THAT 1512 phantom power considerations

[leswatts]

Hi,

First post.

I'm throwing together 6 THAT 1512/1646 based mic pres shortly for my studio. Getting ready to order parts. Before I do i'd like to discuss the phantom blocking circuitry a little.

I've done a search on this and read quite a bit. Seems it boils down to these options:

1) "Menace" configuration (from the AES paper). This is shown in the 15XX development board docs.
It uses rather high 100k input bias resistors that allow a small (2.2uF) non polarized film blocking cap.
I like the idea of no electrolytics, but I wonder about current noise at very low frequencies and improper loading of some dynamic and ribbon mics.

2) polarized 47uF phantom blocking caps with 1K5 bias resistors. This is pretty much standard I guess, but I sure don't like the idea of this when phantom is off...unless it was bypassed. An alternative would be to leave phantom on with dynamics etc so the cap would be polarized.

3) As #2 but with giant, expensive 47uF NP film capacitors. Or a somewhat lower value with higher bias resistors (much less than 100k).

4) Non polarized electrolytics. Seems a bad plan to me, given what i've read about their electrical characteristics.

5) Flying rail with no blocking caps. I read about this, but it seemed to not go anywhere.

So, before I finish my bill of materials I would like to ask for opinions on these options.

Les
L M Watts Technology

 

[Svart]

I like the idea of no electrolytics

Don't fall into that trap.  Proper usage, biasing and spec'ing will make AC coupling caps practically transparent.  It's the improper usage and spec'ing that give them a bad name.

Non polarized electrolytics. Seems a bad plan to me, given what i've read about their electrical characteristics.

This is actually the way I would go for regular AC coupling caps(not phantom blocking caps).  Please post your reasoning behind NOT using them because most of what I've read states that these ARE what you should use for AC coupling duty.

As for phantom DC blocking, I don't think you could pick the film caps over decent 'lytic caps by ear...

 

[leswatts]

Hi Svart....you're not too far from me!

I'm a transducer development engineer, so i'll try to not fall into any audiophool traps. Lord knows I saw enough of that long ago when I designed phono cartridges.

I'm not too concerned about about modern electrolytics when they're properly biased (i.e phantom power ON). It's when phantom is off that is an issue me.

I actually have heard distortion in NP electrolytic caps in speaker crossovers. Might have been defective parts though...they were old.
Of course the ac voltage was much higher than the < 1 volt from a dynamic microphone. If no one else has I could take some distortion measurements of low level ac through a lytic.

Here's what seems to be a reasonable discussion about lytics:

http://yarchive.net/electr/electrolytic_caps.html

Tell me what you think.

Of course a simple solution is to leave phantom on while using dynamic and ribbon mics. The only danger would be from a cable fault, or vintage mic with a center tapped xformer.

Another is to bypass. Not a trivial thing...have to dump the charge controllably and have proper timing with the +48 turn off.

A third thought is to leave phantom on, but switch in a high series resistance such that the current capability would be too low to hurt any dynamic or ribbon mic...but still enough to bias the cap.

I have seen the menace circuit used commercially ( no electrolytics) and that's what I have drawn up, mostly from the THAT demo board ap note. I also have  gain change and output servos, but might not need both.

Perhaps I need to do a current noise calc with the high bias resistors. It's got to be pretty big, but only at a few hertz and below.
There's still the loading issue...many of my mics won't like the 13.6k || 200k diff. input impedance of the menace.

Les
L M Watts Technology

 

[JohnRoberts]

Hi,

First post.

I'm throwing together 6 THAT 1512/1646 based mic pres shortly for my studio. Getting ready to order parts. Before I do i'd like to discuss the phantom blocking circuitry a little.

I've done a search on this and read quite a bit. Seems it boils down to these options:

1) "Menace" configuration (from the AES paper). This is shown in the 15XX development board docs.
It uses rather high 100k input bias resistors that allow a small (2.2uF) non polarized film blocking cap.
I like the idea of no electrolytics, but I wonder about current noise at very low frequencies and improper loading of some dynamic and ribbon mics.

Not familiar with "Menace" but understand the concept. Film "should" be better than electrolytic, all things equal.. but are all things ever equal?

The proper termination for mics can be dealt with by adding shunt resistor on mic side of the blocking caps. There are two issues with relatively small blocking cap values. Rising impedance at low frequency (perhaps noise issue), and sensitivity to matching for CMRR at hum frequencies. Perhaps add a LF trim for CMRR but that might imbalance DC offset..

One would need to breadboard it up to see if 1/F and/or LF noise of devices used is problematic.

2) polarized 47uF phantom blocking caps with 1K5 bias resistors. This is pretty much standard I guess, but I sure don't like the idea of this when phantom is off...unless it was bypassed. An alternative would be to leave phantom on with dynamics etc so the cap would be polarized.

If this causes a problem it should be measurable on a bench. No reason to bias at full 48V under non-phantom applications. This has been widely used with good results. One not so obvious data point, not all electrolytic are created equal. In addition putting most of the mic termination impedance on the mic side, means the caps don't have to work as hard. Less AC current means less expression of several non ideal capacitor terms.

3) As #2 but with giant, expensive 47uF NP film capacitors. Or a somewhat lower value with higher bias resistors (much less than 100k).

Too f'n big... Not sure I even like 2uF film antennas hanging off my inputs.

4) Non polarized electrolytics. Seems a bad plan to me, given what i've read about their electrical characteristics.

why?  Just add two more poles to phantom power switch and bypass input caps when not needed.

5) Flying rail with no blocking caps. I read about this, but it seemed to not go anywhere.

I played with this idea some time ago (discrete-mixed) but never reduced it to a production design, Wayne has done of lot of work more recently, (using THAT chip set) here http://www.picocompressorforum.com/forum/php/viewtopic.php?f=6&t=14

So, before I finish my bill of materials I would like to ask for opinions on these options.

Les
L M Watts Technology

You will get more opinions that actionable advice on this subject. It depends on what you wish to optimize, or which demons you are avoiding hardest.

JR

 

[leswatts]

Hi John. I have read many of your posts.

Well, I do have to balance LF noise and CMRR against electrolytic bad stuff.

In general I don't like to use unbiased electrolytics anywhere, even though they might survive up to a reverse bias of about 1.5v. With drum dynamic mics that can be approached, FWIW.

I'll admit that I am "biased" against non polarized lytics simply because I've heard nasties from them.
That's an unscientific conclusion of course. Lots of them in my machine shop for motor start/run and power factor correction though. They blow up real good.... and regularly too!

With the "menace" circuit I calculate about 200 nV/sqrt Hz noise from the high bias resistors, but only
well below 1/rc. That's high...but at such a low frequency. Have to think about that.

The distortion from various mechanisms with modern electrolytics seems very low as long as the reactance+ ESR << load . Reading up on that.

Right now I'm kinda liking the idea of keeping a small bias potential for the lytic even with phantom "off".
This will kinda depend on leakage of the cap, and how much current a dynamic or ribbon can take in case of a wiring fault. From what I read, anything over 1.5v will keep the cap well formed, or at least not in a degrading condition.

BTW I was looking at your drum stuff. I bought house drums for my studio a while back, and got fairly heavy into tuning/playing. Well, trying anyway. You know what I found? Drummer clients often seem to not know or care much about tuning...but I find it super important to achieving a sound. Actually these preamps i'm rolling are mostly for drum mics!

Anyway I have some time to look at this. I have a specialized microphone design project (about 2 years in) for automotive that has been put on hold for obvious reasons. So I can play with preamps.

Les
L M Watts Technology

 

[JohnRoberts]

Hi John.

With the "menace" circuit I calculate about 200 nV/sqrt Hz noise from the high bias resistors, but only
well below 1/rc. That's high...but at such a low frequency. Have to think about that.

I suggest breadboard and listen to one. I recall a selected for low noise part I used to use, and the part was selected for the spectral balance of the noise floor, not the total noise. Excessive 1/F noise sounds "bad" where whiter or pink noise is not as objectionable.  Our hearing response curves will discriminate against very low frequency. 

The distortion from various mechanisms with modern electrolytics seems very low as long as the reactance+ ESR << load . Reading up on that.

+1.. light loading means less badness

BTW I was looking at your drum stuff. I bought house drums for my studio a while back, and got fairly heavy into tuning/playing. Well, trying anyway. You know what I found? Drummer clients often seem to not know or care much about tuning...but I find it super important to achieving a sound. Actually these preamps i'm rolling are mostly for drum mics!

yup, in the hierarchy of drummers budgets a precision tuner is not high priority.

In studio, not only is tone important, but the same tone from takes recorded today, and takes recorded next week or month.

Les
L M Watts Technology

 

[jdbakker]

In general I don't like to use unbiased electrolytics anywhere, even though they might survive up to a reverse bias of about 1.5v. With drum dynamic mics that can be approached, FWIW.

If you're willing to trade headroom for cap bias you can always bias the 1512-end of the phantom decoupling caps with a negative voltage (as opposed to ground). This is admittedly easier to do (ie: without as much impact on headroom) in a (discrete) implementation of the THAT1512 circuit with a PNP input.

EDIT: Or just shift the supply voltage of the 1512 a corresponding amount. Say 1512 V+ = 10V, input biased to -5V, 1512 V- = -20V. Now you're faced with either a DC offset or reduced headroom at the 1512's output, but that can be fixed too (although not by my brain at 4.30AM, not elegantly that is).

JD 'it's 4:30AM, do you know where your brain is?' B.

 

[squib]

if concerned about distortion in capacitors, then you need to find and read Cyril Batemans series on article in Electronics and Wireless World from a number of years ago. This is the only detained scientific examination of capacitor distortion i've ever come across.

It is illluminating!

The short of it all is that polypropylene capacitors are the best performing. But at the capacitance you require will be physically largish. If worried about this, then shield them. For larger capacitance values the next best performing capacitors are bipolar electrolytics. Nichion brand specifically performed the best. The performance of a good bipolar electro will meet or exceed the distortion levels of the active electronics that follow it.

Should definitely place your chosen Rload on the input side of the capacitor, between hot and cold, keeping the Rbias on the electronics side of the capacitor from each leg to ground to as high a value as possible whilst maintaining the bias of the input stage. This improves the CMRR and keeps the value of the electro to a lower value to achieve a good low freq response and the low I/F noise performance.

Capacitors are not necessarily evil things!!

 

[leswatts]

Thanks for the responses so far!

I suggest breadboard and listen to one. I recall a selected for low noise part I used to use, and the part was selected for the spectral balance of the noise floor, not the total noise. Excessive 1/F noise sounds "bad" where whiter or pink noise is not as objectionable.  Our hearing response curves will discriminate against very low frequency. 

Yeah... I don't expect to hear the noise at a few hertz directly, but I would expect a loss of dynamic range. The "menace" noise spectral density numbers I calculated seemed alarmingly high, but then the bandwidth is very low. Still, it could mean several db degradation in DR I would imagine.

Should definitely place your chosen Rload on the input side of the capacitor, between hot and cold, keeping the Rbias on the electronics side of the capacitor from each leg to ground to as high a value as possible whilst maintaining the bias of the input stage.

Yes, Squib, this makes sense to me. The THAT demo board has a load resistor on the mic side, but peculiarly it's marked as not used, so the dif input impedance would be in the 12k range (from the phantom resistors). As a former Shure development engineer I can tell you that mics like the SM57 would not like loads that high. I think Ernie Seeler said he designed those things for 600 ohms. They are load sensitive.

If you're willing to trade headroom for cap bias you can always bias the 1512-end of the phantom decoupling caps with a negative voltage (as opposed to ground). This is admittedly easier to do (ie: without as much impact on headroom) in a (discrete) implementation of the THAT1512 circuit with a PNP input.

I'll have to think about that a bit, JD. In general I wouldn't want to lose much dynamic range...at high gains the mic pre is already the most limiting thing in the whole recording chain.

I did take note of the SSL always on electrolytic bias mentioned in the flying rail thread. That is two caps aranged -++- with +48 always applied to the center. But it seems to me the first cap would just be floating in the breeze with no bias unless phantom was on and some current was being drawn... or something else providing a little common mode conductance on the mic side.

So far my thoughts are that if the no electrolytic "menace" circuit costs me 6db dynamic range or something...I won't use it. But if I use a lytic I want it properly biased phantom on or off.

I'm finding this discussion very helpful.

Les
L M Watts Technology

 

[JohnRoberts]

Yeah... I don't expect to hear the noise at a few hertz directly, but I would expect a loss of dynamic range. The "menace" noise spectral density numbers I calculated seemed alarmingly high, but then the bandwidth is very low. Still, it could mean several db degradation in DR I would imagine.

?? audible dynamic range or calculated/measured dynamic range?

Yes, Squib, this makes sense to me. The THAT demo board has a load resistor on the mic side, but peculiarly it's marked as not used, so the dif input impedance would be in the 12k range (from the phantom resistors). As a former Shure development engineer I can tell you that mics like the SM57 would not like loads that high. I think Ernie Seeler said he designed those things for 600 ohms. They are load sensitive.

Les
L M Watts Technology

:  Funny how esoteric mic preamp designers play games with the input termination to sound "different". Can't charge more if you sound the same..  They ignore that mics are designed for some expected nominal termination.

JR

 

[leswatts]

?? audible dynamic range or calculated/measured dynamic range?

Perhaps both, John. Although inaudible by itself. a subsonic noise signal would still affect audible dynamic range I think. Imagine a 1 hz sine signal at 0 db FS peak into a converter. If an additional signal at higher frequency were added, it would clip half cycles completely twice a second , right?

This might imply that configurations like the "menace" that have high VLF noise have subsequent high pass filters later to remove it. I think it's a good argument to have input 1/rc very very low too.

Les
L M Watts technology

 

[JohnRoberts]

An interesting hypothetical but not remotely what you will experience with real circuitry.

I would be inclined to listen to it. In my experience the spectral balance of the noise floor matters.  My concern is perceived noise floor quality impression if very low frequency noise is high enough to be audible over mid band noise where our ears are more sensitive.

JR

 

[PRR]

> unbiased electrolytics ...might survive up to a reverse bias of about 1.5v. With drum dynamic mics that can be approached, FWIW.

No. If the R-C product is below the lowest signal frequency, the voltage -across- the cap is very nearly zero. (At -3dB point it is 0.3 times the input at that point; at -1dB it is 0.1, etc.)

With your BJT inputs, even this slight impedance impacts your current noise, so you tend to have the R-C product very-far below your lowest signal frequency.

"Signal frequency" is vague. LPs are not cut below 20Hz, but a phono can have massive 0.55Hz rumble. This is unlikely on studio drums. The mechanics of drum sound severely limits low frequency output (unless the drum is objectionably huge). Also you say "dynamic mike"; these have a trade-off between bandwidth and sensitivity, so you usually get exactly so-much bass and then a 6 or 12dB/oct fall-off.

> I actually have heard distortion in NP electrolytic caps in speaker crossovers

Different. These caps have a real LOAD, and the R-C product is smack in the middle of the audio range. Load gently, R-C far off to the left of anything which can come in, you are sweet. OK, electrolytics ARE nasty chemical contraptions, and they do get weak welds, scuzzy oxide, spewed juice. But modern ones have an -excellent- record of serving well, including the exact application you plan.

1/f noise: -most- rooms, including some fine recording rooms, have rising LF noise due to ventilation. With the best intentions: our concert hall was consulted by one of the old-time big-name acoustics firms, blower ducting and muffling was clearly spelled out, but a sequence of contractor changes and a concrete blow-out gives me 40Hz rumble louder than a harpsichord's main voice. A 1736 stone church with no blowers passes enough rumble from highway trucks to be apparent behind pipe organ. There's all different situations and I'm sure there are rooms where the 1/f of a good mike-preamp matters.

I'm also wondering how 1/f matters when you have Volts of drums.... except I -did- do a session which alternated 118dB SPL bursts with long passages near blower-residual.

> if very low frequency noise is high enough to be audible over mid band

One of our musicians pointed out that ensemble players are constantly aware of beat-tones, indicating slight out-of-tune problems. The huge subsonic noise in our concert hall not only masks that, it leaves them constantly uncertain as one or another random subsonic tone catches their ear.

> Imagine a 1 hz sine signal at 0 db FS peak into a converter. If an additional signal at higher frequency were added, it would clip half cycles completely twice a second, right?

Sure. And even below clipping, XX% simple distortion would give some degree of random intermodulation distortion. However your preamps can be running 0.00X% distortion, with even lower random IM. (This is true for electric noise or for blower noise, tho blower rumble can be reduced by mike fall-off.)

> ...configurations ...that have high VLF noise have subsequent high pass filters later to remove it. I think it's a good argument to have input 1/rc very very low too.

Input R-C is a compromise between passing signal, blocking garbage, and increasing garbage (current noise). Where signal level is very small, as mikes usually are (Volts of drums is borderline), and current-noise is significant, you probably want to let-in a lot of subsonic, gain it up, then high-pass to 40Hz or whatever you feel you need (there's thud to 20Hz but none of your customers have speakers to feel it). Adjustable hi-pass is of course a useful scalpel, but best reserved for mix-down.

You are over-thinking. Because you can. That won't kill you, but can paralyze. Drummer clients won't notice a 7dB noise-figure at 50Hz, like most of them won't notice their tuning is like a watermelon down a bowling alley. Mike placement and mixing (and performance) is what sells a million copies, and a trace of subsonic IM won't hurt sales.

 

[tv]

What I post is totally "from top of my head".

a) thinking about noise - modern production has def. tighter requirement for technically good input - per channel limiting, processing - noise can surface ...
b) I see no reason to NOT just get "inspired" by "menace" 100k/2,2uF config - why not use 47k/2x2,2uF paralleled? or 33k/3x2,2uF paralleled? ETC.

Imo one could find 2,2uF wimas or similar easily.
Wouldn't 63V parts suffice?

Why is the R11 in that's 1500demo.pdf omitted? Any special reason?

 

[JohnRoberts]

"Signal frequency" is vague. LPs are not cut below 20Hz, but a phono can have massive 0.55Hz rumble. This is unlikely on studio drums. The mechanics of drum sound severely limits low frequency output (unless the drum is objectionably huge). Also you say "dynamic mike"; these have a trade-off between bandwidth and sensitivity, so you usually get exactly so-much bass and then a 6 or 12dB/oct fall-off.

While orchestral drums, like tympani and to some extent toms are pretty easy lifting from a LF bandwidth POV, kick drums if you try to capture the "thump" can get pretty low. You are not reading just the head resonance note but also some piston action from the beater smacking the batter. Your LF cut off in that case may be limited by the microphone more than the drum, while you will get a sense of drum size from the after strike decay. 

Timpani are easier than you might expect from their apparent low voice, since they are in fact resonating at two early overtones above the base note, and we are mentally filling in the missing or phantom fundamental.  So when recording you will hear a note lower than what you are actually printing.

It seems this phantom note principle would work in loudspeaker playback to extend the apparent response lower than actual, and some of those bass effects boxes may do this, on purpose or by accident. 

JR

 

[leswatts]

Wow, great posts. Lots to think about.

If the R-C product is below the lowest signal frequency, the voltage -across- the cap is very nearly zero. (At -3dB point it is 0.3 times the input at that point; at -1dB it is 0.1, etc.)

Yes, that's right. Very good point.

Why is the R11 in that's 1500demo.pdf omitted? Any special reason?

No Idea, tv. As previously discussed mic loading matters. And yes, higher cap and lower bias resistor is fair game. It would have to be a big change to matter a lot. I calculate about 6 db increase in noise
for the 100k bias R. As far as 63V...yeah used a lot I guess. I would tend to go to 100V. Just belts and suspenders. A little extra for surges, misplugging, etc. But some size cost.  Also better antenna.

After reading this I went in the studio and took a look at raw recorded mic signal FFTs just to see what was there, particularly with the dynamics. While some of my condensers and ribbons have substantial LF response, the usual dynamics do not.

In the kick drum I had a Shure sm76 omni dynamic. It has an unusually smooth response, and pretty good LF. A version of it was even sold as an analyzer mic.  I found very strong response down to about 30 hz. This was inside the kick... tuned very near the wrinkle point.

Snare and toms were Shure unidyne 546s... Kind of an earlier version of an SM57, but with a copper voice coil rather than aluminum. Strong LF to about 50 hz with floor tom. I think the huge proximity effect from being a couple inches off the head contributes. Also I would think it's in a very reactive field and picks up the fundamental.

You are over-thinking. Because you can.

Ha...had to answer that one, PRR. GUILTY AS CHARGED. As a research engineer I overthink things for a living! 

Well, here's where i'm at... In spite of not being particularly fond of putting those little cans of goo in analog signal paths, I think in this case they could be pretty transparent. I still don't like the idea of an unbiased lytic in any circuit though.

For sure i'm not liking the current noise I calculate for the "menace" circuit. Does it really matter? Not sure at this point, but no reason not to build it up and see (and hear). I had not thought about it much before, because I usually don't build studio mic preamps. I'll admit that I like good numbers in specs though, if they're real. If lytics do nothing bad why add current noise to eliminate them?

Les

 

[JohnRoberts]

I'll admit that I like good numbers in specs though, if they're real. If lytics do nothing bad why add current noise to eliminate them?

Les

Like I said in my very first response, "It depends on what you wish to optimize, or which demons you are avoiding hardest. ". Film caps are better than electrolytic in the extremes. People tend to focus on things they (think they) understand, and like to throw money at circuit upgrades, without rigorous analysis, or sharp pencil budgeting.

I have a capacitor free front end design kicking around in the back of my head... not because I believe it is needed, but because it is arguably better. The thing missing from most of this esoteric mic preamp design discussion, is how good the SOTA already is. If you are familiar with the non-ideal behaviors in microphones, you can appreciate how much more closely mic preamps are to a wire with gain, than either end of the audio chain.


JR

 

[leswatts]

I have a capacitor free front end design kicking around in the back of my head... not because I believe it is needed, but because it is arguably better.

John, it would be one heck of a marketing hook. Well worth doing if for that reason only. Yes, arguably better too.

I'm going to go ahead and order parts. I'll try lytics (biased and unbiased) and film. I'll also try dedicated dynamic/ ribbon channels with no caps or phantom. I'll also consider a phantom sytem where the entire front end is switched... one with phantom and one without phantom or caps. Should be doable as long as I don't intermittently float the 1512 inputs. I can keep a high bias resistor always connected, but shunt it with a lower one in the non phantom front end.


Les

 

[JohnRoberts]

Funny one of my concerns about going DC coupled at the mic preamp  for mass market production, is that the blocking caps will insulate the preamp from the ills of mismatched microphone electronics DC output. If you design a "better" mic preamp that for some reason doesn't work right with popular microphones that work fine in everybody else's console, it becomes your fault not the flaky microphone's, that arguably isn't flaky if used as intended. 

Since I wouldn't use a blocking cap in the gain circuit either (that is actually a bigger problem than the input) , we are looking at dc gains of 1000x. At a minimum this would need some servo action, maybe more than one. (I know servos use caps, but good caps). I doubt phantom powered mic designers anticipate being DC coupled at their outputs.

yes, it is mostly a marketing hook...  part two is to feed directly into A/D convertor that is also floating up at voltage, and then level shift the digital result using opto coupler or whatever..

JR

 

[olafmatt]

5) Flying rail with no blocking caps. I read about this, but it seemed to not go anywhere.

Been there, done that... Actually it works pretty well. Used an opamp to create a 'differential virtual ground' (that only looks like ground to common mode signals, but not so to the differential audio signal). That way you can burn away the phantom offset after the 'flying' input stage. Problem is too keep everything low impedance you need high-power resistors. I did it with 10k (higher than I would usually use), which means 4.8mA DC flowing through a resistor that also carries your audio. Thats about 0.25 Watt to burn. Alternatively, try a zener diode in series to drop the offset down to something the next stage can handle.

There is one reason that speaks against doing it: you will need a filter after the micpre to filter out all the low-frequency rubbish. With the mics shock-mounted on a regular mic-stand I could see passing trains in the wave display of the DAW. And the train line is about 1km away. So unless you come up with a capacitor-free low-cut as well the whole effort is useless.

Olaf