PERSPEX QUASAR ver 2.0 prototype build

Dedicated to those large boxes at one end of the room
JamesD
Old Hand
Posts: 997
Joined: Tue Oct 09, 2007 4:26 pm
Location: North Yorkshire

#181 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by JamesD »

Hi Mark,

Yes you were the first to mention this and to do the research on what valves did this to an objectionable degree and find the very few that didn't - mainly pre 1976 6N6Ps... which is all I had used in the Aikido before moving to pentode frontend Aikidos that greatly reduce this effect...

ciao

James
User avatar
Nick
Site Admin
Posts: 15705
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#182 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by Nick »

Acrylic "casework" ordered...
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
User avatar
IslandPink
Amstrad Tower of Power
Posts: 10041
Joined: Tue May 29, 2007 7:01 pm
Location: Denbigh, N.Wales

#183 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by IslandPink »

:bounce:
Looking forward to seeing this !
"Once you find out ... the Circumstances ; then you can go out"
chris661
Shed dweller
Posts: 2556
Joined: Mon Feb 22, 2010 7:29 am
Location: Sheffield

#184 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by chris661 »

Cool!
Interested to see which crossover you go with.

Chris
User avatar
Toppsy
Shed dweller
Posts: 2400
Joined: Mon Jul 23, 2007 12:10 pm
Location: red rose country

#185 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by Toppsy »

Poste deleted. Fat finger syndrome pressing the wrong keyboard key! :roll:
Last edited by Toppsy on Wed Sep 08, 2021 12:04 pm, edited 1 time in total.
User avatar
Toppsy
Shed dweller
Posts: 2400
Joined: Mon Jul 23, 2007 12:10 pm
Location: red rose country

#186 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by Toppsy »

chris661 wrote: Fri Sep 03, 2021 6:54 pm Cool!
Interested to see which crossover you go with.

Chris
As will I Chris. When last we had my mini meet and measured the test baffles, we now have 3 different OX'ers to play with:
  • James original 1st Order series XO*
  • 2nd order series OX based on actual measured speakers on test baffles and ESP research of series filters
  • 2nd order conventional parallel filter also based on actual speaker measures of the test baffles
Note *. After measuring the speakers Chris was playing around with different filters on his laptop and found that reversing the polarity of the shunted inductor on the Lowthers to the 1st order series filter flattened out a dip in the upper midrange frequency response that was showing with the original designed filter. Subsequent listening tests showed a big improvement in SQ.

Chris, as I mentioned in your new speaker thread, when Nick takes delivery of the acrylic baffle cabinets I shall be organising another small mini meet my place. Hopefully with the help of those attending, Nick can come to some agreement to the final incarnation of the project. My bet is on either the 2nd order filters will win the day. From my initial listening of the filters, the 1st order series filter just doesn't have the same bass extension/integration and lower midrange detail that the 2nd order filters give and without loosing any the Lowther sweetness. Still they are not my speakers and Nick may well choose another option which may or may not include any the above filters. Some interesting times ahead, me thinks.
User avatar
IslandPink
Amstrad Tower of Power
Posts: 10041
Joined: Tue May 29, 2007 7:01 pm
Location: Denbigh, N.Wales

#187 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by IslandPink »

There's a further option.
I discussed this with Nick after getting home and mulling over what I heard and the (surprising) result of needing to swap polarity on one of the drivers.
I think the 100uF cap that was being used on the series crossover was too large, and not dropping the PM6A quickly enough in the bass. So I advised Nick to try something in the range 50 to 70uF with drivers in same polarity.
"Once you find out ... the Circumstances ; then you can go out"
User avatar
Toppsy
Shed dweller
Posts: 2400
Joined: Mon Jul 23, 2007 12:10 pm
Location: red rose country

#188 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by Toppsy »

Yes that can be done very easily and will give a forth and fifth experiment to try out, counting the cap swaps as numbers 4 and 5.

However if memory serves me, I think we started out with something around the 56uF mark and were advised this was perhaps too low a value? But thinking again this may have been when we were using the Li Audio full range driver with the Supravox and might not have been tried with the Lowthers?

I have 56uF and 70uF quality polyprop caps we can try. However, I am yet to be convinced that a 1st order filter i.e. simple shunted cap on bass drivers gives proper control of the LF in a multi-way speaker, be that conventional (parallel) 1st order or series 1st order. I know there are others on this forum that swear by 1st order filters and fine tuning the cap and inductor values by ear and who am I to argue they are wrong. As said this is just my opinion based on the twenty odd 2-way speaker designs I have build over the years.

Purley out of personal interest I might try this experiment in the next couple of days, as I will have to reinstall the drivers back in the test baffles.

The problem we have is one of accurate and true measurements of the drivers Thiele parameters provided by the driver manufacturers to work with to properly design a filter. The data provided by Li Audio was pure fiction and writing to them about the vast differences we got in our measured results to those they officially publish was not helpful in the least, as Nick will no doubt testify. Supravox provide some the Thiele data, but only some that which is provided is of use for a crossover designer and one has to rely more on the frequency plot gained from actual in room measurements and some guess work. Lowther provide the best data.

This is all diverting from the thread of course so I must apologise to Nick for my rant. The good news is when Nick is able to bring over the acrylic baffles we now have five different configurations to experiment with a try out. Something to look forward to.
User avatar
Nick
Site Admin
Posts: 15705
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#189 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by Nick »

As said this is just my opinion based on the twenty odd 2-way speaker designs I have build over the years.
Though it should be said that most of those 20 odd 2 ways have had similar sorts of drivers so maybe not as wide a spectrum as may be thought.
The problem we have is one of accurate and true measurements of the drivers
Though there is no reason why we couldn't measure the drivers to get better info.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
chris661
Shed dweller
Posts: 2556
Joined: Mon Feb 22, 2010 7:29 am
Location: Sheffield

#190 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by chris661 »

Toppsy wrote: Wed Sep 08, 2021 12:03 pm
As will I Chris. When last we had my mini meet and measured the test baffles, we now have 3 different OX'ers to play with:
  • James original 1st Order series XO*
  • 2nd order series OX based on actual measured speakers on test baffles and ESP research of series filters
  • 2nd order conventional parallel filter also based on actual speaker measures of the test baffles
Note *. After measuring the speakers Chris was playing around with different filters on his laptop and found that reversing the polarity of the shunted inductor on the Lowthers to the 1st order series filter flattened out a dip in the upper midrange frequency response that was showing with the original designed filter. Subsequent listening tests showed a big improvement in SQ.

Chris, as I mentioned in your new speaker thread, when Nick takes delivery of the acrylic baffle cabinets I shall be organising another small mini meet my place. Hopefully with the help of those attending, Nick can come to some agreement to the final incarnation of the project. My bet is on either the 2nd order filters will win the day. From my initial listening of the filters, the 1st order series filter just doesn't have the same bass extension/integration and lower midrange detail that the 2nd order filters give and without loosing any the Lowther sweetness. Still they are not my speakers and Nick may well choose another option which may or may not include any the above filters. Some interesting times ahead, me thinks.
Quick note, Colin - it was reversing the polarity on the Lowther driver which improved things. The two drivers weren't summing properly in the lower midrange (where, as noted, there's some overlap), but inverting the Lowthers meant better summation, giving a more even frequency response overall.

It might also be of interest to try an LCR network across the Lowther, to flatten out its impedance around Fs. That would improve the behaviour of the highpass filter - see graph below, borrowed from one of Rod Elliott's articles:

Image
Left is the highpass filter operating into a resistive load, and right is with a "real" (simulated drive unit) load.

The Lowther is likely to be doing something like this to the simpler crossovers. The ones I worked up all put an L-pad around the Lowther, which will smooth out impedance variations a bit, so the bump at Fs wouldn't be so severe.

On the subject of measurements, the ones we did in-room proved pretty useful, but outdoor (or large room) measurements would be useful to cut out the reflections.


Let me know on the mini-meet.

Chris
User avatar
IslandPink
Amstrad Tower of Power
Posts: 10041
Joined: Tue May 29, 2007 7:01 pm
Location: Denbigh, N.Wales

#191 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by IslandPink »

chris661 wrote: Thu Sep 09, 2021 7:01 pm Quick note, Colin - it was reversing the polarity on the Lowther driver which improved things. The two drivers weren't summing properly in the lower midrange (where, as noted, there's some overlap), but inverting the Lowthers meant better summation, giving a more even frequency response overall.

It might also be of interest to try an LCR network across the Lowther, to flatten out its impedance around Fs. That would improve the behaviour of the highpass filter - see graph below, borrowed from one of Rod Elliott's articles:

Image
Left is the highpass filter operating into a resistive load, and right is with a "real" (simulated drive unit) load.

The Lowther is likely to be doing something like this to the simpler crossovers. The ones I worked up all put an L-pad around the Lowther, which will smooth out impedance variations a bit, so the bump at Fs wouldn't be so severe.
The fact that the lower mids improved with polarity swap of the Lowther just tells me that something else is wrong. James's Quasars did not need inverted polarity, he told me this when I was stuck at one point, he even went and double-checked on his own units. So mine, in the end, also did not need this. The problem with this 'fix' is that the upper mids and treble are always going to be out of phase with the bass, when it's finished.
The blend in the lower midrange and upper bass should occur when you get the right cap size in the series cross.
James used around 105uF with a 16R AER unit ; however his choke was bigger, at either 12mH or 15mH . Hence his crossover was more like 150Hz , rather than 200 or 220Hz as Nick and myself get with an 8.2mH choke. This implies you'd need to reduce the cap size, to get the low-pass effect on the midrange driver to be higher-up.
The PM6A on its own has a slightly greater bass extension , I concluded, I haven't got the figures to hand, but maybe it has a lower Fs. This also points to a smaller cap, to pull the bass down more.
When I was listening to Nick's units on the Saturday, with the drivers in the same polarity, I could tell there wasn't enough low bass compared to mine, and this can occur when there is too much bass coming from the midrange unit - it actually cancels the low bass from the Supravox since it's in antiphase. This is also 'fixed' by swapping polarity, but again, it's not the right (global) solution.

So, let's not get too carried away having to introduce LCR filters, and complicating the crossover in all sorts of ways. There were many good reasons for this approach, James thought about it a lot and wanted it to work this way. The point of the series crossover is to get a natural blend in tone and dynamics over about three octaves, rather than hearing the two different drivers. The PM6A is one of the closest current units to the old AER Mk.1, it should work fine.
"Once you find out ... the Circumstances ; then you can go out"
User avatar
Nick
Site Admin
Posts: 15705
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#192 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by Nick »

Useful info Mark. Basically my hope with these drivers is to work with them rather than trying to force them into submission. Time will tell.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
chris661
Shed dweller
Posts: 2556
Joined: Mon Feb 22, 2010 7:29 am
Location: Sheffield

#193 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by chris661 »

IslandPink wrote: Thu Sep 09, 2021 7:32 pm The fact that the lower mids improved with polarity swap of the Lowther just tells me that something else is wrong.
Mark,

I went back to XSim, which was the tool that helped me realise that inverted polarity improved the summation, and tried out every permutation of the crossover components that have been mentioned. In every case, there was more lower midrange & upper bass output with one driver inverted.

I must ask, then: did you or James try inverting one of the drivers? The difference was not subtle, and the improvement in tonal accuracy IMO trumps a very gradual phase shift to 180 degrees.


Having played around a little more in XSim, I've stumbled across something fairly interesting: if you throw out the crossover, wire the Lowther direct to the amp and put the 8.2mH inductor in series with the Supravox (ie, a parallel crossover, if a single component can be counted as a "crossover"), you get decent summation with all polarities positive. 10mH might be a small improvement - we can wire a switch across a 1.8mH inductor and try it both ways.
The down-side is that the Lowther is receiving full power at LF. I don't know how much of a problem that will be, as I don't have a lot of hands-on experience with these drivers.
My suspicion is that it'll be very subject to program material and SPLs, and Nick & Colin will have a better idea of how it might cope.


Chris
User avatar
IslandPink
Amstrad Tower of Power
Posts: 10041
Joined: Tue May 29, 2007 7:01 pm
Location: Denbigh, N.Wales

#194 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by IslandPink »

That's an interesting result !
That's minimalist..
chris661 wrote: Sat Sep 11, 2021 7:00 am Mark,

I went back to XSim, which was the tool that helped me realise that inverted polarity improved the summation, and tried out every permutation of the crossover components that have been mentioned. In every case, there was more lower midrange & upper bass output with one driver inverted.

I must ask, then: did you or James try inverting one of the drivers? The difference was not subtle, and the improvement in tonal accuracy IMO trumps a very gradual phase shift to 180 degrees.
I can't vouch for James, but I did have the driver polarity reversed for a while. This is when I was asking James about how he had them wired up. I couldn't get it to sound quite right, but some things were better ( can't say for sure it was the lower midrange - would have to go digging in the big thread ). In the end I did get a better overall result with more experimentation with cap size ( I documented this about how I looked more at the roll-off slope of the FX120's , then changed the cap size quite a bit ( bigger, in this case ) when it all started to gel, finally).

I did have a Quasar running with Supravox and FF225WK for a while and this sounded well-integrated with same driver polarities.

It may be that there's something specific about the PM6A downslope that is different, and can't work so well with the series cross.
Is there a difference, in your sims, in how the two drivers sum, if you compare between the first-order series cross and a first-order parallel cross ? I know that conventional theory will always show that two drivers with a parallel first-order cross will be in same polarity.

Plenty of experiments for Nick to do, then ! :)

ps. this frequency area is confusing, my measurements of the current FX120-based system show good SPL there, but to the ears, it's the weakest part of the presentation.
"Once you find out ... the Circumstances ; then you can go out"
chris661
Shed dweller
Posts: 2556
Joined: Mon Feb 22, 2010 7:29 am
Location: Sheffield

#195 Re: PERSPEX QUASAR ver 2.0 prototype build

Post by chris661 »

Thinking about it a little more, intuitively the above-mentioned inductor-only filter makes some sense: starting with the (large) assumption that each driver is approximately flat on an infinite baffle, we'd expect a -6dB/octave rolloff towards the bass in the region above Fs on the Lowther. In order to form a first-order crossover, the only thing that's missing is the first-order lowpass on the Supravox.

Reality is a bit more complicated, as always, but it would appear to work in this case. I'm at a gig at the moment (making sure the speakers don't melt too much), but will upload the simulated curves later.


IIRC, the first-order series vs parallel crossovers came out the same with equal crossover components. The home laptop has all the simulations on it, though, so I'll confirm later.
The second-order series filter was a very odd one to optimise - you have to think carefully about which components are contributing to which slope, and adjust so that the components arranged in that way (ie, you end up with two inductors in series, if you follow the correct current path) end up the right values. You then have to split the values to get the slopes in the right places.

I'd highly recommend downloading XSim (it's free) and playing around with some circuits.


Chris
Post Reply