Hello Mike, Thanks for the reply. In your rough calculation you have specified the max. length of the bus to be about 4.8inches running at 50MHz but what about the length differnce among the various bus signals ?. Do you want to say that while routing the bus I need to adhere to 2 rules 1) The maximum length of the bus is constrained to about 4.8 inches and 2) The lenght of the bus signals may lie between 0 to 4.8 inches ?. Secondly, let assume that I am using series terminators with a tolerance of 5% to match the output resistance of the driver to that of transmission line (whose characteristic impedance has a tolerance of about 10% as specified by the fabricator) giving rise to the relection-coefficient of about 0.075 at the driver side. How much conservative should I be in estimating the "no. of round trips" for the bus to quiet down or to settle the signal within the range ?. Can I afford to scale the no.of round trips from 10 to let say 5 ?. Again many thanks for the reply, Adeel Malik -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Michael Khusid Sent: Saturday, August 17, 2002 3:19 AM To: 'Michael Smith ' Cc: 'si-list@xxxxxxxxxxxxx ' Subject: [SI-LIST] Re: Matched Length Constaint Approximation for a bus running between 20-50MHz >I have two questions related to reflective switching. First does reflective switching provide an easy termination scheme by only requiring series terminators at the drivers? This would drive half height waves down the line and full height would be achieved on the reflected pass. Many reflective switching topologies require no discrete resistors in their topologies at all. Sometimes, ASIC vendors place resistors on chip to slow down the drivers, and those series resistors serve to dissipate the energy. Otherwise, the energy dissipates through natural parasitics, so it takes several roundtrips for net to quiet down. >My second question is what do these schemes do for the clock. I wouldn't think you would want a half height clock traversing the line and hovering at the switching threshold. Do they drive a full height clock down the line? If not how do they avoid multiple clock edges? In multiple load reflective switching buses, clocks are normally kept point-to-point. That allows to use same half-height drivers, so that full height wave happens at the far end. So, if you have four devices, you would have four different clock nets of the similar length going to the device clock inputs. Sometimes designers would put two clock inputs at the net. It is acceptable as long as both inputs are at the far end of the line, less than a risetime apart. In this case, a rising edge looks a bit ugly, yet the multiple clock edges do not occur. Also, clock nets typically need terminations. Mike Khusid SI/HF Application Engineer Ansoft Corporation www.ansoft.com >Thanks in advance, >Michael Smith >Hardware Engineer >iZ Technology Corp. -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of Michael Khusid Sent: Friday, August 16, 2002 12:50 PM To: 'Adeel Malik ' Cc: 'si-list@xxxxxxxxxxxxx' Subject: [SI-LIST] Re: Matched Length Constaint Approximation for a bus running between 20-50MHz Adeel, The maximum flight time can be roughly calculated by formula: tflight_time < tperiod - tco - treceiver_setup - tcrosstalk - tjitter where tco is clock to output delay. The question is what is the flight time. Consider a bus with a driver and several receivers. Driver---Receiver1---Recever2--------------Recever3 Most 20-50MHz buses today use reflective switching, so receiver1 has to wait for signal to travel from Driver to Receiver3 and back all the way. That's one full round-trip time delay. Also, such slow speed buses have no terminions (eg. PCI33), so it takes a few round trips for reflections to die out. In a conservative design, consider 5 to 10 round trips for a bus to quiet down. So, let's do a calculation. tperiod = 20ns tco = 2ns treceiver_setup = 2ns (I am guessing) tcrosstalk = 1ns tjigger = 0.5ns (guessing on last two) tflight time < 14.5ns Let's say you want to be conservative and allow for 10 round trips. round_trip_delay = 14.5/10 = 1.45ns line_delay = round_trip_delay / 2 line_length = line_delay / speed = 0.5 * 1.45ns / 0.15 ns/inch = 4.83inches Note that this is a very rough calculation. It gets more much more interesting with Ts/stubs on the line, connectors and terminations. Besides, there are no good ways to estimate time delay caused by crosstalk on the line. If you want to be more accurate, I would recommend using simulation software. Mike Khusid SI/HF Application Engineer Ansoft Corporation www.ansoft.com -----Original Message----- From: Adeel Malik To: si-list@xxxxxxxxxxxxx Sent: 8/16/02 8:14 AM Subject: [SI-LIST] Matched Length Constaint Approximation for a bus running between 20-50MHz Hi All, In order to accurately calculate the maximum lenght difference to meet the setup and hold times among the bus signals (address, control and data) , one needs to find the flight time of the traces, clock-to-output delay of the Flip-Flops and other logic involved ,cycle-time period and other things...... But if someone is designing the bus for a 20-50MHz range, I don't think that there is any need to precisely calculate all the afore-mentioned parameters because I know that the cycle-time period for a bus running at e.g 50MHz is about 20ns while the delay of the outer-layer PCB Track is about 150 ps/inch and clock to output delay of the flip-flop in the memory is about 1-2ns leaving at least 10-15ns of time-margin. So can someone give me any crude approximation to determine the maximum lenght difference among the microstrip traces running in 20- 50MHz range. Regads, ADEEL MALIK, ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu