HF time delay circuit?

[mt490]

I'm under the impression that it can't be done without difficulty using opamp circuitry?

I'm trying to shoehorn a LeCleac'h quasi-optimal crossover onto one of my existing XO boards, but since a delay for the HF driver(s) are needed, it doesn't look to be a very practical thing to do in analogue. Mechanically it's not straightforward either since I'm not using a horn tweeter.

Is there a trick to it? Or am I finally going to have to bite the bullet and change over to DSP?

 

[bwaslo]

You'll probably have to go for DSP. HF delays get tougher, because more cycles are needed to be kept somewhere 'in transit' for a given number of milliseconds delay than at low frequency. If all you have is lumped Rs and Cs and gain blocks (opamp circuits), where is the delayed multi-cycle waveform between the time it goes in and it comes out? It takes lots of poles and zeroes to get anything very useful.
With DSP (and enough memory) it's easy -- just stuff the bits into one end of a long enough shift register and pull them out the other end after a sufficient waiting time.

 

[DcibeL]

You can add delay passively with a ladder delay network. See this design for an example of the implementation.

Small "delay" can also be had by using asymmetrical crossover slopes.

 

[PLB]

mt490,

With a sixth order allpass, using 3 opamps, it is possible to get a good quality constant delay of upto 85uS (=30mm), but I don't know if this is adequate for your needs. If it is, let me know and I'll post the schematic.
If it isn't, bwaslo's recommendation is probably the best one. If this is the case, then the DCX2496 or the miniDSP would be a very convenient way to go.

Regards

Peter

 

[mt490]

You can add delay passively with a ladder delay network. See this design for an example of the implementation.

Small "delay" can also be had by using asymmetrical crossover slopes.

Unfortunately this is not what I'm looking for as it has the same response as the line level active allpass filter. Whilst it's easy enough to delay the driver at the crossover frequency, the delay reduces as frequency goes up which means that the phase alignment is not presented as what the LeCleach crossover aims to achieve. Essentially I have to virtually move the tweeter away from me for all its frequencies, not just near the crossover.

mt490,

With a sixth order allpass, using 3 opamps, it is possible to get a good quality constant delay of upto 85uS (=30mm), but I don't know if this is adequate for your needs. If it is, let me know and I'll post the schematic.
If it isn't, bwaslo's recommendation is probably the best one. If this is the case, then the DCX2496 or the miniDSP would be a very convenient way to go.

Regards

Peter

According to the calculations I need 109us. It would be close, I could make up that difference with physical alignment.

I'm guessing this is made up of three 2nd order allpass networks?

My two main concerns with making a daisy chain are additional noise and space requirements.

 

[mt490]

ED- going insane

 

[speaker dave]

You can add delay passively with a ladder delay network. See this design for an example of the implementation.

Small "delay" can also be had by using asymmetrical crossover slopes.

Zaph achieves 90 degrees phase shift at crossover. 2 more orders of crossover slope would also achieve that, but I can believe that his solution was preferable in this case.

So, mt490, do you need actual broadband delay or just phase shift at the crossover point?

David S.

 

[mt490]

broadband - as I understand the tweeter has to appear as though it's 37mm further away than the woofer, meaning the delay needs to be constant for all frequencies up to 20kHz or greater.

for reference this is what I'm trying to implement

@PLB: After simulating some of the possibilities I'm not sure how you arrived at 85us with only a sixth-order allpass - certainly not all the way up to ~18kHz ?

The miniDSP is looking more and more enticing at the moment, certainly would be convenient having one around as an experimental platform even if I did decide to cut an opamp circuit for the final product.

 

[phase_accurate]

109 us should be feasible with an N-th order Bessel allpass.

I once ran a simu of a 100 us delay line with a bandwidth of 30 kHz. This was a 10th order Bessel Allpass. This 100 us could be easily expanded by 10 % at the expense of 10% bandwidth loss.

Regards

Charles

 

[PLB]

mt490,


"@PLB: After simulating some of the possibilities I'm not sure how you arrived at 85us with only a sixth-order allpass - certainly not all the way up to ~18kHz ?"


I think you may have made an error in your sims. I’ve attached group delay curves for the 6th order allpass I referred to in my last post. With the use of MC33079 opamps, you can obtain an “A” wtd S/N of 109dB wrt 2Vrms I/O.

Peter

 

[CharlieLaub]

active delay PCB in development

I'm under the impression that it can't be done without difficulty using opamp circuitry?

I'm trying to shoehorn a LeCleac'h quasi-optimal crossover onto one of my existing XO boards, but since a delay for the HF driver(s) are needed, it doesn't look to be a very practical thing to do in analogue. Mechanically it's not straightforward either since I'm not using a horn tweeter.

Is there a trick to it? Or am I finally going to have to bite the bullet and change over to DSP?

I'm currently developing a bunch of analog circuits for active loudspeakers that I will offer for sale as kits. Included in these is an analog delay board. Sticking with the principle of full adjustability of all parameters, it will offer continuously adjustable delay up to about 450 uS, when the corner frequency is 13k Hz. 450uS is about 6 inches of pathlength difference for sound in air. This is much better than what SL describes on his web page, because the group delay is completely flat up to almost 10k Hz, and as a result the delay is constant across the entire mid-to-tweeter crossover region. If one does not need such a high corner frequency (e.g. for delay of a midrange) you can reduce the corner frequency and increase the maximum amount of delay possible by the same factor, e.g. 10x reduction in corner frequency corresponds to 10x longer possible maximum delay.

I will be posting info on this circuit, and the others that I am developing, in this thread in the Analog Line Level forum. If you have some feedback, requests, or want more info, please post there.

-Charlie

 

[mt490]

I think you may have made an error in your sims. I’ve attached group delay curves for the 6th order allpass I referred to in my last post. With the use of MC33079 opamps, you can obtain an “A” wtd S/N of 109dB wrt 2Vrms I/O.

Peter

Thanks Peter,

Yeah I'm using FilterPro, I updated it and resimulated and it gave me much more favourable answers. Trouble is that it's a bit inflexible in terms of adjusting individual section F and Q so I can't quite get the results that you're getting there.

I probably should just request a peek at your values

With real world components how much (amplitude) ripple did you get in the pass band? My sims suggest that 0.4dB is a realistic best case.

 

[PLB]

Thanks Peter,

Yeah I'm using FilterPro, I updated it and resimulated and it gave me much more favourable answers. Trouble is that it's a bit inflexible in terms of adjusting individual section F and Q so I can't quite get the results that you're getting there.

I probably should just request a peek at your values

With real world components how much (amplitude) ripple did you get in the pass band? My sims suggest that 0.4dB is a realistic best case.

mt490,

Attached is the schematic, magnitude response and component sensitivity list. I have set the resistor values so that all capacitor values are the same. Changing the value of the capacitors results in a corresponding linear change in the group delay, which has a sensitivity of 25.75uS/1nF.
The input of this circuit should be driven from low impedance source, so I would suggest that if you use the MC33079 quad opamp, make the first stage a buffer.

Please note that I have not used the MC33079 in this circuit before now, so I don’t know if there are any HF stability issues that may arise. With this in mind, I would suggest that you breadboard it before you commit to a PCB layout. On the other hand, I have used NE5532 opamps in this circuit with no problems, provided the supply rails are adequately decoupled.

Regards

Peter

 

[mt490]

I will probably go for duals for flexibility in terms of types available. That and the next PCB I lay will probably have native SMD opamps rather than DIP style opamps.

Additionally I have tried to use the MC33078's before but subjectively the results were poorer than cheap stuff like the NE5532 and LM833. Back then I only really had cursory looks at what was going on in terms of decoupling on the scope though, so it might be that they're a bit more sensitive than the latter items.

ED- actually I remember now... they are definitely more sensitive, but I only ever had a problem with one particular veroboard circuit which had terrible layout and decoupling.