Phantom protection for transformerless output single supply project

[atavacron]

What’s the most robust option for phantom power protection with AC-coupled transformerless line outputs? And does it matter if we’re talking single or dual supply? Trying to cover all the bases of what happens when people inadvertently patch into a mic pre input in a global phantom scenario on longframe and TT bays.

This will be a line amp for 24V single supply but I am getting held up after the driver. Pretty standard dual driven differential stage with 47R isolation, big bipolar caps, and 10K bleeders to 0V.

BAV99 or similar from signal to 24V & 0V, for starters, same as dual supply?

Do capacitors absolutely have to be 63V, or is there some zener/TVS trick to allow a lower rating? It would be wonderful to be able to use 16V, as the interstage coupling is that. Input is 100V C0G, no worries there.

Nichicon UES seems to be the only game in town, so I’m sort of stuck with 50V and crossing my fingers that nobody’s gonna leave the output patched into a mic pre for more than a few minutes. My height limitation is a firm 25mm, and UES 50V in that case maxes out at 220u. It would be really great to find a wide base low profile 470u/63V, but it doesn’t look promising. I’m fine with paralleling. I’d like to avoid polar caps here.

Regular -2dBu single ended / +4dBu differential nominal levels. Max signal is +20dBu/8Vrms/11Vpk single ended, +26dBu/16Vrms/22Vpk differential. I believe typical external load is a little under 1K when feeding something 10K, but of course 600R is possible. Driver is based around an OPA1656 if that matters, but i’m still looking at BJT RRIO options.

And yeah, you guessed it, it’s a Neve line amp replacement for monitoring, but people will probably want to use it for other summing and gain staging applications, outside of the Neve realm. 66dB gain maxed out, not optimized for mic pre purposes, but could be used that way. Mainly just an incredibly flexible, clean drop-in replacement for the 3415, 3416 & associated parts.

Thanks!

 

[metalockpick]

Phantom pwr is a well filtered 48VDC, not a lot of current. A high quality 50VDC cap wouldn't be a stretch here. If you have room on your board, you could lay this one on its side to fit your 25mm limit, or parallel wire >50VDC units of lower capacitance to get your chosen value with de-rating and potentially lower ESR. Do you really need 470uF? At 100 Ohms, 100 uF has a knee of about 15 Hz, so 470 uF @ 600 Ohms seems excessive. Can it work with 35VDC or lower cap? Sure, but for how long?

 

[atavacron]

Phantom pwr is a well filtered 48VDC, not a lot of current. A high quality 50VDC cap wouldn't be a stretch here. If you have room on your board, you could lay this one on its side to fit your 25mm limit, or parallel wire >50VDC units of lower capacitance to get your chosen value with de-rating and potentially lower ESR. Do you really need 470uF? At 100 Ohms, 100 uF has a knee of about 15 Hz, so 470 uF @ 600 Ohms seems excessive. Can it work with 35VDC or lower cap? Sure, but for how long?

This is good.

Yes I can lay caps down, but not ideal. Would need to add hardware for that as zip ties won’t fit and epoxy is forever.

I have the LF rolloffs calc’d for minimum phase disturbance at 20Hz, and to fend off distortion. It’s not about bandwidth or big time sub bass. So my input fc is .5Hz, and with two 330u in parallel per leg, I’m shooting for .25Hz at the output, into the typical just-under-1K-load scenario. If physical parts are the limitation, and I can’t do this, so be it.

There is the Audio Note Kaisei series from Rubycon, but it’s an audiophile product. Seems like some 63V options there, but very spendy. Solen has a huge axial 330u/100V, also $$$, no hardware required.

The biggest UES goes is 330u/50V in 16x30.5. Four of those per board, yikes. Maybe two. Otherwise (assuming that’s a thumbs-up for 50V), 3x 220u per side would totally fit, upright.

 

[atavacron]

Any thoughts on diodes? Just the regular pairs to the rails, or is there a craftier solution a la what The Phantom Menace Returns established for inputs?

 

[metalockpick]

Yours is single supply, so audio can't clip the diodes. They might be there to short out flyback from any attached input transformer power fault. I can't think of anything else, and even if that fault were to happen, the 10k resistors would likely dissipate the excess voltage. Am I missing something?

 

[atavacron]

Am I missing something?

Probably not! Just been a long time since I drew anything with output caps.

 

[ccaudle]

a la what The Phantom Menace Returns established for inputs

Why not just use that scheme for outputs as well?

 

[squib]

Use 2k2 resistors to from the output caps to 0v. This will ensure the phantom voltage gets divided down via the 6k8 feed resistors to below 20 v . Then you can use 25v caps.

 

[atavacron]

Why not just use that scheme for outputs as well?

I’ve been drawing that for direct coupled outputs. Seems like a good idea. I guess I’m unsure if it would be advantageous in conjunction with output caps and bleeders.

 

[atavacron]

Use 2k2 resistors to from the output caps to 0v. This will ensure the phantom voltage gets divided down via the 6k8 feed resistors to below 20 v . Then you can use 25v caps.

Whoaaa….breakin my mind here. Something like 5K6 would be the right bleeder value for 24V single supply, i think, to minimize loading…and to keep the bias about the same when the cap is reverse biased.

5K6 / (5K6+6K8) x 52V = 23.5V.

Or keep it simple and just go with 6K8.

35V caps. Looks like the resistor handles 98mW.

Yes?

[edit: i see that you divided down to 12ish volts, so that the caps are sitting there unbiased and can work fine into a mic pre. interesting.]

 

[atavacron]

Can it work with 35VDC or lower cap? Sure, but for how long?

Now I understand why you used 35V for your example.

10K / (10K + 6K8) x 52V = 31V

31V - 12V = 19V

So maybe the answer to that is “for a little less than the normal life of the cap,” with a 35V cap, or “for a few years” with a 25V cap. The other clear argument for 35V being “what if the driver hits the 24V rail during startup or due to a fault.”

 

[atavacron]

Just to wrap up my capacitor mystery -

I was searching wrong. There are more series. UEP is most common and is 105deg….and in my case, for output, there’s a 1000u/25V in there or 470u/35V that will both fit.

I drew 6K81 in so that I can use the same resistors I use for mic pre phantom, and AFAICT the caps will see their normal 12V forward biased and 14V worst case reverse biased by phantom. No diodes.

Thanks again @squib !

 

[gyraf]

Anyone seen manufacturer-suggested or application-note example of protecting these DRV134/SSM2142 type outputs?

I have a vague feeling that there's some good reason why we never don't protection implemented irl (yes, I've replaced all too many of these in our studios).

Surely, if a good solution existed, it would have been implemented already?

For the above attempt to protect-by-capacitor, I think you run into the same problem as when protecting inputs, that is, the energy stored in a capacitor big enough to pass adequate low-end is always plenty to kill the silicon..

/Jakob E.

 

[atavacron]

For the above attempt to protect-by-capacitor, I think you run into the same problem as when protecting inputs, that is, the energy stored in a capacitor big enough to pass adequate low-end is always plenty to kill the silicon..

single supply, gotta do something about that 12V.

good point though, maybe diodes to 24V and 0V between 50R iso and the cap? or some other current limiting? i guess it would be a good excuse to replicate the Phantom Menace tactics.

 

[ccaudle]

i guess it would be a good excuse to replicate the Phantom Menace tactics.

I didn't really follow up on my previous post of "Why not just use that scheme for outputs" but the more important information to glean from that paper is not just the recommended device arrangement, but the detailed explanation of the causes of high current flow, and explanations of where that current can flow and how it causes damage. The easy path is to just grab the recommended parts arrangement and slap it in, but the more useful long term would be to take an hour or two and look at the diagrams in the paper and the explanations, and see how that applies to an output stage design that you want to use, and determine if the same concerns apply and if the same arrangement is beneficial, or if any modifications would be needed.

 

[user 37518]

Phantom pwr is a well filtered 48VDC, not a lot of current. A high quality 50VDC cap wouldn't be a stretch here. If you have room on your board, you could lay this one on its side to fit your 25mm limit, or parallel wire >50VDC units of lower capacitance to get your chosen value with de-rating and potentially lower ESR. Do you really need 470uF? At 100 Ohms, 100 uF has a knee of about 15 Hz, so 470 uF @ 600 Ohms seems excessive. Can it work with 35VDC or lower cap? Sure, but for how long?

A 100uF cap with a 100 ohm resistor has a 15 Hz knee, but it still provides nearly 2 dB attenuation at 20 Hz and around 0.6 dB at 40 Hz. A 470uF cap with 100 ohms gives 0.12 dB attenuation at 20 Hz, and, with 600 ohms, it gives 0.004 dB att. at 20 Hz. The 470 uF doesn't seem excessive to me.

Also, you need to consider that, when it comes to electrolytics, the lower the AC voltage across them, the less distortion they produce. A general rule of thumb provided by Douglas Self is that the AC voltage across a cap at 20 Hz should be less than 80 mV, in order for it to produce negligible distortion products. Since we are discussing condenser mics, it is not any stretch of the imagination to consider 1 V (or higher) signal levels. A 100 uF cap with a 100 ohm resistor would then have 440 mV across it, with a 600 ohm load it drops 120 mV, but still higher than I would like. On the other hand, the 470 uF cap has 145 mV with 100 ohms and 27 mV with 600 ohms. Again, 470 uF seems fine to me.

One could argue that maybe 330 uF should be sufficient, but that still doesn't make 470 uF a total overkill.

 

[user 37518]

Whoaaa….breakin my mind here. Something like 5K6 would be the right bleeder value for 24V single supply, i think, to minimize loading…and to keep the bias about the same when the cap is reverse biased.

5K6 / (5K6+6K8) x 52V = 23.5V.

Or keep it simple and just go with 6K8.

35V caps. Looks like the resistor handles 98mW.

Yes?

[edit: i see that you divided down to 12ish volts, so that the caps are sitting there unbiased and can work fine into a mic pre. interesting.]

Note that adding the 2.2k resistors as squib suggested would lower the Thevenin equivalent resistance to 1.66 K rather than the standard 6.81 K. This would also have to be accounted for regarding the input impedance of the preamp.

 

[metalockpick]

That's all true. However, you need to consider that allowing excess sub bass through the circuitry (below that which can be reasonably expected to go through transducers at the end of the signal) is likely to have greater negative consequences to what is audible than a wee bit of phase shift down low.

 

[JohnRoberts]

Anyone seen manufacturer-suggested or application-note example of protecting these DRV134/SSM2142 type outputs?

I have a vague feeling that there's some good reason why we never don't protection implemented irl (yes, I've replaced all too many of these in our studios).

Surely, if a good solution existed, it would have been implemented already?

For the above attempt to protect-by-capacitor, I think you run into the same problem as when protecting inputs, that is, the energy stored in a capacitor big enough to pass adequate low-end is always plenty to kill the silicon..

/Jakob E.

I recall from discussions with THAT corp engineering about modifying one of their I/O chips for a different application, they provide robust over voltage clamps for every I/O pin designed into the silicon.

JR

 

[user 37518]

That's all true. However, you need to consider that allowing excess sub bass through the circuitry (below that which can be reasonably expected to go through transducers at the end of the signal) is likely to have greater negative consequences to what is audible than a wee bit of phase shift down low.

Exactly my point, it is not only wee bit of phase shift down below, THD (which extends beyond 20 Hz) is one of the greatest offenders, which, in the case of electrolytics, gets reduced by increasing cap size. I see no negative impact, rather perhaps a larger time to get it charged. I have zero problems with subsonics, if the operator doesn't want them he can use a low cut, but I won't do it for him before hand. Or it is the mic's fault for not tailoring the freq response; the pre should be as transparent as possible, with negligible attenuation at 20 Hz, and low THD. A 2 dB attenuation at 20 Hz, and 0.6 dB at 40 Hz is unacceptable.

 

[atavacron]

take an hour or two and look at the diagrams in the paper and the explanations, and see how that applies to an output stage design that you want to use, and determine if the same concerns apply and if the same arrangement is beneficial, or if any modifications would be needed.

heard. i'm short-handing the reference.

A general rule of thumb provided by Douglas Self is that the AC voltage across a cap at 20 Hz should be less than 80 mV, in order for it to produce negligible distortion products. Since we are discussing condenser mics, it is not any stretch of the imagination to consider 1 V (or higher) signal levels. A 100 uF cap with a 100 ohm resistor would then have 440 mV across it, with a 600 ohm load it drops 120 mV, but still higher than I would like. On the other hand, the 470 uF cap has 145 mV with 100 ohms and 27 mV with 600 ohms. Again, 470 uF seems fine to me.

this is a good breakdown; was it meant to be portable over to outputs?

--

an ƒc of 0.33Hz results in 1° phase shift at 20Hz. 820µF into 600Ω gives an ƒc of 0.32Hz. But good luck finding an 820 in bipolar for short runs. 1000µF into 600Ω gives an ƒc of 0.27. 940µF into 600Ω gives an ƒc of 0.28Hz. Either are overachieving, but 1° seems like a good benchmark. Not sure why, really. @abbey road d enfer mentioned that he uses 100µF into 10K, typically, which well exceeds that benchmark.

470µF into 600Ω gives an ƒc of 0.56Hz, resulting in 1.6° of phase shift at 20Hz. I've seen a whole lotta gear with 470µF in this position if space and $ allows. Does the 0.6° difference matter? Is it all just specsmanship below the 0.1dB at 10Hz (a whopping 4.5° of shift at 20Hz) that an ƒc of 1.59Hz provides? I'm thinking about how multiple ins and outs add up to pretty strong bandwidth limiting, on both the record and mixdown side. And of course THD, as mentioned upthread.

 

[user 37518]

heard. i'm short-handing the reference.

this is a good breakdown; was it meant to be portable over to outputs?

--

an ƒc of 0.33Hz results in 1° phase shift at 20Hz. 820µF into 600Ω gives an ƒc of 0.32Hz. But good luck finding an 820 in bipolar for short runs. 1000µF into 600Ω gives an ƒc of 0.27. 940µF into 600Ω gives an ƒc of 0.28Hz. Either are overachieving, but 1° seems like a good benchmark. Not sure why, really. @abbey road d enfer mentioned that he uses 100µF into 10K, typically, which well exceeds that benchmark.

470µF into 600Ω gives an ƒc of 0.56Hz, resulting in 1.6° of phase shift at 20Hz. I've seen a whole lotta gear with 470µF in this position if space and $ allows. Does the 0.6° difference matter? Is it all just specsmanship below the 0.1dB at 10Hz (a whopping 4.5° of shift at 20Hz) that an ƒc of 1.59Hz provides? I'm thinking about how multiple ins and outs add up to pretty strong bandwidth limiting, on both the record and mixdown side. And of course THD, as mentioned upthread.

I am really not that bothered by the phase shift as much as I am bothered by the attenuation and THD. However, to me, phase shift is not a good indicator; I prefer group delay. I took the liberty of plotting the group delay (or rather the negative of the group delay) of an RC with C = 100uF, 470 uF, R= 100, and 600 ohms in MATLAB. With R = 100 you get around 4 ms of delay at 20 Hz, 2 ms at 30 Hz, and around 1 ms at 50 Hz, are these significant for your ears (or rather your stomach)? you'll be the judge.

 

[user 37518]

And just for the sake of it, here is the unit-step response. Evidently the 470 uF + 600 ohms makes it more sluggish, and the cap will take around 2 seconds to fully charge to its steady-state DC value.

 

[gyraf]

Thanks, JR - found it:

from https://thatcorp.com/datashts/THAT_1606-1646_Datasheet.pdf - page 9

"The 1606 and 1646 each incorporate a
proprietary internal protection scheme, which will
suffice for many situations seen in the field.

However, one might foresee having the line driver's output
mistakenly plugged directly into a microphone
preamplifier input that has +48V phantom power
applied. When this happens, the ac coupling
capacitors on the preamp's input will discharge into
the low-impedance output of the 1606/1646. This can
result in surge currents of over 2 amperes 6 .

The amount of energy stored in these capacitors is
directly proportional to the capacitor value, which is,
of course, not under the 1606/1646 designer's
control. The 1606/1646's internal protection network
will withstand this abuse for coupling capacitors up
to about 33 uF.

To protect against microphone preamplifiers
that incorporate larger values of capacitance, a pair
1N4004 diodes from each output to the supply rails,
as shown in Figure 8, is recommended. This shunts
the discharge current to the power supply bypass
and filter capacitors, thus protecting the output of
the 1606 or 1646."

/Jakob E.

 

[atavacron]

are these significant for your ears (or rather your stomach)? you'll be the judge.View attachment 109920

Uh, yeah, I’d say 1ms at 10Hz would be about all I’d want to add up over the number of times a kick drum or bass DI signal goes in and out of a signal block. Four iterations of the purple curve puts it forward of the yellow curve.