Discrete mic pre project: proto PCBs have arrived.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Would you feel comfortable building an SMD mic pre?

  • Yeah, sure, as long as the parts aren't too small.

    Votes: 83 64.8%
  • Nope. If it's not thru-hole, it's not happening.

    Votes: 39 30.5%
  • Nah, we have enough mic pre designs here as it is.

    Votes: 6 4.7%

  • Total voters
    128

jdbakker

Well-known member
Joined
Nov 24, 2005
Messages
1,431
Location
Amsterdam, The Netherlands
Hi all,

I am working on the design of a discrete microphone preamplifier which I am planning to release here (both as boards and self-etch). It occurred to me that, since all the world is going SMD, it would be interesting to have an easy My First SMDee project here in GroupDIY. Of course the plan would be to not make it too hard: have nice large pads, enough space between parts for (unexperienced) hand solderers and no parts smaller than necessary for best performance. However, before I start laying out a board I'd like to gauge interest as there's little point in setting up a group project if nobody feels comfortable building it. Of course, either in SMD or thru-hole all parts will be easily available from The Usual Suspects, and I expect build complexity to be similar to the Green or the SSL9k, so no crazy dense layout as on some of the DOA designs floating around here.

Thoughts?

JDB
[a bit of context: originally I didn't set out to design Yet Another Mic Pre. I was looking to build a line-in stage with good CMR and enough S/N to be below the noise floor of of modern ADCs. When you start doing the math, it turns out that the design goals are very similar to a wide dynamic range mic pre (<-120dBU noise floor, medium input Z). The design as it stands looks somewhat similar to the SSL9k, only with three fixed gain stages with a pair of ganged attenuators between them. And BTW, no matter whether the end result is SMD or thru-hole, the input transistors will have footprints for low-noise TO-92 transistors like the 2SB737, the MAT-03 and maybe even 2SJ74 P-JFETs. Or you can plonk in a pair of 2N4403s if cost is more important than that last dB of noise performance]
 
I'm not sure I would consider SMT a distinct type of preamp. While there are potential differences from that technology.

In the early days it was difficult getting low Rbb front end transistors in SMT packages presumably because die size was relatively large for the standard GP transistor package. I found it harder to get good film caps in SMT (haven't tried lately).

On the plus side, making a tighter, more symmetrical layout will deliver performance benefits. This was perhaps more noticeable when designing 16 input mixers with a transformer inside  :eek: , but even single sided SMT technology was a noticeable improvement in number of cuts to get an acceptable result (a good thing in business).  The preamp front end could literally be nested under the XLR connector, not possible with thru hole.

I am not a big believer in tweaky resistors but I might be tempted to avoid the smallest parts and check the resistor data sheets. I've encountered voltage coefficient problems with cheap thru hole resistors, so smt will not be immune.

Have fun..

JR
 
JohnRoberts said:
I'm not sure I would consider SMT a distinct type of preamp.

That is/was not my intention, maybe I could have made that clearer.

There are two issues:

1) I'm working on a new discrete mic pre design that might be interesting for other people here

and

2) As long as I'm making a PCB layout for (1), maybe it could be interesting to do it (mostly) in SMT to give people an opportunity to get their feet wet with that technology.

I honestly don't expect the SMT design to perform significantly better or be very much smaller than the thru-hole version; to do that I would want to choose smaller parts and tighter spacing which is incompatible with (2). Also, wherever it doesn't make sense from a DIY perspective to use SMDs (film caps, power resistors/transistors) I'd use regular parts anyway.

JohnRoberts said:
I am not a big believer in tweaky resistors but I might be tempted to avoid the smallest parts and check the resistor data sheets. I've encountered voltage coefficient problems with cheap thru hole resistors, so smt will not be immune.

Indeed. I'd recommend sticking to thin-film parts for best noise and distortion performance, I have good experience with the Susumu range carried by Digi-Key.

JDB.
[BTW John, one of the things I'm considering with this topology is a HV version with no phantom blocking caps. The downside is that many easily available high-hFE transistors have a Vce(max) around 45V, so that might not be best for a simple DIY project]
 
I have been pondering dc coupled phantom power for over a decade, and Wayne has melted a lot of solder on the subject. here: http://www.picocompressorforum.com/forum/php/viewtopic.php?f=6&t=14

He pursued floating a THAT IC up to voltage, all of my mental scratching had been discrete front end, and even one all discrete version.  I'd have to dig out my old notes but IIRC I probably cascoded PNP input devices (because I still have a stash of 737s). A scratch design could also use NPNs and just float a current mirror level shifter above, with appropriate hold off voltage.

One useful point, 48V at the input is not a valid operating condition as that suggests a zero current phantom draw. One could probably set a maximum input voltage of say 24 volts, and allow the voltage feeding the top of the phantom resistors to range higher than that as or if needed. If one is adventurous they could drive also pin one of the XLR negative, but this could definitely cause interface issues with direct boxes, etc as pin one is typically bonded hard to ground.

Lots of ways to skin this cat...  another I have considered is to float an A/D convertor up to voltage and just level shift or optical the digital output. I'm not quite ready to design my own A/D just yet so this is on hold, for now.

IMO the best capacitor is no capacitor, but this is kind of moot with so many mics having caps and electronics inside...  An interesting mental exercise in any case.

JR
 
Hey JDB,

I think this is a great idea for several reasons. 

a) I love the idea of an introductory project for SMT.  So we can just drop 16 channels into the over and come back to a ready baked mixing console ;)
b) All the better that it's not a straight up clone
c) You have considerable experience in both RF and SMT, the idea of a project where you can share some of your design concepts and broader experience sounds great to me.

I'm in and voted accordingly.  Like it or lump it we'll all be poking at SMT gear for repairs in the coming years so let's get our learn on.

Cheers,
Ruairi
 
Because this is for DIY my concern would be on the servicability of it. SMT/SMD is a fragile stuff and does not accept very much mistake no matter how simple the kit is. You solder a wrong resistor or keep the iron a bit longer and it is not a big deal on through hole. I am sure advanced builders would tackle that without even thinking about it but the guys at the lower end of the ladder may not be able to execute it properly. Therefore take that into account when doing the layout and selecting the components.
 
I've made the transition to SMT and it isn't that scary... my biggest problem is dealing with tiny and super dense BGA stuff...  I am currently prototyping using a small SMT couple watt class D amp chip that I can just about see, let alone solder. Some new parts I like are only available in SMT packages.

It is the future, until the next future...

JR
 
SMT is a good way for cheaper board, I like smt, moreover I've a two weller soldering stations for this, but, I think that a studio gear shoud be ph, not smd, so several people can repair their gears with a normal soldering iron. Moreover, smt required manual skills , but many diyers and audio engineers have not these skill becuase them haven't many experiences on smt mounting. 
 
I am actually preferring SMT for prototyping. I use the pad per hole perfboard, with 0805 parts that
just straddle the pads, it's real easy.

It's a bit of a pain with SO-iCs, but I've got some of those adapters from DigiKey that help on that.

For straight construction of pcb's SMT is much easier and faster. Yeah the skill set needs to be improved,
but shouldn't that be a goal anyway?

 
 
I wouldn't hesitate with SMD unless your main goal for this project is to sell as many boards as possible... A great advantage of SMD is the wide availability of cheap dual transistors. Likely a balanced frontend cries for lots of these. Personally I often run into power dissipation limits with SOT-23 parts though.

I'm curious to see the topology you end up with; i've come up with some promising concepts which ain't just a standard transformerless preamp with discrete opamps, but never got around to further develop things.

Samuel
 
Samuel Groner said:
A great advantage of SMD is the wide availability of cheap dual transistors. Likely a balanced frontend cries for lots of these.

Sure; I especially like the newer low-rbb hi-hfe duals produced by Philips NXP and Zetex Diodes Inc.; I use quite a bunch of those in my DOA-in-progress (three dozen transistors and counting...). However, all of those come in SOT-457 (aka SOT-23-6, SOT-26 or SC-74) packages or smaller, with a lead pitch of 0.95mm or less, and I wouldn't want to force first time SMD solderers to deal with those. ATM my lower limit in parts size for this design is 0805 for passives and SOT-23 for transistors.

Samuel Groner said:
Personally I often run into power dissipation limits with SOT-23 parts though.

Absolutely. That's one of the reasons that I still intend to use traditional output transistors like the BD1xx/MJE1xx-series. I find most power SMD packages (SOT-223, TO-263) hard enough to solder by hand even if you do have the right equipment. It's not like using SOT-223s saves you board space, either; it only makes sense for very low profile designs, or hi-volume SMT designs where every thru-hole part comes at a significant P&P cost.

Samuel Groner said:
I'm curious to see the topology you end up with; i've come up with some promising concepts which ain't just a standard transformerless preamp with discrete opamps, but never got around to further develop things.

I'm afraid it's pretty boring. Like I said there are three stages: the first can be reduced to a pair of CFB amps with a fixed gain resistor between their '-' inputs; second and third stages are very similar to the discrete differential building blocks Bruno Putzeys discussed in his AES paper five years ago.

(When designing the line in stage that gave birth to this pre I had been working hard to get away from the beaten path. I was in particular looking for a way to have the first stage reject all CM interference, rather than passing it on to the second stage. I've come up with a dozen schemes, some simple, some complicated, and while they all looked Good On Paper they invariably required tightly matched components or they messed up the HF phase response. Then again, just maybe there's a reason why the tried-and-true solutions are tried-and-true)

The elements that make this design different from most others that I have seen are mostly in the gain setting and in the dynamic range. While most other preamps tend to set gain by changing the feedback resistors in their gain block(s), this design has fixed gain for all three gain blocks, with variable attenuators in between. The advantage is that each gain block can be better tuned for its fixed gain and loading, the lines to/from the pots are less sensitive to capacitive loading and regular log pots can be used to set the gain rather than the rev log pots required by other pres. The downside is, of course, that when the gain pot/switch bounces or scratches, the amp will briefly spike to max gain.

The use of discrete gain blocks allows for some brute-forcing of the pre's dynamic range. It is well known that with all amps you reach a point in setting the feedback network where you need to trade noise for distortion; higher feedback resistances increase noise, whereas lower feedback resistances require more drive current which increases distortion. This is a clear area where no generalist op-amp in DIP8 can win from a custom solution employing a handful of TO-126 transistors.

I still have a few issues to nail down before I can build a prototype, mostly related to biasing and DC offset rejection. About the latter: the original design would drive a dedicated audio ADC, making a software+DAC servo a natural solution. Can't use that now, so I'll have to choose between a microcontroller+DAC (hard to program for DIYers), a few trimmers (might drift too much) and an analog servo or two. If I end up choosing the servo I hope people will forgive me if I implement those with IC op-amps.

JDB.
 
Second and third stages are very similar to the discrete differential building blocks Bruno Putzeys discussed in his AES paper five years ago.

I'd have a few improvements to suggest for these designs if you are interested.

The downside is, of course, that when the gain pot/switch bounces or scratches, the amp will briefly spike to max gain.

Why does it spike to maximum gain? As far as I understand the input becomes disconnected, which produces a transient due to input bias currents. This may be at least partially solved by connecting a large resistor from wiper to ground.

Samuel
 
Samuel Groner said:
Second and third stages are very similar to the discrete differential building blocks Bruno Putzeys discussed in his AES paper five years ago.

I'd have a few improvements to suggest for these designs if you are interested.

Of course. Even though what I have now is more 'inspired by...' than a direct copy, I expect that any suggestions are still sufficiently applicable.

Samuel Groner said:
The downside is, of course, that when the gain pot/switch bounces or scratches, the amp will briefly spike to max gain.

Why does it spike to maximum gain? As far as I understand the input becomes disconnected, which produces a transient due to input bias currents. This may be at least partially solved by connecting a large resistor from wiper to ground.

As it stands the gain control consists of a variable U-pad connected in the differential lines between gain blocks. More specifically, I add series resistors to the output of each gain block, and a shunt potentiometer (wired as a rheostat) after these resistors. Wiper bounce leads to minimum attenuation and thus gain peaking. I have considered other attenuator designs, like a dual pot connected as two L-pads between each of the outputs of the previous gain block, but both pot tracking and pot resistance tolerances would compromise the CMR of the amplifier (which I find more objectionable). I suppose that a gain switch with two decks for each attenuator stage might fix this.

JDB.
 
That is indeed a circuit design consideration and often mitigated with some capacitance in the negative feedback path since the open circuit duration will typically be very brief. That also sounds like perhaps a mechanical issue with the pot design. Pot wipers usually have multiple fingers so one or more will always be making contact while the others bounce. I have seen as many as 6 and as few as 3 fingers used. I have also seen impaired performance if the wiper track isn't well aligned with the screened resist. Finally different resistive media will have more or less surface irregularity to deal with.

One brute force solution is to use a dual pot of 2x the resistance wired in parallel. This should reduce wiper bounce to typically only a 6 dB jump, while reducing hop off resistance and other non ideal pot behaviors. Unless the pots are truly horrible, you should not have both pots lose all wiper contact at the same instant. If yes you really need to source a different pot vendor.

Switches are always nice for precision applications but make before break contacts will also exhibit some gain bounce when switched. Even a perfect gain switch will make some click if switched while signal is present so we must live with some control noise.   

JR

PS: I had to deal with one wein-bridge EQ design where the circuit oscillated during wiper bounce (control sounded scratchy if turned fast). This was because the wiper from one pot manufacturer, was re-purposed from a different application to save tooling cost. The wiper not only started with less fingers than typical, it was also more sensitive to tracking errors. The other pot manufacturer sourcing that pot, tooled a new wiper dedicated for the application. Their pot was silent. For those keeping score at home, Alps was the better pot manufacturer.
 
I add series resistors to the output of each gain block, and a shunt potentiometer (wired as a rheostat) after these resistors.

I see, but this is not a constant load for the driving circuit, right?

With respect to the differential amplifiers: first of all I'd consider if a one-stage topology is really the right thing. Differential IC opamps appear to almost ubiquitously use such to minimise power consumption for a given bandwidth, but this is probably much less of an issue here. Two-stage Miller compensated topologies are simply so much more robust against output loading and other voltage-dependent impedance contributions at the output gain node (i.e. output stage input port) that I've found their use much more beneficial for audio applications. Likely the two-stage approach will also result in a simpler design for a given performance level.

If you want to stay with the one-stage architecture I suggest to improve the output gain node impedance. First if you connect the base of a common-base transistor to the emitter of the folded cascode transistor, feed the emitter of this added transistor with a current ~equal to the collector current of the folded cascode and make the collector of the new transistor track the output voltage, this transistor will inject a current into the folded cascode which cancels the collector-base loss of it. This is not limited to low-frequency hFE loss but also cancels the output C and its voltage dependence. The added power consumption may be re-used for output stage biasing.

Second the active load of the folded cascode needs consideration. If we employ cascodes as Putzeys does it is of great advantage to reference the base/gate of the cascoding transistor to the emitter of the cascoded transistor (and not to the base as shown). This also cancels cascode output impedance as well as its voltage dependence. Alternative approaches include boostrapping the entire collector load (preferably together with the first stage of the output buffer).

That's about what I can say without seeing your schematic.

Samuel
 
I greatly prefer all-smd for prototyping.  Drilling holes is a PITA. Not being able to plate the holes is another problem. But I can make perfect single sided SMD boards quickly from scratch (prototypes). As far as not being repairable, I think that is only the opinion of those who haven't tried. Get some chip-quick ($5) or a hotair station ($130).  Thru-hole dips over 8 pins are very hard to replace cleanly compared to soic once you get the hang of it. Especially if you want it to look as if it hasn't been repaired. Smd rework is a required skill for studio techs in 2009 going foreward. If this is someones first exposure to smd construction, then more power to you!
Mikep
 
mikep said:
Smd rework is a required skill for studio techs in 2009 going foreward. If this is someones first exposure to smd construction, then more power to you!
Mikep

This is what motivates me (not that I plan on becoming a studio tech).  The future is SMD, more and more of the commercial gear we buy will be SMD.  I want to be able to deal with this gear if and when it goes wrong. I'm here to learn and grow as an engineer, not have the easiest option.  If want easy I'll buy commercial gear.

Cheers,
Ruairi

 
SMT/SMD is a fragile stuff and does not accept very much mistake no matter how simple the kit is

Nah.  I've been pretty much designing in SMD for years and years and I've not seen rework be any more problematic for SMD than through hole.  In fact, I've seen through hole designs be more problematic because large parts can sink a lot of an iron's heat and necessitate using higher wattage irons who tend to burn the PCB.

Smd rework is a required skill for studio techs in 2009 going foreward.

Not just studio techs.  If you want to work in any electronics field from now on, SMD is a necessity.

If this is someones first exposure to smd construction....

Better sooner than later.  At least you will have fun with it instead of sweating over it like you would if you were doing it for a job interview...
 
of course.  Just like TH resistors, you have to choose the right ones for the job.  Luckily you now have the choice between thin film, thick film and metal film SMD resistors.
 
Back
Top