Good opamp for that1646

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Any linear pass regulator will have similar heat dissipation. The high tech remedy for excess dissipation is a switching supply.

Its been a while since I talked about this but the Loftech TS-1 power supply I designed back in the 80s used a trick approach where a bipolar transistor switch disconnected the transformer winding from the rectifier after charging up the lower unregulated rail to its required voltage, while still fully charging up the higher rail. In that case I needed to extract a high current 5V rail for digital logic and LED displays, along with modest current rails at +/- 15v. This approach is also too complex for your needs. I didn't design this to reduce heat dissipation but to get the needed voltages from an existing transformer. The energy wasted as heat from driving the 5V regulator directly caused the transformer to sag and not deliver my desired higher +/- 15V rails.
====
Series dropping resistors can scrub off some of the excess voltage inexpensively, without too much complexity. KISS

JR
 
So, I breadboarded the whole thing and at first glance it seems to be working OK butI noticed that I get a difference of around 2.5mV between OUT+ and OUT-.

In fact, I'm reading the following voltages: OUT+ = -0.5mV and OUT- = +3mV.

Is this normal? The +/- voltages powering the 1646 also have around 2.5mV difference so this might be the reason(?) As I mentioned before I'm using resistors to drop the voltage going into the 1646. Should these be matched?

Also, shouldn't I get +mV at OUT+ and -mV at OUT- ?
DC offset has always been a problem with the cross couples circuits. A friend designed a very cool version, where the + out was direct and only the - out caused the increase in gain, he did that because 90% of the time you would only short the - out, and it solved CMMR and the DC offset.

I gave up on using them, as they don't sound that great. I just use an inverting amp for the - and supply a note about lifting pin 2. Through the 30 consoles I built with Tonelux and the 9 with FIX, I've never had a call or email about the loss of gain... or distortion.
 
I haven't thought about it. It was just a quick thought of using diodes instead of regulators.
Shunt regulators always result in more dissipated power. With zeners regulation is obtained by adding current to the load. With a series regulator, regulation is obtained by decreasing current.
I might use LM317/337 and call it a day but still I would like to know what would be the real benefits in using regulators. What improvements should I expect? Better noise figures?
Very likely. Zeners and noisy regulators (78xx and 79xx are noisy) require very careful layout in order to confine their noise currents to a loop that does not impact the ground cleanness.
Better thd?
Possible, since the rails are stiffer with good regulators.
Better CMR?
No.
Better everything?
Better some things.
Another concern I have is that the regulators will have to work hard to drop the voltage from 28VDC to 15VDC (i.e heat, big heatshinks, etc..)
They will have to drop 13V, which they can easily do. One needs to make sure the max dissipation is not exceeded. With the TO220 package, it's about 4W; that would be 300mA. You see there's ample margin. You could probably use the low-power type in TO92 package.
Anyway, this voltage must be dropped somewhere. The suggestion of dropping some of it in additional resistors is not pertinent here, because the regulators can handle the power and regulation is better without resistors.
With zener regulation, you have to dissipate power in resistors and in the zeners. It's always more heat.
 
Surprisingly the LM317/337 circuit seem to have the highest noise.
Three-pin regs are known to be sensitive to layout. In the absence of any info regarding this, I am tempted to suspect the results' validity.
I would have liked to see how the 78xx/79xx fared.
It is true that the simple zener/transistor approach can result in very good noise performance, at the detriment of other aspects, that may not matter so much (regulation, output Z).
There are several VLN (generally discrete) regulators available, e.g. Hypex HNR/HPR12. They are often used as a replacement for 78/79xx or LM317/337.
Turns out they generally provide improvement only when the originals were improperly implemented. Many PCB designers do not understand (or care about) ground current circulation.
 
Three-pin regs are known to be sensitive to layout. In the absence of any info regarding this, I am tempted to suspect the results' validity.
I would have liked to see how the 78xx/79xx fared.
It is true that the simple zener/transistor approach can result in very good noise performance, at the detriment of other aspects, that may not matter so much (regulation, output Z).
There are several VLN (generally discrete) regulators available, e.g. Hypex HNR/HPR12. They are often used as a replacement for 78/79xx or LM317/337.
Turns out they generally provide improvement only when the originals were improperly implemented. Many PCB designers do not understand (or care about) ground current circulation.

Ok that makes sense. FWIW, the article continues with some output impedance and ripple noise measurements under the "Crash course on voltage regulators" section.

As far as the 317/337 are concerned, since the input is DC am I right to believe that the two diodes (from "adj-to-Vo" and from "Vo-to-Vi") are not needed?
 
Ok that makes sense. FWIW, the article continues with some output impedance and ripple noise measurements under the "Crash course on voltage regulators" section.
OK. I'll give it a glance.
EDIT: Done! I can see the guy is quite good; still he admits his methodology was not perfect. We all do mistakes...
I still regret he didn't do any 78/79xx.
As far as the 317/337 are concerned, since the input is DC am I right to believe that the two diodes (from "adj-to-Vo" and from "Vo-to-Vi") are not needed?
These diodes are not intended for rectification; they are there to protect the regulators against fault situations (typically short circuits either at the input or output). It is briefly expalined in the datasheet.
It's important if the output uses a connector with associated risks of misconnection, but in most applications they are not strictly necessary.
 
Last edited:
As far as the 317/337 are concerned, since the input is DC am I right to believe that the two diodes (from "adj-to-Vo" and from "Vo-to-Vi") are not needed?
Absolutely not. You have to feed it DC although it may be 'rough DC'. The diodes are there for protection (nothing to do with rectifying AC). This is clearly stated in most LM317 etc datasheets.
See 9.2.1 in the link below
https://www.ti.com/lit/ds/symlink/lm317.pdf

Just saw Abbey's post above.
 
Does it look OK?

The datasheet calls for 1N4002. Why not 1N4007?

Should I increase C15 to 100uF?

Also, is the input and output filtering an overkill?

BTW, I really appreciate all your help.
 

Attachments

  • 317.jpg
    317.jpg
    41.9 KB · Views: 16
Last edited:
Should I increase C15 to 100uF?
There's no absolutely correct value for that really. If laying out on a pcb I'd tend to allow for up to as large a physical cap as practicable up to, say, 470uF and have multiple footprints to allow for different options. For stripboard type prototyping I'd stick 47u of 100u in there depending what's to hand.
Consider your 28V source. How clean (or not) is it ? Will it be happy charging up a larger cap' on power up ?
I always have 'C19' on the output placed right by the output pin as direct to 0V copper area / plane. Preferably SMT chip X7R.
Probably fine without it but I've done it multiple times and it's my 'goto'.
Doesn't replace the similar cap at the opamp etc.
Also, is the input and output filtering an overkill?
It's along the lines of what I would do. Really it depends on what there is to filter. Same as anything - signal input rfi filtering etc.
 
There's no absolutely correct value for that really. If laying out on a pcb I'd tend to allow for up to as large a physical cap as practicable up to, say, 470uF and have multiple footprints to allow for different options. For stripboard type prototyping I'd stick 47u of 100u in there depending what's to hand.
Consider your 28V source. How clean (or not) is it ? Will it be happy charging up a larger cap' on power up ?
I always have 'C19' on the output placed right by the output pin as direct to 0V copper area / plane. Preferably SMT chip X7R.
Probably fine without it but I've done it multiple times and it's my 'goto'.
Doesn't replace the similar cap at the opamp etc.

It's along the lines of what I would do. Really it depends on what there is to filter. Same as anything - signal input rfi filtering etc.

Thank you Newmarket. Unfortunately space is limited so I don't have the luxury for extra (and possible unused) footprints.

BTW, I think I'm entering the darkweb side of PSUs, haha

http://tech.juaneda.com/en/articles/lm317.html
 
DC offset has always been a problem with the cross couples circuits. A friend designed a very cool version, where the + out was direct and only the - out caused the increase in gain, he did that because 90% of the time you would only short the - out, and it solved CMMR and the DC offset.
I meant to ask - what does "it solved CMMR" mean in this context ? (Not asking what CMMR means - assuming meaning CMRR)
 
How important is for both the 1646 and the 5534 to have identical (+) and (-) supply voltages? Obviously with simple resistors on the regs I can't match exactly the positive and negative rail. I measure +16.8V and -16.6V. Perhaps using trimmers on the regulators is a better idea? However, I'd rather stick to simple resistors is it's not a big deal.
 
How important is for both the 1646 and the 5534 to have identical (+) and (-) supply voltages? Obviously with simple resistors on the regs I can't match exactly the positive and negative rail. I measure +16.8V and -16.6V. Perhaps using trimmers on the regulators is a better idea? However, I'd rather stick to simple resistors is it's not a big deal.
I already gave you an answer on that point. You may want to experiment, and find out that PSU rails unbalance has almost nil effect on offset, which is essentially due to input devices and associated resistors matching.
 
Back
Top