Driving step-up transformer directly from opamp?

Hoping somebody can help clear this up for me.  On an output stage, assuming I'm using either a 5534 or OPA604, and going into a 1:2 step up transformer, and assuming I'll be going into a 600R load (say a pultec, etc), is it ok to drive the transformer directly off of the opamp or will the current demand kill it?  If this is ok, then when is it not ok?  Does the step up ratio affect impedance seen at the primary (opamp side) of the transformer?  I would expect that sure it does, but I'm not really clear there.

I see alot of designs where they're doing just this (like API312, etc), running it directly off a 5534 or DOA, but then consider the LA4 output, why do they have a class AB amplifier there to drive a 1:1 output transformer?  LA4 uses TL074 though, so don't know if that might have something to do with it.

ruckus328 said:

Does the step up ratio affect impedance seen at the primary (opamp side) of the transformer?  I would expect that sure it does, but I'm not really clear there.

Click to expand...

That is the main purpose of a transformer, impedance matching.  Sometimes the "free gain" is a valuable bonus but the original reason to use transformers was impedance matching.  

Better minds than I will be along shorty but the simple version is that the impedance ratio is the square of the turns ratio.  So in this case your turns ratio is 1:2, that means your transformer is "reflecting" impedances with a ratio of 1:4.  

So let's say your opamp stage has an output impedance of 20 ohms for arguments sake (often it's way lower but it will mostly depend on resistors following the opamp) and that the next piece in your chain is the pultec with it's 600 ohm input impedance.  With the 1:4 ratio your output stage is now going to "see" a 150 ohm load rather than and 600 ohm load and your pultec input stage is going to see an 80 ohm source impedance.

150 ohms is too much to ask of a 5532 but an 2520 for example is happy to drive as low a load  as 75 ohms IIRC.

For transformer basics go to the Jensen transformer site and check out some of their  notes - AN002 is a good summary but there are longer articles too with more detailed info.

I should add that the above is a simplified example using a perfect transformer.  There is no such thing of course as winding have their own resistance, capacitance etc that need to be taken into account.  In certain settings these factors are not important, in others they can have quite an effect.

ruckus328 said:

Hoping somebody can help clear this up for me.  On an output stage, assuming I'm using either a 5534 or OPA604, and going into a 1:2 step up transformer, and assuming I'll be going into a 600R load (say a pultec, etc), is it ok to drive the transformer directly off of the opamp or will the current demand kill it?  If this is ok, then when is it not ok?  Does the step up ratio affect impedance seen at the primary (opamp side) of the transformer?  I would expect that sure it does, but I'm not really clear there.

I see alot of designs where they're doing just this (like API312, etc), running it directly off a 5534 or DOA, but then consider the LA4 output, why do they have a class AB amplifier there to drive a 1:1 output transformer?  LA4 uses TL074 though, so don't know if that might have something to do with it.

Click to expand...

A 1:2 step up transformer will change the apparent load impedance by the square of the turns ratio. This means the 600 ohm load after the 1:2 step up will look like 150 ohms reflected back at the opamp. This is low for a 5534, I am not familiar with the specs on the 604. I know there are some new uber opamps that will drive 150 ohms directly, but most older opamps will benefit from transistor buffers to help handle 150 ohms.

JR

And that goes double or quadruple for the TL074, which has a very low-capability output stage.

Driving 600 ohms is not a job for the meek, and driving 150 ohms takes some real juice.

Peace,
Paul

Thanks guys, was my understanding it would see 150R reflected at the primaries but I was unclear what I can get away with, still coming up to par with this stuff.  Looking at the datsheets, both THAT1646 and NE5534 spec they can drive 600R, so that being the case driving a 1:1 transformer with either shouldn't be an issue right?

So I guess either using an A/B transistor output stage to drive a 1:2 or a THAT1646 (or similar) to drive a 1:1 would get me where I need to be, I imagine the transistors would be cheaper, and if using a THAT1646 most of the benefits of using iron are pretty much null at that point anyways, exception being added color of course.

One thing I'm still not clear on though is how to calculate current draw of my output stage if driving 150R?

First, why do you want to drive a 1:2 xfmr?
The only reasons to do that are:
The rails are lower than typical +/- 15-18V, or you want more than +20 dBu output level.

A 5534 powered with +/-17V rails can deliver +21dBu. Is it really not enough?


If you drive a 150R load with +/- 15V peaks (using +/-17V rails), the peak current will be 100mA (15V divided by 150ohms).
In fact the 5534 will current-limit the peaks at ca. 35mA.
But if you put transistors buffers, they will delver these 100mA peaks.

ruckus328 said:

Thanks guys, was my understanding it would see 150R reflected at the primaries but I was unclear what I can get away with, still coming up to par with this stuff.  Looking at the datsheets, both THAT1646 and NE5534 spec they can drive 600R, so that being the case driving a 1:1 transformer with either shouldn't be an issue right?

Click to expand...

Well...

Yes, those chips can drive 600R. How well they drive that load is another story. I don't have information on the THAT chip, but Samuel Groner has done tests on a whole pile of opamps, including the NE5534, and while the 5534 has quite decent performance into higher-Z loads, the midband distortion spectrum gets seriously ugly into 600 ohms. THD at low and mid-frequencies triples.

(Groner's tests are at www.tinyurl.com/opampdistortion.)

So it'll do the job, but not all that well. To be fair, very few IC opamps can drive 600 ohms with aplomb and a nice simple midband HD spectrum. A few discrete opamps could, though, including the John Hardy 990 and one of Groner's own designs.

So if you want good performance into 600 ohms, probably using an all-discrete opamp, or adding a couple of booster transistors to an IC opamp's output, is a good idea. Personally, I'd like to see the booster transistors be Class-A -- and the IC's output biased so that it remains in Class-A when driving the boosters.

Peace,
Paul

Of course there's always the possibillity of putting a bunch of 5532 in parallel á la Self, that is, with small resistor to prevent the outputs from driving each other...

Imho one of the most efficient solutions would be to add the simplest form of a discrete diamond buffer into the nfb loop of the driving opamp.

A humble tl071 with a humble discrete diamond can drive 8Vpp into 32ohm load at +-15V supply at 0.001.. % thd, according to sim. Frequency-phase response isn't shabby either. With better parts, it can only be better.

frequency and phase

thd @ 8Vpp into 32ohm load
(it can drive higher voltage in higher load, this is dependant also on the ratio of the output transistor's emitter resistors compared to the actual load impedance)

Thanks guys.

abbey road d enfer said:

First, why do you want to drive a 1:2 xfmr?

Click to expand...

Well, I don't necessarily want to.  I'm just trying to learn so I know what to do should I find myself in the situation I would need to.

abbey road d enfer said:

The only reasons to do that are:
The rails are lower than typical +/- 15-18V, or you want more than +20 dBu output level.

Click to expand...

Or I need to maintain unity gain.  Isn't typical approach to drop 6db on input stage for extra headroom to ensure that in 99% of practical applications you'd never hit clipping?  Take GSSL for example, if I wanted to put iron on the output wouldn't I need to go into a 1:2 on the output for unity?  Or I guess if I used both halfs of the existing 5532's as is (one of them to one primary winding, one to the other primary winding) configured for 1:1 wouldn't this get me the same results?

abbey road d enfer said:

A 5534 powered with +/-17V rails can deliver +21dBu. Is it really not enough?

Click to expand...

Sure, for me it's waaaay more than enough, but the real issue I'm thinking about again is unity gain, hence needing a 6db step up.  But obvious to me now there's better ways to do this.

abbey road d enfer said:

If you drive a 150R load with +/- 15V peaks (using +/-17V rails), the peak current will be 100mA (15V divided by 150ohms).
In fact the 5534 will current-limit the peaks at ca. 35mA.
But if you put transistors buffers, they will delver these 100mA peaks.

Click to expand...

OK, so it is just as simple as ohms law.  If doing something in 500/51X format, 100mA would obviously be a no go.  I know I can limit the maximum current draw of the transistors though right?  What I'm not sure is, what happens if I do this?  Does it just lower the max signal voltage that can be achieved if I put a limit on the transistors?

The thing that really makes me scratch my head is this:

If I use a THAT1646 Chip (or 5532 as driver) into a 1:1 transformer to drive 600R load, I get my 6db step up, and max current consumption would be... what about 10mA or so right?

If I go into a 1:2 transformer to drive the same 600R load, it's about 100mA as you indicated.

So why is it 10x's more power to do the same task?

ruckus328 said:

Thanks guys.
=====
The thing that really makes me scratch my head is this:

If I use a THAT1646 Chip (or 5532 as driver) into a 1:1 transformer to drive 600R load, I get my 6db step up, and max current consumption would be... what about 10mA or so right?

If I go into a 1:2 transformer to drive the same 600R load, it's about 100mA as you indicated.

So why is it 10x's more power to do the same task?

Click to expand...

Trust Ohms law.  600 ohms driven to 15V peak (1:1) is 25 ma, 375 mW (peak).  Using a 1:2 step up that same 600 ohm driven to 30V peak draws 50 ma, and makes 1.5W (4x peak power).  Since transformer output power is roughly equal to input power (give or take losses), the 15V transformer input driver needs to make that same 1.5W. 

4x difference not 10x difference. 

JR

ruckus328 said:

abbey road d enfer said:

The only reasons to do that are:
The rails are lower than typical +/- 15-18V, or you want more than +20 dBu output level.

Click to expand...

Or I need to maintain unity gain.  Isn't typical approach to drop 6db on input stage for extra headroom to ensure that in 99% of practical applications you'd never hit clipping? 

Click to expand...

That maybe a design choice for an EQ, for example; however, for a compressor, it's not an obligation. In particular, for opto and VCA-based comps, it's quite easy to define an operating level that does not imply reducing the input level. FET compressor always imply reducing the input level.

Take GSSL for example, if I wanted to put iron on the output wouldn't I need to go into a 1:2 on the output for unity? 

Click to expand...

The GSSL uses a unity-gain input. The VCA operates at nominal level. But the I/V converter has 6 dB attenuation, just because the output stage has 6 dB gain.

Or I guess if I used both halfs of the existing 5532's as is (one of them to one primary winding, one to the other primary winding) configured for 1:1 wouldn't this get me the same results?

Click to expand...

that's the way the GSSL is intended to be used, with both halves of the output IC.

abbey road d enfer said:

A 5534 powered with +/-17V rails can deliver +21dBu. Is it really not enough?

Click to expand...

Sure, for me it's waaaay more than enough, but the real issue I'm thinking about again is unity gain, hence needing a 6db step up.  But obvious to me now there's better ways to do this.

Click to expand...

Yes. You can't consider one parameter (gain) and neglect the other (output voltage). In the particular case of a transformer output stage, if you need 6dB gain but you don't need the elevated output voltage, you just use 6dB gain on the opamp and drive a 1:1 xfmr.

I know I can limit the maximum current draw of the transistors though right?  What I'm not sure is, what happens if I do this?  Does it just lower the max signal voltage that can be achieved if I put a limit on the transistors?

Click to expand...

Yes; the max output voltage will be dependant on the actual load impedance.

http://www.picocompressorforum.com/forum/php/viewtopic.php?f=6&t=140

I messed up my post - the 0.001..% THD calculation was done at 8Vp (iow, 16 Vpp) into 32 Ohm load at +- 15V supply. The buffer with values listed starts clipping at 10 Vp (20 Vpp) into 32 Ohm load under same conditions - no nasty waveforms, it's partially current-starved.

At lighter load, it will of course give larger output. It's best to use metal-can transistors for the output pair, the driver pair can be plastic to-92. Or use BD139/140 for the output pair. Under normal conditions it would probably suffice just to bolt them onto the PCB to allow for better heat transfer (without using an actual heatsink).

Another alternative would be to use BUF634:
http://focus.ti.com/docs/prod/folders/print/buf634.html


Both are different variations on the same principle.

Guys, thanks, this has been very informational indeed.  I'm still a little torn on which is the best route though, but based on what I learned here seems to me of the 3 choices:

a) THAT1646 (not possible) or 5532 (with booster transistors) into 1:2 step up would be the worst choice, as I really don't need the extra headroom to begin with, other choices seem better on power, and if in a 500 series module the current demand would almost max the module, so for me this option is out the window.

b) 5532 into a 1:1 with transistors.  Seems a pretty safe bet and guaranteed way to ensure good noise figures into any load.  Would need to limit the max current, but don't really see that being any issue as with say a 20mA limit, if I'm correct would mean 12V output possible, or around 19dbu, which should still be plenty.

c) THAT1646 into a 1:1.  Seems the simplest solution as far as component count is concerned, still not sure what THD figures are going to look like though.  Pico thread seems interesting, only skimmed through it but going to give it a good read through tomorrow.  Thinking maybe I should also see what THAT has to say as datasheet really doesn't go into any details, just that it can drive 600ohm, like was already mentioned here though, how well is another story.

The big thing I'm still not clear on is this:  Let's say I use a THAT1646 or 5532 pair (no transistors) into a 600ohm load with 16V rails, this means the chips are gonna pull 26mA right?  How do I know if what I'm asking of the chips is too much?  I'm assuming in this case it is ok as the datasheets state they can drive 600R, but for other chips/scenareos how do I calculate how much is too much?  The chips are just going to keep putting out whatever I ask them too until they fry right?  Unless I put a limit on the current (which wouldn't be possible with the THAT chips).  But I don't see anything in regards to max current or mW ratings anywhere (unless I'm missing it).

The safest bet for longevity and "muscular" attitude is to use a 5532 and a discrete output buffer.

The second "toughest" is to use a 5532 and a TO-220 version of BUF634.

But if you want to use just a DIP-package opamp into a 600 Ohm, try this (the opamps are pulled/biased into a low-level class-A operation at approx 7.5mA, this results in a fairly "clean" output at low levels):

Note: I used a jelly-bean opamp from LT lib, because I don't have 5532 model here. The same principles apply.

at 2Vp (4 Vpp) into 600 Ohm, this config has pretty low distortion (and part count)

stays fairly clean up to 12Vp (24 Vpp) into 600 Ohm: