|
|
Post by gouge on Apr 5, 2021 1:10:59 GMT -6
I'm embarking on a build of a simple filter box for tracking and mixing. the idea surfaced in the main forum. after looking at the options Matt from Iron Age Audio offered to help me with a simple and cost effective solution specific to my needs. Cannot thank Matt enough. I was chasing something with variable HPF, variable LPF and variable Pad. With the variable pad to -35db so i could incorporate preamps during mixdown. I'm not a tech so if i say something incorrectly please point it out to me. gotta learn as they say. not to quote matt out of context but this is what he was thinking. "I am thinking balanced input, into a level control pot, followed by a discrete class - A buffer. Then an active high pass filter with a slight resonant hump with a -3 dB point from 10-250Hz, where the 10 Hz position puts the resonant peak right at 20 Hz. Then an active low pass filter, 1.8 kHz to 35 kHz, again intentionally with a ~2 dB active resonance at the filter corner, so the 35 kHz puts the peak at 20 kHz. Finally a balanced, cross-coupled output."
the filters are 12db / octave level pot is -inf - +6 each filter is switchable in/out +/- 16v power. Matt let me know some of the circuits are well known solutions been around for a while. Being a luddite, Matt sent me the Gerber files and BOM with instruction on where to upload to get manufactured. i ordered 10 boards (channels) which come with smd parts populated for under $100 usd. i need to buy the case, the pots and switches and the through hole caps. some of those items i already have. i also need to buy the rack unit and the power supply and a drill press. i may have a power supply kit somewhere from JLM. starting with a 4 channel unit. once that's done i'll look towards using the other 4 boards to use with a capi 2 stage line amp kit and transformer output. then i'll have 4 channels of clean and 4 channels of coloured. interestingly, the cost difference between those 2 ideas is significant.  i'll add photos as the build progresses. i ordered the boards today. they may take a while as i'm in australia. |
|
|
|
Post by gouge on Apr 5, 2021 8:07:02 GMT -6
completed a mouser cart tonight for all the remaining bom parts and added items needed for the enclosure. ie. pots, switches, knobs, through hole caps, xlr connections, rack enclosure etc. everything but power for 4 channels.
rack is on back order so i'm guessing ill need to resolve that asap but it was a lot cheaper through mouser than the local supplier.
|
|
|
|
Post by svart on Apr 5, 2021 9:44:51 GMT -6
I'm embarking on a build of a simple filter box for tracking and mixing. the idea surfaced in the main forum. after looking at the options Matt from Iron Age Audio offered to help me with a simple and cost effective solution specific to my needs. Cannot thank Matt enough. I was chasing something with variable HPF, variable LPF and variable Pad. With the variable pad to -35db so i could incorporate preamps during mixdown. I'm not a tech so if i say something incorrectly please point it out to me. gotta learn as they say. not to quote matt out of context but this is what he was thinking. "I am thinking balanced input, into a level control pot, followed by a discrete class - A buffer. Then an active high pass filter with a slight resonant hump with a -3 dB point from 10-250Hz, where the 10 Hz position puts the resonant peak right at 20 Hz. Then an active low pass filter, 1.8 kHz to 35 kHz, again intentionally with a ~2 dB active resonance at the filter corner, so the 35 kHz puts the peak at 20 kHz. Finally a balanced, cross-coupled output."
the filters are 12db / octave level pot is -inf - +6 each filter is switchable in/out +/- 16v power. Matt let me know some of the circuits are well known solutions been around for a while. Being a luddite, Matt sent me the Gerber files and BOM with instruction on where to upload to get manufactured. i ordered 10 boards (channels) which come with smd parts populated for under $100 usd. i need to buy the case, the pots and switches and the through hole caps. some of those items i already have. i also need to buy the rack unit and the power supply and a drill press. i may have a power supply kit somewhere from JLM. starting with a 4 channel unit. once that's done i'll look towards using the other 4 boards to use with a capi 2 stage line amp kit and transformer output. then i'll have 4 channels of clean and 4 channels of coloured. interestingly, the cost difference between those 2 ideas is significant. View Attachmenti'll add photos as the build progresses. i ordered the boards today. they may take a while as i'm in australia. Interesting circuit, although a bit overcomplicated IMHO. The Superbal-type input allows high CMRR by matching input impedances at the input pins more closely on paper but in practice it's shown to have minimal difference from a diff-amp. What's the purpose of a class-A buffer in a circuit like this? The cross-coupled output like this has a nasty habit of seeing DC floating on the upper opamp (U3.2) due to the positive feedback. Resistor tolerance matters here. They do have a positive aspect of being able to be shorted on the output without losing 6dB of gain but they still lose 6dB of headroom in doing so. Not a lot of companies use these anymore with the advent of diff-amp drivers. |
|
|
|
Post by LazyOldSun on Apr 5, 2021 10:10:15 GMT -6
I'll be interested to see this as it progresses. I used to have a Urei 546 and miss the HPF LPF it had.
|
|
|
|
Post by matt@IAA on Apr 5, 2021 10:26:33 GMT -6
svart All the buffer is doing is providing a low source impedance to the high pass filter, since gouge wanted a level pot. I ended up with an odd number of op amps, so I just rolled a quick emitter follower type buffer. I guess I could have used a normal differential input and then used the second op amp to buffer the level. Rev1 ! haha. The reason to use the cross-coupled output was to make it as simple as possible for gouge or whoever to assemble. Everything is done on the board by the pcb smt assembly service except for pots, film caps, and electrolytics. I'd use a diff-driver, but they don't have one in their parts library.
|
|
|
|
Post by svart on Apr 5, 2021 10:50:24 GMT -6
svart All the buffer is doing is providing a low source impedance to the high pass filter, since gouge wanted a level pot. I ended up with an odd number of op amps, so I just rolled a quick emitter follower type buffer. I guess I could have used a normal differential input and then used the second op amp to buffer the level. Rev1 ! haha. The reason to use the cross-coupled output was to make it as simple as possible for gouge or whoever to assemble. Everything is done on the board by the pcb smt assembly service except for pots, film caps, and electrolytics. I'd use a diff-driver, but they don't have one in their parts library. What's the value of the volume pot? I'd think a 2134 should drive something relatively high impedance, ala 10k or so with the following filter circuit, with ease. |
|
|
|
Post by matt@IAA on Apr 5, 2021 10:56:57 GMT -6
The buffer is between the volume pot and the high pass filter cap.
|
|
|
|
Post by svart on Apr 5, 2021 11:21:45 GMT -6
The buffer is between the volume pot and the high pass filter cap. Ah yeah, I see. Sorry I was going from memory. Still though, the buffer likely only sees the two 50K pots, so worst case it'll see ~25K load and without the buffer, the 2134 would see roughly 7K (if the pot is 10K). Just a thought. |
|
|
|
Post by matt@IAA on Apr 5, 2021 11:28:37 GMT -6
It's not to protect the 2134, it's to keep the source impedance low for the hpf. At a lot of attenuation, you'll end up rounding off the corner of the hpf without it. For example - here is changing Zin to the Sallen-Key. This is extreme, showing 1 and 10k Zin, but shows the impact. By buffering after the level pot you avoid this.  Here is comparing the pot at max (no attenuation) with roughly 50% travel, without adjusting the gain..  |
|
|
|
Post by svart on Apr 5, 2021 13:13:32 GMT -6
It's not to protect the 2134, it's to keep the source impedance low for the hpf. At a lot of attenuation, you'll end up rounding off the corner of the hpf without it. For example - here is changing Zin to the Sallen-Key. This is extreme, showing 1 and 10k Zin, but shows the impact. By buffering after the level pot you avoid this. Here is comparing the pot at max (no attenuation) with roughly 50% travel, without adjusting the gain.. I see what you're saying. I'm not sure that in practice your Q will change that much though. I guess you're locked into using 50K due to them being at least an obtainable ALOG value. If reverse logs were easy to get I would have wondered why not just increase the pot value to like 250K and reduce the C to something like 68n for a higher impedance load. I suppose the other thing to do would have been create a setpoint impedance by adding a little gain to the first stage and then using a high impedance divider and a lower impedance load before the filter. Not sure it would have worked as I envisioned, but worth a shot looking at it. then again, I pulled up an old SSL schematic since they also use sallen-keys as their HP/LP and they buffered before the filters, but that buffer was also part of the feedback for two other EQ sections as well, which might be wholly anecdotal since I don't know how their damping looked at extremes. |
|
|
|
Post by matt@IAA on Apr 5, 2021 13:32:16 GMT -6
right on all counts. but you figured the better approach from the get-go - just use a vanilla differential amp on the input and use the now-free second op amp on the 2134 to buffer the gain pot. tada.
that would result in a significantly lower THD profile, because the 2134 is basically flawless. on the other hand, that particular little buffer does have a sound - and I'm a bit partial to it. So I use it kind of all over the place whenever I need a quick and easy way to do what I did here.
|
|
|
|
Post by svart on Apr 5, 2021 13:47:15 GMT -6
right on all counts. but you figured the better approach from the get-go - just use a vanilla differential amp on the input and use the now-free second op amp on the 2134 to buffer the gain pot. tada. that would result in a significantly lower THD profile, because the 2134 is basically flawless. on the other hand, that particular little buffer does have a sound - and I'm a bit partial to it. So I use it kind of all over the place whenever I need a quick and easy way to do what I did here. I do know the 2134 sound well. Clean, slightly scooped. Using one or two doesn't really add up to much but back in the day I exchanged quite a few opamps in my mixer for 2134 and the dozen or so per channel really, really made it sound different. When you use that many the timbre becomes too obvious to ignore, and as they say, once you've heard it you can't unhear it. I ended up back with 5532s in most places simply because I felt that a dozen-ish 5532 didn't change tone as much. |
|
|
|
Post by matt@IAA on Apr 5, 2021 17:31:25 GMT -6
svart what’s the major difference between a cross-coupled output stage and what’s in a DRV134? Just the laser trimmed resistors? I’m curious - if you were going to roll your own balanced output, just for fun, what topology would you used?
|
|
|
|
Post by svart on Apr 5, 2021 20:48:44 GMT -6
svart what’s the major difference between a cross-coupled output stage and what’s in a DRV134? Just the laser trimmed resistors? I’m curious - if you were going to roll your own balanced output, just for fun, what topology would you used? Basically a NI configured unity opamp feeding both the + output and the INV input to a second opamp set to unity for the - output. Honestly unless you're supermatching resistors down to 0.25% you're only going to get cmrr of maybe 40-50db in any circuit, so it may be fruitless to try for anything more fancy. And you're right, those laser trimmed resistors and DC nulling in the diff drivers can allow some degree of positive feedback, but I'm not sure the block diagrams are accurate enough to base assumptions about the design on. Plenty of witchcraft happening on those dies that doesn't get explained. |
|
|
|
Post by matt@IAA on Apr 5, 2021 20:51:27 GMT -6
How do you keep that from being unhappy if someone shorts one side?
|
|
|
|
Post by svart on Apr 5, 2021 21:16:33 GMT -6
How do you keep that from being unhappy if someone shorts one side? Mostly a couple series resistors like 100-200 ohm. Helps decouple capacitive loads too. |
|
|
|
Post by svart on Apr 6, 2021 9:01:26 GMT -6
So here's the output I was thinking of last night. I decided to sim it up and see how good the output would null as a proxy for phase lag issues:  Honestly pretty good with a residual 30pV from 2Vpk/pk input which is probably due to propagation time through the first opamp. CMRR should be pretty decent with 1% resistors too but I haven't sim'd that. But then again, there's always a way to remove any potential phase/delay and see if it'll null a bit better:  Residual of 1500fV, which is arguably better, but it's just a sim. Resistor tolerance will matter here a lot more since the gain of 2 needs to be divided down if you want to maintain relatively similar input impedances or keep the total seen input impedance higher. I suppose you don't really need to divide the NI buffer for unity. I was thinking it might improve impedance matching between the opamps, but then again why would it matter here? It would be especially unimportant if you're using FET input opamps and their naturally high impedances, but then again using the divider resistors would give a more equal noise component:  15pV is pretty great, but it's just a sim. The lack of resistors decreased the null 10x but I'm also guessing that the resistor matching would need to be excellent anyway, so this option is probably more realistic. I think any of these would be great output drivers. Personally I would err on the side of fewer parts as tolerance error builds up when using more resistors. Layout and such tends to have much higher effect on CMRR than differential signal matching anyway. |
|
|
|
Post by matt@IAA on Apr 6, 2021 10:27:36 GMT -6
I'd be interested to see what the distortion looks like driving a balanced vs unbalanced input real-world at higher signal levels. Seems like your shorted amp will dump a ton of current. Plus you lose 6 dB of headroom. Also I'm not sure that in audio we care so much about how differentially balanced the signal is...what counts is matching the impedances for noise rejection, right? Not sure I see the simpler one as better. Also not sure that a discrete cross-coupled is the best solution, either.  |
|
|
|
Post by svart on Apr 6, 2021 11:13:15 GMT -6
I'd be interested to see what the distortion looks like driving a balanced vs unbalanced input real-world at higher signal levels. Seems like your shorted amp will dump a ton of current. Plus you lose 6 dB of headroom. Also I'm not sure that in audio we care so much about how differentially balanced the signal is...what counts is matching the impedances for noise rejection, right? Not sure I see the simpler one as better. Also not sure that a discrete cross-coupled is the best solution, either. I think most DIFF-SE adapter cables typically leave one of the signals unhooked. I don't know that many ground the unused output, at least none of the ones I've bought do, probably because most mass manufacturers assume that it'll be electronically differential since transformer outputs on mass-market gear are practically non-existent. But yes, the vast majority of noise rejection in differential audio signals would be from the impedance balance for common mode noise into a differential input. Differential outputs are much less necessary for noise rejection at audio signal levels, frequencies and distances. I'd go as far as to say that impedance balance is not even really needed in audio. Good shielding and proper ground potential is needed. Perhaps it's a way of cheapening out the shielding in cables and getting around DC offset between devices? CATV, a world that I work in, is completely single-ended and yet has some of the longest transmission lines out in the elements and still seems to work out to many GHz despite the signals being in the mV range and still having 30-40dB SNR. I always see more simple as better unless I have a specific need for something. I guess it's years of being bean-counted to death, but then again, simpler is also generally more reliable too. |
|
|
|
Post by matt@IAA on Apr 6, 2021 11:24:59 GMT -6
I’m just thinking from the user perspective of taking an XLR out and connecting to TS or RCA in. Then you get emails about why does your product sound bad. Or complaints about having to use a “special” cable to float the cold pin.
|
|
|
|
Post by matt@IAA on Apr 6, 2021 11:25:46 GMT -6
Also, sorry gouge. Haha nerd derail.
|
|
|
|
Post by gouge on Apr 6, 2021 15:51:57 GMT -6
All good guys. loving the discussion.
My day job is architecture so i get the design process and the conversation that goes along with it.
End of the day Matt i'm mighty grateful for your help and have a lot of faith in what you can do.
And svart your advice is always welcomed and again very respected too.
|
|
|
|
Post by gouge on Apr 6, 2021 16:06:12 GMT -6
I went through my stash last night and found a jlm world power kit and a jlm ac/dc kit. So I got the power supply covered.
On a side note found a quad eight 127a board i forgot i had...
|
|
|
|
Post by ragan on Apr 6, 2021 23:13:57 GMT -6
Diff/CM gain and CMRR is literally my homework right now. I'm thinking if I just keep an eye on this thread long enough svart and matt@IAA will end up doing it for me. |
|
|
|
Post by svart on Apr 7, 2021 8:28:49 GMT -6
Diff/CM gain and CMRR is literally my homework right now. I'm thinking if I just keep an eye on this thread long enough svart and matt@IAA will end up doing it for me. I did my homework years ago and I'm doing my daywork right now. You do your own homework 'cause I don't want to do anymore work. lol. |
|
