A Direct-Coupled Input-Capacitorless Active Preamp deleted

[PRR]

I thought I posted this a while back, but didn't. If it seems out of context, so be it.

> what exactly does Prodigy* or the manufacturers who watch this forum do with our designs?

I have no clue, and doubt that Prodigy Pro "watches" the forum. (Yes, Ethan peers in, and he is a great server admin, but he is working on his audio understanding.) And if you look at P-Pro's business, it is mostly "selling boxes". They have a website, they take orders, boxes come in from suppliers and go out to customers. Yeah, their President does have a mike-amp project and sells it through the store, and may glance in here on dull days, but he's not "harvesting".

> Show me one.

What, work? Not this week. I work at a school and the semester just started. A clever professor moved his printer and wants me to pull a USB cord out of my rear. Another lost his Address Book. Again....

But as Concept: take the THAT Corp part, run on 0V and +40V. The common-mode input can be +2V to +38V, limited by the 40V process. Use that plan but with 60V parts working off -5V/+55V rails, it should take any common-mode voltage found on P48 systems. If the REF pin is tied to load ground, the output is ground referenced.

Hollow-state devices are hot and noisy but can manage huge input DC.

Of course the other problem is the DC offset. And the common cure is the 4,700uFd cap between FB inputs, of awkward size and dubious sonic integrity. It should be possible to synthesize that with a 0.1uFd cap and couple of op-amps which will not be working at audio rate.

Hmmmm.... it isn't that simple, is it? The "fake cap" must be super low noise.

And I do think it should do a Right Thing, or not a Wrong Thing, when one pin gets grounded. That could mean 48V of "offset".

Here's my reductio ad absurdum. You have small audio signal mixed with large and unknown DC values. You must amplify the signal. You can't amplify the DC hardly at all in the forward path. Amplifying DC in the servo wants to inject audio noise. One way or another, you need a part that is "short" for audio and "open" for DC (or vise-versa).

There is perhaps no other place in the audio studio where we have so much uncontrolled/unknown DC mixed with so little audio.

 

[ruairioflaherty]

I read through all of this thread tonight - fascinating stuff. It reminded me of some recent thoughts, has anyone ever built a self powered solid state microphone? It seems to me that if you had a local DC supply from the mains you could avoid the phantom blocking caps as well as giving the mic a much beefier supply.

The external supply would appear to be a bit of a pain but in reality it's never an issue with tube mics. In fact with that beefier supply one might consider bringing the signal to line level at the mic also - either in the mic or at a floor mounted pre with remote gain control.

I'm sure this has been shot down before but I wouldn't know where to search.

Cheers,
Ruairi

 

[Gus]

The circuit getting to have a lot of parts. It is a cool circuit active CMRR etc..

Question and I mean this in a very good way. How is this better than using polypro input caps with a well thought out preamp circuit?

Poly pros are big but you can get them up to 100uf maybe more

 

[Svart]

I think this is more of "can it be done/proof of concept" design but it's definately cool as hell watching Wayne knock this one out of the park.

 

[JohnRoberts]

[quote author="mediatechnology"]I've been thinking about it and I think the Vactrols will need to be in the lower legs of the bridge to have the CMRR compensation effect.

The Vactrols can null the DC offset located in either the upper arms or the lower legs. But to compensate for an internal phantom takeoff resistor that is slightly too high in value, for example, the Vactrol would need to be in parallel with it to lower it. By doing this, the parallel combination (AC termination) of phantom pullup, phantom takeoff (internal to the mic) and LDR will equalize across the left and right sides of the bridge when DC is brought into balance. Thus, the tip/ring source impedances of the mic (not shown) will be equally loaded.

The LDRs in either position, either arms or legs, should be of such a low percentage of (or high multiple of) the actual resistor they parallel that their current noise contribution should be small.

I shall try it with the LDRs on the legs instead of the arms as well. It may be that this secondary benefit of CMRR tuning isn't that great but it's still worth trying.[/quote]

I'm not sure if balancing input offset this way will help or hurt CMR. In the extreme, a chip with lots of dc offset may imbalance the input impedance to achieve DC null.

I believe I recall mics (probably not many) that used a center tapped output winding, so there are no external resistors to balance in that case. FWIW I designed a studio direct box years ago and it did use the conventional resistor pair to grab phantom power.

Good luck. The CMRR should be pretty straightforward to confirm. You could probably create an intentional dc input offset by changing the current density of just one of the two input devices.

LF CMRR will probably be much superior to cap coupled inputs due to difficulty in getting precision large capacitors. The best cap is no cap, in this case.

JR

 

[JohnRoberts]

Yeah I'm not one to be anti-cap or pro-servo. The first thing the 1510 hits is a film cap on its' back-side. But it just seems like having them at the input of a mic preamp is one place where we should try to eliminate them. Particularly when you can draw sparks with them. :shock:

I'm just very curious about what a preamp sounds like without any reactive components in the input. (Neglecting shunt RFI filter.) We can't do anything about what's internal to the mic reactively, but the goal is to have a "straight wire" from the XLR to the input bases. (OK so there's 4R7 fault current-limiting Rs there but virtually a straight wire.)

Not to be overly contrary the issues with stored charge of input capacitors has been fairly well managed with discrete input clamps. A subtle problem related to this particular fault vector is input devices allowed to inadvertently zener their b-e junction can get noisy. Often they will self anneal over time but still a condition to avoid in high performance discrete designs. I notice the 1510 has anti-zener diodes integrated on chip across their input devices so not a problem here. I guess the on chip input clamps may have difficulty taking the main hit of overly large blocking caps while also maintaining low path impedance.

I am a strong advocate of making individual signal blocks as linear as possible, and let the chips fall where they may (no pun intended) regarding the rest of the path but it may be worth observing that the vast majority of microphones may not be up to this pristine standard.

Of course always a worthwhile pursuit (to not be the weak link). Keep up the good fight.

JR

 

[bcarso]

[quote author="JohnRoberts"] .....A subtle problem related to this particular fault vector is input devices allowed to inadvertently zener their b-e junction can get noisy. Often they will self anneal over time but still a condition to avoid in high performance discrete designs. .....

Of course always a worthwhile pursuit (to not be the weak link). Keep up the good fight.

JR[/quote]

I was amused when I read (was it Motchenbacher?) that you could fix Q's that had been damaged with reverse breakdown by baking them at 500ºC for x hours. Not too practical. I had never heard of self-annealing under normal conditions though.

One thing that comes to mind: the bipolar damage problem might be dealt with using a complementary front end, arranged so there is always one b-e junction being forward-biased in an overload. Clearly there has to be someplace for that current to go as well, but it might push the clamping circuit further into the interstices where perhaps it would have less of a sonic effect (given that we're worried about it to begin with).

 

[JohnRoberts]

[quote author="mediatechnology"]John I've sent you a copy of THAT's AES paper "The Phantom Menace."* They pretty much do require outboard diodes as well around the 1510. Another concern addressed in the paper is not only the damage they do to the preamp device, but what they might accidentally get connected to. I've had to replace a lot of output coupling caps and op amps in devices that got accidentally patched into a "live" mic input. Outputs should also have diode clamps to the rails. Most don't.

When I was commisioning SSL 4K and 6Ks the 82E149 active preamp, on some consoles, was a real PITA compared to the transformer-coupled '02. Ditto for Harrisons. I carried tons of LM394's 'cause they always seemed to get noisy and took forever (in the case of the Harrison) for the electrolytics to get quiet after phantom was turned on. For twnety years I've been telling myself that there's got to be a better way to do this.

I agree that most mics may not be up to the pristine standard. But let them continue to be the weakest link and with a more direct signal path hear them more as they sound. It's pure curiosity on my part.

* The AES is selling this on their website for $20 to non-members. THAT was kind enough to send me this at NC and for those reasons I'm reluctant to post it. If anyone else wants to share their copy please do so.[/quote]

Thanx for the paper... I'm a member of AES but frankly don't keep up on the new stuff (same as old stuff). It looks like THAT explored sundry popular approaches to input clamps. There's one I didn't see them test consisting of two back to back transistors connected c-e with bases n/c, connected as a shunt to ground. Looks like a zener with a diode drop in series. I don't know if they would be any more robust than the other approaches but I've used them and seen them used without major problems.

Indeed outputs need to be protected. It was remarkable in the early days of wireless mics how many receivers were not protected despite being intended for use plugged in there. :roll:


JR

 

[bcarso]

[quote author="mediatechnology"]Brad; Neat trick. We can put it in the oven with the tape (at a lower temp for the tape I suppose.)

Why not just eliminate the charge resevoir altogether? That's my strategy.[/quote]

Employee at the office of transistorepair.com: "Sir, here's your part. It....it's much lower noise now, as good as the day it was made."

Customer, squinting down at an apparently empty small container: "What? What? Where is it??"

Employee: "Do you see that little speck in the corner there, about the size of a large bit of coarse-ground pepper?"

Customer: "WHAT?? THAT's my transistor?"

Employee: "Well, yes it is sir. The case...I'm afraid the case didn't make it. I can assure you that the case is in no pain. However, we can re-encapsulate your transistor for you for an additional fee---Mel, would you please deal with this customer---Sir, step back---SECURITY!!"

 

[JohnRoberts]

[quote author="bcarso"]

I was amused when I read (was it Motchenbacher?) that you could fix Q's that had been damaged with reverse breakdown by baking them at 500ºC for x hours. Not too practical. I had never heard of self-annealing under normal conditions though.

One thing that comes to mind: the bipolar damage problem might be dealt with using a complementary front end, arranged so there is always one b-e junction being forward-biased in an overload. Clearly there has to be someplace for that current to go as well, but it might push the clamping circuit further into the interstices where perhaps it would have less of a sonic effect (given that we're worried about it to begin with).[/quote]

The annealing (per Motchenbacher and Fitchen) can be from temperature or forward current, which probably just causes local heating. The self-annealing at normal operating currents is a personal speculation of mine in an attempt to explain the audiophile hokum about some gear sounding better after it burns-in over time and such. I have never personally observed this, so in hindsight I shouldn't offer it casually as fact.

I guess I don't understand your comment wrt protection. The mechanism for an input device zenering involves one of the pair forward biased with the other in reverse avalanche. Simple small signal diode clamps b-e prevent this and AFAIK are sonically benign.

JR

 

[bcarso]

[quote author="JohnRoberts"][quote author="bcarso"]

I was amused when I read (was it Motchenbacher?) that you could fix Q's that had been damaged with reverse breakdown by baking them at 500ºC for x hours. Not too practical. I had never heard of self-annealing under normal conditions though.

One thing that comes to mind: the bipolar damage problem might be dealt with using a complementary front end, arranged so there is always one b-e junction being forward-biased in an overload. Clearly there has to be someplace for that current to go as well, but it might push the clamping circuit further into the interstices where perhaps it would have less of a sonic effect (given that we're worried about it to begin with).[/quote]

The annealing (per Motchenbacher and Fitchen) can be from temperature or forward current, which probably just causes local heating. The self-annealing at normal operating currents is a personal speculation of mine in an attempt to explain the audiophile hokum about some gear sounding better after it burns-in over time and such. I have never personally observed this, so in hindsight I shouldn't offer it casually as fact.

I guess I don't understand your comment wrt protection. The mechanism for an input device zenering involves one of the pair forward biased with the other in reverse avalanche. Simple small signal diode clamps b-e prevent this and AFAIK are sonically benign.

JR[/quote]


What I mean is if the input is a dual differential pair, PNPs and NPNs, at least one of the b-e junctions is forward-biased in overlaod. The the clamping can be further along the path and not loading the input as much, maybe.

I agree that 4148-ish clamps are pretty benign---we're just pursuing refinements here, possibly approaching medieval scholasticism in the process.

Interesting POV wrt the burn-in improvement reports. I tend to suspect psychoacoustic/perceptual processing at work---familiarity breeding acceptance, unless something is seriously wrong, like the artifacts of badly done digital, in which case then the fatigue factor is invoked, rather than accounting for things degrading with burn-in.

But surely there are some aging effects at work, especially with electrolytics being among the population inside the box. And film caps are not so perfect either, and core material ages in inductors too.

 

[bcarso]

[quote author="mediatechnology"]The result: The CMR error of the microphone itself is corrected.

All for free.

[/quote]

Sounds like a good idea.

 

[JohnRoberts]

congrats... it's nice when a plan comes together...

JR

 

[gkhebert]

Hi Wayne,

One concern I have with this version using the Vactrols is that You are now unbalancing the AC common-mode input impedance as you correct for DC offsets. One can certainly imagine a condensor microphone with a pair of slightly mismatched resistors used to pick off the phantom voltage, and an ac-coupled amplifier with some relatively low, but finite, output resistance to supply the audio signal. The "adjustment" to your input bridge required to correct the DC conditions would not necessarily be the same as that required to maintain CMRR in the audio band.

One can concoct a similar scenario for a transformer-coupled microphone output that picks the phantom voltage off via a center tap. The DC resistance match of the two halves of the winding doesn't necessarily match the ac CM output impedances.

Am I missing something here?

Regards,
Gary[/quote]

 

[JohnRoberts]

[quote author="mediatechnology"]Gary;

Thanks for your comments.

I may be wrong about this but that's why I moved the Vactrols to the lower bridge "legs" rather than having them in the upper "arms" as in my previous post. From a DC perspective the offset compensation results are the same. But I think in the upper arms of the bridge it does imbalance source loading and actually makes it worse as you suggest. In the lower bridge arms my thinking is that it will make it better than the microphone could be on it's own with perfectly balanced preamp loading.

But I may have it backwards...See if my logic makes sense:

Let's assume that we have a really bad microphone where one phantom pick-off resistor is a "perfect" 10K and the other an out-of-tolerance "9K." This produces a large offset. So the servo corrects it by lowering the "good" 10K pick-off resistor to 9K via the Vactrol LDR (it's essentially in parallel with it from an AC perspective) in order to restore DC bridge balance. So after correction, both tip and ring are each loaded with 9K to AC ground.

This really bad microphone will create it's own imbalance even if perfectly loaded. But corrected resistively by sensing the DC offset, the AC loading would also seem to be corrected. I think this may be an equivalent circuit:



Am I missing something?[/quote]

Your choice sounds valid based on the given premise, i.e. unbalanced load resistors.

I have always found myself inadequate to anticipate all the possible ways customers could misuse equipment, hopefully design engineers are less creative in their design mistakes but I'm not overly optimistic.

While it's impossible to correct for badly designed gear, it is important to at least do no harm when interfacing with good gear. The best I can figure from their data sheet the 1510 specs something like +/- 5 mV input offset voltage. This would be added to input offset of servo amp, but if I read that right as an OPA27 the offset voltage should be minimal but even typical input bias and offset current times 100K input resistors is 2.5 mV, worst case a lot more.

So one question is how much does correcting for the 1510 and servo imbalance the input with a perfect mic source?

Perhaps an alternate approach is applying a delta current instead of using a shunt impedance to apply correction but without perfect knowledge of what might get plugged in there who knows?

JR

 

[JohnRoberts]

[quote author="mediatechnology"]With a "perfect" mic source there wouldn't be much offset to correct other than the 1510's input Vos so I think the Vactrols would tend to be in a high resistance condition providing very fine degrees of Vos trimming with Phantom on. (BTW: Only one LED can be illuminated at one time.) But, the servo would try to drive one of them. Phantom being on provides a current source for correction - with phantom off the servo has no means to correct and is open loop. The servo exists only due to phantom and uneven I*R drop due to loading imbalance.

I honestly think I should add a switch pole to open the LED drive current when phantom is off. Without phantom on the LDRs can't provide much if anything in the way of correction. [Or maybe the LDR common, now ground, could be tied to a very low (negative?) potential to provide 1510 input Vos correction.] But I don't think 1510 input Vos correction is even needed with phantom off using reasonably low 6K81 input bias Rs. With the LEDs not driven the LDRs recover to about 10M dark state so they produce insignificant CMR imbalance.

I suppose that the "perfect" phantom-powered mic could be characterized as having an output transformer with low DCR between tip and ring. In this case there would be a Q-point potential for them to provide 1510 input Vos correction but there would be little, if any mic Vos to correct.

So, if phantom is on the Vactrols have something to work with. With phantom off they don't and aren't really needed. In that condition the LEDs shouldn't be driven.[/quote]

Perhaps I wasn't clear.

1510 specs (+/- 5mV) + OPA 27 bias current worst case (+/- 80 nA X 100k= +/- 8 mV (each input?) + offset current (75 nA X 100k (200k?)= 7.5mV. OPA27 offset voltage is insignificant

So worst case physical input could be driven to a differential voltage of 20 mV to make servo happy (or 35mV if OPA27 offsets are per leg). If mic shunt is low resistance might this not imbalance the input?

Admittedly these are worst case numbers but my experience with production is if you ignore published tolerance bands they will reach out and touch you..

JR

 

[JohnRoberts]

[quote author="mediatechnology"]

John: What tolerance resistors did Peavy use for phantom pullup and input bias Rs? Doubt they were 0.1%....

[/quote]

I left Peavey several years ago and can't speak for what they are doing today. There were many different mic inputs with phantom power ranging in price and even the phantom voltage used with inexpensive install amps using maybe 15-20V instead of standard 48V.

My recollection was the better (i.e. more expensive ) products used 1% for the phantom power resistors and some of the premium mic preamp channels used 0.1% resistors in some of the more critical locations. The tolerance of the lower resistance termination and how it is configured will be somewhat more influential on balance than phantom resistors. For example using highest values possible on individual + and - terminals with a shunt between them to bring it down to nominal 2k or so will improve CM matching. Of course in a capacitor less design if the 6.8k resistors are the only resistors providing a DC path for the input they become the primary balance factor. I don’t recall all the details but there were a handful of 0.1% values in the (Peavey) system for use in critical front ends.

Peavey was not exactly the standard bearer for tweaked out mic preamps, and back in the day I would sometimes relax my targets based on the nominal CMRR of twisted pair shielded cable (IIRC maybe 60 DB). The more popular use of "Star-quad" or similar variants these days brings the typical wire CMRR up a notch or two.

I am not trying to make a bigger deal of this than it deserves. The beauty of eliminating those big honker electrolytic capacitors from the input is you eliminate multiple sources of CMRR error (C tolerance at LF, ESL, ESR at HF), not to mention noise from leakage currents (not common but I saw it in one capacitor series we had to remove as a approved part in that application. Note: more C more leakage.).

I am viewing this exercise as a “how good can it be done”, rather than a practical “what is good enough”. I believe there is much merit in the capacitor less approach and mean for my critiques to be supportive overall.

JR

 

[gkhebert]

Gary;

Thanks for your comments.

I may be wrong about this but that's why I moved the Vactrols to the lower bridge "legs" rather than having them in the upper "arms" as in my previous post. From a DC perspective the offset compensation results are the same. But I think in the upper arms of the bridge it does imbalance source loading and actually makes CMR worse as you suggest. In the lower bridge legs my thinking is that it will make CMR better than the microphone could be on its' own even without preamp loading. So yes the preamp input R is imbalanced but complimentary to the mic's self-induced CMR error.

But I may have it backwards...See if my logic makes sense:

Let's assume that we have a really bad microphone where one phantom pick-off resistor is a "perfect" 10K and the other an out-of-tolerance "9K." This produces a large offset. So the servo corrects it by lowering the "good" 10K pick-off resistor to 9K via the Vactrol LDR (it's essentially in parallel with it from an AC perspective) in order to restore DC bridge balance. So after correction, both tip and ring are each loaded with 9K to AC ground.

Wayne,

I think that you have it correct. Upon reflection, it looks like you would indeed be properly compensating for mismatch in the pick-off resistors. You wouldn't, of course, be compensating for any mismatch in the 100 ohm resistors (or the series coupling caps), but you wouldn't be making the situation signficatnly worse, which was my concern.

And getting rid of the large-valued input coupling caps is a certainly a good thing.

Gary
[/quote]

 

[jdbakker]

How about this for a simple direct-coupled mic pre ? Take this design (project #66 from Rod Elliott's collection):

Replace R2 and R6 by ~5mA current sources tied to +60V. Replace the transistors by 2SA970/2SC2240 or other hi-voltage lo-noise types. Ditch C2 and C3. Add trimmers/servos for flavor.

What do you think ? Much tweaking is still needed, but the basic principle should work, no ?

JDB.
[just an appetizer until I can find time to draw the HV-DOA no-cap pre]

 

[JohnRoberts]

[quote author="jdbakker"]How about this for a simple direct-coupled mic pre ? Take this design (project #66 from Rod Elliott's collection):

Replace R2 and R6 by ~5mA current sources tied to +60V. Replace the transistors by 2SA970/2SC2240 or other hi-voltage lo-noise types. Ditch C2 and C3. Add trimmers/servos for flavor.

What do you think ? Much tweaking is still needed, but the basic principle should work, no ?

JDB.
[just an appetizer until I can find time to draw the HV-DOA no-cap pre][/quote]

Indeed, any of the hybrid (discrete front end + opamp back end) may lend themselves to flying just the front end up to phantom. Rather than searching for high voltage low noise inputs I plan to cascode the low noise devices I have already settled on. PNP inputs are more compatible with flying.

An alternative to cascode is synthesizing a flying - rail and adding another pair of electrolytic caps with polarity flipped to block DC between stages.

This topology looks vaguely familiar. Is that similar to the old Soundcraft mic pre?

JR

PS: My new years resolution is to reduce my scratchings on this subject to a presentable schematic. I still have several details to work out.