|
Group Delay Variation - How much is too much?
For various transmission types, how much group delay variation can be
tolerated in the receiver before it causes problems recovering the original signal? I realize that it all depends ... I'm just interested in getting a rough idea. For example, at what point is: 1) SSB voice reception noticeably affected? 2) the ability to receive images from weather satellites affected? 3) various amateur radio modes such as PSK31, RTTY, and MFSK16 affected? 4) HD Radio specs 600 ns as the max for the transmitter, what's the max for the receiver? I'm curious since the group delay variation of SAW filters available for IF filtering seems to vary widely. -- john |
Group Delay Variation - How much is too much?
On Jun 24, 11:46*pm, wrote:
For various transmission types, how much group delay variation can be tolerated in the receiver before it causes problems recovering the original signal? *I realize that it all depends ... I'm just interested in getting a rough idea. For example, at what point is: * 1) SSB voice reception noticeably affected? * 2) the ability to receive images from weather satellites affected? * 3) various amateur radio modes such as PSK31, RTTY, and MFSK16 affected? * 4) HD Radio specs 600 ns as the max for the transmitter, what's the max for the receiver? I'm curious since the group delay variation of SAW filters available for IF filtering seems to vary widely. I have been playing with homebrew crystal filters (following W7ZOI and Bill Carver/K6OLG) for CW, as well as audio filters, and can tell you that on CW the difference between a super-sharp-in-frequency Chesbyshev filter (typical in ham equipment for a long time now) and a more constant-delay (e.g. Gaussian to 6dB or 12dB, or equiripple linear phase) filter is like night and day. My impressions are done "to my ears", not to a spectrum analyzer. The super sharp in frequency Chesbyshev filters have horrible horrible ringing especially on say 40M or 80M in the summer with the QRN. I can hardly listen for a few minutes without getting disgusted. BUT... they do have a real advantage during say a contest when there's competing signals every few hundred Hz. At the other extreme the constant-delay filters sound remarkably clear and transparent. They do not have such a sharp stopband, but my ear makes up for that most of the time. Ringing from summertime band noise is not nearly so tiring. Most of the filter simulation programs (AADE, SPICE, NatSemi's cool new WebBench filter tools, etc.) let you look at not just frequency response but also phase response and (maybe most importantly for summertime QRN) impulse response. Having run the simulations and listened with my ears to my experiments this summer, I cannot emphasize how much more enjoyable it is to use a Gaussian-to-6dB or equiripple linear phase filter in CW. Often when the bands are not crowded but there is QRN, I far far prefer a simple two-pole crystal filter designed for CW in the first place (e.g. my Heath HW-16) to any fancy-pants 8-pole or 12-pole modern filters in my new rigs. I notice you ask about a lot of digital modes but not CW. My ears have been listening to CW for 30-some years now and I can do a lot of processing in my brain. But what my brain cannot remove is horrible filter ringing. I don't know how those other digital modes stack up... maybe computers are better at removing horrible ringing than my brain. I don't think I have a "golden ear" or any other audiophile quality. In fact I'm pretty sure my ears are less good than they were when I was a kid doing CW. Tim N3QE |
Group Delay Variation - How much is too much?
On Jun 25, 9:45*am, Tim Shoppa wrote:
I have been playing with homebrew crystal filters (following W7ZOI and Bill Carver/K6OLG) for CW, as well as audio filters, and can tell you that on CW the difference between a super-sharp-in-frequency Chesbyshev filter (typical in ham equipment for a long time now) and a more constant-delay (e.g. Gaussian to 6dB or 12dB, or equiripple linear phase) filter is like night and day. You wouldn't happen to know the group delay variance of the filters you mentioned? Rough values are okay. I notice you ask about a lot of digital modes but not CW. My ears have been listening to CW for 30-some years now and I can do a lot of processing in my brain. But what my brain cannot remove is horrible filter ringing. I don't know how those other digital modes stack up... maybe computers are better at removing horrible ringing than my brain. I believe that in theory if the exact group delay profile is known, then a digital receiver can perform a certain amount of equalization. What I'm curious about is how much varience can be introduced by the filters in a receiver for various transmission types without needing to equalize. -- John |
Group Delay Variation - How much is too much?
On Jun 25, 2:14*pm, wrote:
On Jun 25, 9:45*am, Tim Shoppa wrote: I have been playing with homebrew crystal filters (following W7ZOI and Bill Carver/K6OLG) for CW, as well as audio filters, and can tell you that on CW the difference between a super-sharp-in-frequency Chesbyshev filter (typical in ham equipment for a long time now) and a more constant-delay (e.g. Gaussian to 6dB or 12dB, or equiripple linear phase) filter is like night and day. You wouldn't happen to know the group delay variance of the filters you mentioned? *Rough values are okay. I haven't actually measured group delay but I'm sure that what I hear is group delay. If you look at published group delay graphs for commercial Chesbyshev filters, a 500 Hz 8th-order crystal filter has a delay around 2ms in the middle of the passband, but within 100 Hz of the edge of the passband the delay peaks up enormously to 4ms and then back down again over the very steep skirt. Maybe you can turn the 2 ms variation into some inter-symbol/intra- symbol limit for some digital modes. It rings like the dickens when hit with QRN, that's for sure! Tim N3QE. |
Group Delay Variation - How much is too much?
test
|
Group Delay Variation - How much is too much?
|
Group Delay Variation - How much is too much?
On Jun 26, 3:32*am, Paul Keinanen wrote:
On Wed, 25 Jun 2008 11:14:03 -0700 (PDT), wrote: On Jun 25, 9:45*am, Tim Shoppa wrote: I have been playing with homebrew crystal filters (following W7ZOI and Bill Carver/K6OLG) for CW, as well as audio filters, and can tell you that on CW the difference between a super-sharp-in-frequency Chesbyshev filter (typical in ham equipment for a long time now) and a more constant-delay (e.g. Gaussian to 6dB or 12dB, or equiripple linear phase) filter is like night and day. You wouldn't happen to know the group delay variance of the filters you mentioned? *Rough values are okay. I notice you ask about a lot of digital modes but not CW. My ears have been listening to CW for 30-some years now and I can do a lot of processing in my brain. But what my brain cannot remove is horrible filter ringing. I don't know how those other digital modes stack up... maybe computers are better at removing horrible ringing than my brain. I believe that in theory if the exact group delay profile is known, then a digital receiver can perform a certain amount of equalization. *What I'm curious about is how much varience can be introduced by the filters in a receiver for various transmission types without needing to equalize. Excuse my ignorance, but why on earth do you do some crude analog filtering and then continue with digital filtering, in which you have much more alternatives ? The only reason that I can think about using sharp IF crystal filters is that the dynamic range of the following stages (product detector and ADC) is not sufficient. In a typical general coverage up converting receiver, the roofing filter will define the bandwidth the ADC must handle. Also some gain control (not necessary automatic) is needed to set the band noise well below one LSB (LF/MF vs VHF/UHF and antenna efficiency on LF). Even when designing an add-on unit for audio processing, why would anyone use the receiver CW filters apart from dynamic range issues ? I kind-of have the same questions too. For a homebrew project, having all the software-designed-radio complexity in addition to the tight- analog-filtering-from-DC-to-daylight complexity seems to just... make everything too complicated and not fun anymore. But the best ham receivers couple impressive front ends with effectively tight roofing filters with SDR aspects effectively (and quite usably) and I can see why someone would want to try their hands at their own competitive design. But, wow, it's a lot of effort. And a lot of ham receivers - especially the first and second generation designs - combined all these technologies into radios that are actually painfully complicated to use. (When the QST review starts contrasting menu option 73 submode 4 with menu option 105 submode 13, that's a real turn-off to me. At the same time, other younger operators just love that sort of complexity!) On the other hand, a truly simple analog front end (e.g. Softrock) combined with a computer is a hell of a lot of fun. You spend a lot more time looking at a computer screen and less listening but that's what some like. Tim N3QE. |
Group Delay Variation - How much is too much?
On Jun 26, 3:32*am, Paul Keinanen wrote:
Excuse my ignorance, but why on earth do you do some crude analog filtering and then continue with digital filtering, in which you have much more alternatives ? To undersample the signal it must be bandwidth limited which means some type of analog filtering. As long as filtering is necessary, it might as well be a narrow as the widest signal of interest and as sharp as possible so long as it's convenient and doesn't distort the signal too much. why would anyone use the receiver CW filters Probably a bit narrower than what I had in mind … I'm currently looking at 500 KHz wide SAW filters. |
Group Delay Variation - How much is too much?
wrote in message
... "As long as filtering is necessary, it might as well be a narrow as the widest signal of interest and as sharp as possible so long as it's convenient and doesn't distort the signal too much." "As sharp as possible" and "doesn't distort the signal too much" are somewhat conflicting goals: In general, the steeper the skirts of a filter, the more group delay variation you get there at the edges (hence, Butterworth has less group delay variation than Chebyshev which has less than Elliptic). Now, you can certainly account for this by widening the passband a bit and then perhaps using even steeper skirts, or you can compensate for it digitally if you can characterize it, but the main point here is that it does get rather complex -- hence the trend to have somewhat "looser" analog filters (and thus low group delay variation) and then do whatever you want digitally. |
Group Delay Variation - How much is too much?
On Jun 26, 12:41*pm, "Joel Koltner"
wrote: "As sharp as possible" and "doesn't distort the signal too much" are somewhat conflicting goals: Understood, part of the point of this thread which was to get an idea of how much group delay variance is acceptable for various types of transmissions without greatly impacting the quality of the received signal. hence the trend to have somewhat "looser" analog filters (and thus low group delay variation) and then do whatever you want digitally. Also understood, it's all about balance. Going narrow impacts group delay variance which distorts signal, going wide impacts dynamic range. Which still leaves me with the notion that you want to go as tight as reasonably possible and no tighter. With that in mind it sounds like what we've determined so far with regards to IF filtering is: transmission type receiver group delay variance ------------------------------------------------------------- CW should be less than 2 ms this is based on Tim Shoppa's posts which were to the point. Does anyone else have data to contribute? -- John |
Group Delay Variation - How much is too much?
On Jun 26, 11:34*am, wrote:
On Jun 26, 3:32*am, Paul Keinanen wrote: why would anyone use the receiver CW filters Probably a bit narrower than what I had in mind … *I'm currently looking at 500 KHz wide SAW filters. As a rough guesstimate, the group delay in a 500kHz wide filter will be 1/500,000 secs, or 2.0 microseconds. Now, depending on shoulder steepness the change in group delay might get to 2, 3, maybe even 5 times 2.0 microseconds. But even at 20 microseconds I don't think any of the HF digital modes you mentioned would be impacted. Most of my comments regarding group delay and ringing in filters were oriented towards narrowish (few kHz or less) filters. Wow, a HF receiver with a 500kHz SAW filter after the mixer. I don't have a clue what you're doing! I thought we were talking about HF receivers for common bandwidths! Tim N3QE |
Group Delay Variation - How much is too much?
On Jun 26, 2:40*pm, Tim Shoppa wrote:
But even at 20 microseconds I don't think any of the HF digital modes you mentioned would be impacted. Thanks ... that's the type of information I was curious about. Wow, a HF receiver with a 500kHz SAW filter after the mixer. I don't have a clue what you're doing! I'm "playing" with something resembling 0 - 175 Mhz up converted to 208 Mhz filtered using a GSM SAW filter sampled at the first IF using a 25 Msps 16 bit ADC. The silly width is because I'm interested in handling broadcast FM including RDS (among other things). I'm also interested in receiving satellite images which in some cases has a bandwidth of 150 Khz. -- John |
Group Delay Variation - How much is too much?
|
Group Delay Variation - How much is too much?
[Let's try this again ... that should be microseconds (us) not (is)]
As another random datapoint there's a MetOp document regarding the LRPT satellite transmissions which says: Frequency range * * * Group delay (kHz) * * * * * * * * * * * *variation (us) [0-40] * * * * * * * * * * * *+/- 2 [40-60] * * * * * * * * * * *+/- 5 -- John |
Group Delay Variation - How much is too much?
wrote in message
... "Does anyone else have data to contribute?" I don't, and I suspect that no one has done a comphrehensive survey of various popular (to hams) modulation formats and their sensitivity to group delay variations. Doing so would definitely be valuable -- it'd be a shoe-in for a QST or QEX article. |
Group Delay Variation - How much is too much?
On Jun 25, 6:45 am, Tim Shoppa wrote:
On Jun 24, 11:46 pm, wrote: For various transmission types, how much group delay variation can be tolerated in the receiver before it causes problems recovering the original signal? I realize that it all depends ... I'm just interested in getting a rough idea. For example, at what point is: 1) SSB voice reception noticeably affected? 2) the ability to receive images from weather satellites affected? 3) various amateur radio modes such as PSK31, RTTY, and MFSK16 affected? 4) HD Radio specs 600 ns as the max for the transmitter, what's the max for the receiver? I'm curious since the group delay variation of SAW filters available for IF filtering seems to vary widely. I have been playing with homebrew crystal filters (following W7ZOI and Bill Carver/K6OLG) for CW, as well as audio filters, and can tell you that on CW the difference between a super-sharp-in-frequency Chesbyshev filter (typical in ham equipment for a long time now) and a more constant-delay (e.g. Gaussian to 6dB or 12dB, or equiripple linear phase) filter is like night and day. My impressions are done "to my ears", not to a spectrum analyzer. The super sharp in frequency Chesbyshev filters have horrible horrible ringing especially on say 40M or 80M in the summer with the QRN. I can hardly listen for a few minutes without getting disgusted. BUT... they do have a real advantage during say a contest when there's competing signals every few hundred Hz. At the other extreme the constant-delay filters sound remarkably clear and transparent. They do not have such a sharp stopband, but my ear makes up for that most of the time. Ringing from summertime band noise is not nearly so tiring. Most of the filter simulation programs (AADE, SPICE, NatSemi's cool new WebBench filter tools, etc.) let you look at not just frequency response but also phase response and (maybe most importantly for summertime QRN) impulse response. Having run the simulations and listened with my ears to my experiments this summer, I cannot emphasize how much more enjoyable it is to use a Gaussian-to-6dB or equiripple linear phase filter in CW. Often when the bands are not crowded but there is QRN, I far far prefer a simple two-pole crystal filter designed for CW in the first place (e.g. my Heath HW-16) to any fancy-pants 8-pole or 12-pole modern filters in my new rigs. I notice you ask about a lot of digital modes but not CW. My ears have been listening to CW for 30-some years now and I can do a lot of processing in my brain. But what my brain cannot remove is horrible filter ringing. I don't know how those other digital modes stack up... maybe computers are better at removing horrible ringing than my brain. I don't think I have a "golden ear" or any other audiophile quality. In fact I'm pretty sure my ears are less good than they were when I was a kid doing CW. Tim N3QE Obviously, both the time and the frequency response are determined by the positions of the poles and zeros of a linear system (filter), but be a bit careful about equating group delay and ringing. It's easy to make an FIR digital filter that has constant group delay, but rings quite nicely. I suspect the people who design RF communications systems using modern modulation schemes know the answers to John's questions. There's probably another newsgroup where you'd get more answers. Or--run some simulations. I can imagine creating ideal signals in Scilab (or Matlab) and feeding them through various filters, and then demodulating them. It shouldn't be terribly difficult to do that, but I'm not volunteering. Cheers, Tom |
Group Delay Variation - How much is too much?
K7ITM wrote on Thurs, Jun 26 2008 2:45 pm
On Jun 25, 6:45 am, Tim Shoppa wrote: On Jun 24, 11:46 pm, wrote: I don't think I have a "golden ear" or any other audiophile quality. In fact I'm pretty sure my ears are less good than they were when I was a kid doing CW. Tim N3QE Obviously, both the time and the frequency response are determined by the positions of the poles and zeros of a linear system (filter), but be a bit careful about equating group delay and ringing. It's easy to make an FIR digital filter that has constant group delay, but rings quite nicely. 'Ringing' and group-delay over a passband are separate things but, in practice they are are related. In passive-component filters they are quite related. 'Ringing' phenomena can be investigated analytically in any time- domain circuit analysis program. I use LTSpice from National Semi- conductor...totally free for download and works on any PC. SPICE compatible, the source stimulus can be set as a pulse of several cycles with the rise-time, fall-time adjustable. If there is electronic-cause ringing, it will show up at the output. Every single passive-component filter has time delay. If the time delay is unequal across the passband, then one will hear the ringing. Such ringing isn't always a physical-electronic thing IN the filter but more in the way the human brain perceives sound. True high- fidelity electronic music systems will have near-equal time delay over its entire passband. Unfortunately, few, if any, of the first Hi-Fi systems makers ever published specifications on group-delay or delay of any kind. [excluding speakers, of course, since those and their interrelationship with a room are so acoustically variable that no common standard could be reasonably adopted] It is very hard to describe sound that is FELT rather than measured by instruments. As Tim said about CW use, ringing causes an actual discomfort. With wideband home music systems there is very little perceived 'ringing' but there exists 'quality' which can only be graded by comparison with another system as 'A-B' testing. The one that FEELS like it sounds best would be the winner. For non-audio use, such as in AM-PM ('QAM') modulation combinations in modems, inter-symbol distortion with/without an 'eye' display can grade things...and group-delay effects aren't the only things to blame there, lots of others in the total path. Simpler FSK systems such as single-channel teleprinter need concentrate on group-delay only over a passband about twice the frequency of its frequency shift between Mark and Space. Group- delay there shows up more on demodulated pulse edge ditortion. Ringing there can be seen easily at the edge transitions. A compromise there is to have group-delay greater beyond the needed passband limits resulting in rounded transitions; demodulator output can be shaped afterwards as desired. NTSC analog video examples do not really apply since the common ACTUAL bandwidth of most smaller TV sets was so limited (down to 1 MHz in some) that group-delay effects are hidden in the resulting video passband distortion of details on objects. Broadcast FM is hard to define in felt-quality since FM's 'quiet' spectrum use is relatively narrow. 'Loud' passages of music uses more spectrum space, thus group-delay effects over a passband are more pronounced on loud passages. ... I can imagine creating ideal signals in Scilab (or Matlab) and feeding them through various filters, and then demodulating them. It shouldn't be terribly difficult to do that, but I'm not volunteering. Machine-in, machine-out systems such as teleprinters and modems can be compared relatively easily with computer analysis programs. Regardless of the computer program, none can substitute for what is FELT in the ear-brain sensory system with acoustic input. I see a valid test as only A-B or A-B-C (or more) comparisons using the same audio or audio-modulation input. That is a LOT more work. Most of what I've done in that delay subject are locked into lab notebooks in corporate ownership and it involves many man-weeks of investigation. The best one can hope for, in my estimation, are general guides on limits from the many and varied radio services. 73, Len AF6AY |
Group Delay Variation - How much is too much?
It just occurred to me that many don't know what "group-delay"
is. It is physically-electronically a time delay from input to output that may vary depending on the input frequency versus the magnitude-phase response of the filter within its passband. Group-delay is defined as the difference in phase delay versus frequency difference over very small differences in frequency. Most analysis programs show that as a matter-of-course using frequency-domain analyses. Using around 50 or more different frequencies of a linear source sweep input will show the actual filter time-delay, input to output, that is very close to the real thing. 73, Len AF6AY |
Group Delay Variation - How much is too much?
On Jun 27, 2:16 pm, AF6AY wrote:
.... I use LTSpice from National Semi- conductor... My friends at Linear Technology will be most interested to hear that, Len. |
Group Delay Variation - How much is too much?
On Jun 27, 10:25�pm, K7ITM wrote:
On Jun 27, 2:16 pm, AF6AY wrote: ... I use LTSpice from National Semi- conductor... My friends at Linear Technology will be most interested to hear that, Len. OOOPS! My bad! :-) Yes, LINEAR TECHNOLOGY CORPORATION did LTSpice...an improvement over their previous SPICE derivative done to promote their switcher ICs. Apologies all around to Linear Technology friends. They make fine ICs. I also have, but have not tried TINA from Texas Instruments, another SPICE derivative. Not enough time here to try everything out that is available. TINA is also free for download but it is harder to get through their web pages to do so. I don't have any friends at TI or Linear or at National Semi, I just use their good products...just like I am currently using a couple of LM337s using National's appnote information in a slightly different way. 73, Len AF6AY |
| All times are GMT +1. The time now is 07:18 AM. |
Powered by vBulletin® Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.
RadioBanter.com