Hello. I understand the basics of xover design but I'm new to the term and meaning of "ladder network". Can anyone educate me on what it means, how they're designed, and why they're helpful? Can you give me an example of one in use and why it was used? I have no background in electrical matters and doing my best to improve my knowledge and my basic xover skills. Thank you.
A ladder network is also called an all-pass network. It is used to change signal phase without altering response, giving the effect of a delay. It does this over a finite band.
It is one of a number of tricks a person can use to bring phase together in a crossover. Different crossover needs require different solutions due to the shape of phase variation involved. This one fits where one of the drivers is physically forward of the other.
An alternative that brings a similar result is the asymmetrical filter.
It is one of a number of tricks a person can use to bring phase together in a crossover. Different crossover needs require different solutions due to the shape of phase variation involved. This one fits where one of the drivers is physically forward of the other.
An alternative that brings a similar result is the asymmetrical filter.
Back in the dark days before cheap DSP, ladder networks were used to alter phase in a passive network.
In 2019, they're completely pointless. The parts are really expensive, easily $50 per channel. For that price, just get a miniDSP.
And this is coming from someone who loves passive networks.
I've just double-checked the cost of the all-pass delay network on my new MTM design (necessary for the LR2 acoustical slopes). £22.61 per channel including VAT and postal charges, aka $28.21 at the current exchange rate, and the UK isn't exactly the cheapest place to buy components. That's with Jantzen 17AWG air core inductors and CrossCap MKPs. So if you love passive networks, you can rest easy when it comes to the cost of all-pass delay circuits -a quality implementation can be achieved without spending a great deal. The cost only goes up if you want to spend more on the components, just as it does for any other part of the filter. While I like the miniDSP, I can't buy a new one for that sort of money, let alone the extra amplifiers (& their linear PSUs) I would then need. Nor do I have the space for them, or the power sockets nearby: I would have to either rewire my room, or run an extension lead almost all the way around the skirting board to provide the three extra sockets required, which I am not keen on doing, especially since the socket that would feed the extension lead is itself required for other purposes.
Last edited:
I understand what you're saying. But where is the fun in that? Where is the creation? Where is the art? For me speaker building is much more than quickly getting good sound from a pair of drivers. I don't want my hobbies to be easy...I want them to challenge my mind, my imagination, my creative skills. I could take out my digital camera and take a perfect shot of the sunset. Or I could place a canvas in front of me with a couple of brushes and some oils and create my own sunset with my own hands. In the long run which of those choices would most likely enhance my life? DSP allows me to quickly achieve good sound. But passive teaches me about electronics and how to read schematics, makes use of my creative skills, allows my hands to build xovers and my nose to smell the solder of my own creation. And I can sit back and look at that painting or listen to the sound that I created with my learned skills and say, "I did that". If good sounding speakers were my only goal I'd simply find and purchase a commercial design that I like. I totally agree with you the miniDSP solves many problems and IS in fact the answer for most speaker builders. And I completely agree that passive is now antiquated. But (for me at least) using miniDSP doesn't really count as something that can be called a hobby. I'd rather LEARN how to create my own ladder network than turning a knob or pushing a button. One is easy, the other is a hobby.
There isn't a lot of information (not for loudspeakers anyway) available on all-pass delay networks; the basics are as you would expect for any electrical filter. Most are 1st order -you can kick that up to 2nd if desired, but 1st order is usually sufficient to provide the necessary level of delay.
In design terms, 'standard' / nominal electrical values don't usually work for the same reason nominal electrical values don't usually work for other parts of the filter: they don't account for impedance variations, or interactions with other components. So to design them, unless you want a lot of soldering & measuring, you need modelling software with sufficient freedom of configuration. Of the popular ones, that means LspCAD, VituixCAD, SoundEasy, XSim or Boxsim (I may have missed some others, but they're the big names). Some have optimisers -to be honest, I tend to just display my target slopes (e.g. LR2), get some approximate values in for the main crossover components, some equally approximate values for the four all-pass components, and then empirically adjust the latter in the software. The two inductors are of equal values of course, as are the two caps, so this isn't particularly arduous. Make sure the software is displaying the phase with the system & individual response curves so you can line things up properly. You may (probably) find this will also require some minor tweaks to the main XO high pass values. It may sound like a rigmarole but most of the time, once you've got a bit of experience in establishing what does what, you can usually zero in on the necessary values quite quickly.
In design terms, 'standard' / nominal electrical values don't usually work for the same reason nominal electrical values don't usually work for other parts of the filter: they don't account for impedance variations, or interactions with other components. So to design them, unless you want a lot of soldering & measuring, you need modelling software with sufficient freedom of configuration. Of the popular ones, that means LspCAD, VituixCAD, SoundEasy, XSim or Boxsim (I may have missed some others, but they're the big names). Some have optimisers -to be honest, I tend to just display my target slopes (e.g. LR2), get some approximate values in for the main crossover components, some equally approximate values for the four all-pass components, and then empirically adjust the latter in the software. The two inductors are of equal values of course, as are the two caps, so this isn't particularly arduous. Make sure the software is displaying the phase with the system & individual response curves so you can line things up properly. You may (probably) find this will also require some minor tweaks to the main XO high pass values. It may sound like a rigmarole but most of the time, once you've got a bit of experience in establishing what does what, you can usually zero in on the necessary values quite quickly.
Back before all the fancy watch-as-you-twiddle crossover software was available, I found it useful to think of the all-pass nature of ladder networks in terms of delay rather than phase. Basically, you are electrically(rather than physically) offsetting the tweeter backward to match the acoustic center of the woofer. Using a single delay network, the amount of delay that can be achieved is tied to how high in frequency the delay is held constant. If you want more delay at a given frequency than a single network provides, you can cascade multiple networks as needed.
Attached is a spreadsheet you can use to get an idea of how much delay is possible and what component values to start with when optimizing your crossover. The input cells are highlighted in blue, where you enter the load resistance and the frequency for -90deg phase shift. An example screen shot is also attached showing values I used to move a tweeter backward by about 1.0” for a 1500Hz crossover.
Note that the ladder networks will only operate as ideal all-pass filters(ie change phase without changing magnitude) if the load impedance is constant and purely resistive. This is another reason you may need to tweak values from theoretical starting points.
Attachments
Last edited:
I just wanted to say that I appreciate your comments there. I meander around this site to read various things.... I might make a comment or two but I've always wondered why someone goes through this turmoil (in my mind).
I'm more of a plug & play so I'm viewing what you do from a totally different perspective and have always wanted to ask others why they do some of these things "to make it harder on themselves"
You essentially answered that which I've never asked, thank you!!
🙂
Getting delay for a given amount of msec requires more and more sections as frequency gets higher. Getting anything very signifant up to high frequency gets impractical pretty quickly. You can play with this in XSim pretty easily, it has a 'circuit block' for both first and second order all-pass, and you can cascade, and look at what it does in phase (that was actually the feature that XSim was originally made for). You MIGHT (but most likely won't without using some physical placement delay as well) be able to use passive at a lower cost than with a miniDSP. But only if you get it right the first time. With a DSP, you can get a number of milliseconds over the entire audio band and change it at will with no further cost, you can't get anywhere remotely near that using passive.
- Status
- Not open for further replies.