Your link to the schematics gives nowhere. Fortunately, they are all available on archive.org

You will see that the wet/dry pot on the DBX290 is fed from the input op-amps (dry send) and a buffer op-amp from the reverb unit (wet return). All the op-amps in those two paths are ancient 4558s, which were very quiet jobs for their day.

The point is that you report the hiss occurs when the wet/dry control is round at the wet end, but is not there when turned to the dry end. Since the dry end is fed from 3x 4558s in series, the conclusion is that the hiss is not coming from anything on the PCB, but from the reverb unit, which is inside the box but external to the PCB.

Hi Boswell and thanks for the reply.

It's just one pcb in the unit. I guess you 're saying about the digital reverb side of the pcb. The schematics shows snd/rtrn (and other parts) but it’s one pcb.

I'm certain that it has nothing to do with the in/out op amps. Also nothing to do with main power supply rails.

It occurs only on when the reverb enters the mix

So I assume it's somewhere around the digital effect side of the unit.

I can try changing the 2 digital side voltage regulators, and the 3-4 simple npn transistors (2n4124) 

Could also be the 2 converter ics (the only smd ones) or something with the ram ics (but there are plenty of those. Or even clocking - crystal issues.

What do you think? If you can see something that could be changed just let me know please.

Below I ‘ve sent a couple of clear pcb pictures.

I hadn't had a view of the inside of a dbx290 until you posted the photos, and the component positioning makes much more sense now. The reverb processor is the Digitech S-Disc VLSI chip. It is fed from the Xilinx FPGA , which does the formatting of the S-Disc serial digital signals to and from the DAC and ADC chips.

To track down your hiss problem, I would start by putting a temporary harmless short circuit across either (but not both) C27 or C28, the 3n3 caps in the DAC reconstruction filters. When you switch on again, set the wet/dry control to fully wet and listen to the stereo output. One channel will still hiss as before, but the other will be different, probably almost silent. If that is the case, the noise problem is in the design of the Digitech chip, and there is not a lot you can do about it.

A few questions for you: Is the hiss the same amplitude in both L and R outputs? Are you putting full amplitude (+4dBu) signals into the unit? Apart from the noise when wet, does the reverb unit work as expected? If so, is the effected input sound similar in frequency content to the dry sound?

Absolutely right Boswell. Thanks for that.

Tried the shorted cap and the result is a dry channel regardless of mix (but when on wet it's like the unit was bypassed, not even the cheap opamp hiss. 

The other channel was problematic,

The reverb unit works flawlessly as a unit. And the recording signal is similar frequency wise. 

Here is a recording of the unit as a send to a very good clean piano instr. I play with the fx mix knob. The decay is very high. The output near unison with the in. The in is not even warm. Try to listen that there is some kind of feedback (or just about to start) in the sound.  Could it be the "sample and hold" section?

If i keep the mix low (around 50%) , the decay around 80%, and in - out unison and put it in the mixer send, it' good as a small reverb, but for bigger settings, the noise makes it like a weird pedal thing. Pretty useless.

Well, I agree with you. If that is a genuine line-level signal out of your dbx290, I can't see how it would have been sold as a professional unit.

How much gain are you having to put on the output signal? You should be able to feed the dbx290 output into a line input on your audio interface and leave the gain near minimum (depending on the make and model of the interface).

What type of cable are you using to make this connection? The dbx290 outputs are only impedance-balanced, not signal balanced, so you can use either a TS lead (guitar lead) or a TRS lead (balanced) with the same result over a short distance.

Yes i have a very good interface - RME. And good pre amps.

The thing is that i can’t have longer tails and big rooms out of the reverb unit (even if some of the programs are made for that sort of sounds), if i jeep the setting as i described. The longer tails start after 70% wet, which is loud noise territory.

I tried with trs patches snd ts ones. Same results.

I don’t think it was sold like that. I mean it’s nothing special, after the early 80s, about having average clean reverb effects. And i can’t see how the digital section can malfunction if there is no other shorted component.

Only if the digital section voltage regulators gave gone wrong, or maybe the two square “computer” ICs gave been placed in a wrong way. But they would be at least shorted, if they were placed in wrong pin sockets. Am I correct?

It’s not out of necessity. More out of experimentation i care about fixing it. 
 

I've had a look at the recording you posted, and there is evidently excess noise when the mix control is near the fully wet end. The maximum noise (sampled at times that did not have any of the piano sound) had a worst-case amplitude of about -25dBFS, compared with around -60dBFS when fully dry. That means there is over 35dB noise difference between wet and dry.

                                 dry piano              wet piano                            dry noise                    various wetness noise    

The only way that I could see that this is other than due to a fault would be if your source audio input (the piano recording) into the dbx290 was at a very low amplitude, and you have had to wind up the gain in the RME line input to get an acceptable piano level for the recording. You didn't reply to the question in my last email about how much gain you applied in your interface. Could it be that you are feeding the dbx290 with microphone level inputs rather than line levels?

It's easy to do a test using the RME TotalMix software to set up a DAC output and an ADC input as an insert loop for the dbx290, and in this way you can see whether the noise is then acceptable. Use as much microphone gain as you can on the RME box before overload (maybe 40-50dB), and then have 0dB gain on the dbx290 return. You would not want to record normally like that as there is an extra DAC-ADC conversion stage in the record chain (in addition to the ADC-DAC in the dbx290), but it would at least help identify where the problem lies.

Thanks Boswell. 

It's always line levels for me.

Haven't got any problem with any other of my much better (korg, roland, yamaha) reverb units.

With the same piano module. Same leads. Same converters, Same channel. Same card. Same levels. Same location.

It;s the unit. But which part of it?

I can't really understand how to do the test. Jumper the in - out from ADC and DAC? Where and how?

Are we talking about a real piano being recorded using microphones, or are you using the wired output of some keyboard?

Please explain what your input is and exactly how you have the dbx290 connected. Where does the signal come from to feed the dbx290 input, and where does the output go. Crucially, how much gain are you having to apply to the dbx290 return to get nominal recording levels? What level (in dBu) are you using as your "line level"?

No. Just a piano synth - piano module

I know you 're talking about RME loopback connection (i just misunderstood the DAC - ADC)

This connection is nothing special.

Just a stereo line level synth (piano module) outputs,

to a stereo line in channel of a mixer,

a couple of mixer outs to a pachbay,

and from there in a couple of ADC inputs

and from there digitally to the RME card

and from there digitally to a couple of DAC outputs, or straight headphones from RME analog phones out. 

I know my connections. And as i've told you, the same path is already tested with other similar equipment, after i realized the problem. Actually I have a korg reverb on the same path now. And all are good

What I realise is that something after the DBX ADC section makes the signal either way too overdriven or plain digitally distorted.

Sorry about forgeting it.

I also have to add that I use the reverbs (like the DBX)  either used  auxiliary send-return on the line mixer and then added with the aux pots of each channel (like in the example) or straight on the patch bay and then make the appropriate in -out connections just before send the channels to the ADC. On that case i play more with the mix pot of the reverb unit. The results are the same noise - wise

What you have to work out is whether the excessive wet noise from the dbx290 is a fault condition, or whether it is normal behaviour. By "normal" I mean that, when the controls are set for optimal signal levels, the noise may be present in the output, but at a level that is not a problem. This is a 30-year old reverb design, and it could be necessary to cut it a little slack.

What I suggest is that you wind the output gain control on the dbx290 down to minimum, and play your keyboard into the unit. Adjust the input gain control so that the overload LED flashes occasionally on the loudest sounds from the keyboard. That step is important. Now set up a medium length decay reverb and turn the mix control to fully wet. Finally, wind up the output control until the keyboard sounds are at a good listening level. Is the noise still present at an irritating level? If so, the conclusion must be that your dbx290 has a fault in the reverb engine, and it is then up to you to decide whether you want to do anything further with the unit.

If under those conditions the reverb sounds are acceptable, then go ahead and use it, always making sure that the input gain is set correctly for the signal levels you are feeding it with.

It's also worth setting the noise gate level in the unit so that the reverb engine is shut off if for input signals that are below a certain amplitude. See page 9 in the manual for details of how to do this.

Positively not “normal” Boswell.

I know quite well good gain staging, so in that case i play on the very safe side. A safe - 15dbVU in. I have tried also around -6dbVU and -20dbVU. Noise gate at the max settings. The results are always this bad signal when the effect is obvious.
Yes it could give a usable small reverb with the large settings set at their minimum, but that’s not what a 90s digital reverb unit, by a normal company, can do. 
Even if it’s a cheap unit, the factory noise figures according to experts are very good. My unit’s are beyond horrible. 
There is obviously a problem. And the area where it lies (digital reverb side- betwen ADV and DAC) is also obvious.

So, the question is, if I want to do something to at least try a probable fix, what your suggestion would be?

The best thing would be to let us hear it in normal use so we can get an idea of the actual level of noise vs undistorted signal. Its feasible I suppose, that the noise is coming from an unexpected source, but the question is hard for you to quantify without access to another good unit, so you can trace the audio path and find exactly where the audio deteriorates? Sadly, with these fetching little on ebay nowadays, id probably just buy another, i no longer enjoy soldering, so always balance repairing gear against my usual pay rate. If a repair takes me a half day, then that looks bad vs ebay replacement.

Thanks for the input Paulears.

After inspection from a much more experienced technician on digital units, it’s likely a bad ram chip (have to replace all 5 of them, improbable to find the bad one, without take all 5 out and then use sockets, or you just throw them and put new in there - slightly time consuming but cheap) or bad dac (slightly more expensive). More likely ram.

if you’re pro i absolutely agree with your approach on cheap common units, but if you ‘re a hobbyist and you do it for the sake of repair and understanding-learning different units, you just insist.

I agree with your technician who inspected the dbx290 that, if it's a fault, the likely culprit is one of the 64Kx4 dynamic RAM (DRAM) chips. Finding which of the five is faulty would be relatively straightforward using an oscilloscope, but there are still ways to diagnose this without.

Looking at your waveform of the fully wet output with no (or very low level) audio input shows that your problem is between -20dBFS and -30dBFS. Given that each bit has 6dB more significance that the next lower one, it's likely to be the 4th or 5th most significant (U36p2 or U27p17), if the fault is a single bit. If all four outputs of a chip are faulty, it's most likely to be the 2nd most significant chip of the five (U27). Note that the DRAM is used only for delay data storage, and is connected to the top 20 bits of the 24-bit data bus.

If you don't have access to a scope, what I would do is set the 290 unit up on a bench, set fully wet with a grounded input signal. Plug in an amp and speaker to the 290 output, then wind up the 290 output gain until you can hear the faulty output noise loudly. Connect one end of a probe lead to the 290 circuit ground, then use the other end of the probe to selectively short-circuit the DRAM chip U27 data pins to ground. These are (in order) pins 17, 15, 3 and 2. Doing this will not damage anything as long as you are careful to touch the probe only on to the data lines. Listen for changes in the noise. If it almost completely stops when you short just one pin, then you have a single-bit fault. If it reduces by different amounts on different data bits, then you have a complete faulty chip. If you detect no difference using the pins of U27, try the same pins on U36 (the most significant chip) or U24.

If you do have a dual-channel scope, look at the CAS signal (any DRAM pin 16) and trigger the scope on the falling edge of that. Use the other scope channel to go round the RAM data bits (pins as previous paragraph), looking at the end of the CAS period for data bits that are significantly different from the most significant (U36p17). This will only hold for the most significant two or three DRAM chips, as the LSBs will naturally be a little noisy in any case.

If you find it's a single-bit fault, you could fudge a work around by cutting a couple of pins and wiring swaps of the data connections on the faulty bit by hijacking the (assumed good) least-significant bit of the data bus (U13p2). Losing this would be inaudible. The work-around would avoid having to extract DRAM chips and find replacements.

Really thanks for the thorough answer Boswell. Will keep it in mind.

For me personally it’s much easier and less time consuming to replace all 5 chips, since i found cheap replacements in china and they will arrive at the same time. I will use sockets to check the old chips as well. In case they don’t work i will try you methods. 
But your suggestions are excellent for reference (not only for this model).

The diagnostic procedure I described was as an alternative to the brute-force replacement of all the DRAM chips.

If replacement of all five DRAM chips does not cure the original problem, there is little point in using the diagnostic method I outlined, as the reason for it not working after chip replacement will be something different.

I saw in an online repair forum that a member had tried replacing the DRAMs in a dbx290, and it created more problems than it solved. The concensus response was along the lines that the design may have been exploiting a property of the original DRAMs that they could perform faster than their marked speed if the OE was held permanently active, which it is in the 290, as it uses early-write timing.

A faulty DRAM chip is not the only possible cause of your problem. Before you go changing the chips, I would try the simple technique of prying the VLSI device half way out of its socket and then pushing it firmly back in. A dirty socket contact can give a similar effect as a dead bit on a DRAM. There are cut-outs in the chip socket walls bottom left and top right that are meant for getting the tangs of an extraction tool into. I would use the twisting of a fine-bladed screwdriver alternately in the cut-outs to cause an upward movement of the chip body. You only need a few mm - don't spring it right out. Use your thumbs to push the chip very firmly home before re-applying power.

Actually that’s one of the first things i tried, since this has happened befire to my line 6 mm4 and it’s apparently very common to happen. The only thing i cannot really tell is if the socket or the chip pins are perfect. Other than cleaning them and pushing them back carefully there is nothing else i could do. I ‘ve also read the same thread in the other forum, but it’s way too technical to get my head around ( i have to admit i don’t enjoy working and dealing with digital electronics in general )

Finally it was one of the 5 dram chips. 

I replaced all 5 with cheap chinese current ones and it worked immediately. Good old noise free digital reverb. Not the greatest but very decent and absolutelly noise free )(i mean clearly audible noise).

And then by removing 1 after the other and using the old ones, i found the faulty original one. Just one faulty d ram chip.caused the whole havoc

Thanks for the help Boswel and all the other members. 

Good - that's what we had predicted. Was it U27?

Yes it was u27 Boswel.