Paralleled swinging op amp

[thor.zmt]

That IRF610/9610 is THT

Nope, it's SMD, just find the right Suffix, DPAK2 IIRC.

and the article you posted is about using an LME49710 with bias current relief via the bootstrap for source impedances up to 1M, so I do wonder if you might not be veering towards a very flexible network to surround a paralleled LM4562 with 1M/220n or something. Y’know, breadboardable. Polypropylene in that case I suppose, though I am recalling one affordable PPS series with leads, maybe Kemet.

I personally strongly tend towards FET's over BJT, not always, but often.

The particular way I do Op-Amp's for AC is down to wanting to avoid coupling capacitors as much as possible.

As long as we use really low distortion Cap's (SMD C0G are usable) this circuit is invisible for AC, DC is reliably blocked without big value electrolytic cap's and lookback impedances for both inputs are always balanced (if we take care of this in the external circuit).

Typically I also have 10pF C0G in parallel with the resistors.

If paralleling multiple Op-Amp, use current sharing resistors and parallel the AC side (if needed with low value isolation resistors.

BTW, to get more output current Sandeman "Class S" aka Technics Class AA is a better choice IMNSHO that brainless paralleling. Even better with a minimally AB biased Class AB output stage on second Op-Amp.

Noise can always be dealt with using a secondary noise cancelling amplifier, which includes bipolar Op-Amp's.

It is possible to make really innovative "compound" op-amp's that are much more interesting than parallel inputs or outputs.

The 5534 is especially interesting because we have various commonly inaccessible nodes brought out.

We could use one 5534 to be the core of a FDA (using a second unit as inverter), using external 1nV|/Hz J-Fet's as input stage, with accessible sources (!!!) and use local feedback loops to the J-Fet drains as both virtual "folded cascode" and to lower overall gain back to stable conditions. Bring up second stage current while we are at it to improve slew rate and linearity.

The use Pin 5 on both 5534 to drive an external buffer, maybe diamond or Diamond Sziklai.

No monolithic Op-Amp or FDA exists that can match what this puppy can do...

Of course, at that point we are only a dozen or so parts away from fully discrete.

Thor

 

[thor.zmt]

You’re a 1678 user. Would it like 3M3 for each inner loop and 470n between each inverting input and a shared attenuator?

I did not parallel anything, but in an actual product I used 6M8 // 10pF & 100nF (PPS) + 10nF (C0G).

Sort of a fancy way to save parts

Not meant to save parts.

but I see that it would ensure dead nuts AC accuracy.

And minimum DC to avoid AC coupling on outputs.

And it avoids the need for servo's and/or big value electrolytic capacitors in the feedback divider, all of which have their issues.

The principle can be extended to AC couple the output as well and keep the output coupling capacitor in the feedback loop.

Thor

 

[JohnRoberts]

Understood. But my question is really whether this actually offers a benefit (rather than a feature) to a user. Don't get me wrong - I like it technically. Just not sure about the merit of higher voltage operation in general. Esp when that level is unlikely to be supported throughout the signal chain.

A fair question in the context of studio recording. Back in the 80s I dedicated one of my "Audio Mythology" magazine columns to comparing +4dBu to -10dBV hardware. Even by the 80s semiconductor noise floors were low enough to not dominate. From memory IIRC I concluded that the semi-professional -10dBV platforms offered more headroom and lower cost. For both the noise floors were dominated by mic preamps and ambient room noise. [edit- there is a slight benefit from reduced interface noise between sundry SKUs while this depends on implementation./edit]

Yes - what I'm questioning is the reason for wanting those higher voltage rails (beyond technical curiosity)

this was a marketing hook used back in the day promoting discrete API consoles running hotter rail voltages. The fact that the rest of the industry didn't follow them down that rabbit hole suggests minimal real world benefit.

+1



Well "Better is Better" wrt CM distortion

Perception of better, is always valuable...

JR

 

[user 133392]

I'm glad to see you using duals in parallel. I've always wanted to try that. That second op amp is along for the ride no matter what and the output current is certainly needed to drive the added load current.

The oldest example of this circuit I can find is 1974. "A Few Extra Components 741 Op Amps For High-Voltage-Swing Applications," Surjan Dogra, Electronic Design, April 26, 1974. Bootstrapped Op Amp Articles - Pro Audio Design Forum

I built Dogra's circuit right after it was published and the concept of floating rails stuck with me. Graeme cited Dogra and we all took it from there on. I have a folder full of variations on this theme.

I realize they have a highish threshold voltage but what about the SMT versions of the BS170 (N) and BS250 (P) for the bootstraps?

They are a little hard-to-find right now, particularly the PNP, but the ZTX851/951 are rugged as hell. I changed more than my fair share of the plastic PN5679/PN5681 pairs in ±32/36V MCI JH-series consoles. ZTX have several SMT PNP/NPN compliments worth looking at.

In my current-boosted version I shifted the position of the resistor and diode so that it served double-duty to bias the output stage.
The C-B capacitors for the bootstrap transistors are a must.
I found local Cfb around the op amp to be the best compensation method. In my OP that comment is a couple of posts down.
In this application it looks like the best op amp bypass is differential with a single C from the Vcc to Vee pins rather than two Cs to ground. From the perspective of the bootstrap transistors the single differential bypass C is bootstrapped. That C "sees" a constant 30V or whatever differential supply voltage for the OA is chosen.

One comment I'll make about using dual bipolar op amps (non Ib-compensted) are that the bias current differences between sections are usually well matched. Vos match is random but Ib is usually pretty close. I mention this as a point of reference. In the OP, with the small resistor values, the delta Vos will dominate circulating currents between outputs.

 

[atavacron]

I realize they have a highish threshold voltage but what about the SMT versions of the BS170 (N) and BS250 (P) for the bootstraps?

Good direction, specs are notably better. MMBF170 was pretty straight ahead to figure out, nothing special about OnSemi generally but the numbers are right. Vishay does have a BS250 in SOT but it doesn’t seem to be quite the complement. Hmm. If I’m gonna go wading through data sheets, I’m looking for a 60V/-60V N & P that have a low Vth, Crss and…low Rds on? Anything else? It is *wild* how the exact same part from different manufacturers will have significantly different typical specifications - not just min or max.

 

[atavacron]

I did not parallel anything, but in an actual product I used 6M8 // 10pF & 100nF (PPS) + 10nF (C0G).



Not meant to save parts.



And minimum DC to avoid AC coupling on outputs.

And it avoids the need for servo's and/or big value electrolytic capacitors in the feedback divider, all of which have their issues.

The principle can be extended to AC couple the output as well and keep the output coupling capacitor in the feedback loop.

Thor

Let me make sure I’m on the same page. Riso & bootstrap biz not drawn yet. Yea?


I’ll add in the bootstrap biz and do a DC coupled version as well, using OPA1692 as the closest BJT amp, so we can look at pros & cons and any supporting components we might not have thought of yet. Just want to make sure I have the right topology for FET amps first. I’m okay with the natural Vos of the driver exiting the unit; not so hot on ‘lytics myself.

 

[user 133392]

Good direction, specs are notably better. MMBF170 was pretty straight ahead to figure out, nothing special about OnSemi generally but the numbers are right. Vishay does have a BS250 in SOT but it doesn’t seem to be quite the complement. Hmm. If I’m gonna go wading through data sheets, I’m looking for a 60V/-60V N & P that have a low Vth, Crss and…low Rds on? Anything else? It is *wild* how the exact same part from different manufacturers will have significantly different typical specifications - not just min or max.

You may have noticed that the BS250 is also a TP0610 and is available in SMT SOT-23 https://www.vishay.com/docs/71411/tp0610k.pdf https://www.vishay.com/docs/70209/70209.pdf

 

[atavacron]

You may have noticed that the BS250 is also a TP0610 and is available in SMT SOT-23 https://www.vishay.com/docs/71411/tp0610k.pdf https://www.vishay.com/docs/70209/70209.pdf

Bingo. And widely available.

[big edit]

Let’s jump straight to complementary pairs, as their specs are way closer. Lower capacitance and threshold too. Do all (or none) of these fit the bill?

Diodes DMC31D5UDJ (30V SOT-963) heavily stocked

Panjit PJT7603 (50V SOT-363) almost kinda hand-solderable

Panjit PJX8603 (50V SOT-563)

 

[atavacron]

What do you guys think about this for each side of the input buffer? No need to specify the amp, lots of options there.



Gate HPF might be overkill but I’m just keeping the rolloff consistent. Also maybe we don’t need the 10u/100n on the 16V lines?

 

[marcuswilson007]

In my opinion, the more headroom the better! Linearity of active electronics gets worse the more current you ask of it.

 

[thor.zmt]

Let me make sure I’m on the same page. Riso & bootstrap biz not drawn yet. Yea?

View attachment 109295

I would use individual RC per Op-Amp on + Input and add a 10pF 0402 C0G cap directly at the Op-Amp for RFI suppression

I’m okay with the natural Vos of the driver exiting the unit; not so hot on ‘lytics myself.

Sure. Or you can use a single sided supply, fully ACcouple and keep all AC coupling except input inside the feedback loop... ;-)

Thor

 

[atavacron]

I would use individual RC per Op-Amp on + Input

can do

and add a 10pF 0402 C0G cap directly at the Op-Amp for RFI suppression

across which nodes?

 

[thor.zmt]

What do you guys think about this for each side of the input buffer? No need to specify the amp, lots of options there.

View attachment 109297

Gate HPF might be overkill but I’m just keeping the rolloff consistent. Also maybe we don’t need the 10u/100n on the 16V lines?

Gate stopper is needed, Mosfet followers love to oscillate. No caps needed on +/-16V at all is only used for bias.

You can use 6M8 and 100nF, additionally also around the Op-Amps if not using them as follower.

If using OPA1678, why not OPA1679 4-Way parallel compound?

6dB noise advantage over a single, plenty of current drive, competitive cost. And OPA1644 as upgrade.

Thor

 

[thor.zmt]

across which nodes?

+In to gnd 6M8 // 10pF per Op-Amp
-In to out 6M8 // 10pf per Op-Amp

Coupling caps to each input 100nF minimum. To be more sure of isolation between inputs and correct RFI suppression add 100 Ohm series resistors per input, per Op-Amp.

200 Ohm will be negligible for noise on FET/J-Fet types. The turnover of 100Ohm & 10pF is 160MHz, so it will not impact on anything stabilitywise, but may help.a lot with 2.4GHz & 5GHz interference.

Thor

 

[thor.zmt]

Next Stop, super high input impedance and super low noise balanced input using a Quad 1679 and we can publish, once build and measured.

And we earn eternal gratitude and unlimited free samples from TI...

Thor

 

[atavacron]

@thor.zmt I'm drawing your version of the output. Can you look at Diodes DMC31D5UDJ please and LMK if it cuts the mustard. Thanks!

Also, you iterated the need for a gate stopper when we were looking at the unity gain (input) version. Were you thinking any safety or stability components might be necessary in addition to the 1K resistors? That complementary pair has a variety of diodes for gate protection as well.

 

[user 37518]

I find MCI-like circuits extremely good to obtain a large headroom, however, it is also the best way to fry transistors and generate heat. The MCI JH-500, which used these sort of topologies, is notorious for being a barbeque grill with knobs.

 

[atavacron]

I find MCI-like circuits extremely good to obtain a large headroom, however, it is also the best way to fry transistors and generate heat. The MCI JH-500, which used these sort of topologies, is notorious for being a barbeque grill with knobs.

I'll be posting BJT and FET versions of the output driver in a day or two, and would appreciate any insight on them you may have. There is also this (below), from someone who's built high-voltage swinging amps and was commenting privately on floating supply topologies like the first one I posted. I am not sure whether all of it is transferable to the dual-dual-supply approach, which is what this has turned into (for now).

There’s a couple drawbacks to that approach to keep in mind:

1. Power-up can often cause latching situations because the output voltage of many op amps is not well defined during start-up.
2. There can be stability issues:
   1. The miller cap in op amps is referenced to the power supplies, and if the power supplies are moving with the output signal, the miller cap is no longer being charged/discharged in the expected way.
   2. The power supplies are now forming an additional feedback loop. At higher frequencies, where PSRR is rolling off, the feedback from the output is increasing.

 

[user 37518]

There’s a couple drawbacks to that approach to keep in mind:

1. Power-up can often cause latching situations because the output voltage of many op amps is not well defined during start-up.
2. There can be stability issues:
   1. The miller cap in op amps is referenced to the power supplies, and if the power supplies are moving with the output signal, the miller cap is no longer being charged/discharged in the expected way.
   2. The power supplies are now forming an additional feedback loop. At higher frequencies, where PSRR is rolling off, the feedback from the output is increasing.

Those are all really good points. In my opinion, unless you have a really good reason to do this, I would advice not to. The MCI JH-500 consoles (A and B revisions) did it because, initially, they used the MCI2002 (an Intersil rebranded part) op-amp, which was able to run with high voltages (+/- 40V IIRC). This was prior to the NE5534 hitting the market. After the NE5534 was released, the engineers wanted to use the 5534 (relabeled in a metal can as MCI2003) for its superior performance and low noise, however, I guess that they wanted to retain the big headroom that they used to obtain with the 2002 (or perhaps as a marketing gimmick, since they actually used to place magazine ads boasting the bootstrapped 5534), so they used the transistor bootstrapped arrangement, which (according to one of the designers) was taken straight out from an 'op-amp cookbook".

This was also the era were recording consoles with only IC op-amps were not that popular and frowned upon, so perhaps the idea of having a +/-15V console would've sounded as sub-par to those used to at least +/-24V or the original +/-36V MCI JH-500 with MCI2002 op-amps. Be it as it may, they did it.

Initially, the same +/- 36V PSUs that were originally used on the A and B rev desks with the MCI2002s were used and released with bootstrapped 5534s, but they had to lower the voltages to +/- 32V because the op-amps kept getting fried. The 5534-bootstrap was done on the C revisions and some of the D rev, although, IIRC the last of the D revisions had all 5534s (and maybe some 5532s) with +/-15V and no transistor bootstrap. In the final desk ever produced by MCI: the JH-600 desk series, the entire bootstrap idea was completely abandoned; probably due to all negative things it carried around, namely, driving the 5534s to their limits, fried parts and an excessive amount of heat, which not only had the downside of being a nuisance for the operator, but also because it drastically lowered the electrolytic capacitors' lifespan to a few years. On top of that, MCI used the infamous 'orange Siemens' capacitors, which, to make things worse, were rated at 85 °C and had a tendency to spill out nasty electrolytic all over the PCB. I dare you to find a JH-500 with any modules that are not covered with an almost impossible-to-clean layer of gunk dating back to the late 70's/early 80s covering the entire PCB.

In my experience dealing with the JH-500 for years, I can say that the 5534-bootstrap is one of its main appeals, but it is also its curse. I would really avoid that configuration, especially in a desk (where it is perhaps where it's most attractive), you can do it in a preamp or similar, although I don't know if everyone in today's record-straight-to-the-PC era would want a preamp with such a headroom that it overloads the input of 99.9% of ADCs.

 

[thor.zmt]

Can you look at Diodes DMC31D5UDJ please

Looks ok for the job, watch power dissipation with shorted output.

Also, you iterated the need for a gate stopper when we were looking at the unity gain (input) version. Were you thinking any safety or stability components might be necessary in addition to the 1K resistors?

No, 1k Resistors should be ok. But without oscillation is near certain. And put them close to the FET Pin's.

Thor