Design of a sound card "Front end" for audio test set

[blue_luke]

Hello all!
JR said that "back in the 80s I designed a piece of semi-pro test equipment (TS-1) and that was harder than I expected."
Well!
I am swimming quite vigorously these days!
Understanding exactly what he means!
It is true, I find designing test equipment is an art in itself.
So here is my offering for the input section of the test set I am trying to design.
I am talking here from the output of DUT to the line input of the Sound Card.

Looking at this schema, I think I have felt to the "while we're at it" syndrome
Mind you, as a guitarist, after 21 guitars and basses I still need a few more and I am also afflicted with G.A.S, the Gear Acquisition Syndrome .

Boaf! it's only money, and having fun implies money. Sometimes rather lot of it!

So here it goes.
The schema is a mash of many ideas gathered here in our discussions, from Pete Millet and AbbeyRoad D'enfer point of views, some recommendations by Ruffrecord and all... And of course some of my own ideas!
I build my preamps, microphones, speakers, amplifiers etc...
So the versatility of the test set is rather important.

Let's beat this schema to pieces and see what we can come up with.

For AbbeyRoad_d'enfer, I retain your Idea that there should be a filter in the system. I have looked at your own schema and I think I can, maybe, get a bit further. I am not saying anything now, still in the conceptualization mode for this important section that will come later.
But I fully agree that there should be a set of filters in the circuit.

Also everybody, remember I wanted to keep this thing 'simple'
Well, I guess this has gone out the window!

Note: not all the components are not calculated as of yet. Since I use KiCad for my projects, I found that it is quite faster to select a part, assign a PCB footprint and then simply copy/paste the part as needed on the schematic, instead of constantly go thru the process of selecting a part, assigning a value, then a footprint. Resistors are resistors, and caps are caps... on with it. So you may find that some values don't make sense.
Don't fret!

Luc

 

[blue_luke]

Let me start right here.
I think the protection diodes, D12-13-14-15, are at the wrong place.
They should be placed directly at the '+' inputs of U3A/B pins.
There is also probably a better, more elaborate protection network to be thought out here ?
We are dealing with a very large spectrum of signals, level and frequency wise here.

 

[Dimitree]

Please forgive me ignorance, but what is the purpose of U4 and U5? Signal before U4 is already balanced, why do you need to make it single ended and then balance it again?

 

[jp8]

You write you you want a universally usable frontend for anything between microphones (i.e. low-noise, low level, phantom power etc.) to power amplifiers (high power levels, need protection etc.). More or less the same requirements that I had. Did you read the articles that I attached to post #19? There you'll find some suggestions for input circuit protection in the "Stepped Attenuator and Line-Input - Modimications.pdf" document. I'm not going to review your design proposals into the smallest detail and do the engineering work for you, but what immediately struck me:

• No input RF filtering.
• Jacks can short out to ground when inserting the jack. If already connected to a power amp under power, it wont like it...
• There's no power rating mentioned with the load resistors. Should they be able to handle the full output swing of e.g. a 2000W amp?
• C15/16/17/18: No need to parallel a small film cap to and elcap. Much has been said already about this in other threads and forums. It will not help reducing the THD measured.
• I don't see the need for DC coupling. The Audio Interface will block it anyway. And we measure audio signals, which are AC by default.
• Protection: see my documents uploaded in post #19. Should be low resistance (<500R) and clip to a voltage level below the power supply rails of the op-amps for best protection (as in Rohde & Schwarz UPL). Use a lamp or polyfuse to limit the current, whicle keeping resistance low.
• As Dimitree already pointed out: no need to go from differential to single-ended and back. I had to go to single ended because the UMC202HD ADC filter was configured for single-ended use, but preferably I had kept it differential. If you want to include filters in the circuit, then I can imagine you want to keep the path between input diff amp and output stage single ended, though.
• The attenuator is not a balanced, differential/symmetrical design. This will negatively affect CMRR. Suggest to change to a fully symmetrical attenuator as posted by AbbeyRoad D'enfer and me.
• The attenuator impedance is quite high. This will cause a low-pass filter in combination with stray capacitances, which may make it necessary to add trimming caps on the attenuator resistors, depending on the highest frequencies you expect to measure and how much attenuation you can accept.
• For ultra-low noise, many of the resistor values are too high (R10, R26, R32, gain resistors.) Again, I would like to refer to the docs I uploaded.

I hope this helps.

Jan

 

[blue_luke]

You write you you want a universally usable frontend for anything between microphones (i.e. low-noise, low level, phantom power etc.) to power amplifiers (high power levels, need protection etc.). More or less the same requirements that I had. Did you read the articles that I attached to post #19? There you'll find some suggestions for input circuit protection in the "Stepped Attenuator and Line-Input - Modimications.pdf" document. I'm not going to review your design proposals into the smallest detail and do the engineering work for you, but what immediately struck me:

• No input RF filtering.

OOPS! Of course there should be RF filtering, just forgot about it.

• Jacks can short out to ground when inserting the jack. If already connected to a power amp under power, it wont like it...
• There's no power rating mentioned with the load resistors. Should they be able to handle the full output swing of e.g. a 2000W amp?

The load selector is meant to set the proper load for the most used standards, so the power rating could be say 0,6W?
I have made provision for a 5-ways binding post so we can terminate with any load value we need, but also connect as input pretty much anything. When I test an amplifier, I built a long time ago a dummy load for 2-4-8-16 ohms with a BNC connector to send a portion of the signal to outer instruments.

• C15/16/17/18: No need to parallel a small film cap to and elcap. Much has been said already about this in other threads and forums. It will not help reducing the THD measured.

Noted.... Out they go!

• I don't see the need for DC coupling. The Audio Interface will block it anyway. And we measure audio signals, which are AC by default.
• Protection: see my documents uploaded in post #19. Should be low resistance (<500R) and clip to a voltage level below the power supply rails of the op-amps for best protection (as in Rohde & Schwarz UPL). Use a lamp or polyfuse to limit the current, whicle keeping resistance low.

That's a neat trick! I will see about it

• As Dimitree already pointed out: no need to go from differential to single-ended and back. I had to go to single ended because the UMC202HD ADC filter was configured for single-ended use, but preferably I had kept it differential. If you want to include filters in the circuit, then I can imagine you want to keep the path between input diff amp and output stage single ended, though.

Indeed, this is where I want to include a filter, which I have not decided the topology as of yet. You will agree that we develop these projects one section at a time.

• The attenuator is not a balanced, differential/symmetrical design. This will negatively affect CMRR. Suggest to change to a fully symmetrical attenuator as posted by AbbeyRoad D'enfer and me.

I agree

• The attenuator impedance is quite high. This will cause a low-pass filter in combination with stray capacitances, which may make it necessary to add trimming caps on the attenuator resistors, depending on the highest frequencies you expect to measure and how much attenuation you can accept.

There is an excellent article by Rod Elliot on the design of attenuators for measuring instruments https://www.sound-au.com/articles/meter-atten.htm
My reasoning was to make the input Z 1Mohm, but now thinking about it some more, I don't do valve audio equipment except for guitar amplifiers, which we don't really test for SNR THD and such! And my older test equipment is perfect for this type of work.
So I will probably revert to 100K input Z.

• For ultra-low noise, many of the resistor values are too high (R10, R26, R32, gain resistors.) Again, I would like to refer to the docs I uploaded.

I downloaded the docs you posted in post #19 but for some reason no link work. Do you have this posted on a site?

I hope this helps.

YESSSSS! I will 'steal' ideas from your design for sure!

Luc

 

[jp8]

The files are pdf copies of my website, which I'm working on already quite some time now, but it's still offline. That's why those internal links to pages on my website wont work. I made pdf files of the pages about my DIY Audio Analyzer that are more or less finished. I'm a slow writer and English is not my native language, which is a double handicap.

Rod Elliot's site is one of my favorite sites on audio related stuff. Loads of information, well written and no-nonsense. A real gold mine.

Jan

 

[blue_luke]

The files are pdf copies of my website, which I'm working on already quite some time now, but it's still offline. That's why those internal links to pages on my website wont work. I made pdf files of the pages about my DIY Audio Analyzer that are more or less finished. I'm a slow writer and English is not my native language, which is a double handicap.

Got it! No problems. And English is not my native language either. I am one of those French Canadian that even the European French claim they can't understand!

Rod Elliot's site is one of my favorite sites on audio related stuff. Loads of information, well written and no-nonsense. A real gold mine.

Jan

Same here! I wrote to Rod a few times and the guy is very knowledgeable, very smart and quite witty!
Right now I am revising my version 0.2 of the input attenuator and I must admit I am 'stealing' quite a bit of ideas from you, and I am a bit ashamed I forgot about the RF filtering!
I will post later today and for the rest of the weekend I will be absent probably.
I am recording a concert Sunday afternoon and tonight is the general rehearsal, which I insist to be present, tomorrow is the set-up and Sunday is the recording/breakdown.

 

[blue_luke]

Please forgive me ignorance, but what is the purpose of U4 and U5? Signal before U4 is already balanced, why do you need to make it single ended and then balance it again?

My idea is (was, see further posts)) to have a rather high Z impedance, in the order of 1Mohm or so....
But today with a bit of insight from JP8 I realise that unless you work with valve equipment, this hi-Z is unnecessary.
A scope is still the best instrument to work with valves equipment.
I do build valve stuff, but mostly for guitar amps where distortion and noise is almost irrelevant, in fact IT IS irrelevant!

So I have recalculated the input attenuator to be completely symmetrical, to keep a better CMRR and also lowered the Zin to 100Kohms to keep a better SN/ratio.
But I still want to convert to unbalanced line so I can insert Hi and LO pass filters following AbbeyRoadD'enfer excellent suggestion.
The only thing is that I want to think this filter a bit further than just copying his schema.
I am thinking of having one or two inflexion point in the High-Pass and the Low-pass sections.
As RuffRecord mentioned, there is a lot of stuff going on above 20Khz and the audio interfaces I have allows me to go up to 192Ks/24bit, so I can measure(?) up to 96khz bandwidth. I don't think this will work, I will have to test, I am dubious, but definitively, 48khz bandwidth is achievable with a bit of care!
And that's all is probably necessary for meaningful testing.

I like the idea of taking on a project like this, keeping everybody posted on my evolution and contemplating all the inputs from some very knowledgeable persons in this group.
We all learn, and we all get better from it.
Thanks ya'll
Luc

 

[blue_luke]

• Use a lamp or polyfuse to limit the current, whicle keeping resistance low.

I think I get the gist of it, but can you elaborate?
Never thought of using a lamp bulb in this application but it is a very neat trick!
I tink I prefer it than the polyfuse option.
ValveWizard uses MosFets to protects the inputs (see here: https://www.valvewizard.co.uk/soundcardinterface.html)
He uses LND150 and I think this is clever.
So, we all agree we have to protect the inputs, we found a few scheme, polyfuse, lamp bulb, MosFets, what should we use?
Luc

 

[blue_luke]

Here is the "soupe du jour" (daily soup) on the input section of our audio test set.
Never mind the high number of the parts nomenclature.... there will be far less than 150 resistors in this thing!
This is just a byproduct of KiCad annotation tool that I don't master completely as of yet.
Per JP8 suggestions I dropped a few things, went from a fully balanced attenuator, went from 1Mohm input Z to 100Kohm, added his very well tought in put protection and added RF input filtering.
There is still some calculations to be done, like the value and type of L1 and L2 for the RF filtering.
This is still not a definitive schema, but in the spirit of how this project is going, submitting my daily work/thinking has become like the design reviews we did when I was still active (I am an old retired 70 years old fart!
Check the file, this is version .02. Version 1.0 will be when I commit to a PCB design.
Luc

 

[MidnightArrakis]

English is not my native language

Jan

>> Should you be at all interested.....I had studied both "Radio/TV Broadcasting" and "Journalism" while I was in college and then I also ended up being the "College Yearbook Editor" before being shipped out to Southeast Asia during the Vietnam War. If you would want to send me your writings as -- WORD -- files, so they are easily editable, I could review your work and make certain that your sentence and paragraph structures are clear, your terminology is correct, etc., etc., etc. ..... Or, not.

Just a suggestion.

/

 

[blue_luke]

>> Should you be at all interested.....I had studied both "Radio/TV Broadcasting" and "Journalism" while I was in college and then I also ended up being the "College Yearbook Editor" before being shipped out to Southeast Asia during the Vietnam War. If you would want to send me your writings as -- WORD -- files, so they are easily editable, I could review your work and make certain that your sentence and paragraph structures are clear, your terminology is correct, etc., etc., etc. ..... Or, not.

Just a suggestion.

/

You seem to be quite a generous person, offering your time and effort to verify schematics and PCB designs, and now our writings.
This is precious and very appreciated.
BUT, as any precious resources, it should be used judiciously
Even if for Jan and I, English is not our native language, we are quite fluent, enough to convey our toughs quite clearly.
If this project of the design of a front end comes to fruition, be prepared to get involved in revising the PCB drawings, the tech notes that will come with it etc...
It would be nice to make this a 'build it thread' complete with instructions, manuals, etc... and Gerber files for those who would like to roll their own.

The other day you gave me the idea that making PCBs for whoever is interested in building this thing could be quite practical.
I agree.
If it was just for me and a 'one off' I would simply use veroboard which I am quite adept at.
But I feel there is some virtues in making a nice PCB and front plates and putting it in a nice enclosure.
So yes, probably there will be some PCB's made and since the Chinese board makers are very affordable, and they requires quantities in multiple of 5, I am willing to finance the production of a small batch of say between 10 to 20 boards that I would resell for a little profit and S/H.
Being from Canada, every shipment would be by Post Canada with no tracking number, unless someone is willing to pay for the surcharges incurred by courier.
I guess if someone in Europe or elsewhere is willing to have a batch made, I am willing to supply the Gerbers and schematics
This is no national or army secret here. Most of the ideas thrown around here are public domain and basic state of the art practices.
Also, lately I had boards made by JLPCB and I can say that the quality is very good, the delivery was very quick, like 10 days! and the price was astoundingly low!!
And I think, I have to verify, that if I submit a PCB design, they have a place on their website where it seems I can deposit the Gerber files and then anybody on the planet can order the boards, except that it may have to be in quantity of 5 or multiple of 5. I have to check about this.

Luc

 

[MidnightArrakis]

Even if for Jan and I, English is not our native language, we are quite fluent, enough to convey our toughs quite clearly.

I "could" make a comment here, but.....I'm not going to. I clearly understand your "toughs" here.

The other day you gave me the idea that making PCBs for whoever is interested in building this thing could be quite practical.

COOL!!!.....

So yes, probably there will be some PCB's made

Please do everyone a favor when you are setting up your CAD PCB-design program up, OK??? >> SET UP YOUR PCB-DESIGN SOFTWARE PROGRAM TO USE --- >> MILS << --- AS THE DIMENSION SETTING AND -- NOT -- EITHER "ENGLISH" OR "METRIC" AS THE STANDARD!!!

As I have reviewed countless numbers of GERBER and N/C Drill files on this forum, nobody on here is actually aware of how "messed-up" these files really are should you open them up in a "Text Editor" (-- NOT -- a "GERBER Viewer") and you see what the actual dimensions of these files really are!!! (As an example..... a 20-mil wide track is - NOT - 0.508mm, but actually ends up being 0.482637mm)!!! MOST ALL of the N/C-Drill files I have seen on this forum are - NOT - actual drill-sizes, but instead are "pretty close" to standard drill-sizes!!! This means that the PCB-fabricator has to either "round-down" or "round-up" to a nearest actual drill-size!!! So, set your PCB-design program to -- MILS -- and then your files can be imported into any other program -- without -- having to go through a conversion.

I am willing to finance the production of a small batch of say between 10 to 20 boards

There is a PCB-fabricator that advertises within one of my industry magazines and they have a running "Special" in every issue stating, "BUY 10 PCB'S AND RECEIVE 20 AT NO EXTRA CHARGE"!!!

/

 

[blue_luke]

I "could" make a comment here, but.....I'm not going to. I clearly understand your "toughs" here.

Yeah, touché!

COOL!!!.....

Please do everyone a favor when you are setting up your CAD PCB-design program up, OK??? >> SET UP YOUR PCB-DESIGN SOFTWARE PROGRAM TO USE --- >> MILS << --- AS THE DIMENSION SETTING AND -- NOT -- EITHER "ENGLISH" OR "METRIC" AS THE STANDARD!!!

I have been using "mills" since my first version (V2.0) of Protel for DOS in the mid '80s. Protel is now Altium and in those days the schematic and the PCB were two separate packages! 'twas interestin' !

There is a PCB-fabricator that advertises within one of my industry magazines and they have a running "Special" in every issue stating, "BUY 10 PCB'S AND RECEIVE 20 AT NO EXTRA CHARGE"!!!

/

Who is that?
I have used JLPCB lately, for the first time, and I am very satisfied with the quality of the boards and the promptness in execution and delivery.

Luc

 

[jp8]

>> Should you be at all interested.....I had studied both "Radio/TV Broadcasting" and "Journalism" while I was in college and then I also ended up being the "College Yearbook Editor" before being shipped out to Southeast Asia during the Vietnam War. If you would want to send me your writings as -- WORD -- files, so they are easily editable, I could review your work and make certain that your sentence and paragraph structures are clear, your terminology is correct, etc., etc., etc. ..... Or, not.

Just a suggestion.

/

Thanks for your generous offer! Very, very kind of you! But I have decided not to take advantage of your offer, as I do not want to continuously edit the texts. I'd rather spend more of my time on generating the content, i.e. working in my workshop.

Jan

 

[jp8]

I think I get the gist of it, but can you elaborate?
Never thought of using a lamp bulb in this application but it is a very neat trick!
I tink I prefer it than the polyfuse option.
ValveWizard uses MosFets to protects the inputs (see here: https://www.valvewizard.co.uk/soundcardinterface.html)
He uses LND150 and I think this is clever.
So, we all agree we have to protect the inputs, we found a few scheme, polyfuse, lamp bulb, MosFets, what should we use?
Luc

Yeah, I should have mentioned that one on valvewizard as well. I actually even have a link to this website on the introduction page of my DIY Audio Analyzer, but apparently I forgot about it... Very clever indeed and will not add noticeable noise. I ran into that circuit after I built and finished my own, so I decided to keep the circuit as it was. I used a series combination of a Bourns resettable fuse type CMF-RL50a-10-0 and a 100R 2W wirewound resistor. Whether it works just as well as the MOSFETs, I don't know. It's a bit cheaper than the MOSFETs, but that is probably the least of your concerns. The lamp solution is a known method for tweeter protection, but AP employed this method in their System 1. And maybe later models too, but I only have the service manual of the System 1. The lamp works in the same way as a Resettable Fuse (aka Polyfuse): at low currents, the resistance is low. Which is what we want for low noise contribution. In an overload situation, as the current increases, the lamp/fuse heats up, which causes the resistance to rise. As this is an overload situation, the higher noise contribution is of no concern. The idea is that the lamp/fuse limit the current to a value the clamping diodes can handle continuously. From LTspice simulations, I know the fuse limits the current to a safe value. For a lamp, I wouldn't know. You would have to characterise the specific lamp you intend to use to see if it would limit the current to a safe value. The only pitfall of such a circuit is the trip time of lamp/fuse. Both take time to heat up and rise in temperature. The Polyfuse probably taking a little longer. During the trip-time, the clamping diodes and zeners should be able to handle the temporary overload. The BAR99 diodes that I employed, should be able to handle higher surge currents compared to the 1N4148. Again, I did not test this. It is based on assumptions and datasheet information.

If I were to design the circuit again, I would probably choose the MOSFET solution. Btw, it has been described in this application note: https://ww1.microchip.com/downloads.../ApplicationNotes/ApplicationNotes/AN-D11.pdf

Jan

 

[MidnightArrakis]

I have decided not to take advantage of your offer, as I do not want to continuously edit the texts.

I am guessing that my written offer must not have been as clear as it needed to be because.....what I was offering to you was that -- I -- would have done > ALL < of the editing for you, including checking for terminology, grammatical structure, clarity of conveyance and so on and so forth. But, sure.....go ahead and knock yourself out!!! After all.....it's your baby!!!

My composing and presenting 30-page technical reports on various details of design engineering projects in order to convince corporate management to spend -- tens-of-thousands -- of dollars for new equipment and software certainly could not have helped me learn how to write technical information in such a manner so non-engineering types could easily understand stuff they didn't know anything about now, would it?

/