feedback can also be used to burn off
excess OL gain, at the expense of a super-low
output impedance. Which you IMO don't really
want anyway.

A series resistor between opamp and pass transistor
base promotes stability.

Running the opamp at high gain also amplifies its
own noise into the reg's output.
 

rjm

Member
Joined 2004
Paid Member
Looking more and more like mine now ;)

Please consider it a complement! :)

A series resistor between opamp and pass transistor
base promotes stability.

Isolate the op amp from the capacitance of the transitor base, right? I was wondering about that... I'll give it a try on the bench to see if it makes a difference.

Running the opamp at high gain also amplifies its
own noise into the reg's output.

This I knew, and actually you can see the effect of it on the scope trace. Still it seems for the present like a reasonable trade off for getting a sensible bandwidth.


@d to the g

You'll have to download and install Eagle www.cadsoftusa.com to view and manipulate the files. You may need the latest version, 5.3, I'm not sure.


I'm going to christen the circuit (inverting voltage amp + split supplies + crossed reference voltage + split output) the "Xreg". 'cause it sounds cool. A bit of dotting i's and crossing T's on the weekend and we will be about ready to proceed to the beta test stage.
 

Attachments

  • r1026068s.jpg
    r1026068s.jpg
    90.7 KB · Views: 3,766
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.