DBX 286A Modifications

[ecc83]

the problem with measuring with the computer is the variance of adding a converter in the circuit will give inaccurate results. Audio converters are not electronic lab/instrumentation grade in their coupling. So you would get what is capable for that signal chain.

Well, since you don't HAVE lab/instrumentation grade kit, do the best with what you have! A decent modern AI will give a noise floor of better than -100dBFS, I doubt the mic pre will get close to that.

Were you UK based I would loan you my Wayne/Kerr/Radford microvoltmeter, gladly!

Dave.

 

[innkeeper]

too bad no one has la DSA or an audio precision analyzer etc..
if the meter can limit the measured spectrum to 20k it wold give me an idea if its in spec.

oh and not so sure about the AI being quieter lol.....

take the DBX286a for example.
at +30db gain, the noise floor is at -92.7dBu (measured on my HP DSA)

most AI's FS is +28dBu so using that -100 dBfs that would be -72dBu noise.

Ever notice AI makers always spec ther products at the gain that gives the best S/N ration. Focusrite in example specs -122EIN at 60DB gain, sounds great until you realize the noise level is -62DB at 60db gain... which makes it useless for like dynamic mics for voiceover (unless you use an external booster)

but that is not a surprise to anyone who has tried it.

 

[innkeeper]

On the A model I would look to see if you can get the +/- 15 to be exact. But the power supply caps can do that along with other things.

But I do notice that going on inside the S model on a smaller scale. I only have an o-scope but that will be good enough to look at the noise and help determine its origin.

been there done that - this is "scoobis" from the other forum.

 

[ecc83]

too bad no one has la DSA or an audio precision analyzer etc..
if the meter can limit the measured spectrum to 20k it wold give me an idea if its in spec.

oh and not so sure about the AI being quieter lol.....

take the DBX286a for example.
at +30db gain, the noise floor is at -92.7dBu (measured on my HP DSA)

most AI's FS is +28dBu so using that -100 dBfs that would be -72dBu noise.

Ever notice AI makers always spec ther products at the gain that gives the best S/N ration. Focusrite in example specs -122EIN at 60DB gain, sounds great until you realize the noise level is -62DB at 60db gain... which makes it useless for like dynamic mics for voiceover (unless you use an external booster)

but that is not a surprise to anyone who has tried it.

Hmm? Well if I was noise testing a PRE AMP I would run its OUTput into the LINE inputs of my KA6 at around +4dBu and I know the noise floor is better than -100dBFS .

There are many people here who run F'rites and other AIs with SM57/58s and seem quite happy?

Dave.

 

[drtechno]

been there done that - this is "scoobis" from the other forum.

Hi scoobis this is drtechno/Audiospecific/Audiospecific2 on gearslutz. There is a moderator there that doesn't like me personally, and for the past year or so he's gotten into banning me without giving me a reason.

the A model does have a weird grounding scheme and both models are unbalanced stages. There might be some noise reduction we could try to apply, but I want to optimize this thing starting with the mic preamp first.

It might be debatable weather the SMT discrete mic preamp or a transistor array would be a better design than a instrument op-amp. But in all instances, a transformer input would be a better design than a transformer less mic preamp because removal of those phantom blocking caps can be a great improvement. Those transformer less with adjustable termination resistor that they labelled "input impedance" does change the loading what the microphone sees, but changes the passband response of the phantom blocking caps at the same time.

 

[drtechno]

Hmm? Well if I was noise testing a PRE AMP I would run its OUTput into the LINE inputs of my KA6 at around +4dBu and I know the noise floor is better than -100dBFS .

There are many people here who run F'rites and other AIs with SM57/58s and seem quite happy?

Dave.

well the dynamic mics you've mentioned have step up transformers in them and they establish a fixed impedance That is why putting a tab t-58 transformer on a SM7 improves its gain/noise floor performance. Doing this on the mic preamp side will improve its compatibility with all microphones because it can establish the loading without the loss associated with the phantom blocking caps. That is why a transformer output condenser mic doesn’t sound good on those interfaces.

An interface's signal to noise is directly associated with amplitude mainly because the virtual ground on the Vref is not as low of an impedance because of the esr of the capacitors in series. Now if the interface was like a DAV or a Burl, that is transformer input; its virtual grounds are derived off the center tap of the secondary and not in series with a capacitor. That is why they have a wider bandwidth response and better signal to noise and very little variances of it with amplitude.

 

[innkeeper]

Hi scoobis this is drtechno/Audiospecific/Audiospecific2 on gearslutz. There is a moderator there that doesn't like me personally, and for the past year or so he's gotten into banning me without giving me a reason.

the A model does have a weird grounding scheme and both models are unbalanced stages. There might be some noise reduction we could try to apply, but I want to optimize this thing starting with the mic preamp first.

It might be debatable weather the SMT discrete mic preamp or a transistor array would be a better design than a instrument op-amp. But in all instances, a transformer input would be a better design than a transformer less mic preamp because removal of those phantom blocking caps can be a great improvement. Those transformer less with adjustable termination resistor that they labelled "input impedance" does change the loading what the microphone sees, but changes the passband response of the phantom blocking caps at the same time.

i kind of gathered there is a bit of animosity toward you over there.
i jumped over here as you were the only one rally participating on that thread.. well participating and knowing what there talking about.
kind of got tired of being trolled by people 1/2 understanding of what is going on.

i already had this account here... i should see if i can get the name changed... but that is probably more problems then it is worth though.

I've given it some thought on the transformer. on the A model this would be a pretty easy thing to set up and test, making a module to plug into the existing socket i have for the instrumentation amp. and i just happen to be sitting on a crap load of nos GE darlingtons made for preamps, I might try and make something up using those.. just cause i have a lot of them on hand (literaly 1000's of them).

but at the moment i need to be a bit fugal, so that is going have to wait till a deal falls in my lap for a transformer.

id probably work on the A model as the S model i have is defiantly buggered. there is clipping in the processing section and something osculating at a low level. so not as simple a fix as i was hoping, and without the schematic, its a lot of tracing things out. i looked at it long enough to decide that its one of those snowy day projects lol.

 

[drtechno]

[MENTION=196623]innkeeper[/MENTION], Yea the s model does seem to have a gain structure issue as the noise floor goes up almost 40db when processing is engaged. I will address this. Now on your A model, the input transformer I would use would be the same one as the API312 and change the stage to what is listed on the datasheet of the op amp.

 

[innkeeper]

[MENTION=174130]drtechno[/MENTION]
are you able to do have the tools to do some calibrated noise measurements on your unit.
looking at the DBX286S vs the A model the noise appears to be a lot higher with processing bypassed
though, could just be this unit needing some work, as it is looks like the processor section is oscillating, so just could be related, noise leakage and all.
i defiantly see some weird stuff.

I just realized the way i put it i may not have communicated what i intended.
i meant to say the noise on the A model is significantly lower then the S mode.

but again ... it might be jsut a bad unit.. still need to fix the S model

 

[innkeeper]

[MENTION=196623]innkeeper[/MENTION], Yea the s model does seem to have a gain structure issue as the noise floor goes up almost 40db when processing is engaged. I will address this. Now on your A model, the input transformer I would use would be the same one as the API312 and change the stage to what is listed on the datasheet of the op amp.

I'll have to see what i can get my hands on

for noise with processing in i see about ~30db higher noise on the A model with processing on and the output gain at 0db.
though the complete story is in the table below.

there is a gain balance adjustment for processing enabled vs disabled. so before you think it is that bad, make sure its not just introducing extra unexpected gain.
i recently realized this myself, so not sure if my A model gain is balanced between bypass and processed levels or not. I'll ahve to check that and maybe re measure.

 

[innkeeper]

Hmm? Well if I was noise testing a PRE AMP I would run its OUTput into the LINE inputs of my KA6 at around +4dBu and I know the noise floor is better than -100dBFS .

There are many people here who run F'rites and other AIs with SM57/58s and seem quite happy?

Dave.

its easy for me to just see it on the DSA which makes it really easy to see if changes actually make any differences.

i got started with this madness when i had nise using a MXL BCD-1 dynamic into a focusrite at high gains doing voice over.
don't get me wrong i got around it by sending the mic though a cloudlifter type inline amp and had dealt with it before that by using a software noise gate.

on AI's ... yea true plenty are happy but i see a lot of complaints too when running lower sensitivity dynamics into an AI. guess it also depends what your using it for. the people i hear complaining are the ones recording voice overs. they typically get around it by getting a fethead or cloudlifter etc. many running sm7b's use them. so its easy enough to deal with.

I tell ya that extra ~25DB of clean gain before it hits the strip or the AI really makes a difference.

i do want to use the strip, and not post process. and i see the flaws at higher gains. Basically though i am wanting to get the strip quieter at high overall gain, and maybe do away with the cloudlifter and still maintain the same functionality, gain range and phantom power for my condensers. there is a lot of noise in the processing section though. but the cleaner the input is the easier that will be to deal with

I was against the idea of using a transformer, but the more i mull this over, the more it is making sense.

 

[drtechno]

It should make since because the transformer will step up the Mic voltage without a noisy gain stage. The trick is going to be selecting the two input resistors low enough for the signal without destabilizing the input stage bias that is formed from the resistor that goes to input to ground. I stated it would be possible to para-feed the input transformer just like some of the other solid state designs (like wsw/tab/telefunken). Right now I'm on my phone and can't access the off, but the 286a's output is on the compressor. I did noticed you noted a change in noise. This is because of power supply. It does makes me wonder what load resistance they picked for the output of the preamp. Because that part of it seems they are trying to get the Mic pre to develop high gain and at the same time have it drive a lower impedance output. Which would result in a high noise floor

 

[ecc83]

well the dynamic mics you've mentioned have step up transformers in them and they establish a fixed impedance That is why putting a tab t-58 transformer on a SM7 improves its gain/noise floor performance. Doing this on the mic preamp side will improve its compatibility with all microphones because it can establish the loading without the loss associated with the phantom blocking caps. That is why a transformer output condenser mic doesn’t sound good on those interfaces.

An interface's signal to noise is directly associated with amplitude mainly because the virtual ground on the Vref is not as low of an impedance because of the esr of the capacitors in series. Now if the interface was like a DAV or a Burl, that is transformer input; its virtual grounds are derived off the center tap of the secondary and not in series with a capacitor. That is why they have a wider bandwidth response and better signal to noise and very little variances of it with amplitude.

You seem to state a lot of things as 'facts' and yet I struggle to accept them. "Capacitor loss"? A typical pre amp might use 47mfd capacitors as 48V blocks and they have a reactance at 50Hz of about 68 Ohms. Thus the loss into to a typical 1k5 pre amp is less than 1dB (at 50Hz). Not, I would aver, very different from the transformer losses?

ESR trouble? Are these capacitors coming out of the 1950s junk box? Modern electrolytic capacitors are extremely low loss and the old stunt of shunting them with a 100nF or so is no longer needed at audio frequencies.

I am not qualified to judge but all I have read in the last 60 years or so tells me you are more in the Russ Andrew's camp than Ethan Winners.

I respectfully urge other readers to take this stuff carefully with a handful of salt and check other, reliable sources.

Dave.

 

[drtechno]

You seem to state a lot of things as 'facts' and yet I struggle to accept them. "Capacitor loss"? A typical pre amp might use 47mfd capacitors as 48V blocks and they have a reactance at 50Hz of about 68 Ohms. Thus the loss into to a typical 1k5 pre amp is less than 1dB (at 50Hz). Not, I would aver, very different from the transformer losses?

ESR trouble? Are these capacitors coming out of the 1950s junk box? Modern electrolytic capacitors are extremely low loss and the old stunt of shunting them with a 100nF or so is no longer needed at audio frequencies.

I am not qualified to judge but all I have read in the last 60 years or so tells me you are more in the Russ Andrew's camp than Ethan Winners.

I respectfully urge other readers to take this stuff carefully with a handful of salt and check other, reliable sources.

Dave.

I can Illustrate what I'm referring to, btw different dielectrics have different Xc values, but for argument's sake we can stick to the simpler formulas. Also since we are talking about the capacitive reactance, it changes in resistance at different frequencies. This results in changes in phase response at that frequency.

In the drawing attached: Using ECC83's Xc value of 68 ohms reduces the output of the microphone by a ratio of over 2/3 thus requiring to boost the gain that in turns amplify the preamp's noise floor and other interference from the foil phantom blocking cap that are acting like antennas.

 

[ecc83]

68 Ohms per 'leg' into the USUAL mic input of 1k5 does NOT result in a 3.7dB loss.

If you insist on using low load resistances then all bets are off but you could just use bigger caps! (and shunt them with foils if you are bothered but there is no need).

Dave.

 

[drtechno]

68 Ohms per 'leg' into the USUAL mic input of 1k5 does NOT result in a 3.7dB loss.

If you insist on using low load resistances then all bets are off but you could just use bigger caps! (and shunt them with foils if you are bothered but there is no need).

Dave.

I think you looking the wrong way, as the voltage source impedance is used and not the load resistance to determine signal attenuation from the source. After all I'm looking at signal drop from the source.

no matter how you look at, it its an attenuator. the source resistance is smaller than the series pi resistors so a lot of attenuation happens.


I'll use real numbers to put this in perspective. Using the pi formula, the dc resistance of the mic is 45.0817 ohms , the Xc is 68 ohms, which makes R2 approx. 136 which is an attenuation of approx 18 db of attenuation from the net source impedance of 35

If you want to use something like an online calculator to look at this, you could use this: Pi Attenuator Calculator - Electrical Engineering & Electronics Tools input the numbers 18db attenuation and 35 ohm impedance, hit calculate, and see what you get.

Diffrerence between the unbalanced PI and balanced PI is that balanced pi has two R2 that each one is 1/2 of the R2 in the unbalanced PI.


Keep in mind this attenuator changes with frequency, as capacitors are not linear in its AC resistance. which makes this attenuation variable (more or less) depending on what frequency is passed.

 

[drtechno]

its easy for me to just see it on the DSA which makes it really easy to see if changes actually make any differences.

i got started with this madness when i had nise using a MXL BCD-1 dynamic into a focusrite at high gains doing voice over.
the people i hear complaining are the ones recording voice overs. they typically get around it by getting a fethead or cloudlifter etc. many running sm7b's use them. so its easy enough to deal with.

Not familiar with the Mxl mic. But for curiosity sake, what is the resistance in ohms across pins 2&3?

On the SM7's I have I installed the same type of up transformer used in the sm58 to get it to behave like an sm58 on whatever mic preamp I use. If the mic pre sounds good with an sm57 or sm58, it sounds good with the modified sm7b. No cloudlifter needed.

 

[ecc83]

I am merely treating the mic as a zero resistance source because it is the CAPACITORS we are discussing.

Yes, of course Xc varies with frequency but in a perfectly predictable way and does not seem to bother most audio designers.

Dave.

 

[drtechno]

getting back with the 286A for scoobis/innkeeper, I see the problem area in the insert circuit. Its not configured like the usual dc coupled transformer-less insert.

Typically the insert is usually AC coupled, but since the board is not provisioned with the solder places for these capacitors, I would just remove the ceramic caps so it would be the same as a valley people 400 strip.
VP 400:


DBX 286A
View attachment 286 (Project 1) RevA4.pdf

 

[drtechno]

I am merely treating the mic as a zero resistance source because it is the CAPACITORS we are discussing.

Yes, of course Xc varies with frequency but in a perfectly predictable way and does not seem to bother most audio designers.

Dave.

Well capacitors are the nessary evil if you want to go the cheap way of ac coupling. Higher end designers will use transformers (like Rupert Neve) or find a way to dc couple it ( like the Jim Williams EE (rip) and Jim Williams the audio tech/modifier)