connecting input directly to ucd, no input buffer

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
If instructions for this exist, please point me to them, the only information I have seen has been regarding removing decoupling caps and changing the op amp power supply.

I am gearing up for fully active loudspeakers. Shortly, I will be posting crossover pcb, +/-12 v power supply pcb, project writeup, etc.

The next step will be to put the crossover circuit board and ucd amp in one box. I have a spot for adjustable gain on the crossover, so I can jack it up to 5, and hopefully connect the output directly to the UCD, bypassing the op amps. The OPA2134 in my board are low dc offset, I don't believe they will be a problem except for maybe the bass due to an LT circuit potentially amlifying any interstage DC emitted (in that case I'll use decoupling caps).

Any directions on connecting the circuit boards directly, bypassing the op amps. Without any further instruction, I may just try connecting the signal ground to amp ground and soldering input directly to the existing opamp output. I anticipate dealing correctly with the grounds will be a problem due to my unbalanced inputs.

I will report back in two to three weeks if everything goes well.

Lee
 
Hi Lee,

If it helps, the AC coupling caps are _after_ the op amp buffers. So if you remove them you can solder your inputs directly to the comparator, or output side, of the coupling cap via, and you're good to go.

Use the signal ground to connect your supply to, not the power/amp ground.

Trace the inputs back to the op amp and you'll then know which coupling cap was acting as inverting and non inverting, then you should know all you need to make a go of it.
 
So I can just remove the ac coupling caps and connect the input to the non-inverting cap, and the ground to the the inverting cap, as shown in the picture? I located the caps, and circled their output pins.
 

Attachments

  • topost.jpg
    topost.jpg
    90.6 KB · Views: 599
Yeap, I'm zombie brain at this hour but it looks right :)

You can also try being super slick and tweaky (....ahah..... sigh) and trace those pins over to the header pins of the daughter card, solder them direct to that, all for the sake of saving ~1cm of signal length and totally avoiding those wimpy lil traces.

I'd make a real effort to get them soldered fast though, and heatsink the header pins you're soldering to on the other side of the board (probably don't want too much heat getting into the daughter card!).

Whatever you feel most comfortable doing, but I think that will work.

Disclaimer time.... I haven't tried any of this myself :D

Let me know how it turns out.
 
Thanks, sounds like a bit of a warning about the daughter card. I think I'll be safer soldering to the main pcb. The other option sounds pretty extreme, and I'm out of money.

I should get around to doing this early next week, my girlfriend is going on vacation with her family, and I will be playing electronics a lot. I've got circuit boards lined up for the day she leaves, and they have an adjustable gain so I'll be able to mod one channel only and compare. Though I hesitate to make any sweeping comments other than works/not works, good/bad, I'll see if I can get it done in time for the columbus diy meet so more experienced people can give their equally subjective but more reputable opinion.
 
Yeah it's a little bit of a warning, they never hurt. Proceed with caution and all. I don't think you'd want enough heat getting into the daughter card to reflow anything, or even create a poor joint with the header pins going into it. You can visually inspect that and touch up if need be though. Some forceps clamped to the pin on the top side of the board would make an excellent heatsink while you solder it.

Probably the real advantage of doing it this way, far more so than avoiding that little bit of trace length, is that you can keep your wires tightly twisted right up to where you're soldering. If you do it the other way, at the cap solder pads, you're going to have a nice little loop of greatly increased area, be a great way to pick up noise.

If you trace it out, you'll see the pins you need are the top 2. I think that's how I'll do it when I try it.. eventually.
 
I see you've given this some thought, I suppose I'll try the daughter board. It will be my payment for free advise, which I will inevitably nag for again in the future.

You think radiated noise is more important as the connection is closer to the switching components? If you think so I may go get some proper shielded microphone cable, I was just going to try twisted pair for experimentation and sub in better stuff if necessary.

At least if my goal is to remove the op amp, and I break it, I can buy one of the cheaper versions with no worries about op amp flavor.
 
Hm, I think experimenting with normal twisted pair first would be the best idea, why complicate it if you don't need to?

The HF carriere wave will appear at these inputs since you're skipping the buffer stage, that's likely to be more of a concern than proximaty to the switching node.

What good a shield would be since you're driving it single ended I'm uncertain of, probably twisting will do the job nicely.

Another thing you could try would be a ferrite choke or clamp on the twisted pair, close to the module, I'd probably opt for that before trying to add a shield.

If you do decide to try it with a shield, I think you'll also conveniantly find the signal ground maybe just one pin down :)

Pop the coupling caps off, like you were going to do anyway, and you should be able to leave the op amps in place for all your testing, can't break em that way! It might be a good idea to tie both their inputs to ground though (at the usual connector) to ensure they don't their outputs flying all over the place.
 
I'm going to try this tonight, or tomorrow night. Does the ucd need to be connected to a low impedence output.

i.e, my crossover has a 200 ohm output resistor. Is it better to remove this or leave it in? There is approx 1.5 feet of cabling between crossover and amp boxes.
 
mazurek said:
I'm going to try this tonight, or tomorrow night. Does the ucd need to be connected to a low impedence output.

i.e, my crossover has a 200 ohm output resistor. Is it better to remove this or leave it in? There is approx 1.5 feet of cabling between crossover and amp boxes.

I'd replace the 200 ohm with a lower value, say 22R or 47R.

The direct input impedance on the + input is 8.2k afaik, so it's a safe bet to have an as low as possible impedance on the source side.
 
Member
Joined 2005
Paid Member
The idea of the input buffer on the UcD modules is threefold, firstly they offer a bit of gain and secondly they make sure the UcD modulator stage is fed from a low impedance source. Thirdly, but very important is the input buffer improves CMRR and PSRR of the whole amplifier considerably since it is a full balanced design. Bypassing the input buffer will only work if you make sure the alternative does just that. Personally I think it is a fruitless exercise, the input buffer featuring the AD8620 with CRDs is extremely transparant and neutral. The only improvement you could possibly make is a full discrete input buffer that has substantially better CMRR, THD, SNR, bias current and slew rate, no small task I might add, considering the AD8620's excellent specifications.

Best regards,

Sander Sassen
http://www.hardwareanalysis.com
 
I have completely removed the input stages on all my ucd modules. They still work. More comments when I have time.

SSassen - I hope to exploit some of the benefits you mention by placing allowing the active crossovers to function as input buffer, allowing them to operate at higher gain. Also, if I decide the system works satisfactorily (looks good so far), I can save money on the expensive version of ucd. I'm trying to work this out from a system perspective, and hopefully with the help of knowledgable people like yourself, I can achieve a better system. I agree that this mod is not necessarily appropriate for those without a comparable gain stage in their preamp/crossover in close proximity to their modules. If necessary, its not hard to balance my crossover output, or make any modifications, but I'd like to eventually package the crossover and amps together and use them to their fullest potential.
 
Hi Sander,

There's just one thing I disagree with you on. The AD8620 with CRD is extremely clean sounding, but it's far from being transparent or neutral. If you can get away with driving the pure comparator directly it could prove worthwhile, especially in the case of mazurek, you save alot by not having to experiment swapping parts constantly and risking the PCB... time, money, risk. Even if you are up for that game it's worthwhile trying just to see the kind of neutrality you should be aiming for. I agree with the rest though, the input stage is a brilliant aspect of the design and worth trying to make the most of.

Regards,
Chris
 
Here's a brief writeup of the procedure and some results.

[Modding the UCDs]

I removed the capacitors as suggested. Cheap desoldering solutions like solder wick/sucker were too big/destructive for the small space, so I just heated one pin at a time and rocked the capacitors out, taking care to rotate around and let everything cool.

I traced the pins on the ucd daughter board. Top away from board is inverting input, top closest to board is non-inverting. I began using some forceps as heatsink, but I found that the + input did not have any suitable attachment point, so I went without them. The header did a good job of dissipating heat, and the daughter board did not subjectively get very hot (of course I was careful not to spend too much time). I did not end up using a heatsink for any solders.

I didn't have enough tools/hands to hold everything in place, so I did a sloppy initial solder job on both inputs, then twisted the wires together. After twisting the wires, they had some support, and I reflowed the solder.

The new inputs are very conveniently placed. I left the old inputs as they were. In future installations, I will probably just short + to - on the op amp at the board, and only connect ground and 'on'.

[Performance]

I measured two ucd modules side by side, fed by my standard stereo setup with everything on and nothing playing. The modded module had a DC offset of 8 or 9mV. The stock one had a DC offset of 4 mV. Switching on and off the amplifier still yields no audible popping noise. I am using stock hypex power supplies.

I tried to measure the noise level, but even practically touching the cone, its below the measurement ability of my RS meter, and I didn't feel like getting out a measurement mike. Subjectively, the system noise level dropped very significantly at all preamp volume levels when I decreased the amplifier gain. However, when I made up for it in my crossover, the noise level went up back to the original level. I believe this just points to background noise in the preamp/source as compared to my amp and crossover. I have not spent the time to trace its origins, my system is slightly too sensitive anyhow due to a +4.5 dB baffle step gain, and perhaps all 4 ohm drivers. So I may reduce the crossover gain slightly.

I tried the system first on an older driver. After that success I tried it on a speaker. It worked with and without an output resistor on the crossover. I was a little nervous that I should wait until the systems are in the same chassis so the wiring is reduced. I have not checked radiation.

I do not have the tools nor the memory to objectively measure a change in system character. Suffice it to say that it did not noticably get worse. I believe that this would be a good project for someone building an integrated amp, or a fully active loudspeaker.

If anyone has the time to stop by columbus oh, and compare, you are welcome.
 
Alright, classd, you're forcing my hand, its gotten hard for me to share these types of opinions lately, over the past months I've read and analyzed thousands of pages of peer reviewed scientific literature, and had the voids pointed to me by high class researchers.

I like it, I would say it is more of the same, and a little more fluid. It removes a little bit of the subconscious edge. Its not like the old version is bad by any means. My criteria for an excellent system is one that can play your typical boring or progressive audiophile junk, but can also smoothly play for example old Hendrix cds, and while the notes are piercing they're still just barely smooth. It moves the amp slightly more in that direction if you know what I mean. Similar effect with other music, just a slightly more convincing fluidness.

I had no intentions to, but I have been listening to my stereo ever since I got home from work, and listening at higher spls than usual when studying. I'm definitely happy with my system. Keep in mind though, in the past couple of days I redid the crossover and replaced the drivers, so there is some confounding here.

troystg. I'm convinced that if your crossover is close enough to the amp, its not a big penalty to run that connection single ended. And putting the balanced stage in front of the crossover would regain some of its benefits I bet. As far as total noise level in the system goes, the crossover off and its outputs muted to ground leaves an ultraquiet noise level. I think this is a fair evaluation of the crossover to amp connection because the output impedance should not be much different when it is on (no output resistance). The crossover itself does generate a little hum, but I may just remove the psu connection to ground. I suppose doubling your crossover boards for a fully balanced system could also take care of that, but sure would be expensive.
 
Mr. Erath is working on integrating his Feedback controller network (which has a cross-over built in) into a pair of bi-amped speakers consisting of the UcD180AD's on top driving the Scan Speak Revelator tweeters and B&W 6.5" mids and UcD180ST driving 2 ten inch Aluminum woofers for bass.

I love his FCN but hate not using the balanced out of my pre for the 20 ft runs to the active powered speakers.
 
Supposedly, I believe BobEllis stated his next active crossover boards to be printed have balanced in. Also, look at jensen transformers for some input buffer circuits, and balanced / unbalanced converters. Sound.westhost.com and Hypex also have some drawings of balanced receivers and transmitters. I think it would not be too hard to get a small, saw 2x3" perfboard and put a balanced reciever on the input to the crossover, then put a hard switch to switch between balanced and unbalanced input. You could decide yourself whether or not it is worth it to balance the output.

That project is along the lines of where I'm going. I'm trying to convince my dad he needs the awesomist loudspeakers ever, and I'll put a revelator tweeter and woofer, and 2 10's in there. Any more info on that feedback network, that would be an interesting next technology to look at.
 
Another note.

I believe a line is balanced if it has equal ouput impedence on the + and - inputs. If you remove the output resistance of the op amp, they are more equal. Then you can borrow the chassis ground as the sheild. So non-inverting input is op amp output, inverting input is crossover ground, and sheild is chassis ground.

This argument comes from my understanding based upon reading the documents at jensen transformers.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.