I managed to get the setup for the LXmini into the MiniDSP 4x10 HD so I could use it with my speakers. It sounds good to my ears.
In a completely unscientific "test" I put my ear against the tweeter. I do hear a faint hiss. As far as I can tell by listening the MiniDSP 4x10 HD is about as quiet as the Focusrite Saffire PRO 24 sound card I've been using until now.
Tom
In a completely unscientific "test" I put my ear against the tweeter. I do hear a faint hiss. As far as I can tell by listening the MiniDSP 4x10 HD is about as quiet as the Focusrite Saffire PRO 24 sound card I've been using until now.
Tom
From the Grimm LS1 White paper:
5.2 Clock
The clock circuit is the same as that used in the CC1 except that the sampling rate is set to 93.75kHz instead of one of the more traditional audio rates. This is specifically done to improve the performance of the SRC chip. An uncommon clock frequency reduces the odds that mix products between the incoming clock and the internal clock fall inside the PLL loop bandwidth of the SRC.
That sounds reasonable. I would think that if all clocks are either the same or integer multiples of each other you'd get the same effect. Basically any mixing products would then overlap with the phase detector spur and go undetected.
Good to know that Bruno et al. have thought about that stuff, though.
Tom
Good to know that Bruno et al. have thought about that stuff, though.
Tom
Cool, thanks for sharing. The slightly improved best case SnR with the higher output swing follows what I've seen in output buffer design. MiniDSP's found the same physics as anyone else, no clever tricks.
Focusrite's spec is 29.5 uV RMS A. That's for what I think MiniDSP would call a 6.36V setting. So the equivalent figure of merit for MiniDSP would be about 31.4 uV RMS A, putting Focusrite a little over 0.5 dB quieter. Probably about half that is attributable to slightly quieter NJM part selection by Focusrite and probably the other half's down to somewhat cleaner buffer implementation on Focusrite's part.
In cost no object design it's possible drive the buffer almost completely out of the picture and have output noise dominated by the DAC. For the CS4272 used by Focusrite and the CS42528 used by MiniDSP that offers 8-9 dB improvement, closer to 12 if one opens up DAC selection but keeps the design realistic for delivering signal across an interconnect (rather than optimizing for headline numbers at the DAC outputs). More reasonable but what I'd still consider somewhat borderline buffer design---typical of RME price points, say---is about 6 dB quieter, though one can do a couple dB better if the BOM stretches to quieter DAC selection. Hypex's DLCP is a bit of an exception as it hits this latter performance point at a more reasonable price.
Perhaps the most interesting part of this is there's hints, mainly the AK4495, pressure on headline numbers is motivating DAC manufacturers towards higher analog rails. That's cool as it's both the most effective gain structure and the most cost effective. But I'm not sure anyone besides TI and maybe Analog has a mixed signal process which could go to ±15. AKM's already 10 dB or so beyond what most power amps can take advantage of anyhow.
To the best of my knowledge that's the case.tomchr said:
Nice!tomchr said:
"Hypex's DLCP is a bit of an exception as it hits this latter performance point at a more reasonable price"
Selecting parts is not the most important thing, how you implement them has more impact. And don't forget the circuitboard design: in digital audio and class D amps it is the most important part. Most of the design effort at Hypex goes in the latter.
De dac and analog part in the DLCP takes 1/3 of the board space with 6 local regulators, current sources, leds for voltage references and capacitors for power decoupling.
Selecting parts is not the most important thing, how you implement them has more impact. And don't forget the circuitboard design: in digital audio and class D amps it is the most important part. Most of the design effort at Hypex goes in the latter.
De dac and analog part in the DLCP takes 1/3 of the board space with 6 local regulators, current sources, leds for voltage references and capacitors for power decoupling.
That seems well intentioned but naïve regarding precision analog design. It's more effective to reason about performance tradeoffs in context rather than assigning significance a priori, though that does require putting in the effort to analyze design tradeoffs and understand the solution space well enough to identify the optimum solution for a given price point.
I think basically what you're trying to say is Hypex is a company which does this effectively. That's correct but do take a closer look at Bruno's posts and other writings. For example, like other skilled class D amp designers Bruno's quick to point out the layouts required for amps with hi-fi performance figures can't be obtained without the compactness of surface mount parts. Assertions attempting to treat these as independent considerations are, frankly, kind of pointless.
Similarly, the notion circuit implementation and layout can deliver DAC performance better than datasheet figures hinges on the assumption the product designer is able to find meaningful optimizations which the manufacturer missed. In practice it's far more likely that if one's got a design around a DAC specified at 114dB DnR whose output noise is DAC limited then the most effective way to lower the output noise is to switch to, say, a DAC of 120dB DnR. Those aren't arbitrarily chosen numbers. You'll likely find it helpful to refer to the datasheets for the Cirrus parts Focusrite and MiniDSP use along with the AK4396 on the DLCP and spend some time noise modeling in Spice.
Explains why the noise in my completely unscientific test appeared the same. I could of course just measure the Focusrite...
The technical performance of the DLCP is impressive. The performance of the software, apparently not so much (see comments on that a few pages back). That will improve over time, though.
Tom
I have to say, if I was buying another active setup would seriously consider the hypex. However as I got a 2x8HD, vol-fp and digi-fp for €250 value for money wise miniDSP has it for me. Assuming I can drive it successfully. A lot of new ways to screw up when you go digital
I found the 4x10 HD pretty easy to get going. My TV remote uses one of the formats the MiniDSP recognizes, so getting input select and volume up/down going was a couple of button clicks.
For the XO, I had screen shots of the MiniDSP 2x4 setup for the LXmini saved, so transferring the information to the 4x10 HD was pretty easy. The only snag was that the 2x4 has six biquads in the EQ whereas the 4x10 only offers five. The LXmini uses a pretty high-Q notch around 13 kHz for the tweeter. I just moved that to the input EQ instead. That does mean that the woofer gets a 13 kHz notch as well, but by then it's pretty far down the LR2 XO slope and combined with the dropoff in driver efficiency at HF, it has no measurable impact on the end-to-end system performance.
Tom
You had a setup to start with. I'm starting from scratch with all my inexperience in speaker design and a slightly unique setup. It will be a learning experience. At least the garden is big enough to do some measurements in if it stops raining long enough
You know with the 4x10 you can drop sample rate and get more PEQs? Not sure it's a trade-off I would make though.
You know with the 4x10 you can drop sample rate and get more PEQs? Not sure it's a trade-off I would make though.
If you don't mind using analog in on the 4x10, here's a trick for getting more PEQ bands:
I2S "Loopback" on the 2x8 / 4x10
-- Jim
Equalising gain
I intend to use the Mod 86 that I'm currently building as part of an active system - for the treble channels in fact. However the other amps I'll be using have 26db gain. So, I'm wondering which is the better option; I can either change a resistor in the Mod 86 to increase the gain to 26db, or I could use the 8v output on the 2 channels of my 4x10HD that will feed the Mod 86 but use the 4v output for the other channels that will feed the 26db amps. ( The difference between the 4v and 8v outputs being 6db according to the manual.)
Is one of these options better than the other ?
Edit- a third option would be to adjust the gain in the digital domain using the DSP output control.
I intend to use the Mod 86 that I'm currently building as part of an active system - for the treble channels in fact. However the other amps I'll be using have 26db gain. So, I'm wondering which is the better option; I can either change a resistor in the Mod 86 to increase the gain to 26db, or I could use the 8v output on the 2 channels of my 4x10HD that will feed the Mod 86 but use the 4v output for the other channels that will feed the 26db amps. ( The difference between the 4v and 8v outputs being 6db according to the manual.)
Is one of these options better than the other ?
Edit- a third option would be to adjust the gain in the digital domain using the DSP output control.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.