Hello All,
I've been very happy using a Rane MA6S amp to biamp my LCR speakers. Plenty of power and headroom. They're a little noisy, but not too bad. In an attempt to improve their performance, I replaced the input opamps with LME49713 and the noise was reduced a bit. My main question is, is it possible to bypass the input section and use as little of the amplifier circuitry as possible, with the intention of reducing noise and distortion by avoiding unnecessary opamps and capacitors. I'm using a highly modified Behringer DCX2496 as my crossover. The output section of my DCX2496 has been replaced with high quality film capacitors connected directly to the DAC output pins. Since there is no DC on the outputs (2.5Vp-p) is it possible to bypass the input opamps, high pass filter and opto/servo clip protection circuitry and connect the input directly to Q10 on the MA6S schematic? Or is the Z2 opamp (5534) required?
Any advice on moding is welcome. My other option is to build a 6 channel LM3886 based amplifier. Opinions are encouraged
Thanx
Chris Mitchell
I've been very happy using a Rane MA6S amp to biamp my LCR speakers. Plenty of power and headroom. They're a little noisy, but not too bad. In an attempt to improve their performance, I replaced the input opamps with LME49713 and the noise was reduced a bit. My main question is, is it possible to bypass the input section and use as little of the amplifier circuitry as possible, with the intention of reducing noise and distortion by avoiding unnecessary opamps and capacitors. I'm using a highly modified Behringer DCX2496 as my crossover. The output section of my DCX2496 has been replaced with high quality film capacitors connected directly to the DAC output pins. Since there is no DC on the outputs (2.5Vp-p) is it possible to bypass the input opamps, high pass filter and opto/servo clip protection circuitry and connect the input directly to Q10 on the MA6S schematic? Or is the Z2 opamp (5534) required?
Any advice on moding is welcome. My other option is to build a 6 channel LM3886 based amplifier. Opinions are encouraged
Thanx
Chris Mitchell
Yes, Z2 is necessary. The amp will not work without Z2.
If you wish to bypass everything, connect to R15, the input of Z2.
However, if your source has no volume control, you will have no choice but to connect to R12. Then upgrade C7 to BlackGates or something similar.
Regards
Mike
Your noise is coming mostly from R13 and R14 once you've turned down input gain a bit (which I would expect to be doing - gain's a tad high at 33.5 dB and far more than needed when you're running anything close to a full nominal +4 dBu into it). Apparently those are used for dropping gain in BTL mode?
If you can live without that, remove the optocoupler Z3, jumper one of these resistors and replace the other by 220 ohms or so.
Note that the amp is giving away some practical input CMRR due to the choice of R2 and R4. I'd use something more like 100k (1% MF should be OK), which would increase common-mode input impedance to 50k.
If you can live without that, remove the optocoupler Z3, jumper one of these resistors and replace the other by 220 ohms or so.
Note that the amp is giving away some practical input CMRR due to the choice of R2 and R4. I'd use something more like 100k (1% MF should be OK), which would increase common-mode input impedance to 50k.
It must be my eyes but I'm not seeing any noise at R13 and R14.
Z3 is a Vactrol VTL5C2/2. This opto-coupler is a compressor. It does not increase gain. It is only activated when the output of the opamp Z2 hits a certain level. The compressor route is via R38.
There is a 2x gain when the input is driven in the balanced mode (R5,R6,R7,R8). I seriously doubt a 2x gain is the cause for noise unless the source is noisy to begin with.
The most likely cause is the gain setting at the feedback network. R24, R25 sets the amplifier's gain at 46x (33.4dB), which I feel is rather high. Unless one is experienced enough, I would not recommend lowering the gain as it could lead to instability.
Z3 is a Vactrol VTL5C2/2. This opto-coupler is a compressor. It does not increase gain. It is only activated when the output of the opamp Z2 hits a certain level. The compressor route is via R38.
There is a 2x gain when the input is driven in the balanced mode (R5,R6,R7,R8). I seriously doubt a 2x gain is the cause for noise unless the source is noisy to begin with.
The most likely cause is the gain setting at the feedback network. R24, R25 sets the amplifier's gain at 46x (33.4dB), which I feel is rather high. Unless one is experienced enough, I would not recommend lowering the gain as it could lead to instability.
It's the resistors themselves. At 5k1 each, they make up the largest part of source impedance at the opamp input. Thermal noise voltage of 10k at 295 K in 20 kHz BW is 1.8 µV. At the output, that's 83 µV from these resistors alone, with volume turned down all the way. Including the 5534 and feedback network, we're at 100 µV - by no means particularly noisy, but a value that decent integrated amps can reach as well, with an extra 10 dB of total gain.
Replace 2x 5k1 by 220R, and minimum noise floor drops to less than 40 µV. If you have speakers of 100 dB SPL / 2.83 V / 1 m sensitivity, you can probably live with 3 dB of noise at 1 m anechoic easily...
(You can treat the amplifier as a big opamp and stuff part values into a noise calculator, with e_n and i_n as given for the 5534.)
As a nice extra, that should also improve distortion resulting from impedance unbalance at the inverting and noninverting opamp inputs. Stock, unbalance is 9k-11.5k, the modified version would improve this to about 1k8 or less, with a null at an input gain setting slightly below -20 dB or so.
(With .33R emitter resistors and I don't know how much quiescent current, I would expect practical distortion to be output stage limited though. 8 ohm speakers definitely preferred.)
Ah yes, that makes more sense. That's what I was suspecting initially, but I couldn't find a signal input.
Actually I wouldn't call that a 2x gain. Overall gain for both inverting and noninverting inputs is unity. Of course the contributions of both are added up, but both of them already counted in balanced signal levels as well. So if you send 0 dBu (balanced) into the input circuitry, you'll get 0 dBu (unbalanced) out.
I would expect the source plus input circuitry to be dominating noise in many cases when input level is cranked up all the way.
The combiner opamp already sees 10k total (btw, you can reduce R5-R8 values with a beefier Z1 - 4k7 should be easily doable with the '49713, but you need precision MF parts) - that's 1.8 µV + x.
Consumer soundcard, 106 dB SNR @ 1 Vrms (no PGA): 5 µV.
Ditto @ 2.2 Vrms: 11 µV.
Semi-pro soundcard, 115 dB SNR @ 9 Vrms: 16 µV.
So even in stock form, you may have to turn down input gain by 20 dB for source noise to drop below amplifier noise.
I therefore suggest that the OP do exactly that. Chances are that noise will be pretty much eliminated, while obtainable output levels still are more than sufficient. And it's free.
I've looked at the stock DCX output stage. It's not that bad actually. Granted, it does not take advantage of distortion cancellation in balanced inputs (only on the DAC side), and with all the output amplitude delivered by a single opamp, distortion could probably be a bit better, but the output is also adapted to unbalanced inputs easily. Bypassing it potentially allows high-frequency hash to wreak havoc in following amplifier stages if those are not up to snuff. But to each their own. At the very least I would leave the 121 ohm series resistors in (or install extra ones, precision MF) - capacitive load driving is not specified for the AK4393 outputs, and stability with more than a few kOhms of load may be marginal. (The datasheet is, unfortunately, rather tight-lipped when it comes to these things, which is frustrating. The quality of the output stages suggested also seems rather dubious, what with stupidly low feedback network impedance in Example 2. Apparently someone noticed later on, and the same circuit in the AK4396 datasheet looks a fair bit more reasonable. The properties of the DAC's analog sections might have changed as well, of course.)
Last edited:
Resistor noise is very minute. Noise to me is when you amplify a microphone signal 1000x from 1mV to 1V.
If I send a 1V balanced signal into the input, there will be 2V at the unbalanced out.
I appreciate the info provided. Since the cable length from my DCX2496 is <18", I'm confident that unbalanced is fine. Also, I don't need the HP filter, gain reduction vactrol or gain control potentiometer.
The output of my DCX2496 is my version of the passive film cap mod done by so many. I use 2 capacitors in parallel, a 10uf poly film and a .01 oil paper Sprague. Could I connect this directly to the input of R15 and Z2 pin 3?
The output of my DCX2496 is my version of the passive film cap mod done by so many. I use 2 capacitors in parallel, a 10uf poly film and a .01 oil paper Sprague. Could I connect this directly to the input of R15 and Z2 pin 3?
You can connect directly to R15 but a better way is to remove C7. This will disconnect the front end completely.
Then, solder a 1K 1% metal film resistor to R15. The other end of the 1K connects to your DCX2496 output caps. The 1K with 270pF (C31) will help to prevent RF from entering the input.
Then, solder a 1K 1% metal film resistor to R15. The other end of the 1K connects to your DCX2496 output caps. The 1K with 270pF (C31) will help to prevent RF from entering the input.
You can sim it if you don't believe me.
But just to be sure that we agree on the definitions, 1 Vrms balanced is 2.83 Vpp between "hot" and "cold" (which may mean 1.41 Vpp to ground for both of them), and 1 Vrms unbalanced is 2.83 Vpp between signal and ground.
The inverting side has a gain of -(10k/10k) = -1. The noninverting side has a gain of 1 + (10k/10k) = 2, but the input signal is attenuated to half its original amplitude by a 10k/10k resistor divider, so overall gain is 1/2 * 2 = 1 there as well.
(In order to calculate these gains, always assume the other input remains tied to ground.)
If the input signal now is unbalanced at either input, overall gain obviously is unity, and anything in between is a linear superposition of these two extremes - including the case of minus half the signal on inverting and plus half the signal on noninverting.
You do often find a gain of 2 in balanced output stages, which may consist of a unity-gain buffer for the noninverting side and an inverting amplifier with a gain of -1 for the inverting side. If you send 1 Vrms unbalanced into one of these, you'll get 1 Vrms "hot" plus 1 Vrms "cold" = 2 Vrms balanced.
Doesn't mean you can't have a ground loop.
The inputs on this amp are always balanced and should prove some CMRR as-is, balancing source impedance merely gives much higher CMRR still. If I have my math right, 20k121 on one leg vs. 20k0 on the other will give a CMRR of about 44 dB, while even a worst-case 0.1% mismatch on the 10k resistors in the input stage would still give 54 dB.
(If you rewire the amp for direct unbalanced input, the CMRR you get is none - 0 dB. And minus the gain control, you wouldn't even have a way of attenuating the ground loop crap - or source noise, for that matter.)
As I outlined above, if you truly didn't need the gain control you probably wouldn't have mentioned noise.
This "less is more" paradigm is stupid audiophoolery. It's quite definitely not universally true. Try building a 60 dB amplifier with negligible distortion and a bandwidth of at least 100 kHz or so using just one ordinary audio opamp. Good luck, you'll need it. The same with 3 stages in series - easy as pie.
The DAC provides a differential (balanced) output. The amplifier provides a balanced input. If I were looking at mods, I'd definitely want to exploit that, fancy caps or not. Actually you wouldn't even need caps then, as +2.5 V would be well within the MA-6S' input stage's common-mode range. (I assume both the amp and the DCX more or less agree about ground potential by means of a central safety earth connection.) The input stage is AC-coupled anyway.
BTW, you can hardly even measure the effect of generously-sized bipolar electrolytic coupling caps, let alone hear it.
You could. I hope I have made clear why I'd consider that a misimprovement though. Unless you enjoy shooting yourself in the foot, obviously.
Last edited:
There's unity gain on the inverting and non-inverting inputs. We can agree on that.
My point is if there is a 1Vrms at the inverting input and a 1Vrms at the non-inverting input, the output of the opamp is 2Vrms (all referenced to 0V of course).
But if there is 1 Vrms on the noninverting and another 1 Vrms of opposite phase on the inverting, that's 2 Vrms worth of differential mode signal! Hence why I'm saying 2 Vrms in --> 2 Vrms out.
So we agree on what the circuit does, just not on the definition of balanced levels. That's something, isn't it?
Let's see whether we can't sort the definition issue out.
I see common mode and differential mode as an orthogonal coordinate system that you are projecting ground-referred voltages to (i.e. it's a coordinate transformation thingy):
Vdm = Vni - Vi
Vcm = (Vni + Vi) / 2
Vni = Vcm + Vdm / 2
Vi = Vcm - Vdm / 2
The small difference of a factor of two (which ultimately is what we're arguing about) is needed for consistency, as we'll see.
My definition of balanced levels is Vbal = Vdm.
We can study this at DC to make things clearer (this would even work in practice if one left out the coupling caps).
Let's say Vni = + 1 V, Vi = -1 V - a balanced / differential voltage. Then we get
Vdm = 1 V - (-1 V) = +2 V
Vcm = (1 V + (-1 V)) / 2 = 0 V
What about an unbalanced voltage of +2 V? I.e. Vni = +2 V, Vi = 0 V. This yields
Vdm = +2 V
Vcm = +1 V
As you can see, Vdm is the same. A voltmeter between the two lines will display the same. Connect a load between the two lines, and it'll draw the same current and dissipate the same power.
Now you would be calling the first case "1 V balanced", and the second case "2 V unbalanced". Oops.
Now let's go back to AC and use a 1:1 isolation transformer to convert a balanced signal of 2x 1 Vrms to unbalanced. The result obviously is 2 Vrms unbalanced. So your "1 Vrms balanced" would become "2 Vrms unbalanced". Double oops.
If, however, you define Vbal = Vdm, it is "2 Vrms" on both sides - consistency is restored.
So why does Vcm get a factor 1/2? That's simple:
Assume Vni = Vi = +1V. Then
Vdm = 0 V
Vcm = +1 V
Without the factor 1/2, Vcm would become +2 V. And you'll probably agree that it would make no sense to define it like this. A plain ol' average with the factor 1/2 in seems a lot more sensible.
Last edited:
So we agree on what the circuit does, just not on the definition of balanced levels. That's something, isn't it?
Regardless of definitions, the fact still remains that there's always a gain in a balanced input. It's basically a summing amp with two inputs, one in absolute phase and the other, in anti-phase. Noise, which is in common mode, gets cancelled in the summing.
So is your point that the low amount of noise introduced by the balanced input circuit is negligible as compared to to loss of CMRR? I think I understand. I've already swapped the opamp for a LME49740, would swapping those resistor values and losing the vactrol theoretically reduce the noise even further?
I'm not sure if it matters, but the ADC and DAC in the DCX2496 have been swapped for the AKM5394 & AKM4396.
Sent from my iPhone using Tapatalk
I'm not sure if it matters, but the ADC and DAC in the DCX2496 have been swapped for the AKM5394 & AKM4396.
Sent from my iPhone using Tapatalk
I think this is becoming an exercise in learning more about noise as it relates to gain structure. A subject I would like to know more about. There is a very faint (<55db@1ft) white noise hiss coming from the high frequency drivers. There is a slight difference when the pot is off, but no more the a couple db at best.
Would the noise floor benefit from changing the rectifiers to a different version?
I know this is a mid range pa amp of some vintage, but I'm enjoying the process as much as the result.
Sent from my iPhone using Tapatalk
Would the noise floor benefit from changing the rectifiers to a different version?
I know this is a mid range pa amp of some vintage, but I'm enjoying the process as much as the result.
Sent from my iPhone using Tapatalk
That's not too bad. Not great but tolerable. I doubt you can hear the hiss at your listening distance (say 6ft~10ft) but it's still annoying knowing it there. Usually, I have to put my ear right at the tweeter to hear any hiss.
What you are hearing with the volume pot in the off position is noise from the amplifier itself. When the volume is full clockwise, you are hearing the amplifier with the section preceding the volume pot.
Since there is only a small difference, we can safely deduce that most of the noise is from the amplifier itself.
My guess the main culprit is the gain structure at the feedback. A gain setting of 46.4x is rather high, resulting in noise. I set mine at 33x and my amps are very silent.
Looking at the schematic, I can see why your amp has such a high gain. R13 and R14 in series is approx 10K. With R15 at 35.7K, the signal is attenuated quite a fair bit. So, to make up for the loss, the feedback gain is increased.
To reduce the noise, you will have to isolate the section preceding R15. This will remove R13 and R14. The next step is to change the amplifier's gain to between 30x ~ 33x. The problem with changing the feedback gain is the possibility of instability. You will need a scope, signal gen and someone with the experience to mess around with those resistors (R24, R25).
Hmm. That's less difference than I'd have expected. It would obviously depend on how much input noise you have, of course. With the input shorted, I would expect no more than a few dBs indeed. Can you try that? If that gives you virtually no difference at all, the amp would be noisier than expected.
If the difference remains about the same, however, tackle R13/R14/Z3 next. For a start, you could just replace both Rs by 100-220 ohms each or so, whatever you've got floating around. Standard 1% cheapie metal films will do just fine.
Your speakers are some of these, I guess? They're all fairly sensitive, typically 89 or 91 dB (per 2.83 V / 1 m, I assume). It wouldn't take more than about 180-200 µV of speaker output noise for a noticeable hiss - 250 µV will do on 88 dB speakers that I have. "8 ohms compatible", har har - that's as useless and meaningless as a "4-8 ohm" spec. Methinks someone wanted to cover up the fact that impedance dips a fair bit below 6.4 ohms in parts.
Definitely not - but the regulators may be worth a shot if we find that amp noise is higher than expected. Opamp power supply rejection typically is quite high but by no means infinite. 78xx/79xx regs are pretty generic, but that doesn't mean they're all created equal in terms of noise. On Semi parts (MC7815/MC7915) are supposed to be quite good, IIRC.
It also seems that the regs are followed by very little capacitance that could swallow noise - I'm seeing only 10µ in the power supply + 10µ per amplifier channel (and only 16 V types at that, rather tight for 15 V). While space on the amp cards is tight, it looks like there ought to be enough space around C49 and C50 to fit a few hundred µF each.
Note that removing any of the amplifier cards requires removing the output transistors from the heatsink, and replacement of the insulator pads may be needed upon reinstallation. Read disassembly instructions carefully.
That loss is a whopping 2.6 dB worst-case, 2.1 dB min. The designers may not even have thought about it much. A 33.5 dB gain actually is quite common in pro amps, or was back in the day at least. It allows for an input sensitivity of <0 dBu in a 150 wpc amp. Consumer amps rarely go higher than 30 dB, 26 dB is more typical, and even lower gains can be found. Typical input sensitivities for these tend to be 1-1.5 V.
I find it odd that DC resistance is not matched between the 5534 inputs. Maybe the value of R15 was a compromise between unbalanced operation with just R24 (44k2) on the other input, and BTL operation where that is reduced to 22k1 (R24||R23). It's not like that makes a huge difference in terms of output offset (just a few millivolts), but good DC balance tends to be beneficial for distortion at AC as well. Not like it matters much on this amp - with a bootstrapped VAS and .33 ohm emitter resistors, it's hardly a king of load invariance. And then you can still run the thing in BTL - with 16 ohm speakers only, I assume. I think those are pretty uncommon even in the 'States.
And R23.
R24 only (or R24 = R23). You'd need to make sure that the amp doesn't break into oscillation even with a severe capacitive load of 10 or even 47 nF.
Speaking of stability, L3 is shown as a pure inductor. Is that a pure air coil or is it would around a resistor in the 10 ohm vicinity? It is common to put a resistor in parallel with the coil in order to reduce ultrasonic impedance peaking, though I don't think that this would affect stability here since the RC zobel network comes before the inductor.
Last edited:
PS:
I forgot one detail here - it would probably be necessary to install a protection diode from reg input to output as outlined in the datasheet. ON Semi's MC7900 datasheet states that none is generally required up to 10 µF, but the stock amp already is at 10µ + 6x 10µ = 70µ, and adding a few hundred more would certainly be pushing things.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.